CHEMICAL HAZARD & SAFETY THE BASIC CHEMICAL HAZARD CLASSIFICATION.
Automated Methods in Chemical Risk Assessment€¦ · value. Chemical properties relevant to hazard...
Transcript of Automated Methods in Chemical Risk Assessment€¦ · value. Chemical properties relevant to hazard...
Automated Methods in Chemical Risk Assessment
by
Trevor N. Brown
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Chemistry University of Toronto
© Copyright by Trevor N. Brown 2011
ii
Automated Methods in Chemical Risk Assessment
Trevor N. Brown
Doctor of Philosophy
Chemistry
University of Toronto
2011
Abstract
In this thesis, topics in chemical hazard and risk assessment are explored through the use of
multimedia mass balance models and high-throughput chemical property prediction methods.
Chemical hazard metrics, as calculated by environmental fate and transport simulations, are
investigated to determine the validity of two common simplifying assumptions in the underlying
models; the use of octanol as a surrogate for organic matter and the use of environmental
parameters that do not vary in time. A major finding is that the use of these common simplifying
assumptions in multimedia mass balance models has little effect on chemical risk assessment,
provided that chemicals are ranked relative to each other rather than a predetermined cutoff
value. Chemical properties relevant to hazard and risk are collected, and applied in a large-scale
chemical hazard assessment to derive a short list of potential Arctic contaminants. Several
further data needs are identified; these are widely applicable and easily calculable metrics for
chemical biodegradation, toxicity and emissions. A new method of predicting chemical
properties is presented to assist in meeting these data needs. The method automatically creates
predictive, quantitative relationships between the structures and properties of chemicals that are
comparable to similar relationships created with expert judgement.
iii
Acknowledgments
First and foremost I would like to thank my supervisor Frank Wania; I am grateful for his
unfailing support and patience. I will always appreciate the excitement and interest he showed in
the projects we worked on together, and how he always made time to discuss ideas about
environmental chemistry, european travels and many other topics. I would also like to thank my
committee members Jamie Donaldson and Derek Muir for their support and interest over the
years, as well as the additional members of my defense committee George Arhonditsis, and my
external examiner Matt MacLeod for being interested and taking the time participate.
I would like to acknowledge the two CEFIC-LRI grants that helped to fund much of my Ph.D.
work, and thank the members of the RLT for their interest and understanding in the projects.
Special thanks especially to the chairmen and regular attendees at the annual project meetings
Andrew Riddle and Dolf van Wijk.
Thanks to my collaborators and coauthors for their insight, helpful comments and hard work.
Michael McLachlan, Knut Breivik, Kai-Uwe Goss, Gerjte Czub and Emma Undeman have my
gratitude for all their help and for making the annual meetings during the IMPS project so
enjoyable, with the addition of Jon Arnot and Xianming Zhang in the Screen-POP project.
I am privileged to have two of the most loving and supportive parents in the world. I will never
be able to thank them enough for their unwavering support through many, many years of school
with all of the related false starts, wrong turns and long distance moves. I am grateful for the
support of both my brothers, my sister inlaw, and far too many other relatives to mention here.
On the topic of family I‘d also like to thank the other mother figures in my life while in Toronto;
Ying Lei and Olga Petrovski for checking to make sure that I was happy, healthy and well-fed.
Over the last five years I have met so many great people I would be forced to add another chapter
here to mention them all by name. Thank you to both the UTSC and downtown chemistry grad
students I‘ve known over the years, especially members of the Wania Group, past, and present.
Thank you to my climbing partners who have been great friends and introduced me to a lifelong
passion, especially the original partners-in-climb Johnny, Erin and Magda.
Finally, thank you to Rachel Chang for greatly enriching my life and for being so incredibly
supportive during the writing of this thesis.
iv
Table of Contents
Acknowledgments .......................................................................................................................... iii
Table of Contents ........................................................................................................................... iv
List of Tables ............................................................................................................................... viii
List of Figures ................................................................................................................................ xi
Chapter 1 Introduction .................................................................................................................... 1
1 Chemical Hazard and Risk Assessment ..................................................................................... 2
2 Environmental, Bioaccumulation and Exposure Modelling ...................................................... 6
3 Automated Chemical Screening and Prioritization .................................................................. 10
4 Objectives and Outline ............................................................................................................. 12
Chapter 2 Screening Chemicals for the Potential to be Persistent Organic Pollutants: A Case
Study of Arctic Contaminants .................................................................................................. 14
1 Introduction .............................................................................................................................. 15
2 Methods .................................................................................................................................... 16
2.1 Outline of the Approach .................................................................................................... 16
2.2 Compilation of Data .......................................................................................................... 17
2.3 Screening for Long Range Transport to the Arctic and Bioaccumulation in Humans ..... 19
2.4 Screening for Persistence in Air ....................................................................................... 22
2.5 Screening for Resemblance with the Structural Profile of Known Arctic Contaminants . 22
3 Results and Discussion ............................................................................................................. 23
3.1 Preliminary Screening ....................................................................................................... 23
3.2 Chemicals Matching the Structural Profile of Known Arctic Contaminants ................... 25
3.3 Potential Arctic Contaminants .......................................................................................... 27
3.4 High Risk Chemicals ........................................................................................................ 28
4 Acknowledgements .................................................................................................................. 34
5 Appendix .................................................................................................................................. 34
5.1 Compiling a List of Known Arctic Contaminants ............................................................ 34
5.2 Profiling Arctic Contaminants .......................................................................................... 40
5.3 Chemical Profiling Criteria ............................................................................................... 41
5.4 Application of Chemical Profiles ...................................................................................... 42
5.5 Interpretation of Figure 9 .................................................................................................. 47
5.6 Definition of the POP score .............................................................................................. 49
5.7 Interpretation of the POP Score ........................................................................................ 50
v
Chapter 3 Development and Exploration of an Organic Contaminant Fate Model Using Poly-
Parameter Linear Free Energy Relationships ........................................................................... 85
1 Introduction .............................................................................................................................. 86
2 Methods .................................................................................................................................... 88
2.1 PP-LFER Equations .......................................................................................................... 89
2.2 Implementation of PP-LFERs in CoZMo-POP2. ............................................................. 90
2.3 Solute Descriptors. ............................................................................................................ 91
2.4 Model Parameters. ............................................................................................................ 93
2.5 Model Comparison. ........................................................................................................... 94
2.6 Linking Solute Descriptors to Chemical Fate. .................................................................. 95
3 Results and Discussion ............................................................................................................. 96
3.1 Comparison of KOC Values. .............................................................................................. 96
3.2 Comparison of KPA Values. .............................................................................................. 97
3.3 Comparison of Model Results. ......................................................................................... 98
3.4 Linking Solute Descriptors to Chemical Fate. ................................................................ 102
4 Acknowledgements ................................................................................................................ 105
5 Appendix ................................................................................................................................ 106
5.1 Derivation of Equations for Enthalpies of Phase Change ............................................... 106
5.1.1 Derivation of PP-LFER Equations for ΔHAW, ΔHOA and ΔHOW ........................ 106
5.1.2 Derivation of PP-LFER Equations for ΔHHA and ΔHHW .................................... 107
5.1.3 Derivation of a PP-LFER Equation for ΔHPA ..................................................... 107
5.2 Calculation of Z-values in CoZMo-POP 2 ..................................................................... 109
5.3 PP-LFER for Sorption to Forest Canopy ........................................................................ 111
5.4 PP-LFER Equation for Hexadecane/Air Partitioning ..................................................... 113
5.5 Dataset Reduction Method .............................................................................................. 114
5.6 Description of Model and Parameters ............................................................................. 124
5.7 Interpolation Method ...................................................................................................... 129
5.8 Validity of Showing Results in the Chemical Space ...................................................... 134
Chapter 4 Investigating the Time Resolution of Water and Organic Carbon Balances and
their Effect on Contaminant Fate Modeling ........................................................................... 147
1 Introduction ............................................................................................................................ 148
2 Methods .................................................................................................................................. 150
3 Results .................................................................................................................................... 156
4 Discussion and Conclusions ................................................................................................... 164
vi
5 Acknowledgements ................................................................................................................ 168
6 Appendix ................................................................................................................................ 169
6.1 Variable Mass Balances of Air, Organic Aerosol, Water and Particulate Organic
Carbon. ............................................................................................................................ 169
6.1.1 General Description. ........................................................................................... 169
6.1.2 Air Balance. ........................................................................................................ 170
6.1.3 Organic aerosol balance. ..................................................................................... 171
6.1.4 Water balance. ..................................................................................................... 172
6.1.5 POC balance. ....................................................................................................... 176
6.2 Meteorological Data and Climatic Scenario Definition. ................................................. 181
6.2.1 Data source and selection. ................................................................................... 181
6.2.2 Heat Transfer Model. .......................................................................................... 181
6.2.3 Insolation Model. ................................................................................................ 182
6.2.4 Precipitation Model. ............................................................................................ 184
6.2.5 Primary Productivity Model. .............................................................................. 185
6.2.6 Atmospheric OH Radical Concentration Model. ................................................ 186
6.3 Derivation of Empirical Relationship for Predicting Temperature Dependence. ........... 191
6.4 Definition of Environmental Half-lives. ......................................................................... 194
Chapter 5 Creating a Fragment-Based QSAR with Iterative Fragment Selection (IFS) ............ 199
1 Introduction ............................................................................................................................ 200
2 Methods .................................................................................................................................. 201
2.1 Data Processing ............................................................................................................... 201
2.2 Overview of the Method ................................................................................................. 202
2.3 Generation of Descriptors ............................................................................................... 202
2.4 Model Selection .............................................................................................................. 204
2.5 Goodness of Fit Metric ................................................................................................... 205
2.6 Fragment Selection Rules ............................................................................................... 206
2.7 Cross Validation .............................................................................................................. 207
2.8 External Validation ......................................................................................................... 210
2.9 Experimental Datasets .................................................................................................... 211
3 Results .................................................................................................................................... 212
3.1 Fitting and External Validation ....................................................................................... 212
3.2 Fragment Coefficients ..................................................................................................... 217
3.3 Fragment Selection ......................................................................................................... 218
vii
4 Discussion and Conclusion .................................................................................................... 219
4.1 Model Averaging and Model Uncertainty ...................................................................... 219
4.2 Comparison to Arnot et al. QSAR .................................................................................. 225
4.3 Fragment Analysis .......................................................................................................... 227
5 Acknowledgements ................................................................................................................ 231
6 Co-Author Contributions ....................................................................................................... 231
7 Appendix ................................................................................................................................ 232
Chapter 6 Conclusions ................................................................................................................ 264
1 Summary of Major Conclusions ............................................................................................ 265
2 Recommendations and Future Directions .............................................................................. 267
References ................................................................................................................................... 269
viii
List of Tables
Table 1 - Summary of highproduction volume chemicals which are predicted to become Arctic
contaminants or which match the structural profile of known Arctic contaminants. ................... 30
Table 2 - Names, CAS numbers, and the overall POP score for a list of known Arctic
contaminants ................................................................................................................................. 36
Table 3 - CAS number, chemical class, name and POP score of chemicals meeting the elevated
AC-BAP, atmospheric oxidation half-life, and POP score criteria. ............................................. 52
Table 4 - CAS number, chemical class, name and POP score of chemicals meeting the elevated
AC-BAP and atmospheric oxidation half-life criteria, but not meeting the POP score criteria. .. 65
Table 5 - Explanation of the Acronyms for Chemical Class Used in Tables S2 and S3 .............. 83
Table 6 - Chemical classifications of the chemicals which meet the elevated AC-BAP and
atmospheric oxidation half-life criteria. ........................................................................................ 84
Table 7 - Phase Descriptors for PP-LFER Equations Used in this Study. .................................... 90
Table 8 - Real Chemicals Dataset ............................................................................................... 115
Table 9 - Hypothetical Chemicals Dataset .................................................................................. 122
Table 10 - Solute Descriptor Inter-Correlations (R2): Real Chemicals ...................................... 124
Table 11 - Solute Descriptor Inter-Correlations (R2): Hypothetical Chemicals ......................... 124
Table 12 - Summary of Model Parameters ................................................................................. 126
Table 13 - Statistical Ratios of Interpolated Solute Descriptors ................................................. 138
Table 14 - Sub-sampling Statistics ............................................................................................. 146
Table 15 - Mass Balance Compartments and Fluxes: their Interdependence and Fitting. ......... 179
Table 16 - Relative Precipitation Intensities. .............................................................................. 188
Table 17 - Average, Minimum and Maximum Characteristic Travel Distances of Persistent and
Labile Chemicals at Hourly, Daily, Weekly, Monthly and Seasonal Environmental Parameter
Resolutions, in Kilometers. ......................................................................................................... 197
Table 18 - Summary of Statistics for Fitted and External Validation Datasets. ......................... 213
ix
Table 19 - 2D Fragment Based IFS-HLN QSAR Averaged from Three Cross-Validations. ...... 221
Table 20 - Chemicals with log HLN Residuals Greater than 1 in IFS-HLN and EPI HLN .......... 226
Table 21 - Qualitative Prediction Accuracy for Fish log HLN Fit on Dataset Splitting from this
Work Using Average Model from Three Different Cross-Validations. ..................................... 227
Table 22 - Comparison of Fragment Coefficients from IFS-HLN and EPI-HLN. ....................... 229
Table 23 - Training and Validation Datasets Summary. ............................................................. 232
Table 24 - 2D Fragment Based QSAR from IFS-HLN with a 1:2 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 245
Table 25 - 2D Fragment Based QSAR from IFS-HLN with a 1:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 246
Table 26 - 2D Fragment Based QSAR from IFS-HLN with a 2:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 248
Table 27 - 2D Fragment Based QSAR from IFS-EPI with a 1:2 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 250
Table 28 - 2D Fragment Based QSAR from IFS-EPI with a 1:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 251
Table 29 - 2D Fragment Based QSAR from IFS-EPI with a 2:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 253
Table 30 - 2D Fragment Based QSAR from IFS-yscr with a 1:2 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 255
Table 31 - 2D Fragment Based QSAR from IFS-yscr with a 1:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 256
Table 32 - 2D Fragment Based QSAR from IFS-yscr with a 2:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 257
Table 33 - 2D Fragment Based QSAR from IFS-KOW with a 1:2 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 259
Table 34 - 2D Fragment Based QSAR from IFS-KOW with a 1:1 Internal Training to Internal
Validation Ratio During Cross-Validation. ................................................................................ 260
x
Table 35 - 2D Fragment Based QSAR from IFS-KOW with a 2:1 Internal Training to Internal
Validation Ratio during Cross-Validation. ................................................................................. 262
xi
List of Figures
Figure 1. Schematic Diagram of the Factors Considered in Chemical Risk Assessment. .............. 3
Figure 2 – Chemical Space plot of elevated Arctic contaminant potential. .................................. 20
Figure 3 - Schematic diagram of the screening process. The 105,584 chemicals in the EPI Suite
database were screened for their potential to accumulate in the Arctic food chain. ..................... 24
Figure 4 - Locations of all chemicals in the EPISuite database which match the structural profile
of known Arctic contaminants. Plotted using the "HLC data set". ............................................... 26
Figure 5 - Plot of all 105,581 chemicals in the EPI Suite database. ............................................. 39
Figure 6 - Frquency distribution of the number of structural atoms (SA). distributions are shown
for five different chemical "populations"...................................................................................... 44
Figure 7 - Frequency distribution of the degree of halogenation (CX). distributions are shown for
five different chemical "populations" ........................................................................................... 45
Figure 8 - Frequency distribution of the degree of internal connectivity (XS). distributions are
shown for five different chemical "populations" .......................................................................... 46
Figure 9 - Mean and standard deviation of the total number of structural atoms (SA), degree of
halogenation (CX), and degree of internal connectivity (XS) for five different chemical
"populations". The deviation from zero is shown in orange ......................................................... 47
Figure 10 - Locations of the 1460 chemicals for which solute descriptors were obtained in the log
KOA/log KAW chemical space. Values for log KOA and log KAW are calculated using equation (9)
and the phase descriptors from Table 7. The reasonable domain of model applicability is shown
in red. ............................................................................................................................................ 92
Figure 11 - (A) Correlation between the log KOC values obtained by the SP-LFER equation and
values obtained from the PP-LFER obtained for humic acid. (B) Correlation between the logKPA
values obtained by the SP-LFER and values obtained from the PP-LFER obtained for whole
aerosol. .......................................................................................................................................... 97
xii
Figure 12 - Chemical space plots of phase distribution of the PP-LFER based model and the sum
of absolute differences in the percent phase distribution between the SP-LFER and PP-LFER
based models for each of the three emission scenarios. .............................................................. 100
Figure 13 - Vector plots of movement in the log KAW/log KOA chemical space that can be related
to the contributions of each phase descriptor. Vector magnitudes are the value of the upper
quartile minus the value of the lower quartile of the distribution of solute descriptors multiplied
by the corresponding phase descriptor. ....................................................................................... 103
Figure 14 - Linear Regressions of Plant Sorption with log KOA ................................................. 113
Figure 15 - Solute Descriptor Plots ............................................................................................. 137
Figure 16 - KOC and KPA Correlations of Hypothetical Chemicals ............................................. 138
Figure 17 - Götz et al. PP-LFER KPA versus SP-LFER KPA ....................................................... 139
Figure 18 - Aspvreten PP-LFER KPA versus SP-LFER KPA ....................................................... 139
Figure 19 - Chemical Space Plots for Hypothetical Chemicals .................................................. 140
Figure 20 - Relative Difference in Model Outputs for Air Emissions ........................................ 141
Figure 21 - Relative Difference in Model Outputs for Soil Emissions ....................................... 142
Figure 22 - Relative Difference in Model Outputs for Water Emissions ................................... 143
Figure 23 - Chemical Space Plots for Degrading Real Chemicals ............................................. 144
Figure 24 - Characteristic Travel Distance of Degrading Real Chemicals ................................. 145
Figure 25 - Schematic of the Major Pools and Fluxes of the Four Environmental Mass Balances
in CoZMoMAN-EV and their Interconnectivity. ....................................................................... 151
Figure 26 - Air temperature (A) and primary biological production rate in fresh water (B) at
hourly, daily, weekly, monthly, and seasonal resolutions. ......................................................... 154
Figure 27 - Percentage of Hourly Resolved Variability Captured with Daily, Weekly, Monthly
and Seasonal Resolved Data for Labile and Persistent Chemicals in all Model Compartments.
Each Cell is a Chemical Space Plot Showing the Fraction of Intra-Annual Variability Captured at
Each Point; Labile and Persistent Environmental Distributions are Provided as a Guide. ......... 160
xiii
Figure 28 - Hourly Characteristic Travel Distances (CTDs) in Kilometers for Labile (A-E) and
Persistent (F-J) Chemicals using Hourly (A&F), Daily (B&G), Weekly (C&H), Monthly (D&I),
and Seasonal (E&J) Resolved Environmental Parameters. ........................................................ 163
Figure 29 - Moles of Chemical in Air Simulated using Hourly Resolved Environmental
Parameters for Labile (A) and Persistent (B) Chemicals, Ranging from a log KOA of 4 in Red to
a log KOA of 10 in Blue, and a log KAW of -5 to 1. ................................................................. 164
Figure 30 - Hourly Resolved Air, Water and Soil Temperatures in Kelvin. .............................. 187
Figure 31 - Hourly Resolved Insolation and Corrected Insolation as Fraction of Maximum. ... 187
Figure 32 - Hourly Precipitation Amounts as a Fraction of Annual Precipitation. .................... 188
Figure 33 - Primary Biological Production Rate as a Fraction of Annual Production. .............. 189
Figure 34 - Time Averaging of Environmental Parameters Soil Temperature (A), Marine Water
Temperature (B), Marine Primary Biological Production Rate (C), Atmospheric OH Radical
Concentration (D), Rainfall (E) and Wind Speed (F). ................................................................ 190
Figure 35 - Correlation of Enthalpies of Phase Change in kJ/mole with the Logarithm of their
Corresponding Partition Coefficients. ........................................................................................ 193
Figure 36 - Annual Pattern of Variation in the Molar Amount of Chemical in Air, Cultivated
Soil, and Fresh Water for Hypothetical Labile Chemicals with the original CoZMoMAN Model
(Red Lines) and the New Model (Blue Lines) with Monthly Resolved Environmental Parameters.
..................................................................................................................................................... 195
Figure 37 - Annual Pattern of Variation in the Molar Amount of Chemical in Air, Cultivated
Soil, and Fresh Water for Hypothetical Persistent Chemicals with the original CoZMoMAN
Model (Red Lines) and the New Model (Blue Lines) with Monthly Resolved Environmental
Parameters. .................................................................................................................................. 196
Figure 38 - Extrapolation to Steady State Characteristic Travel Distance for Labile Chemicals
(A) and Persistent Chemicals (B). .............................................................................................. 198
Figure 39 - Model fit and predictions from IFS-HLN for (A,B) 1:2, (C,D) 1:1, and (E,F) 2:1 ratios
between internal training and internal validation datasets during cross-validation. ................... 214
Figure 40 - Model fit and predictions for the averaged IFS HLN QSAR for training (A) and
external validation (B) dataset. ................................................................................................... 224
xiv
Figure 41 - Model fit and predictions for fish HLN using IFS-EPI for (A,B) 1:2, (C,D) 1:1, and
(E,F) 2:1 ratios between internal training and internal validation datasets during cross-validation.
..................................................................................................................................................... 242
Figure 42 - Model fit and predictions for fish HLN using IFS-yscr for (A,B) 1:2, (C,D) 1:1, and
(E,F) 2:1 ratios between internal training and internal validation datasets during cross-validation.
..................................................................................................................................................... 243
Figure 43 - Model fit and predictions for log KOW using IFS-KOW for (A,B) 1:2, (C,D) 1:1, and
(E,F) 2:1 ratios between internal training and internal validation datasets during cross-validation.
..................................................................................................................................................... 244
1
Chapter 1
Introduction
2
1 Chemical Hazard and Risk Assessment
A large and ever-increasing number of chemicals are used in commerce; the chemical
abstracts service (CAS) reports more than 282 thousand inventoried or regulated substances,1
and the OECD reported that member countries produced more than 4600 chemicals in
quantities of 1000 tonnes or more in 2007.2 It is impossible to determine the exact number of
chemicals with significant commercial use by searching publicly available data, because
many countries do not gather data on chemical production, or report production to the
OECD. Some of these chemicals are likely to pose a human health risk, but chemical risk
assessment is a lengthy, data intensive process. Many factors must be considered when
determining if a chemical is a human health risk; human exposure to a chemical is a function
of the amount produced, the usage pattern, the route of exposure, and the properties of the
chemical.3 If the potential for human exposure exists this still only constitutes a hazard. In
order to translate into risk the chemical must additionally be toxic and have properties that
allow it to resist metabolism and accumulate to dangerous levels in human tissues.4
Furthermore, if a chemical does not pose a human health risk it may still be an environmental
risk, which poses the daunting challenge of assessing the risk for a myriad of different
organisms in different environments. At the same time there has never been more public
pressure on governments to protect their citizens and the environment from the adverse
effects of chemical pollution, and regulators have responded at the national and international
level with new chemical regulations and risk assessment programs.5-7
To ensure the success
of these programs regulators have called on academia and industry to assist them with new
sources of data and new tools to allow for rapid screening and prioritizing of chemicals that
may be a human health or environmental risk.
Figure 1 shows a simplified schematic of the factors that must be considered when
performing a chemical risk assessment for a single anthropogenic parent chemical acting on a
single target organism. The schematic approximately follows the course of a chemical as it
moves through the environment and into an ecosystem, and covers most of the fields that
make up environmental organic chemistry. In order to constitute a risk a chemical must make
its way from Production through the schematic to Adverse Effects. Assessing a chemical for
only a subsection of steps between Production and Adverse Effects is a hazard based
assessment rather than a risk based assessment. There are multiple pathways for a chemical
3
to follow on its way from Production to Adverse Effects, some short and others more
circuituous, but all of them move through Exposure, and end in Adverse Effects. These two
factors make up the accepted definition of risk and form the basis of various risk metrics,
where risk is a multiplicative function of Exposure and Adverse Effects measured or
predicted in comparable units.3
Figure 1. Schematic Diagram of the Factors Considered in Chemical Risk Assessment.
4
Chemicals which take the shortest path in Figure 1 from Production to Adverse Effects;
flowing through Usage and Emission directly to Exposure, and then directly from Exposure
to Adverse Effects, are of course the most regulated chemicals, i.e. acutely toxic chemicals
with the potential for direct consumer exposure.8 In the past any chemical with either one of
these attributes, a high potential for direct consumer or occupational exposure (―near-field‖
exposure) or acute toxicity, were more closely regulated than those chemicals where
exposure occurs indirectly through the environment (―far-field‖ exposure) or those chemicals
which bioaccumulate to a threshold level before adverse effects are observed.8 There have
been relatively recent changes in this regard though as international agreements such as the
Stockholm Convention,5 and national programs such as REACH in the European Union
6
have been brought into force seeking to regulate chemicals taking a more circuituous path
from Production to Adverse Effects. These regulatory efforts fall into two related categories
that seek to regulate POPs (Persistent Organic Pollutants) or PBTs (Persistent,
Bioaccumulative and Toxic chemicals). Both paradigms seek to regulate chemicals that are
persistent in the environment, bioaccumulate in target organisms and have adverse effects,
with the POPs oriented regulatory efforts further seeking to regulate chemicals susceptible to
long range transport.9,10
This combination of properties creates the potential for chemicals to
follow the more circuituous path in Figure 1 through Environmental Fate and Transport to
Exposure, and then through Bioaccumulation to Adverse Effects.
If a chemical is determined to be a risk then the aim of regulators is to cut the flow of
chemical off somewhere in the schematic of Figure 1 before it reaches Adverse Effects.
Regulatory action is, however, restricted to acting within the first three steps, Production,
Usage and Emission, and Environmental Fate and Transport by banning production, banning
or restricting further use and by requiring remediation of environmental contamination,
respectively.8 If the only route of exposure is direct exposure from usage then a usage ban is
effective in mitigating the risk a chemical poses. For chemicals with significant
environmental emissions however shutting the process down at the production or usage steps
may not eliminate the risk. If a chemical is persistent in the environment then there will be a
remaining pool of chemical causing continued exposure. Additionally, if a chemical is
bioaccumulative then target organisms will have a pool of the chemical stored in their bodies
5
that may continue to cause adverse effects even if regulation is effective in halting exposure.
Some classic examples of these complications are the legacy POPs targeted by the
Stockholm Convention such as PCBs (polychlorinated biphenyls).11
At every one of the stages included in Figure 1 the properties of the chemical that is being
assessed for risk play a role. Chemicals are produced and used because they have properties
that are useful. Some examples are PCBs, produced and used because of their high resistance
to heat and stress,12
perfluorinated surfactants and polymers, produced for their non-stick and
stain-repellent properties,13
and chlorinated pesticides, of course, were produced and used
specifically because they were toxic to certain organisms, and persistence was desirable
because it meant that they were effective for longer. As the previous examples demonstrate,
it is sometimes the very properties that make a chemical useful that also make the chemical a
risk. Chemical Emissions, and Environmental Fate and Transport are dependent on the
interaction between chemical properties, the usage of the chemical, and the properties of the
environment. How a chemical distributes itself among environmentally relevant phases and
resists degradation by physical processes and native microbes are important factors in
determining how the chemical enters the environment, where it goes and how long it stays
there.14,15
Exposure, Bioaccumulation and Adverse Effects are also functions of chemical
properties, and of the characteristics of the target organism. Chemical properties and the
presence of specific functional groups play a role in determining the absorption efficiency of
chemicals through different exposure routes, resistance to metabolism and toxicity.4
The direct pathways in Figure 1; direct exposure to chemical from usage and acute toxicity,
are only loosely linked to chemical properties or are very difficult to predict from known
chemical properties. Direct exposure can instead be estimated from industry and usage data,
assuming that these data exist and are accessible,16
and while it is difficult to predict which
organisms will experience acute toxicity from which chemicals the effect is often easily
observable and can be attributed to a specific chemical if the route of exposure is
determined.4 However, the pathway followed by POPs and PBT substances through Figure 1
is intrinsically linked to their properties; namely resistance to environmental degradation and
6
metabolism and the correct combination of partitioning properties that allow for
environmental mobility, high exposure potential and bioaccumulation.17
If a chemical has
properties consistent with those of other POPs and PBTs it may be considered a candidate for
POP or PBT related regulation.18
Risk assessment is, however, rarely an easy task because Figure 1, while a useful
demonstrative tool, is a gross oversimplification of all the factors involved in chemical risk
assessment. Many of the difficulties in chemical risk assessment are due to large possible
variations in chemical usage, emissions, environmental effects,19
and the behaviour and
physiology of the target organism.20
Additionally, many factors are not so clearly delineated
as Figure 1 suggests; for example in many cases environmental degradation or metabolism in
target organisms can transform a chemical into secondary products that may be more or less
harmful than the parent chemical, breaking the chain between Production and Adverse
Effects.21,22
Bioaccumulation comprises far more than chemical uptake in a single organism;
to accurately determine the chemical concentration in a single organism its food chain and
other effects such as maternal transfer must be considered, so bioaccumulation involves
really a series of organisms being exposed to a chemical.23
By far the largest difficulty in
chemical risk assessment though is simply gathering enough information to make an
informed decision. Experimental measurement of chemicals in the environment and in target
organisms has become routine and sensitive, but gathering experimental data at every point
in Figure 1 is a lengthy expensive process, especially if large lists of chemicals need to be
assessed for risk. For this reason many regulatory efforts have turned to the use of
environmental, bioaccumulation and exposure models to assist in predicting if chemicals
constitute a risk.3,24-26
2 Environmental, Bioaccumulation and Exposure Modelling
Environmental chemical fate and transport models are tools used to understand, explain and
predict the movement of chemicals through the environment. Foodweb bioaccumulation
models are similarly applied to study the accumulation of chemicals in biota, and together
these computer models are used to assess the potential exposure of humans, and other
7
organisms, to chemical contaminants.3,24
Environmental models explicitly consider the major
storage reservoirs of chemicals, also referred to as sinks or pools, and movements, referred to
as transport, or dispersion. Long term storage reserviors of organic chemicals may be soils,
sediments or deep waters, and typical movements are by diffusion, or in moving air and
water which act as carriers.14
Bioaccumulation models also consider storage reservoirs and
movements of chemical within an organism; with fatty tissues generally acting as storage
reservoirs and blood circulation, respiration, ingestion and egestion being typical movements
of chemical within an organism.23
Models are attractive tools in chemical risk assessement
because they help to fill in knowledge gaps about the environmental transport, fate and
bioaccumulation of chemicals using information about the chemical‘s properties and
parameters describing the relevant storage reservoirs and movements in the environment and
in organisms of interest, without the need for direct measurements of chemical
concentrations in the environment and biota.
Some of the most useful tools in assessing environmental exposure potential are multimedia
mass balance models which simulate the transport and fate of chemicals in the physical
(abiotic) environment.24,27
Multimedia models divide the environment into compartments,
with typical compartments including the atmosphere, aquatic environments, sediments, soil
and possibly plant matter. Each compartment is composed of one or more subcompartments
made up of pure phases such as air, water, mineral matter and organic matter. A compartment
is essentially a unit of the world which is assumed to be homogenous in both its physical
properties and the distribution of chemical within it, meaning that chemical equilibrium is
assumed within compartments. Compartments vary in size depending on the spatial
resolution of the environment being modelled. Although equilibrium is assumed within a
compartment this may not be the case in different parts of the environment depending on the
complexity of the model. Using the nomenclature of Mackay,15
environmental models may
be Level I, II, III, or IV models, with Level I models assuming equilibrium in a closed
system with no variance in time; and at the other end of the spectrum are Level IV models
that employ a non-equilibrium, non-steady state, time-variant description of the chemical
mass balance. Chemicals move between compartments either by diffusion or in a carrier
phase, and these movements, along with emissions and degradation are called fluxes. The
8
mass balance of chemical in a multimedia model essentially keeps track of the amount of
chemical in each compartment of the model and adjusts the amounts by summing the fluxes
which add or remove chemical in each of them.
Many mass balance models use fugacity as an equilibrium criterion to descibe how chemicals
partition between different parts of the environment and are often referred to as Mackay
models, after the author of the first manuscript that described their definition and use.28
Fugacity is somtimes referred to as the ―escaping tendency‖, because chemicals tend to have
a net movement from areas of high fugacity to areas of low fugacity, and is linearly related to
the concentration as shown in equation (1), where C is the molar concentration, f is the
fugacity in Pa, and Z is the fugacity capacity in mol m-3
Pa-1
.
C = f · Z (1)
When a chemical is at equilibrium between two locations its fugacity will be equivalent, so
the assumption of equilibrium within compartments in mass balance models is an assumption
of equi-fugacity, and fugacities between compartments will be different if the compartments
are not at equilibrium. Using fugacity as an equilibrium criterion simplifies the formulation
of the mass balance equations by removing the need for a partitioning coefficient between
each pair of compartments. The fugacity capacity, or Z value, essentially represents one half
of a partition coefficient, so the mass balance for each compartment can be calculated
separately with its fugacity and Z value instead of needing to simultaneously calculate the
equilibrium condition between every pair of comparments.15
Chemical fluxes are calculated
as shown in equation (2), where N is the flux in mol h-1
, and D is in units of mol Pa-1
h-1
.
N = f · D (2)
The calculation of the D values varies depending on the type of flux, diffusive exchange,
degradation or advection, but are all directly additive in the mass balance equation if they
occur in parallel and apply to phases having equivalent fugacities. In each case the relevant
fugacity is used to calculate the flux in moles per hour; the fugacity of the chemical in the
compartment in the case of degradation and the fugacity of the chemical in the source
9
compartment in the case of transport fluxes.15
Mass balances are constructed for each
compartment using the fluxes, concentrations, and compartment volumes.
Mass balance models of bioaccumulation and food webs work in much the same way as they
do for the abiotic environment.29
Each organism may be treated as a compartment, with sub-
compartments composed of water, lipids and important tissues in equilibrium. A more
sophisticated model may treat each tissue as a compartment on its own and calculate fluxes
between the tissues via blood flow. An important distinction however is that instead of
chemical emissions, as in the environmental mass balance models, biota experience chemical
exposure. Exposure is through three main routes; respiration (either air or water), ingestion of
water and food (other biota in the case of a food web), and by dermal contact.30
Fluxes of
chemical losses from biota are through respiration, egestion and metabolic biotransformation.
The concentration of chemical in biota may also be reduced by growth dilution, where the
amount of chemical remains constant but the organism grows to a larger size.29
Human
chemical exposure and bioaccumulation modelling is especially of interest for the purposes
of chemical risk assessment, and a number of models exist to predict the extent of chemical
exposure from the abiotic environment and the human food chain.23,31-33
Recent research
efforts have worked towards linking abiotic and human exposure mass balance models.
These linked models allow a chemical to be traced from the point of emission, through the
environment to exposure and simulate the bioaccumulation in the food chain and
humans,25,31,34
essentially covering most of Figure 1 in a single model.
There is one primary difficulty with using mass balance models for chemical risk assessment;
not surprisingly this difficulty is a lack of information. If a chemical is being assessed for risk
it is likely to be a relatively new chemical, although this is not always the case, and therefore
the chemical property data required to assess the chemical for every step in Figure 1 is
unlikely to be available or will be uncertain. Parameters describing the influence of the
physical environment and bioaccumulation may also be unavailable or uncertain. Many
simplifying assumptions have been made in both the abiotic and bioaccumulation models
with regards to the sorption capacity of various phases and the variability, both spatially and
10
temporally, in many parameters.14,15
These assumptions simplify the data requirements and
avoid unnecessary model complexity, but the effect of these simplifications on their
effectiveness as risk assessment tools is not always known.35
3 Automated Chemical Screening and Prioritization
Environmental, exposure and bioaccumulation models are used to calculate a number of
chemical hazard metrics, which along with data on adverse effects contibute to chemical risk
assessments. Hazard metrics calculated from the abiotic environmental models fall under two
main categories; long range transport metrics,36,37
and environmental persistence metrics. 37,38
Long range transport metrics give a measure of the exposure potential in locations that are
remote from the point of emission, which is important for the international regulation of
chemicals such as that called for by the Stockholm Convention.5 Environmental persistence
metrics measure how long a chemical will remain in the environment; a long environmental
residence time means that exposure will continue long after a chemical‘s use has been
discontinued due to storage in environmental reservoirs.11
Bioaccumulation and food web
models can be used to calculate a chemical‘s bioaccumulation potential from different routes
of exposure and with time variable dosing regimes. Knowing the peak and average internal
chemical concentrations (doses) is a useful piece of information when judging the possibility
of adverse effects.39
Linked environmental and bioaccumulation models can be used to
calculate overall exposure metrics such as intake fractions and human body burdens that
directly relate the amount of chemical in an organism to the amount of emissions released
into the environment.26,31,34,40
Chemical risk assessment is a difficult, data-intensive process that is expensive both in terms
of time and in terms of the tests which must be performed for proper assessment. Because of
the large number of chemical that are used in commerce it is virtually impossible to perform
a detailed risk assessment on all of them. This has led to the development of chemical hazard
screening and prioritization methods which are used to reduce the length of chemical lists by
automatically filtering out chemicals which are likely to be a low risk based on their hazard
metrics. A number of tools exist which can perform automated hazard metric screening on
11
chemicals given a set of chemical properties or a chemical structure.41,42
A common practice
is to set cut-off values for specific chemical properties or hazard metrics and focus only on
chemicals that meet these criteria; a good example is the logarithm of the octanol-water
partition coefficient (log KOW) being screened for values greater than five as a surrogate for
bioaccumulation potential.43
Chemicals may also be ranked by their properties or hazard
metrics, with the top-priority chemicals being those with the hazard metrics most likely to
make them the highest risk to human or environmental health.3,26
These methods are not
without problems however, because if data inputs for chemical properties are unavailable it
may not be possible to prioritize some chemicals, and if the properties are available but
uncertain there may be false negatives and false positives for chemical screening methods.
There are many chemical property prediction methods available to deal with the problem of
missing properties when performing a chemical risk assessment. Paritioning of chemicals
between environmentally relevant phases can be predicted by a large number of methods
with a wide range of sophistication. Sorption of chemicals to organic matter in the
environment and biota can be predicted with single-parameter regressions with partitioning
into n-octanol,44
with multi-parameter empirical equations,45
or with quantum chemical
calculations.46
Paritioning of a chemical between its pure form, octanol, water and air also
have a variety of prediction methods available, some also quite sophisticated such as
qunatum chemical predictions,47
and others quite simplistic such as quantitative structure
property relationships (QSPRs) that predict chemical properties based only on the 2D
structure of a chemical.41
Property prediction methods also exist for degradation in the
environment,48
biotransformation in fish,49
and adverse effects,50
but these methods are
generally more simplistic and more uncertain due to the chemical-specific nature of these
properties and the difficulty in measuring them experimentally. The prediction methods
mentioned here have varying degrees of uncertain. Unfortunately, the usual trade-off for
chemical property prediction methods is that the faster and easier to use methods are by far
the least accurate. Two very important parameters related to the chemical cannot be predicted
from chemical structure at all, these are the production volume and the usage profile of the
chemical. These data must be gathered from the few publically available databases, solicited
12
from industry, or estimated; if none of these options are possible then the exact level of
exposure cannot be known, and only a hazard-based assessment is possible.
4 Objectives and Outline
The overall objective of this thesis is to explore the feasibility of automated calculation of
hazard metrics for chemicals that have little existing data available, and to explore the
uncertainties involved in the models used to calculate these hazard metrics. Chapter 2 of this
thesis is a chemical screening exercise which involves a preliminary chemical hazard
assessment of a large list of chemicals, evaluating them for the potential to be contaminants
that fit into the POP paradigm and that may be found in the Arctic environment. Important
issues related to chemical risk that are discussed in Chapter 2 are the lack of publicly
availabile chemical production values, the uncertainties associated with using simplistic
chemical property predictions, chemical hazard screening, and the hazardous attributes of
POPs.
Chapter 3 and Chapter 4 explore the uncertainties associated with the application of
multimedia mass balance models due to the description of environmental phase distribution,
and the description of environmental temporal variability, respectively. Essentially, there
exist more sophisticated methods for describing environmental distribution and
environmental variability than are commonly used in multimedia models. Incorporating these
more sophisticated methods may improve the accuracy of the model outputs, but there is a
corresponding increase in the complexity of the model and in the data required to run the
model. Chapter 3 and Chapter 4 examine if this trade-off is worth making in the context of
automated chemical risk and hazard assessment.
As discussed throughout the Introduction, one of the greatest challenges when performing a
chemical risk or hazard assessment is often the lack of information regarding a chemical, its
uses, how it behaves in the environment and in biota. Chapter 5 of this thesis is a first step
towards providing another chemical property prediction tool intended to provide information
13
required for high-throughput chemical hazard and risk assessments. This tool is first applied
to the problem of predicting rates of metabolic transformation of chemicals, because this has
been shown to have a large potential impact on the human health risk of some chemicals.
14
Chapter 2
Screening Chemicals for the Potential to be Persistent Organic
Pollutants: A Case Study of Arctic Contaminants
Trevor N. Brown, Frank Wania
Environmental Science & Technology 2008 42 (14), 5202-5209
Reproduced with permission from Environmental Science and Technology
Copyright 2008 American Chemical Society
15
1 Introduction Considering the large number of chemicals in commerce, the complication of varied and
unknown chemical degradation pathways in the environment, and the highly selective nature
of most analytical detection systems, it is conceivable, if not highly likely, that the majority
of contaminants present in the environment, in wildlife and in humans, remains unidentified.
Even though there are more than 82,000 substances covered by the US Toxic Substances
Control Act (TSCA)51
, the US Centers for Disease Control monitor less than 120 substances
in human blood52
. Contaminants whose environmental occurrence has not previously been
documented are neither detected, studied, monitored, nor evaluated, let alone regulated. For
example, even though synthesized and used for more than half a century, the ubiquitous
presence of perfluorinated alkyl-compounds in people, wildlife and the physical environment
was not recognized until a decade ago. Now these substances are of the highest priority to
chemical regulators around the world 53
. Discovery of previously undetected environmental
contaminants is rarely the result of a directed search, but often owes much to serendipity.
Clearly, there is a need for a more rational approach with a greater promise to succeed in
identifying undetected contaminants.
Based on a quantitative understanding of the relationship between chemical properties and
environmental behavior, it is now becoming possible to identify the properties of substances
that qualify as contaminants of concern. For example, we now can predict what chemical
characteristics make an organic substance susceptible to accumulation in the Arctic physical
environment54
. Similar approaches seek to constrain the properties of substances
bioaccumulating in human food chains23
. It should thus be possible to screen the multitude of
chemicals of commerce for those that do possess such properties, and which therefore are
most likely to be found in environmental and human tissue samples.
Prioritizing chemicals with respect to environmental concern has been an ongoing effort for
several decades. Muir and Howard55
provide an in-depth review of screening methods and
projects aimed specifically at identifying persistent organic pollutants (POPs). The
motivation for such screening has intensified in response to the United Nations Environment
16
Program‘s Stockholm Convention on POPs5, which calls for the identification and global
regulation of chemicals that fulfill the four criteria of persistence, bioaccumulation, long
range transport and toxicity. Nearly all screening initiatives include parameters that address
these criteria when searching for new potential POPs55
.
The task of developing an effective screening method is difficult and complex. A method
must be highly discriminatory to eliminate the many chemicals that are of little concern, and
must simultaneously identify those chemicals which constitute a large hazard. In order to be
considered effective a method must demonstrate a strong selectivity for those chemicals
which cause the highest concern, capturing as many high risk chemicals and as few low risk
chemicals as possible. We present here an approach which is specifically designed to be
highly selective for chemicals which may become Arctic contaminants. The rationale is that
traditional dietary habits place Northern indigenous residents among the human
subpopulations most vulnerable to elevated exposure to POPs40
, and many of the traditional
POPs are of especially high concern to this specific population.
2 Methods
2.1 Outline of the Approach
The approach consists of two parallel screening methodologies; one methodology screens
chemicals based on substance properties and the other screens chemicals based on a
structural profile of known Arctic contaminants. Since substance properties are determined
by molecular structure, both methodologies might be expected to flag the same chemicals as
being of high concern. However, comparing the sets of substances identified by either
methodology can provide valuable insight into their relative strengths and limitations.
The pathway a chemical must follow to become a contaminant in the upper trophic levels of
the Arctic food chain is circuitous; emission, long range transport, deposition, exposure,
bioconcentration and biomagnification all play a role. The primary screening methodology
applied here, and in most screening exercises, examines each step along this pathway and
17
defines a set of criteria which chemicals must meet. Here we rely on three such criteria; the
chemical must have distribution properties that allow it to reach the Arctic marine food
chain, it must have distribution properties that allow it to bioaccumulate in humans, and it
must be sufficiently persistent in the atmosphere to reach the Arctic without degrading. This
methodology rests on three main assumptions; the first is that the potential to accumulate in
the Arctic marine food chain depends on chemical distribution, meaning chemicals with
certain combinations of partitioning properties will have a higher potential to accumulate,
and the second assumption is that long range transport happens mainly via the atmosphere;
and therefore atmospheric oxidation is the primary factor limiting long range transport to the
Arctic. Although a chemical must also be resistant to metabolism to achieve high levels in
top predators of the Arctic food chain, we do not include a metabolic criterion because of the
difficulty of predicting susceptibility to metabolism.
The second screening methodology rests upon the observation that chemicals which have
already been identified as contaminants in the higher trophic levels of the Arctic food chain
must have the correct combination of substance properties to traverse the entire pathway
from emission to human. A chemical profiling method is employed which compares a
chemical's structure to the structures of known Arctic contaminants, and those chemicals
which match the structural profile are assumed to have a higher likelihood of being Arctic
contaminants. Cases where these two paradigms overlap and where they differ are used to
identify possible false positives and negatives, and to assist in classification. These screening
methodologies test a chemical for the potential to be an Arctic contaminant, but to be a risk
the chemical must also be emitted. Ideally usage profiles and emissions data should be
consulted but this is impractical for a large number of chemicals and the data is largely
unavailable, so it is assumed that chemicals with a higher production volume will have a
greater likelihood of significant emissions.
2.2 Compilation of Data
Identifying information (CAS number, SMILES string), physical-chemical properties (molar
mass, octanol-water partition coefficient KOW, Henry's law constant, vapor pressure, aqueous
18
solubility), and the atmospheric oxidation half-lives for 105,584 individual chemicals were
compiled and calculated using the extensive CAS/SMILES string database of the EPISuite
software package41
. These data served in the first screening method. In all cases experimental
values were used where available. It should be noted that EPISuite assumes that all
molecules are in their neutral form, so it is not possible to assess the effects of any
dissociation equilibria. Two separate sets of partitioning properties were calculated from
these data, with the same log KOW used in both cases. In the first set (―HLC data set‖) a value
for the air-water partition coefficient KAW was derived from the Henry's Law constant (KAW =
HLC/(RT)) and a value for the octanol-air partition coefficient KOA was obtained by applying
a thermodynamic cycle with the KOW (KOA = KOW/KAW). If the Henry's Law constant was
unavailable then it was estimated using the ratio of vapor pressure and aqueous solubility. In
the second set of partitioning properties (―VP data set‖) the vapor pressure was used to
estimate a value for the KOA based on an empirical equation from Xiao et al.56
, then a value
for KAW was calculated from KOW (KAW = KOW/KOA).
Ideally experimental values should be used for all partitioning properties and OH oxidation
rates, because the applicability of the EPISuite prediction methods to a very wide range of
organic compounds is untested and doubtful57-59
. However, the list of organic substances with
reliable experimental data is much smaller than the number of chemicals with SMILES
strings in the EPISuite database. Experimental values are also much less likely to be
available for chemicals which have not already been identified as harmful, and screening will
therefore not identify any new potential POPs. Properties predicted from chemical structure
alone, such as those available in the EPISuite database, are of limited and unknown accuracy
but at this time they are the most extensive available and can help identify high concern
chemicals which have not yet been closely examined.
For the second methodology based on structural resemblance, a list of 86 known Arctic
contaminants was compiled. This list (Table 2), references for concentrations found in Arctic
biota, the procedure and the criteria used in compiling the list, as well as a short description
of the six major structural classes of known Arctic contaminants is given in the Appendix.
19
SMILES strings were found in the EPISuite database or composed and used to quantify the
86 chemicals‘ structural properties.
Five different lists were used to determine if a chemical has a high production volume
(HPV): the Canadian Domestic Substances List which flags substances having an annual
production volume greater than 1,000 t (metric tonnes)7, the US EPA's HPV Challenge
Program which includes chemicals that are produced or imported into the US in quantities of
454 t (1 million pounds) or more per year60
, the HPV list compiled by the European
Chemical Bureau‘s European Chemical Substances Information System which includes
chemicals produced or imported to the European Union in quantities of 1000 t or more61
, the
OECD's list of HPV chemicals produced in quantities of 1,000 t or greater in the EU or
another member country62
, and any chemical listed under the US Toxic Substances Control
Act (TSCA) produced in a quantity of 454 t (1 million pounds) or more in any year for which
data is available63
. Chemicals were further checked against three databases of registered
pesticides; the Canadian Domestic Substances List7, the US EPA's list of registered
pesticides64
, and the World Health Organization's list of current use pesticides65
.
2.3 Screening for Long Range Transport to the Arctic and Bioaccumulation in Humans
By linking global transport calculations66
with a food chain bioaccumulation model23
, Czub
et al.40
have delineated the combination of partitioning properties KAW, KOW, and KOA, which
result in a high potential for organic chemicals to be transported to the Arctic and to
accumulate in the Arctic human food chain. They illustrated these combinations by plotting
an Arctic Contamination and Bioaccumulation Potential (AC-BAP) after seventy years for a
multitude of hypothetical perfectly persistent chemicals as a function of the chemical
partitioning space defined by KOA and KAW (Figure 2). The area of elevated AC-BAP is
defined here as exceeding at least 10% of the maximum AC-BAP70 value (red outline in
Figure 2). Over half of the 86 known Arctic contaminants fall within the defined area of
elevated AC-BAP. Some, such as heavily chlorinated PCDD/Fs and PCBs, fall outside of the
red outline because the global transport calculations predict a low potential to reach the
20
Arctic for particle-bound substances with log KOA values above 1140,66
. This subject is
explored further below. The area of elevated AC-BAP comprises the following thresholds:
log KOW ≥ 3.5
log KOA ≥ 6
0.5 ≥ log KAW ≥ -7
log KAW ≤ -1.78 · log KOA + 14.56
(3)
Figure 2 – Chemical Space plot of elevated Arctic contaminant potential.
The red outline delineates chemical partitioning property combinations that result in a
high (i.e. in excess of 10 % of the maximum) potential for accumulation in the Arctic
physical environment and the Arctic human food chain as described by Czub et al.40
.
21
Values for log KOA and log KAW of 86 known Arctic contaminants (experimental values
referenced in the Appendix, and predicted with SPARC67
if not available) are also
shown.
The selection of the 10% line is arbitrary and not very conservative, but was deemed to be a
compromise between selectivity and the possibility of excluding chemicals which should be
flagged as of concern. Using the 1% line instead would greatly improve the agreement
between the model results and the monitoring data, as all but two of the chemicals fall within
the line indicating 1% of the maximum AC-BAP70. However, doing so would essentially
neglect the effects of long range transport because the 1% line is nearly equivalent to the area
of elevated Arctic bioaccumulation40
. The contours of the plot of AC-BAP suggests that the
elevated area extends below log KAW = -5. There are computational reasons for not extending
the plot below this line but no mechanistic reasons, so we have extrapolated two log units
into this area to ensure we capture as many potential Arctic contaminants as possible.
By checking if the partitioning properties of an organic chemical fall within the red outline in
Figure 2, the potential for environmental transport to the Arctic and bioaccumulation in the
Arctic food chain are being screened for simultaneously. Whereas the screening process is
simple, the definition of the area of elevated AC-BAP is based on detailed mechanistic
models and represents a sophisticated and effective exclusion method. Using an enclosed
chemical space has intuitively a higher power of discrimination than using cut-off values for
only one of the three partitioning properties, such as the common use of a log KOW threshold
of 523
, which in itself may not even be valid (Equation (3),ref 23). The partitioning properties
of the entire chemical data set were screened using the thresholds of Equation (3). Both the
HLC data set and the VP data sets were checked and chemicals were considered to meet the
criterion of elevated AC-BAP, if either data set placed the chemical within the red
boundaries.
22
2.4 Screening for Persistence in Air
In the screening step involving the partitioning space, chemical persistence is not being
considered. To filter out chemicals which are quickly degraded in the atmosphere and
therefore have low long range transport potential, an atmospheric oxidation half-life cut-off
is used. Hydroxyl radicals are the primary atmospheric oxidant for most substances68
; so a
hydroxyl radical atmospheric oxidation half-life (OH thalf) was used as the screening
parameter. Some experimental OH thalf values are available in the EPISuite database and the
rest were predicted using EPISuite41
. There will be contributions from other oxidants to the
total atmospheric oxidation rate but the reaction rates with OH are the most reliable69,70
.
Reaction with other oxidants will only increase the overall rate of oxidation, so OH thalf
values may be viewed as the maximum half life for a chemical in air. A cut-off value of two
days was used as suggested by the Stockholm Convention5 and Muir and Howard
55. We
should caution that this criterion could eliminate chemicals that may undergo long range
transport to the Arctic in the oceans, or have a longer oxidation half-life when sorbed to
atmospheric particles.
2.5 Screening for Resemblance with the Structural Profile of Known Arctic Contaminants
There is a distinct set of combinations of partitioning and persistence properties which meet
all the criteria of the first screening methodology, but these properties are themselves
governed by chemical structure. This implies that there may be a limited number of possible
combinations of structural features that give rise to the properties of an Arctic contaminant.
A structure property relationship (SPR) was developed to predict if a chemical has the
potential to be an Arctic contaminant based on chemical structure. The vast majority of SPRs
rely on multiple linear regressions but that approach is impractical in this case as no
continuous scale of the potential to become an Arctic contaminant exists that can be used to
parameterize a regression equation. Instead the list of 86 known Arctic contaminants is
compared to the full list of chemicals and deviations from a "typical" molecule are quantified
for 3 structural parameters; halogenation, internal connectivity and molecular size. A detailed
23
description of the derivation and application of the structural profile of Arctic contaminants
is provided in the Appendix (Section 5.1).
3 Results and Discussion
3.1 Preliminary Screening
Figure 3 shows a schematic representation of how the two screening methodologies classify
the screened chemicals. From the list of 105,584 chemicals 16,888 chemicals fall within the
boundaries of elevated AC-BAP using either the HLC or VP data sets (see Figure 5 of the
Appendix). A total of 13,905 chemicals are identified by the HLC data set, 14,957 are
identified by the VP data set and 11,974 chemicals are identified by both data sets. Neither
data set shows a significant bias towards over- or underestimating the log KAW or log KOA
values so it can not be immediately determined which data set is more accurate. A total of
16,151 chemicals have experimental or estimated OH thalf values greater than two days. There
are 2,025 chemicals which both fall within the chemical space and have an OH thalf value
greater than two days. There are 3,088 chemicals which match the structural profile of known
Arctic contaminants.
More than two thirds of the chemicals screened (74,016 chemicals or 70.1%) did not meet
any of the three screening criteria and are unlikely to reach the Arctic or bioaccumulate in
humans in this region. A further 12,669 (12.0%) chemicals do not match the structural profile
of known Arctic contaminants and fall outside of the area of elevated AC-BAP, but are
persistent in air. These chemicals may possibly reach the Arctic environment but because the
area of elevated AC-BAP is strongly influenced by the potential to bioaccumulate they are
unlikely to lead to significant human exposure. An additional 14,608 (13.8%) chemicals fall
within the area of elevated AC-BAP but are not persistent in air and do not match the
structural profile of known Arctic contaminants. It is unlikely that a significant fraction of
these chemicals will reach the Arctic and become contaminants. Eliminating these three
groups of chemicals from further consideration excludes 101,293 chemicals (95.9%).
24
Is the chemical
persistent in air?
no
yes
Do the chemicals have
partitioning
properties that allow
for accumulation in
the physical
environment and in the
human food chain of
the Arctic?
no
yes
Does the chemical match
the structural profile of known
Arctic contaminants?
yesno
yesno
yesno
yesno
Very unlikely to be an
Arctic contaminant.
Possible false positive by profiling criteria
or both unusual exposure route and
bioaccumulation mechanism
Chemical is unlikely to reach the Arctic
human food chain in significant amounts
or bioaccumulate there
Possible false negative by
LRT/B/P methodology due to unusual
bioaccumulation or transport mechanism
Chemical will likely degrade en route to
the Arctic human food chain
Possible false negative by LRT/B/P
methodology due to direct emissions or
non-gas phase persistence
Possible false positive by
LRT/B/P methodology or possible
non-POP like Arctic contaminant
Very likely to have the properties that
cause a chemical to become an
Arctic contaminant
1,203 (1.1%, 32 HPVC) 822 (0.8%, 27 HPVC)
255 (0.2%, 6 HPVC)14,608 (13.8%, 535 HPVC)
1,457 (1.4%, 49 HPVC)12,669 (12.0%, 1,042 HPVC)
554 (0.5%, 6 HPVC)74,016 (70.1%, 2,626 HPVC)
Figure 3 - Schematic diagram of the screening process. The 105,584 chemicals in the EPI Suite database were screened for
their potential to accumulate in the Arctic food chain.
25
3.2 Chemicals Matching the Structural Profile of Known Arctic Contaminants
There are 2,266 chemicals (2.1%) which despite matching the structural profile of known
Arctic contaminants fail to meet one or both of the other two screening criteria. Of these
there are 554 (0.5%) that fail to meet both the persistence in air and chemical space criterion
and 255 (0.2%) that fall within the area of elevated AC-BAP but are not persistent in air.
Due to low air persistence these chemicals are unlikely to reach the Arctic, however it is
possible that their degradation products will be of concern. The oxidation products could be
more persistent and their partitioning properties may be different, but predicting the major
oxidation products of the 809 chemicals identified is beyond the scope of this study.
An additional 1,457 (1.4%) chemicals match the structural profile of known Arctic
contaminants and are persistent in air but fall outside of the area of elevated AC-BAP. Figure
4 shows the location in the chemical space of all 3,088 chemicals which match the structural
profile of known Arctic contaminants. There are thus large numbers of substances that are
structurally similar to known Arctic contaminants, but nevertheless do not possess the right
combination of partitioning properties to become Arctic contaminants based on the model
predictions and assumptions used in this study. A closer inspection of Figure 4 reveals three
types of chemicals outside of the red boundaries. It is enlightening to explore what prevents
these chemicals from being classified as potential Arctic contaminants, because it could
possibly reveal the limitations of both the screening and the chemical profiling technique.
Type 1: Volatile Chemicals. Many chemicals are located to the upper left of the high AC-
BAP area, indicative of high log KAW (> 0.5) and low log KOA (< 8) values. These chemicals
are similar to known Arctic contaminants in size and degree of halogenation but are
considerably more volatile, and thus will not bioaccumulate in air-breathing organisms40
.
The explanation is that these are mostly highly fluorinated chemicals, which experience
weak van der Waals interactions in comparison to their highly chlorinated analogs, and so
partition relatively weakly to condensed phases71
. Perfluorinated alkanes are examples of
such chemicals.
26
Figure 4 - Locations of all chemicals in the EPISuite database which match the
structural profile of known Arctic contaminants. Plotted using the "HLC data set".
Type 2: Highly Sorptive Chemicals. A significant number of chemicals which match the
structural profile of known Arctic contaminants lie to the right of the high AC-BAP area,
having a predicted log KOA above 8 and a log KOW above 5. These are low-volatility and
sparingly water-soluble chemicals that sorb predominately to particles in both the
atmosphere and aqueous media. Particle settling processes are predicted to prevent efficient
long range transport in both atmosphere and oceans66
. Examples are octachlorodibenzo-p-
dioxin and decabromodiphenyl ether. It is noteworthy to point out that these chemicals are
predicted to have considerable potential for bioaccumulation23
and that some known Arctic
contaminants indeed have such partitioning properties (Figure 2, Table 2). These chemicals
27
have the potential to become Arctic contaminants if they are emitted close to the Arctic. It is
also possible that particle-bound atmospheric transport is more efficient in delivering
contaminants to the Arctic than is currently described in global transport calculations; one
reason may be because the effect of intermittent rain is ignored72
. Another possible long
range transport mechanism for these chemicals is biotransport, as migrating organisms may
take up these contaminants in their southern wintering grounds73
.
Type 3: Small, Polar Chemicals. The majority of the chemicals which match the structural
profile of known Arctic contaminants and fall outside of the area of elevated AC-BAP do so
because their log KOW values are too low. Many of these chemicals are small and polar due
to the placement of polar functional groups on the "back-bone" of the molecule, making
them too hydrophilic to bioaccumulate.
3.3 Potential Arctic Contaminants
There are 2,025 (1.9%) chemicals which both fall within the area of elevated AC-BAP and
are persistent in air; 822 (0.8%) of these also match the structural profile of known Arctic
contaminants. All 822 chemicals that have partitioning properties corresponding to high AC-
BAP, are persistent in air, and match the structural profile of known Arctic contaminants are
listed in Table 3 of the Appendix. The majority of these chemicals have already been
detected in Arctic biota; the general chemical classes and the number of chemicals flagged
are shown in Table 6 of the Appendix. Major chemical groups of environmental concern,
such as PCBs, PCNs, PCDDs, PCDFs and PBDEs are identified, along with a number of
homologues which are halogenated with fluorine, bromine, iodine or combinations thereof
with chlorine atoms. There are also a large number of homologues of various chemical
classes which are substituted with halogenated alkyl side chains. One of the larger chemical
classes is polychlorinated diphenylethers (PCDEs) which have been previously quantified in
environmental samples74
. Chlorinated benzenes, phenols, nitrophenols and nitrobenzenes are
named along with other chlorinated benzene derivatives such as toluene, aniline and others.
An almost equally large number of these are flagged which contained halogens other than
chlorine. A structurally diverse group of halogenated or halo-alkyl triazine derivatives are
28
also among the potential POPs. Organochlorine pesticides designated by the Stockholm
Convention (mirex, chlordanes, toxaphene, HCH, HCB) are also identified along with a
number of other obsolete and current-use pesticides.
Remaining are 1,203 (1.1%) chemicals which both fall within the area of elevated AC-BAP
and are persistent in air but do not match the structural profile of known Arctic contaminants.
This group thus includes chemicals that we would not necessarily suspect to become Arctic
contaminants based on structural similarity, but are predicted to have the respective
properties. It would thus contain any truly novel type of Arctic contaminant. Out of the 1,203
chemicals 793 have some degree of halogenation, and 483 of these fall into one of the
chemical classifications that comprise the majority of the list of 822 POP-like chemicals. For
example there are 20 mono- and dichloro- PCBs which are excluded from the list of POP-
like chemicals because of their low degree of chlorination. Almost all of the 1,203 non POP-
like chemicals have at least some portion which is aromatic. A large number of the chemicals
that are not halogenated are nitro-substituted aromatics or contain a silane or siloxane
substituent. There are also 43 current-use or obsolete pesticides which fall into this category.
3.4 High Risk Chemicals
There are 4,291 chemicals which are identified as potential Arctic contaminants by one or
both screening methodologies. These chemicals were checked against the five HPV chemical
lists and 120 were found to be HPV chemicals; the identified HPV chemicals are
summarized in Table 1. A number of chemicals which were included in the set of 86 known
Arctic contaminants (PBDEs, HBCD, endosulfan, DDT, -HCH and HCB) appear in the list
of 120 HPV chemicals. Some of these have been banned or severely restricted in North
America but appear on the European HPV list, which was most recently updated in 2000.
There are three chemicals structurally related to Endosulfan (CAS 115275, 115286,
2157199) among the non-persistent molecules identified, including two of the chemicals
which also fall outside of the area of elevated AC-BAP. Representatives of the three types of
29
chemicals that fall outside of the area of elevated AC-BAP (Figure 4) are present.
Perfluoroundecane, which is intended to represent perfluoroalkanes with chain length
between 5 and 18 carbons in the EPISuite SMILES database (CAS 86508421), falls under
the volatile category (type 1), PBDEs are particle-bound substances (type 2) and the three
pesticides chlorothalonil, picloram and nitrapyrin are small, polar chemicals (type 3).
Twenty-seven chemicals are identified by both screening methods, three of which are in the
training set. Exact usage profiles for some of these chemicals are difficult to obtain but most
appear to be used as intermediates in the synthesis of halogenated pesticides or polymers.
Some may not be a major environmental concern, for example the two
trichloromethylpyridines (CAS 1817136, 69045836) and pentachloropyridine (CAS
2176627) are reportedly used only in closed systems75
. The brominated species (CAS
632791) is used as a flame retardant in polymers76
. Perhaps of more concern is the presence
of a current use pesticide, pentachloronitrobenzene (CAS 82688). Also of concern are the
chlorinated cyclopentanes and cyclopentenes which are highly chlorinated but have virtually
no information available regarding their usage profiles.
30
Table 1 - Summary of highproduction volume chemicals which are predicted to become
Arctic contaminants or which match the structural profile of known Arctic
contaminants.
CAS AC-
BAP(a)
Persistent
(b) Profile
(c) Pesticide
(d) Name
115286 no no yes no 1,4,5,6,7,7-hexachloro-5-norbornene-2,3-dicarboxylic acid
1691992 no no yes no N-ethyl-1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-N-(2-hydroxyethyl)-1-octanesulfonamide
2157199 no no yes no Endosulfan alcohol
25637994 no no yes no hexabromocyclododecane (1,3,5,7,9,11-hexabromocyclododecane)
(e)
27905459 no no yes no 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluorodecyl, 2-propenoic acid ester
61262531 no no yes no 1,1'-[1,2-ethanediylbis(oxy)]bis-2,3,4,5,6-pentabromobenzene
77474 no yes yes yes 1,2,3,4,5,5-hexachloro-1,3-cyclopentadiene
87616 no yes yes no 1,2,3-trichlorobenzene
87683 no yes yes no 1,1,2,3,4,4-hexachloro-1,3-butadiene
98157 no yes yes no 1-chloro-3-(trifluoromethyl)-benzene
98464 no yes yes no 1-nitro-3-(trifluoromethyl)-benzene
98566 no yes yes no 1-chloro-4-(trifluoromethyl)-benzene
108770 no yes yes no 2,4,6-trichloro-1,3,5-triazine
115253 no yes yes no Octafluorocyclobutane
120821 no yes yes yes 1,2,4-trichlorobenzene
307357 no yes yes no 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-1-octanesulfonyl fluoride
311897 no yes yes no 1,1,2,2,3,3,4,4,4-nonafluoro-N,N-bis(nonafluorobutyl)-1-butanamine
328847 no yes yes no 1,2-dichloro-4-(trifluoromethyl)-benzene
329011 no yes yes no 1-isocyanato-3-(trifluoromethyl)-benzene
335422 no yes yes no heptafluoro-butanoyl fluoride
338841 no yes yes no 1,1,2,2,3,3,4,4,5,5,5-undecafluoro-N,N-bis(undecafluoropentyl)-pentanamine
423507 no yes yes no 1,1,2,2,3,3,4,4,5,5,6,6,6-tridecafluoro-1-hexanesulfonyl fluoride
428591 no yes yes no trifluoro(trifluoromethyl)-oxirane
647427 no yes yes no 3,3,4,4,5,5,6,6,7,7,8,8,8-tridecafluoro-1-octanol
678397 no yes yes no 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluoro-1-decanol
719324 no yes yes no 2,3,5,6-tetrachloro-1,4-benzenedicarbonyl dichloride
865861 no yes yes no 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,12-heneicosafluoro-1-dodecanol
1163195 no yes yes no decabromodiphenylether (BDE-209)
1737935 no yes yes no 3,5-dichloro-2,4,6-trifluoropyridine
1897456 no yes yes yes Chlorothalonil
1918021 no yes yes yes Picloram
31
1929824 no yes yes yes Nitrapyrin
2043530 no yes yes no 1,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8-heptadecafluoro-10-iododecane
2043541 no yes yes no 1,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10-heneicosafluoro-12-iodododecane
2043574 no yes yes no 1,1,1,2,2,3,3,4,4,5,5,6,6-tridecafluoro-8-iodooctane
2402791 no yes yes no 2,3,5,6-tetrachloropyridine
3194556 no yes yes no 1,2,5,6,9,10-hexabromocyclododecane
3825261 no yes yes no Ammonium perfluorooctanoate
5848931 no yes yes no 5-chloro-3-(trichloromethyl)-1,2,4-thiadiazole
10469097 no yes yes no 3,4,5,6-tetrachloro-2-pyridinecarboxylic acid
14143603 no yes yes no 4-amino-3,5,6-trichloro-2-pyridinecarbonitrile
17824838 no yes yes no 3,4,5,6-tetrachloro-2-pyridinecarbonitrile
32534819 no yes yes no pentabromodiphenylether (BDE-99)(e)
32536520 no yes yes no octabromodiphenylether (BDE-203)(e)
36483600 no yes yes no hexabromodiphenylether (BDE-167)(e)
40088479 no yes yes no tetrabromodiphenylether (BDE-55)(e)
52314677 no yes yes no 3-(2,2-dichloroethenyl)-2,2-dimethyl-cyclopropanecarbonyl chloride
59808785 no yes yes no tetrachlorocyclopentane (1,2,3,4-tetrachlorocyclopentane)
(e)
60825265 no yes yes no (3,5,6-trichloro-2-pyridinyl)oxy-acetic acid, methyl ester
63936561 no yes yes no nonabromodiphenylether (BDE-206)(e)
68928803 no yes yes no heptabromodiphenylether (BDE-173)(e)
69045789 no yes yes no 2-chloro-5-trichloromethylpyridine
86508421 no yes yes no perfluoro compounds C5-18 (perfluoroundecane)
(e)
138495428
no yes yes no 1,1,1,2,2,3,4,5,5,5-decafluoropentane
163702076
no yes yes no 1,1,1,2,2,3,3,4,4-nonafluoro-4-methoxy-butane
101053 yes no yes yes Anilazine
115275 yes no yes no 1,4,5,6,7,7-hexachloro-5-norbornene-2,3-dicarboxylic anhydride
115297 yes no yes yes Endosulfan
3278895 yes no yes no 1,3,5-tribromo-2-(2-propenyloxy)-benzene
3734483 yes no yes no Chlordene
24448097 yes no yes no 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-N-(2-hydroxyethyl)-N-methyl-1-octanesulfonamide
78637 yes yes no no 1,1'-(1,1,4,4-tetramethyl-1,4-butanediyl)bis[2-(1,1-dimethylethyl) peroxide
80079 yes yes no no 1,1'-sulfonylbis[4-chlorobenzene]
80104 yes yes no no Dichlorodiphenylsilane
81141 yes yes no no 1-[4-(1,1-dimethylethyl)-2,6-dimethyl-3,5-dinitrophenyl]-ethanone
81152 yes yes no no 1-(1,1-dimethylethyl)-3,5-dimethyl-2,4,6-trinitrobenzene
32
84515 yes yes no no 2-ethyl-9,10-anthracenedione
88062 yes yes no yes 2,4,6-trichlorophenol
88857 yes yes no yes Dinoseb
98737 yes yes no no 4-(1,1-dimethylethyl)-benzoic acid
101633 yes yes no no 1,1'-oxybis 4-nitrobenzene
102363 yes yes no no 1,2-dichloro-4-isocyanatobenzene
115322 yes yes no yes Dicofol
118796 yes yes no no 2,4,6-tribromophenol
136607 yes yes no no benzoic acid, butyl ester
144796 yes yes no no chloromethyldiphenylsilane
147820 yes yes no no 2,4,6-tribromobenzenamine
320729 yes yes no no 3,5-dichloro-2-hydroxybenzoic acid
540976 yes yes no no 2,2,4,4,6,6,8,8,10,10,12,12-dodecamethylcyclohexasiloxane
1185097 yes yes no no 1,1,2,2-tetrachloro-ethanesulfenyl chloride
1539044 yes yes no no 1,4-benzenedicarboxylic acid, 1,4-diphenyl ester
1836755 yes yes no yes Nitrofen
2116849 yes yes no no 1,1,1,5,5,5-hexamethyl-3-phenyl-3-[(trimethylsilyl)oxy]-trisiloxane
2392485 yes yes no no 4-chloro-1-(2,4-dichlorophenoxy)-2-nitrobenzene
3457612 yes yes no no 1,1-dimethylethyl 1-methyl-1-phenylethyl peroxide
13472087 yes yes no no 2,2'-azobis-2-methylbutanenitrile
34893920 yes yes no no 1,3-dichloro-5-isocyanatobenzene
35578473 yes yes no no bis(4-bromophenyl)-ethanedione
50594440 yes yes no no 5-[2-chloro-4-(trifluoromethyl)phenoxy]-2-nitrophenol, 1-acetate
50594779 yes yes no no 3-[2-chloro-4-(trifluoromethyl)phenoxy]-phenol, 1-acetate
52270447 yes yes no no Neodecanoic acid, cobalt(2+) salt
63734623 yes yes no no 3-(2-Chloro-4-trifluoromethylphenoxy)-benzoic acid
64667330 yes yes no no 4,6,6,6-Tetrachloro-3,3-dimethylhexanoic acid methyl ester
50293 yes yes yes yes p,p'-DDT
58899 yes yes yes yes -HCH
68360 yes yes yes no 1,4-bis(trichloromethyl)-benzene
82688 yes yes yes yes pentachloronitrobenzene
87843 yes yes yes no 1,2,3,4,5-pentabromo-6-chlorocyclohexane
95943 yes yes yes no 1,2,4,5-tetrachlorobenzene
117088 yes yes yes no 4,5,6,7-tetrachloro-1,3-isobenzofurandione
118741 yes yes yes yes HCB
133493 yes yes yes no pentachlorobenzenethiol
626391 yes yes yes no 1,3,5-tribromobenzene
632791 yes yes yes no 4,5,6,7-tetrabromo-1,3-isobenzofurandione
634662 yes yes yes yes 1,2,3,4-tetrachlorobenzene
1134049 yes yes yes no 2,3,4,5-tetrachloro-6-(trichloromethyl)-pyridine
33
1203867 yes yes yes no 2,2-dichloro-1-(2,4,5-trichlorophenyl)-ethanone
1817136 yes yes yes no 3,6-dichloro-2-(trichloromethyl)-pyridine
2176627 yes yes yes no pentachloropyridine
5216251 yes yes yes no 1-chloro-4-(trichloromethyl)-benzene
17700093 yes yes yes no 2,3,4-trichloronitrobenzene
29091096 yes yes yes no 2,4-dichloro-1,3-dinitro-5-(trifluoromethyl)-benzene
30554724 yes yes yes no tetrabromodichlorocyclohexane (1,2,3,4-tetrabromo-5,6-dichlorocyclohexane)
(e)
62111471 yes yes yes no heptachlorocyclopentene (1,2,3,3,4,4,5-heptachloro-1-cyclopentene)
(e)
68258902 yes yes yes no heptachlorocyclopentane (1,2,3,3,4,5,5-heptachlorocyclopentane)
(e)
68258913 yes yes yes no hexachlorocyclopentane (1,2,2,3,4,4-hexachlorocyclopentane)
(e)
69045836 yes yes yes no 2,3-dichloro-5-trichloromethylpyridine
72030263 yes yes yes no hexachlorocyclopentene (1,2,3,4,5,5-hexachloro-1-cyclopentene)
(e)
73588428 yes yes yes no 1,3-dichloro-5-(1,3,3,3-tetrachloro-1-methylpropyl)-benzene
75147205 yes yes yes no 2,2,3,4,4-pentachloro-3-butenoic acid, butyl ester
(a) Indicates if the chemical falls within the area of elevated AC-BAP.
(b) Indicates if the chemical has an OH thalf greater than 2 days.
(c) Indicates if the chemical matches the structural profile of known Arctic contaminants.
(d) Indicates if the chemical appears on one of the three pesticide lists.
(e) The CAS is for an unspecified isomer or analogue, the specific chemical noted in brackets corresponds to
the SMILES string found in the EPISuite database.
The majority of the HPV chemicals which are persistent in air and fall within the area of
elevated AC-BAP but do not match the structural profile of known Arctic contaminants are
halogenated to some degree. Three chlorinated pesticides are included in this group; 2,4,6-
trichlorophenol, dicofol and nitrofen. An interesting result is the presence of a non-
halogenated pesticide Dinoseb, and dodecamethylcyclohexasiloxane (D6) which is an
ingredient found in cosmetic and personal hygiene products77
.
The selectivity of the screening method presented here could be varied by changing the
criteria thresholds. For example, the net could be cast wider by defining the thresholds for
the partitioning properties (Equation (3)) based on 1 % of maximum AC-BAP70 or it could
be made more selective by raising the threshold for the atmospheric oxidation half life. The
thresholds for partitioning and atmospheric oxidation could be made less stringent, i.e. the
danger of false-negatives could be reduced, if reliable screening steps for biotransformation
34
and environmental persistence could be devised and added, which would limit the number of
false positives. Another research need is the development of methods to predict the
partitioning, environmental fate and food chain accumulation of charged chemical species.
We suggest the next steps with respect to the identified potential Arctic contaminants include
the verification of the property data used in this study through laboratory measurements or
more sophisticated estimation techniques as well as laboratory tests of bioaccumulation to
assess the potential for biotransformation. If a substance should indeed be shown to be
resistant to metabolism, Arctic monitoring programs should search for it.
4 Acknowledgements We acknowledge funding from the Long-range Research Initiative of the European Chemical
Industry Association (CEFIC), and Michael McLachlan for insightful comments.
5 Appendix
5.1 Compiling a List of Known Arctic Contaminants
To minimize possible "sampling error" in constructing the list of known Arctic contaminants
four criteria were applied: (1) Chemicals must have been positively identified and quantified
in the tissues of Arctic biota, preferably in biota at the higher trophic levels of the food chain.
(2) Congeners with consistently higher concentrations in Arctic biota are selected over
congeners with lower concentrations, (3) Chemical structure groupings should be balanced to
contribute approximately equal numbers of chemicals to the list, and (4) in cases where some
chemicals must be excluded to balance the list, the remaining chemicals should be
representative of the structural diversity of the chemical group. Note that all of the
contaminants selected are traditional POP-like contaminants, meaning references to "known
Arctic contaminants" here and elsewhere in the Appendix refer specifically to "POP like
Arctic contaminants".
Three recent review papers identified major POPs present in marine78
, freshwater79
and
terrestrial biota80
from the Canadian Arctic. The anthropogenic contaminants in higher
35
trophic organisms (fish, birds and mammals) identified in these reviews included PCBs,
HCHs, DDTs, dieldrin, endosulfan, chlordanes, chlorobenzenes, toxaphenes, PCDDs,
PCDFs, PBDEs, PCNs, perfluorinated alkyl acids (PFAs) and chlorinated paraffins (CPs).
All of the identified chemicals are halogenated and have "backbone" structures composed
primarily of carbon. Sample chemicals containing all three of the above halogens (F, Br, Cl)
are included, but the chemical structure grouping is based on the "backbone" of the
molecule. This is not an exhaustive list of known Arctic contaminants and there are many
that have been detected but are not listed here.
The first chemical group is biphenyls and naphthalenes. Coplanar biphenyls and
naphthalenes are structurally similar and have similar chemical and toxicological
properties81,82
. Ortho-substituted biphenyls are also included in this group to represent the
structural diversity of group. The second chemical group is bridged diphenyls which includes
PBDEs and DDTs. DDT has different steric factors which affect the orientation of the rings
than those that affect PBDEs83,84
, but they are structurally similar enough for the purposes
here because SMILES string contain no information regarding steric factors. The third
chemical group is three-ring heterocycles which includes PCDDs and PCDFs. The fourth
chemical group is monocyclics which includes HCBs, HCHs and polychlorinated styrenes
(PCSs). The fifth chemical group is straight-chain alkyls which include PFAs and CPs. The
sixth and final chemical group is chemicals containing a norbornene or norbornane structure
which includes dieldrin, endosulfan, chlordanes and toxaphenes. Another chemical, mirex, is
also included in this group. Though mirex does not contain a norbornene or norbornane
structure it is more similar to this chemical group than the other groups.
36
Table 2 - Names, CAS numbers, and the overall POP score for a list of known Arctic
contaminants
Also given is the contribution of the three chemical profiling criteria SA (structural
atoms), CX (degree of halogenation), XS (degree of internal connectivity) to the POP
score. The references provide concentrations for the contaminants in Arctic biota and
physical-chemical properties. See page S2 for description of how this list was compiled.
biphenyls and naphthalenes
Arctic Reference
Property Ref.
CAS SA
Cont. CX
Cont. XS
Cont. SCORE
(tetra) PCN-28 85 86 53555638 -0.591 2.749 1.875 4.033
(tetra) PCN-29 85 86
149864788 -0.591 2.749 1.875 4.033
(tetra) PCN-34 85 86
67922218 -0.591 2.749 1.875 4.033
(tetra) PCN-38 85 86
149864802 -0.591 2.749 1.875 4.033
(penta) PCN-52 85,87 86
53555650 -0.591 3.208 1.875 4.492
(penta) PCN-54 85,87 86
150224161 -0.591 3.208 1.875 4.492
(penta) PCN-61 85 86
150224229 -0.591 3.208 1.875 4.492
(hexa) PCN-66 85,78 86
103426966 -0.591 3.609 1.875 4.893
(hexa) PCN-69 85 86
103426944 -0.591 3.609 1.875 4.893
(hexa) PCN-71 85 86
90948280 -0.591 3.609 1.875 4.893
(tetra-o) PCB-52 87 81
35693993 -0.235 2.406 1.563 3.733
(tetra-o) PCB-66 88 89
32598100 -0.235 2.406 1.563 3.733
(tetra-o) PCB-74 88 89
32690930 -0.235 2.406 1.563 3.733
(penta-o) PCB-99 90 89
38380017 -0.235 2.830 1.563 4.157
(penta-no) PCB-118 88 81
31508006 -0.235 2.830 1.563 4.157
(penta-no) PCB-126 87,91 89
57465288 -0.235 2.830 1.563 4.157
(hexa-o) PCB-138 88,90 81
35065282 -0.235 3.208 1.563 4.535
(hexa-o) PCB-153 88,90 81
35065271 -0.235 3.208 1.563 4.535
(hexa-no) PCB-169 87,91 89
32774166 -0.235 3.208 1.563 4.535
(hepta-o) PCB-170 88,90 89
35065306 -0.235 3.545 1.563 4.872
(hepta-o) PCB-180 88,90 81
35065293 -0.235 3.545 1.563 4.872
bridged diphenyls
p,p'-DDD 91 92
72548 -1.062 2.138 1.339 2.416
p,p'-DDE 91,93 92
72559 -1.062 2.138 1.339 2.416
p,p'-DDT 91,93 92
50293 -1.062 2.532 1.339 2.810
BDE-28 94 95,96
41318756 -0.649 1.804 1.442 2.598
BDE-47 94,97,98,99 95,96
5436431 -0.649 2.264 1.442 3.058
BDE-99 94,98,99 95,96
60348609 -0.649 2.673 1.442 3.467
BDE-100 94,98,99 95,96
189084648 -0.649 2.673 1.442 3.467
BDE-153 94,98,99 95,96
68631492 -0.649 3.039 1.442 3.833
BDE-154 94,98,99 95,96
207122154 -0.649 3.039 1.442 3.833
three-ring heterocycles
2,3,7,8-tetraCDD 87 100
1746016 -1.062 2.138 2.009 3.086
1,2,3,7,8-pentaCDD 87 100
40321764 -1.062 2.532 2.009 3.479
1,2,3,4,7,8-hexaCDD 87 100
39227286 -1.062 2.887 2.009 3.834
37
1,2,3,6,7,8-hexaCDD 87 100
57653857 -1.062 2.887 2.009 3.834
1,2,3,7,8,9-hexaCDD 87 100
19408743 -1.062 2.887 2.009 3.834
1,2,3,4,6,7,8-heptaCDD
87 100 35822469 -1.062 3.208 2.009 4.155
OCDD 87 100
3268879 -1.062 3.499 2.009 4.446
2,3,7,8-tetraCDF 87 100
51207319 -0.649 2.264 2.164 3.779
1,2,3,7,8-pentaCDF 87 100
57117416 -0.649 2.673 2.164 4.188
2,3,4,7,8-pentaCDF 87 100
57117314 -0.649 2.673 2.164 4.188
1,2,3,4,7,8-hexaCDF 87 100
70648269 -0.649 3.039 2.164 4.554
1,2,3,6,7,8-hexaCDF 87 100
57117449 -0.649 3.039 2.164 4.554
2,3,4,6,7,8-hexaCDF 87 100
60851345 -0.649 3.039 2.164 4.554
1,2,3,7,8,9-hexaCDF 87 100
72918219 -0.649 3.039 2.164 4.554
1,2,3,4,6,7,8-heptaCDF
87 100 67562394 -0.649 3.368 2.164 4.883
1,2,3,4,7,8,9-heptaCDF
87 100 55673897 -0.649 3.368 2.164 4.883
OCDF 87 100
39001020 -0.649 3.666 2.164 5.181
monocyclics
PeCB 93,101 92
608935 -2.244 4.374 1.563 3.693
HCB 93,98,101,102 92
118741 -2.244 4.811 1.563 4.130
-HCH 90,93 103
319846 -2.244 4.811 1.563 4.130
-HCH 90,91,93 103
319857 -2.244 4.811 1.563 4.130
-HCH 93 103
58899 -2.244 4.811 1.563 4.130
HBCD 94,99 67
25637994 -0.235 3.208 0.781 3.753
PtCS (pentachlorostyrene)
104 67 14992815 -1.418 3.701 1.172 3.456
HxCS (hexachlorostyrene-
beta-trans) 104 67
90301921 -1.418 4.124 1.172 3.878
HpCS (heptachlorostyrene-
beta,beta) 102(a),104 67
29082755 -1.418 4.491 1.172 4.245
OCS (octachlorostyrene)
102,104,105 67 29082744 -1.418 4.811 1.172 4.566
alkyls
perfluorooctane sulfonate (PFOS)
90,106 58 45298906 -0.235 5.641 0.000 5.406
perfluorooctane sulfonamide (PFOSA)
91 58 754916 -0.235 5.641 0.000 5.406
perfluorooctanoic acid (PFCA-8)
106 58 335671 -0.591 5.774 0.000 5.183
perfluorononanoic acid (PFCA-9)
90,106 58 375951 -0.178 5.842 0.000 5.665
perfluorodecanoic acid (PFCA-10)
90,106 58 335762 -0.235 5.898 0.000 5.662
perfluoroundecanoic acid (PFCA-11)
90,106 58 2058948 -0.649 5.944 0.000 5.295
chlorinated paraffin (C10,Cl7)
b
107 67 32534784 -0.591 3.962 0.000 3.371
chlorinated paraffin (C10,Cl8)
b
107 67 214327669 -0.591 4.277 0.000 3.686
chlorinated paraffin 107 67
219697082 -0.591 4.558 0.000 3.967
38
(C10,Cl9)b
chlorinated paraffin (C11,Cl7)
b
107 67 none
assigned -0.178 3.742 0.000 3.564
chlorinated paraffin C11,Cl8)
b
107 67 36312819 -0.178 4.052 0.000 3.874
chlorinated paraffin (C11,Cl9)
b
107 67 none
assigned -0.178 4.330 0.000 4.152
chlorinated paraffin (C12,Cl6)
b
107 67 214327681 -0.235 3.208 0.000 2.972
chlorinated paraffin (C12,Cl7)
b
107 67 214327692 -0.235 3.545 0.000 3.310
chlorinated paraffin (C12,Cl9)
b
107 67 214327716 -0.235 4.124 0.000 3.889
norbornenes-norbornanes
oxychlordane 90,93,98 100
27304138 -0.178 4.052 3.409 7.283
Heptachlor-exo-epoxide
91,98 92 1024573 -0.178 3.742 3.409 6.974
trans-nonachlor 91,98 100
39765805 -0.591 4.558 2.813 6.780
MC 6 98 67
98318979 -0.591 4.558 2.813 6.780
Aldrin 101(a) 92
309002 -0.235 3.208 3.125 6.097
Dieldrin 91,93,102,101(a) 92
60571 -0.649 3.039 3.606 5.996
Endrin 91 92
72208 -0.649 3.039 3.606 5.996
a-endosulfan 93(a),101(a) 92
959988 -0.649 3.039 2.164 4.554
b-endosulfan 93(a) 92
33213659 -0.649 3.039 2.164 4.554
endosulfan-sulfate 93 67
1031078 -1.062 2.887 2.009 3.834
Toxaphene 26 91 67
142534712 -0.591 4.277 1.875 5.561
Toxaphene 50 91 67,108
66860808 -0.591 4.811 1.875 6.095
Toxaphene 62 91 67,108
154159065 -0.591 4.558 1.875 5.842
Mirex 91,101 67,108
2385855 -0.591 5.249 4.688 9.346
a The authors present concentrations for chemical groups which include this congener, but do not
explicitly list the concentrations for individual congeners.
b There is no chemical structure available for an "average" chlorinated paraffin, chemical structures
were made by taking a carbon chain of the correct length and adding the correct number of chlorine
atoms evenly spaced along the chain.
39
Figure 5 - Plot of all 105,581 chemicals in the EPI Suite database.
Experimental values are used for the plot where available, otherwise predicted values
from the "HLC data set" are used. Chemicals with an OH oxidation half-life less than
two days are highlighted in green, and the area of elevated AC-BAP is delineated in
red.
40
5.2 Profiling Arctic Contaminants
All of the data needed to create a profile of known Arctic contaminants is theoretically
contained in the SMILES strings of the EPISuite database. Parameters such as the number
and types of atoms and the first-order bond connectivity are easily extracted by simple
parsing algorithms. A more difficult task is selecting the parameters to use in the chemical
profile. Multiple linear regression is not possible because we do not have a continuous and
universal scale against which to correlate a multiple linear regression; we can only say with
certainty whether a chemical accumulates in Arctic biota or not, which is discreet data. The
basic approach taken to profiling Arctic contaminants is as follows; first a hypothetical
"typical" molecule is characterized using the list of all available SMILES strings, then this
molecule is compared with a ―typical‖ Arctic contaminant, derived from the list of 86
contaminants mentioned above, to determine how the structural attributes of known Arctic
contaminants deviate from the hypothetical "typical" molecule. The observed deviations are
finally combined to create a "POP score" which indicates how closely a chemical matches
the profile of known Arctic contaminants. Several criteria must be met in order to ensure that
the POP score is robust and has predictive power. The most important factor is the selection
of known Arctic contaminants which has already been discussed. There should be as few
chemical parameters as possible, each should have a sound mechanistic reason for being
included, and the parameters should not be inter-correlated.
Examining the list of 86 known contaminants some general observations can be made about
the structural characteristics they share and which therefore may be important in rendering a
chemical a potential Arctic contaminant. A widely accepted characteristic used to help
identify POPs by structure is chlorination109
and many known Arctic contaminants have a
relatively high degree of chlorination. More recently perfluorinated organic acids and
brominated flame retardants have become a concern and halogenation in general can be
viewed as a possible indicator that a chemical may be a POP99,110
. All of the 86 known Arctic
contaminants fall within a fairly narrow range of molecular size, suggesting that there is an
optimum. If a chemical is too small it will be too volatile to be deposited in the Arctic and it
will be exhaled by organisms; if it is too large it will not be subject to long range
atmospheric transport and may not bioaccumulate. A third common attribute of the known
Arctic contaminants is a tendency towards cyclic structures. Many contain aromatic rings
and some of the most toxic, persistent and bioaccumulative chemicals such as those
identified by the Stockholm Convention have highly cyclized structures, e.g. mirex and
41
chlordane. Perfluorinated acids are obviously the exception to this rule, but they also have a
higher degree of halogenation. Based on this examination, molecular size, halogenation and
internal connectivity were selected as the criteria for profiling POP-like structures.
5.3 Chemical Profiling Criteria
The criterion of molecular size is quantified as the total number of structural atoms (SA),
where structural atoms are defined as carbon, oxygen, nitrogen, silicon, sulfur and
phosphorous atoms.
SA = Σ C, O, N, Si, S and P atoms (4)
SA does not include hydrogen atoms. SMILES strings do not contain hydrogen atoms in
most cases and determining the valence of each atom and adding hydrogen atoms needlessly
increases the complexity of the parsing algorithms. Furthermore, hydrogen atoms contribute
relatively little mass or size to molecules compared to other atoms, so can be safely
neglected. Halogen atoms are also not included in SA, because otherwise the parameter
indicating the degree of halogenation would be correlated with the molecular size parameter,
which should be avoided. SA only refers to the structural "back-bone" of a molecule, it will
not be strongly correlated with the degree of halogenation, and it will have the same value
for groups of homologous molecules (PCBs for example) which is desirable.
The criterion quantifying the degree of halogenation of a molecule is annotated CX. The
number of halogen-carbon bonds is used to define CX instead of the number of halogen
atoms, because there are a number of salts in the EPISuite database and only halogenated
organics are of interest. To prevent any weak correlation with the size of a molecule (larger
molecules have more possible halogenation sites) CX is normalized to the total number of
non-hydrogen atoms in the molecule (TA):
number of halogen-carbon bonds = CX
TA (5)
42
This also restricts the possible values of CX to between zero and approximately one which
makes the degree of halogenation of large and small molecules more comparable.
The third criterion of internal connectivity could be defined as the number of rings, but
identifying and counting rings again needlessly complicates the SMILES string parsing
algorithms. A more readily obtained measure of internal connectivity is the number of bonds;
for example n-hexane has five carbon-carbon bonds and cyclohexane has six carbon-carbon
bonds. This is a characteristic of the structural back-bone of a molecule so only structural
atoms as defined by Equation (4) and bonds between structural atoms as defined by Equation
(6) will be used.
SB = Σ bonds between C, O, N, Si, S and P atoms (6)
The degree of internal connectivity is defined as the ratio of "excess structural bonds" (the
number of structural bonds minus the number of bonds that would be present in an n-alkane
with the same number of atoms) to the number of structural atoms (XS) as shown in Equation
(7).
[ - ( -1)] =
SB SAXS
SA (7)
As with CX, using a ratio restricts the possible values of XS to between zero and
approximately one.
5.4 Application of Chemical Profiles
All three of the chemical profile criteria SA, CX, and XS were calculated for the 105,584
molecules and statistically analyzed. distribution graphs which show the relative frequency
of values were constructed for five different "populations" of chemicals; the full list of
chemicals, only those chemicals which fall within the area of elevated AC-BAP (only
accumulative), only those chemicals which have an OH thalf value greater than two days
(only persistent), the chemicals which are both accumulative and persistent, and finally the
43
86 known Arctic contaminants. The distribution graphs are Figure 6 to Figure 8. The SA has
a normal distribution only slightly skewed to larger values and so can be adequately
described by the mean and standard deviation of the values. For CX and XS however the
distributions are dominated by an enormous number of chemicals with a CX or XS value of
zero which has a large influence on the mean and standard deviations of the values. For the
full, only accumulative and only persistent populations a more statistically sound method of
analyzing the variation in the population is to use the mode (the most common value, in this
case zero) instead of the mean and calculate the deviation from the mode. This is not valid
for the populations of both accumulative and persistent chemicals or for the known Arctic
contaminants, but in these cases the deviation from zero becomes very similar to the mean of
the population as the distribution of CX and XS values migrate further from zero. The mean
and standard deviation of SA, CX and XS as well as the deviation from zero of CX and XS are
shown in Figure 9. Comparisons of the CX and XS values of the different populations are
primarily based on the mean and standard deviation to make the comparisons easier, but
subsequent calculations are based on the deviation from zero.
44
Figure 6 - Frquency distribution of the number of structural atoms (SA). distributions
are shown for five different chemical "populations"
The full list of chemicals in the EPI Suite database (All), chemicals which fall within the
defined chemical space (Accumulative), chemicals with an OH oxidation half-life of
greater than 2 days (Persistent), chemicals which are both Accumulative and Persistent
(Both), and the list of 86 known Arctic contaminants (Known).
0 20 40 60 80 100 120 140 160
02468
10121416182022
SA
Kn
ow
n
050
100150200250300350
Bo
th
0200400600800
100012001400
Pe
rsis
ten
t
0
500
1000
1500
2000
Accu
mu
lative
01000200030004000500060007000
All
45
Figure 7 - Frequency distribution of the degree of halogenation (CX). distributions are
shown for five different chemical "populations"
The full list of chemicals in the EPI Suite database (All), chemicals which fall within the
defined chemical space (Accumulative), chemicals with an OH oxidation half-life of
greater than 2 days (Persistent), chemicals which are both Accumulative and Persistent
(Both), and the list of 86 known Arctic contaminants (Known).
0.0 0.2 0.4 0.6 0.8
02468
10
CX
Kn
ow
n
0
100
200
300
400
500
Bo
th
0
2000
4000
6000
8000
Pe
rsis
ten
t
02000400060008000
10000120001400016000
Accu
mu
lative
0
20000
40000
60000
80000
100000
All
46
Figure 8 - Frequency distribution of the degree of internal connectivity (XS).
distributions are shown for five different chemical "populations"
The full list of chemicals in the EPI Suite database (All), chemicals which fall within the
defined chemical space (Accumulative), chemicals with an OH oxidation half-life of
greater than 2 days (Persistent), chemicals which are both Accumulative and Persistent
(Both), and the list of 86 known Arctic contaminants (Known).
0.0 0.2 0.4 0.6 0.8
02468
1012141618
XS
Kn
ow
n
050
100150200250300
Bo
th
0100020003000400050006000
Pe
rsis
ten
t
0
1000
2000
3000
4000
Accu
mu
lative
05000
1000015000200002500030000
All
47
Figure 9 - Mean and standard deviation of the total number of structural atoms (SA),
degree of halogenation (CX), and degree of internal connectivity (XS) for five different
chemical "populations". The deviation from zero is shown in orange
5.5 Interpretation of Figure 9
The mean number of structural atoms in accumulative molecules is virtually the same as for
the whole population, but the mean value for persistent molecules is significantly lower
(Figure 9). For the molecules which are both accumulative and persistent the mean value of
SA is much closer to the "only persistent" population, and the mean value for known Arctic
contaminants is encouragingly close to this SA value as well. Also notable is the very tight
clustering of the SA values for known Arctic contaminants; the standard deviation is only
2.42. This seems to confirm the conjecture that there is an optimum size for Arctic
contaminants.
0 5 10
15 20 25 30
SA
0.0 0.1 0.2 0.3 0.4 0.5
CX
0.0
0.1
0.2
0.3
Population
XS
Full Population
Only Accumulative
Only Persistent
Accumulative & Persistent
Known Contaminants
48
Figure 9 shows that the pattern of the mean values is closely approximated by the deviations
from zero. Figure 7 shows that while most chemicals have a CX value of zero, there is a large
amount of variation in the degree of halogenation for those chemicals with CX greater than
zero. For the full population as well as for the accumulative and the persistent sub-
populations a pattern similar to SA is observed; the accumulative population has a mean CX
value similar to the entire population, but the value for persistent molecules is significantly
different, much higher in this case. For molecules which are both accumulative and
persistent CX is much higher again, and for the known Arctic contaminants the mean CX is a
full order of magnitude higher than the mean of the full population. The deviation from zero
increases by a factor of three from the full population to the known Arctic contaminants.
Much higher CX values for the known Arctic contaminants suggest that halogenation
contributes significantly to both persistence and bioaccumulation of individual molecules,
however regarding the population of molecules as a whole, halogenation contributes more to
persistence. Another notable piece of information is that the magnitude of the standard
deviation of the known contaminants is comparable to the standard deviation of the whole
population; this implies that while there is not an optimum degree of halogenation,
increasing the amount of halogenation increases the persistence and accumulation of a
molecule.
Variations in the mean values of XS are more complex than SA and CX. The distribution of
XS is dominated by a large number of chemicals with no internal connectivity (Figure 8), but
Figure 9 shows that the deviation from zero is a good approximation of the mean for all
populations. Looking only at accumulative molecules slightly increases the mean XS value,
but for the persistent molecules the mean XS value slightly decreases. However for
molecules which are both persistent and accumulative the mean XS value is much higher,
and is higher again for known Arctic contaminants. There are most likely a number of factors
affecting the results, for example methyl groups are more resistant to OH oxidation than
secondary or tertiary carbons69
which may account for the decrease in XS for the persistent
sub-population. There are many molecules containing aromatic rings in the sub-population
which is both accumulative and persistent, so it appears that increasing the degree of
aromaticity may increase the persistence of a molecule and this in turn increases the mean
value of XS in this specific sub-population of accumulative molecules. However, there are
many molecules in the list of known Arctic contaminants which contain non-aromatic rings
so it could also be concluded that increasing the degree of internal connectivity of any kind
contributes to the potential to become an Arctic contaminant. The magnitude of the standard
49
deviation of XS is similar for all populations and slightly increases for the list of known
Arctic contaminants so it does not appear that there is an optimal value. Clearly there is a
finite limit to how many internal rings there can be in a stable molecule; for a group of
molecules with the same number of atoms there will be an optimal number of internal
connections which maximizes stability. However it must be remembered that the molecules
in EPI Suite database are chemicals in commercial use and a tendency to more stable
molecules must exist. Given these assumptions we conclude that the optimal degree of
internal connectivity is not generally exceeded in our population of molecules, so increasing
the XS value tends to increase the bioaccumulation of a molecule, and persistence is also
often increased due to the correlation of XS with the number of aromatic rings.
5.6 Definition of the POP score
The most important conclusions of the previous analysis are that SA has an optimal value,
indicated by the small standard deviation of SA for the population of known Arctic
contaminants (Figure 9), and that increases in CX and XS make a molecule more persistent
and bioaccumulative. The latter is obvious from greatly elevated mean CX and XS values for
known Arctic contaminants (Figure 9). One difficulty remaining is how to weight the
contribution of each of these three criteria to the POP score. As stated before this cannot be
done using multiple linear regression so we instead look at the deviations from a hypothetical
"typical" molecule. The first restrictions we make are that the hypothetical "typical"
molecule must have a lower POP score than a molecule fitting the profile of known Arctic
contaminants. We define the contribution of CX and XS to the POP score as the number of
standard deviations above the mode of the full population. The CX or XS value of a molecule
is divided by the standard deviation of the full population. This means that a "typical"
molecule with the most common CX and XS values of zero would have a POP score of zero
if only these two criteria were used. The contribution of SA to the POP score is defined as the
negative of the absolute number of standard deviations from the optimal value, meaning that
deviations from the optimal value of SA decrease the POP score. The mean value of SA for
known Arctic contaminants is assumed to be the optimal value, and is subtracted from the SA
of a molecule and then the absolute value is divided by the standard deviation of SA for
known Arctic contaminants. The complete calculation of the POP score is shown in Equation
(8).
50
| - 11.43|POP score = + -
0.1039 0.1067 2.420
CX XS SA (8)
5.7 Interpretation of the POP Score
When interpreting the POP scores some discretion must be exercised. As was stated
previously there is no universal scale of susceptibility to becoming an Arctic contaminant,
and the POP score is not nearly sophisticated enough to be interpreted as such. When
comparing molecules the one with a higher POP score is not necessarily more likely to
accumulate in Arctic biota. Qualitative data was used to create the POP score, therefore only
qualitative data can be provided by it. Interpretation of the score should keep in mind the
molecules that were used to parameterize it. All 86 known Arctic contaminants are
halogenated and the lowest contribution that CX makes to the POP score is 1.809, so any
chemical with a CX contribution to the POP score of less than 1.8 should be excluded. For all
but two of the 86 known Arctic contaminants the CX makes a larger contribution to the POP
score than XS (Table 2), and these two chemicals (dieldrin and endrin) have an unusually
high level of cyclization, a likely explanation for this is that halogenation stabilizes the
highly cyclized structures. Based on these observations there does not seem to be a need to
restrict the interpretation of the POP score based on the XS value. Finally, the number of
structural atoms for the list of known Arctic contaminants ranges between six and fourteen
so any molecule too far outside these bounds should be regarded with skepticism; a molecule
one standard deviation below 6 (SA <= 3) or one standard deviation above 14 (SA >= 17)
should not be considered a potential POP.
The mean POP score calculated for the 86 known Arctic contaminants was 4.447 ± 1.122.
The lower limit of the 95% prediction interval of the POP score is 2.57 and the lowest POP
score for a known Arctic contaminant is 2.416 (DDE). To ensure that all possible Arctic
contaminants are captured by the POP score all chemicals with a score of 2.4 or greater,
while also meeting the SA and CX requirements previously mentioned, are defined as
matching the profile of known Arctic contaminants.
Although the POP score should not be interpreted quantitatively, it is worthwhile to look at
which chemicals have particularly high POP scores. Any molecule with a POP score greater
than the mean score of known Arctic contaminants (4.447) was defined as a chemical which
51
has an above average likelihood of being an Arctic contaminant. There are 79 chemicals out
the 822 chemicals which are both persistent and fall within the area of elevated AC-BAP
which exceed this cut-off (identified in Table 3). Among these are known Arctic
contaminants such as mirex, toxaphenes, chlordanes, hexachlorobiphenyls, and a number of
perfluorinated organic acids, as well as variations of known Arctic contaminants, such as
some highly fluorinated biphenyls. Included are also a series of chlorinated cyclopentanes,
pentenes and pentadienes, a number of highly halogenated benzimidazoles, several triazines
substituted with trichloromethyl or trichloroethyl groups, two chlorinated PAHs and a
number of different halogenated bridged diphenyls.
52
Table 3 - CAS number, chemical class, name and POP score of chemicals meeting the
elevated AC-BAP, atmospheric oxidation half-life, and POP score criteria.
A “yes” in the column entitled “Hi POP” implies that the chemical has a POP score
greater than 4.447.
CAS CLASS NAME HI-POP SCORE
82688 CUP benzene, pentachloronitro- 3.475 87865 CUP phenol, pentachloro- 3.519 88824 CUP benzoic acid, 2,3,5-triiodo- 2.443
117180 CUP 2,3,5,6-tetrachloronitrobenzene 2.999 634662 CUP benzene, 1,2,3,4-tetrachloro- 3.168
2678219 CUP OLPISAN 2.470 6379460 CUP Vancide PB 2.470 7159344 CUP Pyroxychlor 2.999 8003461 CUP trichlorodinitrobenzene 2.470 25167833 CUP phenol, tetrachloro- 3.008 58138082 CUP Oxirane, 2-(3,5-dichlorophenyl)-2-(2,2,2-trichloroethyl)- yes 4.534 297789 OP Isobenzan yes 6.498
1344327 OP benzene, trichloro(chloromethyl)- 3.008 1715408 OP bromociclen yes 4.776 2550756 OP HERCULES426 yes 5.608 3495429 OP chlorquinox 4.033 3513937 OP Hercules 426D yes 5.738 41318756 PBDE 2,4,4'-tribromodiphenyl ether 2.598 49690940 PBDE tribromodiphenyl oxide 2.598 1336363 PCB PCBs 3.733 2437798 PCB 2,2',4,4'-PCB 3.733 2542292 PCB pentachlorobiphenyl 4.157 7012375 PCB 2,4,4'-PCB 3.251 11097691 PCB AROCLOR 1254 4.157 11126424 PCB AROCLOR 5460 3.733 12672296 PCB AROCLOR 1248 3.733 12674112 PCB Aroclor 1016 3.251 15862074 PCB 2,4,5-PCB 3.251 15968055 PCB 1,1'-biphenyl, 2,2',6,6'-tetrachloro- 3.733 16606023 PCB 2,4',5-PCB 3.251 18259057 PCB 2,3,4,5,6-PCB 4.157 25323686 PCB trichlorobiphenyl 3.251 25429292 PCB pentachlorobiphenyl 4.157 26914330 PCB tetrachlorobiphenyl 3.733 31508006 PCB 2',3,4,4',5'-PCB 4.157 32598100 PCB 2,4,3',4'-PCB 3.733 32598111 PCB 2,3',4',5-PCB 3.733 32598122 PCB 1,1'-biphenyl, 2,4,4',6-tetrachloro- 3.733 32598133 PCB 3,3',4,4'-PCB 3.733 32598144 PCB 1,1'-biphenyl, 2,3,3',4,4'-pentachloro- 4.157 32690930 PCB 2,4,4',5-PCB 3.733 33025411 PCB 2,3,4,4'-PCB 3.733 33284525 PCB 3,3',5,5'-PCB 3.733 33284536 PCB 2,3,4,5-PCB 3.733 33284547 PCB 2,3,5,6-PCB 3.733 33979032 PCB 2,2',4,4',6,6'-PCB yes 4.535 35693926 PCB 2,4,6-PCB 3.251 35693993 PCB 2,2',5,5'-PCB 3.733 36559225 PCB 1,1'-biphenyl, 2,2',3,4'-tetrachloro- 3.733 37680652 PCB 2,2',5-PCB 3.251 37680663 PCB 2,2',4-PCB 3.251 37680685 PCB 1,1'-biphenyl, 2,3',5'-trichloro- 3.251 37680696 PCB 3,3',4-trichlorobiphenyl 3.251 37680732 PCB 2,4,5,2',5'-PCB 4.157 38379996 PCB 2,2',3,5',6-PCB 4.157 38380028 PCB 2,3,4,2',5'-PCB 4.157
53
38380039 PCB 2,3,3',4',6-PCB 4.157 38380040 PCB 2,2',3,4',5',6-PCB yes 4.535 38411222 PCB 2,3,6,2',3',6'-PCB yes 4.535 38444734 PCB 2,2',6-PCB 3.251 38444767 PCB 2,3',6-trichlorobiphenyl 3.251 38444778 PCB 2,4',6-PCB 3.251 38444789 PCB 2,2',3'-PCB 3.251 38444814 PCB 2,3',5-PCB 3.251 38444847 PCB 2,3,3'-PCB 3.251 38444858 PCB 2,3,4'-PCB 3.251 38444869 PCB 2',3,4-PCB 3.251 38444870 PCB 3,3',5-PCB 3.251 38444881 PCB 3,4',5-trichlorobiphenyl 3.251 38444905 PCB 3,4,4'-PCB 3.251 38444938 PCB 2,2',3,3'-PCB 3.733 39485831 PCB 1,1'-biphenyl, 2,2',4,4',6-pentachloro- 4.157 39635331 PCB 3,3',4,5,5'-pentachlorobiphenyl 4.157 41464395 PCB 2,2',3,5'-PCB 3.733 41464408 PCB 2,2',4,5'-PCB 3.733 41464419 PCB 2,2',5,6'-tetrachlorobiphenyl 3.733 41464420 PCB 1,1'-biphenyl, 2,3',5,5'-tetrachloro- 3.733 41464431 PCB 1,1'-biphenyl, 2,3,3',4'-tetrachloro- 3.733 41464464 PCB 1,1'-biphenyl, 2,3',4',6-tetrachloro- 3.733 41464475 PCB 1,1'-biphenyl, 2,2',3,6'-tetrachloro- 3.733 41464486 PCB 1,1'-biphenyl, 3,3',4,5'-tetrachloro- 3.733 41464497 PCB 1,1'-biphenyl, 2,3,3',5'-tetrachloro- 3.733 41464511 PCB 2,2',3',4,5-PCB 4.157 52663588 PCB 1,1'-biphenyl, 2,3,4',6-tetrachloro- 3.733 52663599 PCB 2,2',3,4-PCB 3.733 52663602 PCB 2,2',3,3',6-PCB 4.157 52663613 PCB 2,2',3,5,5'-PCB 4.157 52663624 PCB 2,2',3,3',4-pentachlorobiphenyl 4.157 53469219 PCB Aroclor 1242 3.733 53555661 PCB 3,4,5-trichlorobiphenyl 3.251 54230227 PCB 2,3,4,6-tetrachlorobiphenyl 3.733 55215173 PCB 2,2',3,4,6-pentachlorobiphenyl 4.157 55312691 PCB 2,2',3,4,5-pentachlorobiphenyl 4.157 55702459 PCB 2,3,6-PCB 3.251 55702460 PCB 1,1'-biphenyl, 2,3,4-trichloro- 3.251 55712373 PCB 1,1'-biphenyl, 2,3',4-trichloro- 3.251 55720440 PCB 2,3,5-trichlorobiphenyl 3.251 56558168 PCB 2,2',4,6,6'-pentachlorobiphenyl 4.157 56558179 PCB 1,1'-biphenyl, 2,3',4,4',6-pentachloro- 4.157 56558180 PCB 1,1'-biphenyl, 2,3',4,5',6-pentachloro- 4.157 57465288 PCB 3,3',4,4',5-pentachlorobiphenyl 4.157 60145202 PCB 2,2',3,3',5-pentachlorobiphenyl 4.157 60145213 PCB 1,1'-biphenyl, 2,2',4,5',6-pentachloro- 4.157 60233241 PCB 2,3',4,6-tetrachlorobiphenyl 3.733 60233252 PCB 1,1'-biphenyl, 2,2',3,4',6'-pentachloro- 4.157 62796650 PCB 1,1'-biphenyl, 2,2',4,6-tetrachloro- 3.733 65510443 PCB 2,3,4,4',5-pentachlorobiphenyl 4.157 65510454 PCB 2,2',3,4,4'-PCB 4.157 68194047 PCB 1,1'-biphenyl, 2,2',4,6'-tetrachloro- 3.733 68194058 PCB 2,3,6,2',4'-PCB 4.157 68194069 PCB 2,2',4,5,6'-pentachlorobiphenyl 4.157 68194070 PCB 1,1'-biphenyl, 2,2',3,4',5-pentachloro- 4.157 68194105 PCB 1,1'-biphenyl, 2,3,3',5',6-pentachloro- 4.157 68194116 PCB 1,1'-biphenyl, 2,3,4',5,6-pentachloro- 4.157 68194127 PCB 2,4,5,3',5'-PCB 4.157 70362413 PCB 2,3,3',4,5'-pentachlorobiphenyl 4.157 70362457 PCB 1,1'-biphenyl, 2,2',3,6-tetrachloro- 3.733 70362468 PCB 1,1'-biphenyl, 2,2',3,5-tetrachloro- 3.733 70362479 PCB 2,2',4,5-tetrachlorobiphenyl 3.733 70362480 PCB 2,3',4',5-tetrachlorobiphenyl 3.733 70362491 PCB 3,3',4,5-tetrachlorobiphenyl 3.733 70362504 PCB 3,4,4',5-tetrachlorobiphenyl 3.733 70424678 PCB 2,3,3',5-tetrachlorobiphenyl 3.733 70424689 PCB 2,3,3',4',5-pentachlorobiphenyl 4.157 70424690 PCB 2,3,3',4,5-pentachlorobiphenyl 4.157 70424703 PCB 2',3,4,5,5'-pentachlorobiphenyl 4.157
54
73575527 PCB 2,3',4,5'-tetrachlorobiphenyl 3.733 73575538 PCB 1,1'-biphenyl, 2,3',4,5-tetrachloro- 3.733 73575549 PCB 2,2',3,6,6'-pentachlorobiphenyl 4.157 73575550 PCB 2,2',3,5,6'-pentachlorobiphenyl 4.157 73575561 PCB 1,1'-biphenyl, 2,2',3,5,6-pentachloro- 4.157 73575572 PCB 1,1'-biphenyl, 2,2',3,4,6'-pentachloro- 4.157 74338231 PCB 2,3',5',6-tetrachlorobiphenyl 3.733 74338242 PCB 1,1'-biphenyl, 2,3,3',4-tetrachloro- 3.733 74472336 PCB 2,3,3',6-tetrachlorobiphenyl 3.733 74472347 PCB 1,1'-biphenyl, 2,3,4',5-tetrachloro- 3.733 74472358 PCB 2,3,3',4,6-pentachlorobiphenyl 4.157 74472369 PCB 1,1'-biphenyl, 2,3,3',5,6-pentachloro- 4.157 74472370 PCB 2,3,4,4',5-pentachlorobiphenyl 4.157 74472381 PCB 1,1'-biphenyl, 2,3,4,4',6-pentachloro- 4.157 74472392 PCB 2,3',4',5',6-pentachlorobiphenyl 4.157 76842074 PCB 2',3,3',4,5-pentachlorobiphenyl 4.157
50748 PCBD 2,3,4,5-tetrachlorobenzoic acid 2.999 51398 PCBD benzoic acid, 3,4,5-trichloro- 2.443 58902 PCBD phenol, 2,3,4,6-tetrachloro- 3.008 82622 PCBD 3,4,6-trichloro-2-nitrophenol 2.567 87876 PCBD 1,4-benzenediol, 2,3,5,6-tetrachloro- 2.962 95943 PCBD benzene, 1,2,4,5-tetrachloro- 3.168
117975 PCBD benzenethiol, pentachloro-, zinc salt 3.210 133493 PCBD benzenethiol, pentachloro- 3.519 484673 PCBD phenol, 2,3,5,6-tetrachloro-4-methoxy- 2.999 527208 PCBD pentachloroaniline 3.519 606075 PCBD pentachloroethylbenzene 3.456 608935 PCBD benzene, pentachloro- 3.693 634833 PCBD 2,3,4,5-tetrachloroaniline 3.008 634902 PCBD benzene, 1,2,3,5-tetrachloro- 3.168 781157 PCBD benzene, 1,2,3,4-tetrachloro-5,6-dinitro- (6CI,7CI,8CI,9CI) 2.952 875401 PCBD benzene, 1,2,3,5-tetrachloro-4-methyl- (9CI) 3.008 877098 PCBD benzene, 1,2,3,5-tetrachloro-4,6-dimethyl- (9CI) 2.962 877101 PCBD 1,2,4,5-tetrachloro-3.6-dimethylbenzene 2.962 877112 PCBD pentachlorotoluene 3.519 879390 PCBD 1,2,3,4-tetrachloro-5-nitrobenzene 2.999 935955 PCBD phenol, 2,3,6-trichloro- 3.008 938227 PCBD benzene, 1,2,3,5-tetrachloro-4-methoxy- 2.962 938863 PCBD benzene, 1,2,3,4-tetrachloro-5-methoxy- 2.962 944616 PCBD benzene, 1,2,3,4-tetrachloro-5,6-dimethoxy- 3.096 944785 PCBD 1,4-dimethoxytetrachlorobenzene 3.096 994785 PCBD tetrachloro-1,4-dimethoxybenzene 3.096
1006311 PCBD 2,3,5,6-tetrachlorotoluene 3.008 1006322 PCBD 2,3,4,5-tetrachlorotoluene 3.008 1012846 PCBD benzoic acid, pentachloro- 3.475 1198556 PCBD 1,2-benzenediol, 3,4,5,6-tetrachloro- 2.962 1441027 PCBD pentachlorophenyl acetate 3.554 1825190 PCBD pentachloro(methylthio)benzene 3.456 1825214 PCBD pentachloroanisole 3.456 1897412 PCBD 1,4-benzenedicarbonitrile, 2,3,5,6-tetrachloro- 3.096 1953997 PCBD 1,2-benzenedicarbonitrile, 3,4,5,6-tetrachloro- 3.096 2438882 PCBD tetrachloronitroanisole 3.241 2539175 PCBD 2-methoxytetrachlorophenol 2.999 2668248 PCBD 4,5,6-trichloroguaiacol 2.443 2877147 PCBD pentachlorophenoxyecetic acid 3.682 3215654 PCBD (2,3,6-trichlorophenyl)-acetonitrile 2.443 3481207 PCBD 2,3,5,6-tetrachloroaniline 3.008 3714623 PCBD 2,3,4,6-tetrachloronitrobenzene 2.999 4824720 PCBD 4-nitro-2,3,5,6-tetrachlorophenol 3.096 4901513 PCBD 2,3,4,5-tetrachlorophenol 3.008 5324685 PCBD benzene, 1,3,5-trichloro-2,4,6-trimethyl- 2.443 6284839 PCBD benzene, 1,3,5-trichloro-2,4-dinitro- 2.470 6936409 PCBD benzene, 1,2,4,5-tetrachloro-3-methoxy- 2.962 7476826 PCBD Butanoic acid, 2,3,4,6-tetrachlorophenyl ester (9CI) 2.952 10460330 PCBD phenol, 2,3,4,6-tetrachloro-5-methyl- 2.962 12408105 PCBD tetrachlorobenzeneS 3.168 13061284 PCBD ethanone, 1-(2,4,5-trichlorophenyl)- (9CI) 2.443 13608872 PCBD 2,3,4-trichlor-acetophenon 2.443 13801508 PCBD 1,2-benzenedithiol, 3,4,5,6-tetrachloro- 2.962 16022698 PCBD benzenemethanol, 2,3,4,5,6-pentachloro- 3.456
55
16766293 PCBD benzene, 1,2,3-trichloro-4,5-dimethoxy- 2.567 17700093 PCBD 2,3,4-trichloronitrobenzene 2.443 18708708 PCBD benzene, 1,3,5-trichloro-2-nitro- 2.443 19303889 PCBD phenol, 2,3,4,5-tetrachloro-6-methyl- (9CI) 2.962 20098388 PCBD benzene, 1,2,4,5-tetrachloro-3,6-dinitro- 2.952 20098480 PCBD benzene, 1,2,3-trichloro-5-nitro- 2.443 20404028 PCBD 2,3,6-trichloro-4-nitrophenol 2.567 20925853 PCBD benzonitrile, pentachloro- 3.456 23399908 PCBD 2,4,6-trichlorophenylacetate 2.567 27735644 PCBD phenol, pentachloro-, sodium salt, monohydrate (8CI,9CI) 3.519 27864137 PCBD benzene, 1,2,4-trichloro-3-nitro- (6CI,8CI,9CI) 2.443 29450633 PCBD benzaldehyde, pentachloro-, oxime (8CI,9CI) 3.475 29733708 PCBD tetrachloromethylbenzene 3.008 33715627 PCBD 3,4,5-trichloroacetanilide 2.567 42138727 PCBD benzenamine, 2,3,4,5-tetrachloro-6-methoxy- 2.999 53014418 PCBD benzene, tetrachloro(methylthio) 2.962 53452816 PCBD tetrachloroanisoles 2.962 54965707 PCBD benzene, 1,3,5-trichloro-2-(1-methylethyl)- 2.443 55538733 PCBD phenol, 3,4,6-trichloro-2-propyl- 2.567 56680650 PCBD phenol, 2,3,5,6-tetrachloro-4-methyl- (9CI) 2.962 57057837 PCBD phenol, 3,4,5-trichloro-2-methoxy- 2.443 61465790 PCBD 2,3,6-trichlorocumene 2.443 61465814 PCBD trichlorodimethoxybenzene 2.567 68671909 PCBD benzene, 1,2,4,5-tetrachloro-3-(methylthio)- 2.962 69576803 PCBD benzene, 1,2,3,4-tetrachloro-5-methoxy-6-nitro- 3.241 69911611 PCBD pentachlorotoluene 3.519 70439962 PCBD benzenamine, 2,3,5,6-tetrachloro-4-methoxy- 2.999 70833635 PCBD ar,ar,ar-trichlorobenzeneacetonitrile 2.443 74313001 PCBD benzene, 1-methoxy-4-methyl-, trichloro DERIV. 2.443 81686422 PCBD 2,3,6-trichloro-p-cymene 2.567 81686433 PCBD 2,3,5,6-tetrachloro-p-cymene 3.096 85298073 PCBD 1,2,5-trichloro-3,4-dimethoxybenzene 2.567
107409529 PCBD 1,2,4,5-tetrachloro-3-(methylsulfinyl) 3.096 1746016 PCDD 2,3,7,8-TCDD 3.085 30746588 PCDD 1,2,3,4-tetrachlorodibenzo-p-dioxin 3.085 33423926 PCDD 1,3,6,8-tetrachlorodibenzo-p-dioxin 3.085 36088229 PCDD pentachlorodibenzo-p-dioxin 3.479 40581940 PCDD 1,4,7,8-tetrachlorodibenzo-p-dioxin 3.085 62470535 PCDD 1,3,7,9-tetrachlorodibenzo-p-dioxin 3.085 67028197 PCDD 1,2,3,4,6-pentachlorodibenzo-p-dioxin 3.479 71669255 PCDD 1,2,3,6-tetrachlorodibenzo-p-dioxin 3.085 71669266 PCDD 1,2,3,9-tetrachlorodibenzo-p-dioxin 3.085 22274426 PCDE 2,3,4,5,6-pentachlorodiphenylether 3.467 24910734 PCDE 3,4',5-trichlorodiphenyl ether 2.598 28076735 PCDE 2,2'4,4'-tetrachlorodiphenyloxide 3.058 31242930 PCDE hexachlorodiphenyl oxide 3.832 31242941 PCDE tetrachlorodiphenyl ether 3.058 42279298 PCDE pentachlorophenyl ether 3.467 52322802 PCDE benzene, 1,2,4-trichloro-5-phenoxy- 2.598 56348722 PCDE 3,3',4,4'-tetrachlorobiphenyl ether 3.058 57321638 PCDE trichlorophenyl ether 3.832 59039213 PCDE 2,4,4'-trichlorodiphenyl ether 2.598 60123640 PCDE 2,2',4,4',5-pentachlor-diphenylether 3.467 60123651 PCDE 2,3',4,4',5-pentachlorodiphenyl ether 3.467 61328447 PCDE 1,2-dichloro-4-(2-chlorophenoxy)benzene 2.598 61328458 PCDE 2,4,4',5-tetrachlorodiphenyl ether 3.058 63553300 PCDE 2,4,4',6-tetrachlorodiphenyl ether 3.058 63646515 PCDE 3,4,4'-trichlorodiphenyl ether 2.598 63646559 PCDE 2,3,4',5,6-pentachlorodiphenylether 3.467 65075005 PCDE 2,4',5-trichlorodiphenyl ether 2.598 65075016 PCDE 2,3,4,4'-tetrachlorodiphenyl ether 3.058 66794603 PCDE 3,3',4-trichlorodiphenyl ether 2.598 68914976 PCDE 2,2',4-trichlorodiphenyl ether 2.598 71585370 PCDE 2,2',3,4,4'-pentachlorodiphenylether 3.467 71585381 PCDE 2,2',3,4,4',5-hexachlorodiphenylether 3.832 71585392 PCDE 2,2',3,3',4,4'-hexachlorodiphenylether 3.832 71859308 PCDE 2,2',4,4',5,5'-hexachlorodiphenylether 3.832 85918316 PCDE 2,3,3',4,4'-pentachlorodiphenylether 3.467 85918338 PCDE 2,3,4,6-tetrachlorodiphenyl ether 3.058 85918350 PCDE 2,2',3,4,6'-pentachlorodiphenylether 3.467
56
85918372 PCDE 2,2',3,4',5',6-hexachlorodiphenylether 3.832 94339590 PCDE 3,3',4,4',5-pentachlorodiphenylether 3.467
104294168 PCDE 2,2',4,4',6-pentachlorodiphenylether 3.467 106220819 PCDE 2,2',4,4',5,6'-hexachlorodiphenylether 3.832 113464178 PCDE 2,3,4,4',5-pentachlorodiphenylether 3.467 116995201 PCDE 2,2',3,4',6-pentachlorodiphenylether 3.467 124076660 PCDE 2,2',3,3',4,6'-hexachlorodiphenylether 3.832 130892669 PCDE 2,3',4',6-tetrachlorodiphenyl ether 3.058 130892670 PCDE 2,2',4,5,6'-pentachlorodiphenylether 3.467 131138211 PCDE 2,2',4,5,5'-pentachlorodiphenylether 3.467 147102634 PCDE 2,2',3,4'-tetrachlorodiphenyl ether 3.058 147102645 PCDE 2,3',4,5'-tetrachlorodiphenyl ether 3.058 155999921 PCDE 2,2',4,5'-tetrachlorodiphenyl ether 3.058 155999932 PCDE 2,3',4-trichlorodiphenyl ether 2.598 155999976 PCDE 2,3,3',4',5,6-hexachlorodiphenylether 3.832 157683711 PCDE 2,3,4'-trichlorodiphenyl ether 2.598 157683722 PCDE 2,4',6-trichlorodiphenyl ether 2.598 157683733 PCDE 2,2',3,4',5-pentachlorodiphenylether 3.467 157683744 PCDE 2,3',4,4',6-pentachlorodiphenylether 3.467 159553681 PCDE 2,3,3',4',5-pentachlorodiphenylether 3.467 159553692 PCDE 2,3,3',4',6-pentachlorodiphenylether 3.467 160282042 PCDE 2,3',4,5,5'-pentachlorodiphenylether 3.467 160282053 PCDE 2,3,4,4',6-pentachlorodiphenylether 3.467 160282075 PCDE 2,3,3',4,5'-pentachlorodiphenylether 3.467 160282086 PCDE 2,2',3',4,5-pentachlorodiphenylether 3.467 160282097 PCDE 2,2',3,4,5'-pentachlorodiphenylether 3.467 160282100 PCDE 2,2',3,3',4-pentachlorodiphenylether 3.467 162853249 PCDE 2,3,5-trichlorodiphenyl ether 2.598 162853250 PCDE 2,3,6-trichlorodiphenyl ether 2.598 162853261 PCDE 2,2',4,5-tetrachlorodiphenyl ether 3.058 162853272 PCDE 2,3,3',4'-tetrachlorodiphenyl ether 3.058 162853283 PCDE 2,2',3,4',5,5'-hexachlorodiphenylether 3.832 24478726 PCDF dibenzofuran, 1,2,3,4-tetrachloro- 3.779 24478737 PCDF 1,2,4-trichlorodibenzofuran 3.319 30402143 PCDF 1,2,3,4-tetrachlorodibenzofuran 3.779 30402154 PCDF pentachlorodibenzofuran 4.188 43048006 PCDF trichlorodibenzofuran 3.319 51207319 PCDF 2,3,7,8-tetrachlorodibenzofuran 3.779 54589718 PCDF 2,4,8-trichlorodibenzofuran 3.319 55722275 PCDF tetrachlorodibenzofuran 3.779 57117325 PCDF 2,3,8-trichlorodibenzofuran 3.319 57117336 PCDF 2,3,6-trichlorodibenzofuran 3.319 57117347 PCDF 2,3,4-trichlorodibenzofuran 3.319 57117358 PCDF 1,3,7,8-tetrachlorodibenzofuran 3.779 57117369 PCDF 1,3,6,7-tetrachlorodibenzofuran 3.779 57117370 PCDF dibenzofuran, 2,3,6,8-tetrachloro- 3.779 57117381 PCDF dibenzofuran, 2,4,6,7-tetrachloro- 3.779 57117392 PCDF dibenzofuran, 2,3,6,7-tetrachloro- 3.779 57117405 PCDF 3,4,6,7-tetrachlorodibenzofuran 3.779 57117427 PCDF dibenzofuran, 1,2,3,6,7-pentachloro- 4.188 57117438 PCDF 2,3,4,6,7-pentachlorodibenzofuran 4.188 58802145 PCDF 2,4,6-trichlorodibenzofuran 3.319 58802156 PCDF 1,2,4,7,8-pentachlorodibenzofuran 4.188 58802167 PCDF 1,3,4,7,8-pentachlorodibenzofuran 4.188 58802178 PCDF 2,3,7-trichlorodibenzofuran 3.319 58802189 PCDF 1,7,8-trichlorodibenzofuran 3.319 58802190 PCDF 2,4,6,8-tetrachlorodibenzofuran 3.779 58802203 PCDF dibenzofuran, 1,2,7,8-tetrachloro- 3.779 62615081 PCDF 1,2,3,8-tetrachlorodibenzofuran 3.779 64126870 PCDF 1,2,4,8-tetrachlorodibenzofuran 3.779 64560141 PCDF 1,4,8-trichlorodibenzofuran 3.319 64560152 PCDF 1,2,6-trichlorodibenzofuran 3.319 64560163 PCDF 1,3,7-trichlorodibenzofuran 3.319 64560174 PCDF dibenzofuran, 1,3,7,9-tetrachloro- 3.779 66794590 PCDF 1,4,6,7-tetrachlorodibenzofuran 3.779 67481225 PCDF dibenzofuran, 2,3,4,6,8-pentachloro- 4.188 69433007 PCDF 1,2,6,7,8-pentachlorodibenzofuran 4.188 69698573 PCDF 1,2,4,6,8-pentachlorodibenzofuran 4.188 70648134 PCDF 1,4,9-trichlorodibenzofuran 3.319 70648156 PCDF 1,3,4,6,9-pentachlorodibenzofuran 4.188
57
70648167 PCDF dibenzofuran, 1,3,4,7-tetrachloro- 3.779 70648189 PCDF 1,2,6,9-tetrachlorodibenzofuran 3.779 70648190 PCDF 1,4,6,9-tetrachlorodibenzofuran 3.779 70648203 PCDF 1,3,4,7,9-pentachlorodibenzofuran 4.188 70648214 PCDF 1,3,6,7,8-pentachlorodibenzofuran 4.188 70648225 PCDF 1,2,8,9-tetrachlorodibenzofuran 3.779 70648236 PCDF 1,2,4,8,9-pentachlorodibenzofuran 4.188 70648247 PCDF dibenzofuran, 1,2,4,6,9-pentachloro- 4.188 70872821 PCDF dibenzofuran, 1,2,6,7,9-pentachloro- 4.188 71998726 PCDF dibenzofuran, 1,3,6,8-tetrachloro- 3.779 71998737 PCDF 1,2,4,6-tetrachlorodibenzofuran 3.779 71998748 PCDF 1,2,4,7,9-pentachlorodibenzofuran 4.188 76621120 PCDF 1,3,8-trichlorodibenzofuran 3.319 82911588 PCDF 1,4,6,8-tetrachlorodibenzofuran 3.779 82911599 PCDF 1,6,8-trichlorodibenzofuran 3.319 82911602 PCDF 1,4,6-trichlorodibenzofuran 3.319 82911613 PCDF 1,3,4-trichlorodibenzofuran 3.319 83636479 PCDF 1,2,3-trichlorodibenzofuran 3.319 83690986 PCDF 1,3,6,9-tetrachlorodibenzofuran 3.779 83704216 PCDF 1,2,3,6-tetrachlorodibenzofuran 3.779 83704227 PCDF 1,2,3,7-tetrachlorodibenzofuran 3.779 83704238 PCDF 1,2,3,9-tetrachlorodibenzofuran 3.779 83704249 PCDF 1,2,4,9-tetrachlorodibenzofuran 3.779 83704250 PCDF 1,2,6,7-tetrachlorodibenzofuran 3.779 83704261 PCDF 1,2,7,9-tetrachlorodibenzofuran 3.779 83704272 PCDF dibenzofuran, 1,3,4,6-tetrachloro- 3.779 83704283 PCDF 1,3,4,9-tetrachlorodibenzofuran 3.779 83704294 PCDF 1,4,7-8-tetrachlorodibenzofuran 3.779 83704307 PCDF dibenzofuran, 2,3,4,6-tetrachloro- 3.779 83704318 PCDF 2,3,4,7-tetrachlorodibenzofuran 3.779 83704329 PCDF 2,3,4,8-tetrachlorodibenzofuran 3.779 83704330 PCDF 1,6,7,8-tetrachlorodibenzofuran 3.779 83704341 PCDF 1,2,8-trichlorodibenzofuran 3.319 83704363 PCDF 1,3,4,6,7-pentachlorodibenzofuran 4.188 83704374 PCDF 1,2,7-trichlorodibenzofuran 3.319 83704385 PCDF 1,2,9-trichlorodibenzofuran 3.319 83704396 PCDF 1,3,6-trichlorodibenzofuran 3.319 83704409 PCDF 1,3,9-trichlorodibenzofuran 3.319 83704410 PCDF 1,4,7-trichlorodibenzofuran 3.319 83704421 PCDF 2,4,7-trichlorodibenzofuran 3.319 83704432 PCDF 3,4,6-trichlorodibenzofuran 3.319 83704443 PCDF 3,4,7-trichlorodibenzofuran 3.319 83704454 PCDF 2,6,7-trichlorodibenzofuran 3.319 83704465 PCDF 1,6,7-trichlorodibenzofuran 3.319 83704501 PCDF 1,2,4,6,7-pentachlorodibenzofuran 4.188 83704512 PCDF 1,2,3,6,8-pentachlorodibenzofuran 4.188 83704534 PCDF 1,2,3,7,9-pentachlorodibenzofuran 4.188 83704545 PCDF 1,2,3,8,9-pentachlorodibenzofuran 4.188 83704556 PCDF 1,3,4,6,8-pentachlorodibenzofuran 4.188 83710070 PCDF 1,2,6,8-tetrachlorodibenzofuran 3.779 83719408 PCDF 1,2,4,7-tetrachlorodibenzofuran 3.779 92341043 PCDF 1,3,4,8-tetrachlorodibenzofuran 3.779 1321648 PCN naphthalene, pentachloro- yes 4.492 1321659 PCN naphthalene, trichloro- 3.505 1335871 PCN naphthalene, hexachloro- yes 4.893 1335882 PCN naphthalene, tetrachloro- 4.033 20020024 PCN 1,2,3,4-tetrachloronaphthalene 4.033 31604281 PCN 1,3,5,8-tetrachloronaphthalene 4.033 50402512 PCN 1,2,4-trichloronaphthalene 3.505 50402523 PCN 1,2,3-trichloronaphthalene 3.505 53555638 PCN 1,2,3,5-tetrachloronaphthalene 4.033 53555649 PCN 1,3,5,7-tetrachloronaphthalene 4.033 53555650 PCN 1,2,3,5,7-pentachloronaphthalene yes 4.492 55720371 PCN 1,3,7-trichloronaphthalene 3.505 55720406 PCN 2,3,6-trichloronaphthalene 3.505 55720439 PCN 1,4,6,7-tetrachloronaphthalene 4.033 67922218 PCN 1,2,4,7-tetrachloronaphthalene 4.033 67922263 PCN 1,2,3,4,6-pentachloronaphthalene yes 4.492 90948280 PCN 1,2,4,5,6,8-hexachloronaphthalene yes 4.893
103426977 PCN 1,2,3,5,6,7-hexachloronaphthalene yes 4.893
58
150224241 PCN 1,2,3,5,8-pentachloronaphthalene yes 4.492 29086382 PCS Trans-Heptachlorostyrene 4.246 29086393 PCS Cis-Heptachlorostyrene 4.246 61255810 PCS Heptachlorostyrene 4.246
355464 PFA 1-hexanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,6-tridecafluoro- yes 4.849
375928 PFA 1-Heptanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,7-pentadecafluoro- yes 5.375
423541 PFA Octanamide, 2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro- (7CI,8CI,9CI) yes 5.184
1763231 PFA 1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro- yes 5.407
2795393 PFA 1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-, potassium salt yes 5.218
41006411 PFA 1-Butanesulfinyl chloride, 1,1,2,2,3,3,4,4,4-Nonafluoro- 3.170
45298906 PFA 1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-, ion(1-) yes 5.407
51735843 PFA 1-Butanesulfinamide, 1,1,2,2,3,3,4,4,4-Nonafluoro- 3.583
68259074 PFA 1-Heptanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,7-pentadecafluoro-, ammonium salt yes 5.112
68259085 PFA 1-hexanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,6-tridecafluoro-, ammonium salt yes 5.036
68259096 PFA 1-pentanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,5-undecafluoro-,ammonium salt yes 4.451
68259121 PFA 1-Nonanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,9-nonadecafluoro- yes 5.066
68555679 PFA 1-Octanesulfinic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-, sodium salt yes 5.464
50293 POP p,p'-DDT 2.810 53190 POP o,p'-DDD 2.416 57749 POP chlordane yes 6.498 58899 POP gamma-HCH 4.131 72548 POP p,p'-DDD 2.416
118741 POP HCB 4.131 143500 POP chlordecone yes 8.666 319846 POP alpha-HCH 4.131 319857 POP beta-HCH 4.131 319868 POP delta-HCH 4.131 608731 POP hexachlorocyclohexane 4.131 789026 POP o,p'-DDT 2.810
1024573 POP Heptachlor epoxide yes 6.973 2385855 POP Mirex yes 9.345 3224826 POP o,p'-DDE 2.416
4329071 POP (m,p'-DDT) benzene, 1-chloro-3-[2,2,2-trichloro-1-(4-chlorophenyl)ethyl]- (9CI) 2.810
5103719 POP cis-chlordane (alpha-) yes 6.498 5103731 POP Cis-Nonachlor yes 6.780 5103742 POP trans-chlordane yes 6.498
6108107 POP Cyclohexane, 1,2,3,4,5,6-hexachloro-, (1.alpha.,2.alpha.,3.alpha.,4.beta.,5.beta.,6.beta.)- 4.131
6108118 POP ZETA-hexachlorocyclohexane 4.131 6108129 POP ETA-hexachlorocyclohexane 4.131 8001352 POP Toxaphene yes 5.561 8017343 POP p,p'-DDT 2.810 12789036 POP chlordane yes 5.964 21161580 POP 1,2,3,6,9,10,10-Heptachloropentacyclodecane yes 8.058 27154445 POP hexachlorocyclohexane 4.131 27304138 POP Oxychlordane yes 7.283 33442830 POP Photoheptachlor yes 8.058 39765805 POP Trans-nonachlor yes 6.780
39801144 POP 1,3,4-metheno-1H-cyclobuta[cd]pentalene, 1,1a,2,2,3,3a,4,5,5,5a,5b-undecachlorooctahydro- (9CI) yes 9.137
55570848 POP 1,3,4-metheno-1H-cyclobuta[cd]pentalene, 1a,2,2,3,3a,4,5,5,5a,6-decachlorooctahydro- (9CI) yes 8.907
115384946 POP 1,2,3,4,5,6,7,8,8,-Nonachlorooctahydro-4,7-METH* yes 6.780 87821 PXBD benzene, hexabromo- 4.131 87832 PXBD benzene, pentabromomethyl- 3.519
319879 PXBD benzene, pentachlorofluoro- 4.131 392574 PXBD benzene, 1,2,4,5-tetrafluoro-3,6-diiodo- 4.131 576556 PXBD phenol, 2,3,4,5-tetrabromo-6-methyl- 2.962
59
607954 PXBD phenol, 2,4,6-tribromo-, acetate (8CI,9CI) 2.567 608219 PXBD benzene, 1,2,3-tribromo- 2.526 608297 PXBD benzene, 1,2,3-triiodo- 2.526 608902 PXBD pentabromobenzene 3.693 615543 PXBD benzene, 1,2,4-tribromo- 2.526 626391 PXBD benzene, 1,3,5-tribromo- 2.526 626448 PXBD 1,3,5-triiodoenzene 2.526 634684 PXBD 1,2,3,4-tetraiodobenzene 3.168 634899 PXBD 1,2,3,5-tetrabromobenzene 3.168 634924 PXBD 1,2,3,5-tetraiodobenzene 3.168 636282 PXBD benzene, 1,2,4,5-tetrabromo- 3.168 636317 PXBD 1,2,4,5-tetraiodobenzene 3.168 832531 PXBD benzenesulfonyl chloride, pentafluoro- (7CI,8CI,9CI) 3.245
2338207 PXBD benzoic acid, 3,4,5-triiodo- (7CI,8CI,9CI) 2.443 2401210 PXBD benzene, 1,2-dichloro-3-iodo- (7CI,8CI,9CI) 2.526 2708976 PXBD benzene, 1,2,3,4-tetrafluoro-5,6-diiodo- (9CI) 4.131 2810697 PXBD benzene, 1,2,3,4-tetrabromo-5,6-dimethyl- (9CI) 2.962 3032813 PXBD benzene, 1,3-dichloro-5-iodo- 2.526 5075616 PXBD phosphine, dimethyl(pentafluorophenyl)- 3.475 13075019 PXBD benzene, 1,2,4-tribromo-3,5,6-trichloro- (6CI,7CI,8CI,9CI) 4.131 13311723 PXBD phenol, 4-bromo-2,3,6-trichloro- (7CI,8CI,9CI) 3.008 14400943 PXBD 2,3,4,6-tetrabromophenol 3.008 14862523 PXBD benzene, 1,3-dibromo-5-chloro- 2.526 15396368 PXBD benzoic acid, 2-chloro-3,5-diiodo- (8CI,9CI) 2.443 17274123 PXBD benzoic acid, 2,3,5-triiodo-, sodium salt (8CI,9CI) 2.443 19393921 PXBD 1-bromo-2,6-dichlorobenzene 2.526 19752557 PXBD benzene, 1-bromo-3,5-dichloro- 2.526 20020193 PXBD benzenamine, 2,4,6-tribromo-N-nitro- 2.567 20555913 PXBD benzene, 1,2-dichloro-4-iodo- 2.526 22311257 PXBD 1,2,3,4-tetrabromobenzene 3.168 23488382 PXBD benzene, 1,2,4,5-tetrabromo-3,6-dimethyl- 2.962 25376389 PXBD phenol, tribromo- 2.526 28779080 PXBD tribromobenzene 2.526 29682415 PXBD benzene, 1,4-dichloro-2-iodo- (8CI,9CI) 2.526 29682448 PXBD benzene, 1-bromo-2,4,5-trichloro- (6CI,8CI,9CI) 3.168 29898326 PXBD benzene, 2,4-dichloro-1-iodo- (8CI,9CI) 2.526 35488176 PXBD 2,3,4,5-tetrabromo-6-methoxyphenol 2.999 36059219 PXBD tetrabrom-O-xylol 2.962 38926851 PXBD 2,3,4-tribromo-6-methoxyphenol 2.443 39569216 PXBD tetrabromo-2-chlorotoluene 3.519 41424366 PXBD benzene, 1,3,5-tribromo-2-methoxy-4-methyl- 2.443 46495663 PXBD benzene, 1-chloro-2,3,5-trifluoro-4,6-diisocyanato- (9CI) 2.952 56759569 PXBD 1,2-benzenediol, 3,4,6-tribromo-5-methyl- (9CI) 2.443 56961774 PXBD 1-bromo-2,3-dichlorobenzene 2.526 58169996 PXBD phenol, 2,3,4,6-tetrabromo-5-methyl- 2.962 58683708 PXBD benzene, 1,2,4-tribromo-5-(1-methylethyl)- (9CI) 2.443 61878555 PXBD triiodobenzene 2.526 62778192 PXBD benzene, 1,3-dichloro-2-iodo-5-nitro- 2.443 63884435 PXBD 1,4-benzenediol, 2,3,5-tribromo-6-methyl- (9CI) 2.443 68084297 PXBD phenol, pentabromo-, aluminum salt 3.210 90077784 PXBD bromotetrachlorobenzene 3.693
115172162 PXBD dibromodichlorobenzene 3.168 434902 PXBP Decafluorobiphenyl yes 5.701
3883861 PXBP 1,1'-biphenyl, 2,2',3,3',5,5',6,6'-octafluoro- yes 5.176
10386842 PXBP 1,1'-biphenyl, 4,4'-dibromo-2,2',3,3',5,5',6,6'-octafluoro- (9CI) yes 5.701
59080330 PXBP 2,4,6-tribromobiphenyl 3.251 50585427 PXDD 2,3-dichloro-7,8-difluorodibenzo-p-dioxin 3.085 84761819 PXDF 1,2,8-tribromodibenzofuran 3.319 84761820 PXDF 2,3,8-tribromodibenzofuran 3.319
107227565 PXDF bromotrichlorodibenzofuran 3.779 18134582 silane silane, trichloro(1,4,5,6,7,7-hexachloro-5-norbornen-2-yl)- 4.323
68360 XAB benzene, 1,4-bis(trichloromethyl)- 3.879 81196 XAB benzene, 1,3-dichloro-2-(dichloromethyl)- 3.008
134258 XAB benzene, 2,4-dichloro-1-(dichloromethyl)- 3.008 401774 XAB benzene, 1-fluoro-3-(trichloromethyl)- (9CI) 3.008 402426 XAB benzene, 1-fluoro-4-(trichloromethyl)- (9CI) 3.008 488982 XAB benzene, 1-fluoro-2-(trichloromethyl)- 3.008 881992 XAB alpha,alpha'-hexachloro-m-xylene 3.879
1079170 XAB benzene, 1,2,4,5-tetrachloro-3,6-bis(chloromethyl)- (9CI) 3.879
60
1133579 XAB benzene, 1,2,3,5-tetrachloro-4,6-bis(chloromethyl)- (9CI) 3.879 1424799 XAB benzene, 1,2,4-trichloro-3-(chloromethyl)- 3.008 1592310 XAB benzene, 1,4-bis(dibromomethyl)- 2.962 2136814 XAB benzene, 1-chloro-3-(trichloromethyl)- 3.008 2136892 XAB benzene, 1-chloro-2-(trichloromethyl)- 3.008 2136950 XAB benzene, pentachloro(dichloromethyl)- (9CI) 4.321 2741573 XAB benzene, 1-(dichloromethyl)-2-(trichloromethyl)- (9CI) 3.456 3955268 XAB benzene, 1,2,4-trichloro-5-(chloromethyl)- 3.008 4960489 XAB benzene, 2-(bromomethyl)-1,3,4-trichloro- (9CI) 3.008 5216251 XAB benzene, 1-chloro-4-(trichloromethyl)- 3.008 7398825 XAB benzene, 1,4-bis(dichloromethyl)- (9CI) 2.962 10541716 XAB benzene, 1,4-dichloro-2-(trichloromethyl)- (9CI) 3.519 13014181 XAB benzene, 2,4-dichloro-1-(trichloromethyl)- 3.519 13014249 XAB benzene, 1,2-dichloro-4-(trichloromethyl)- 3.519 13209159 XAB benzene, 1,2-bis(dibromomethyl)- 2.962 13911029 XAB benzene, 1,2,3-trichloro-4-(chloromethyl)- (9CI) 3.008 17293037 XAB benzene, 1,3,5-trichloro-2-(chloromethyl)- (9CI) 3.008 17299977 XAB 1,3,5-tri(alpha-chloromethyl)benzene 2.443 23691272 XAB benzene, (1,3,3,3-tetrachloropropyl)- 2.999 24653858 XAB benzene, 1,2,4-trichloro-3-(dichloromethyl)- (9CI) 3.519 25393980 XAB benzene, 1,4-bis(1,2-dibromoethyl)- (9CI) 3.096 25641990 XAB benzene, 1,2-bis(dichloromethyl)- 2.962 25850491 XAB benzene, 1,3-bis(1,2-dibromoethyl)- (9CI) 3.096 28744458 XAB benzene, 1,3,5-trichloro-2-(dichloromethyl)- (9CI) 3.519 30359536 XAB benzene, 1-(2,2,2-trichloroethyl)-3-(trifluoroM* 3.887 31904184 XAB Toluene, .alpha.,.alpha.,.alpha.,ar,ar-pentachloro- 3.519 33429708 XAB benzene, 1,2,4-trichloro-5-(dichloromethyl)- (9CI) 3.519 36323281 XAB benzene, 1,3-bis(dibromomethyl)- 2.962 41999842 XAB benzene, 1,4-dichloro-2,5-bis(dichloromethyl)- (9CI) 3.879 54965401 XAB benzene, 1-chloro-4-(2,2-dichloro-1-methylethyl)- 2.443 56682872 XAB 1,3,5-triS(dichloromethyl)benzene 3.887 56762233 XAB benzene, (1,2,3-tribromopropyl)- (9CI) 2.443 56961821 XAB benzene, 1,2,3-trichloro-4-(dichloromethyl)- (9CI) 3.519 56961832 XAB benzene, 1,4-dichloro-2-(dichloromethyl)- (9CI) 3.008 56961843 XAB benzene, 1,2-dichloro-4-(dichloromethyl)- 3.008 56961854 XAB benzene, 1,3-dichloro-5-(dichloromethyl)- (9CI) 3.008 68238937 XAB benzene, dichloro(dichloromethyl)- 3.008
73588428 XAB benzene, 1,3-dichloro-5-(1,3,3,3-tetrachloro-1-methylpropyl)- 3.956
93962664 XAB benzene, 1-(3-bromopropyl)-2,4-dichloro- (9CI) 2.443 94022987 XAB benzene, 1-(2-iodoethyl)-2-(trifluoromethyl)- (9CI) 2.999 95998644 XABP chloro-AR,AR'-bis(trifluoromethyl)-1,1'-biphenyl 3.485 95998666 XABP trichloro-AR,AR'-bis(trifluoromethyl)-1,1'-biphenyl* 4.043 99686529 XABP dichlorobis(trifluoromethyl)biphenyl 3.777 346554 XAH Quinoline, 4-chloro-7-(trifluoromethyl)- 4.093 613536 XAH Quinoline, 2-(tribromomethyl)- 3.589 949428 XAH 1,3,5-triazine, 2-methyl-4,6-bis(trichloromethyl)- 3.887 956387 XAH 2-(2-bromoethyl)-4,6-bis(trichloromethyl)-S-TR* 4.309
1128161 XAH Pyridine, 3,5-dichloro-2-(trichloromethyl)- 3.519 1129197 XAH Pyridine, 2,4-dichloro-6-(trichloromethyl)- 3.519 1134049 XAH Pyridine, 2,3,4,5-tetrachloro-6-(trichloromethyl)- 4.321 1145444 XAH 2-(2-chloroethyl)-4,6-bis(trichloromethyl)-S-T* 4.309 1201305 XAH Pyridine, 3,4,5-trichloro-2-(trichloromethyl)- 3.950 1817136 XAH Pyridine, 3,6-dichloro-2-(trichloromethyl)- 3.519 2338309 XAH 4567-tetraBr-2-CF3 benzimidazole yes 5.247 3393597 XAH 4,5,7-triCL-2-CF3-benzimidazole yes 4.893 3599711 XAH 2-ethyl-4,6-bis(trichloromethyl)-S-triazine 3.956 3599744 XAH 2,4-bis(trichloromethyl)-S-triazine 3.879 3599766 XAH 2-methylthio-4,6-bis(trichloromethyl)-s-triazine 3.956 3671601 XAH 2-CF3-5-bromobenzimidazole 4.033 5311251 XAH 2,4-dimethoxy-6-trichloromethyl-S-triazine 2.737 5516483 XAH 2-methylthio-4,6-bis(1,1,2-trichloroethyl)-S-T* 3.754 6542672 XAH 2,4,6-triS(trichloromethyl)-S-triazine yes 4.850 6587219 XAH 2-CF3-5,6-dibromo-benzimidazole yes 4.492 7041249 XAH Pyridine, 2,3,5-trichloro-6-(trichloromethyl)- 3.950 7682328 XAH benzimidazole-4,5,6-triBr-2-CF3 yes 4.893 10243831 XAH N,N'-(6-chloro-S-triazine-2,4-diYL)bis(2,2,2-*) yes 4.670 13704900 XAH 2-chloro-4-phenyl-6-trichloromethyl-S-triazine 3.058 14946180 XAH 2,4-bis(methylthio)-6-trichloromethyl-S-triazine 2.737 21384338 XAH 2-(1-aziridinyl)-4,6-bis(trichloromethyl)-S-triazine* yes 4.923
61
22652148 XAH Pyridine, 2,6-dichloro-4-(trichloromethyl)- 3.519 23779977 XAH Quinoline, 4-chloro-8-(trifluoromethyl)- (8CI,9CI) 4.093 24481332 XAH 2-(1,1,2-trichloroethyl)-4,6-bis(trichloroMET*) yes 4.905 24481376 XAH 2-PenTYL-4,6-bis(trichloromethyl)-S-triazine 3.112 24481401 XAH 2-isobutyl-4,6-bis(trichloromethyl)-S-triazine 3.754 24481412 XAH 2-butyl-4,6-bis(trichloromethyl)-S-triazine 3.754 24481423 XAH 2-isopropyl-4,6-bis(trichloromethyl)-S-triazine 4.071 24481434 XAH 2-propyl-4,6-bis(trichloromethyl)-S-triazine 4.071 24481558 XAH 2-(O-chlorophenyl)-4,6-bis(trichloromethyl)-S-* 3.485 24504221 XAH 2-phenyl-4,6-bis(trichloromethyl)-S-triazine 3.164 24802855 XAH 2-(tert-butylamino)-4,6-bis(trichloromethyl)-S* 3.112 24803052 XAH 2-dimethylamino-4,6-bis(trichloromethyl)-S-triazine* 4.071 24803245 XAH 2-methylamino-4-propyl-6-trichloromethyl-S-triazine* 2.470 24803267 XAH 2-isopropyl-4-methylamino-6-trichloromethyl-S-triazine* 2.470 24803643 XAH 2-methylamino-4,6-bis(trichloromethyl)-S-triazine* 3.956 30339505 XAH 2-amino-4-(1,1,2-trichloroethyl)-6-trichloroMET* 3.956 30339572 XAH 2-amino-4,6-bis(1,1-dichloroethyl)-S-triazine 3.241 30339583 XAH 2,4-bis(1,1-dichloroethyl)-6-methylamino-S-triazine* 2.952 30339607 XAH 2-amino-4,6-bis(1,1,2-trichloroethyl)-S-triAZI* 4.071 30339709 XAH 2-amino-4-butyl-6-trichloromethyl-S-triazine 2.470 30339732 XAH 2-amino-4-(SEC-butyl)-6-trichloromethyl-S-triazine* 2.470 30339798 XAH 2-amino-4-isobutyl-6-trichloromethyl-S-triazine 2.470 30361978 XAH 2-(chloromethyl)-4,6-bis(trichloromethyl)-S-TR* 4.248 30362313 XAH 2-(dichloromethyl)-4,6-bis(trichloromethyl)-S-* yes 4.566 30362448 XAH 4,6-bis(trichloromethyl)-S-triazine-2-Methanol* 3.112 30362459 XAH alpha-methyl-4,6-bis(trichloromethyl)-S-trazine 2.495 30362620 XAH 2,4-diethyl-6-(trichloromethyl)-S-triazine 2.737 30362744 XAH 2,4,6-triS(1,1-dichloroethyl)-S-triazine 3.754 30362788 XAH 2,4,6-triS(1-bromoethyl)-S-triazine 2.470 30369154 XAH 2-amino-4-(2,4-dichlorophenyl)-6-(trichloroME*) 2.810 30377243 XAH 2-amino-4-(P-chlorophenyl)-6-trichloromethyl-S-triazine 2.416 30863004 XAH 2,4-bis(1,1-dichloroethyl)-6-(methylthio)-S-triazine 2.952 30863220 XAH ethyl 4,6-bis(trichloromethyl)-S-triazine-2-carboxylic acid* 3.112 30863253 XAH 4,6-bis(trichloromethyl)-S-triazin-2-YL isothiozole 4.071 30863515 XAH 2-methoxy-4,6-bis(trichloromethyl)-S-triazine 3.956 30894894 XAH 2-chloro-4,6-bis(trichloromethyl)-S-triazine 4.246 30894907 XAH 2-chloro-4,6-bis(1,1-dichloroethyl)-S-triazine 3.554 30894918 XAH 2-chloro-4,6-bis(1,1,2-trichloroethyl)-S-triazine 4.309 30894985 XAH 2,4-dibromo-6-(1,1,2-trichloroethyl)-S-triazine 3.456 30894996 XAH 2-bromo-4,6-bis(1,1,2-trichloroethyl)-S-triazine 4.309 31120237 XAH 2,4-dimethoxy-6-(1,1,2-trichloroethyl)-S-triazine 2.470 51492014 XAH Pyridine, 2,3-dichloro-6-(trichloromethyl)- 3.519 55366308 XAH 2,6-dichloro-3-(trichloromethyl)pyridine 3.519 69045836 XAH 2,3-dichloro-5-trichloromethylpyridine 3.519 70788544 XAH Pyridine, 2,3,4-trichloro-6-(trichloromethyl)- 3.950
108030779 XAH Pyridine, 3,5-dichloro-2,4-dimethoxy-6-(trichloromethyl)- 3.682 1573575 XAlka 1,2,2,3,3,4-hexachlorobutane 2.705 2431552 XAlka 1,1,2,2,3,4-hexachlorobutane 2.705 6820742 XAlka Perchloroisobutane 3.805 18585381 XAlka 1,2,3,4,5,6-hexachlorohexane 2.569 18791190 XAlka 1,1,1,2,3,4,4,4-Octachlorobutane 3.346 20338265 XAlka 1,1,2,2,3,3,4,4-Octachlorobutane 3.346 21483625 XAlka 1,1,1,2,2,3,3,4,4-Nonachlorobutane 3.593 26523637 XAlka Butane, hexachloro- (6CI,7CI,8CI,9CI) 2.705 32694761 XAlka 1,1,1,2,3,3,4,4-Octachlorobutane 3.346 34973416 XAlka 1,1,2,2,3,4,4-Heptachlorobutane 3.054 68512163 XAlka Decane, brominated chlorinated 2.617 79458541 XAlka 1,1,1,4,4,4-hexachlorobutane 2.705 83682284 XAlka 2,2,3,4,5,5-hexachlorohexane 2.569 83682295 XAlka 1,2,2,5,5,6-hexachlorohexane 2.569 83682693 XAlka 1,1,2,2,3,3-hexachlorobutane 2.705 83682706 XAlka 1,1,1,2,2,3,3-Heptachlorobutane 3.054
87843 XAlkaC Cyclohexane, 1,2,3,4,5-pentabromo-6-chloro- 4.131 707551 XAlkaC Cyclohexane, 1,1,2,3,4,5,6-heptachloro- (6CI,7CI,8CI,9CI) yes 4.501
1837918 XAlkaC Cyclohexane, 1,2,3,4,5,6-hexabromo- 4.131 3194578 XAlkaC 1,2,5,6-tetrabromocyclooctane 2.962 3322938 XAlkaC Cyclohexane, 1,2-dibromo-4-(1,2-dibromoethyl)- 2.962 22138392 XAlkaC 1-pentachlorocyclohexane 3.693 30554724 XAlkaC Cyclohexane, tetrabromodichloro- 4.131 30554735 XAlkaC Cyclohexane, tribromotrichloro- 4.131
62
31454485 XAlkaC tetrabromocyclooctane 2.962 33489279 XAlkaC 3-I-pentachlorocyclohexane 4.131 33489280 XAlkaC 1-I-pentachlorocyclohexane 4.131 36635026 XAlkaC 2-Br-pentachlorocyclohexane 4.131 36635037 XAlkaC 1-Br-pentachlorocyclohexane 4.131 51962631 XAlkaC 4-(1,2-dichloroethyl)-1,2-dichlorocyclohexane 2.962 55265695 XAlkaC 2,4-diBr-tetrachlorocyclohexane 4.131 55298458 XAlkaC 1,2-diBr-tetrachlorocyclohexane 4.131 55332893 XAlkaC 2,3-diBr-tetrachlorocyclohexane 4.131 56421444 XAlkaC 1-methylpentachlorocyclohexane 3.519 56421455 XAlkaC 1,4-dimethyltetrachlorocyclohexane 2.962 68258902 XAlkaC Cyclopentane, heptachloro- yes 4.832 68258913 XAlkaC Cyclopentane, hexachloro- yes 4.467 83682626 XAlkaC 1,2-bis(1,2-dichloroethyl)cyclobutane 2.962 83682648 XAlkaC 1,2,4-trichloro-4-(1,1,2-trichloroethyl)cyclohexaNE 3.879 376794 XAlke 1-Hexene, 1,3,4,5,6,6-hexachloro-1,2,3,4,5,6-hexafluoro- 4.173
1725742 XAlke 1,2,3,4,5,6-hexachloro-3-hexene 2.569 2482680 XAlke 2-Perchlorobutene 3.346 3050428 XAlke Perchlorobutene 3.346 36678452 XAlke 2-Butene, 1,1,2,3,4,4-hexabromo- 2.705 83682319 XAlke 2,2,3,4,5,5-hexachloro-3-hexene 2.569 83682400 XAlke 1,1-dichloro-2,3-bis(trichloromethyl)propene 3.266 706785 XAlkeC Octachlorocyclopentene yes 5.140
6317255 XAlkeC 1,3-cyclopentadiene, 1,2,3,4-tetrachloro-5-(dichloromethylene)- 4.131
6928570 XAlkeC 1,3-Cyclopentadiene, 1,2,3,4,5-pentachloro-5-(trichloromethyl)- (8CI,9CI) yes 4.818
55044467 XAlkeC cyclobutane, 1,2-dichloro 3,4-bis(dichloromethylene)- 4.131 57722153 XAlkeC hexachlorocyclohexene 4.131 57722164 XAlkeC hexachlorocyclohexene 4.131 57722175 XAlkeC hexachlorocyclohexene 4.131 62111471 XAlkeC Cyclopentene, heptachloro- yes 4.832 72030263 XAlkeC Cyclopentene, hexachloro- yes 4.467 737202 XAN Heptafluoro-2-(trichloromethyl)naphthalene yes 6.110 345904 XBDP benzene, 1,1'-(bromomethylene)bis[4-fluoro- (9CI) 2.598 361079 XBDP benzene, 1,1'-(2,2,2-trifluoroethylIdene)bis[4-chloro- 2.810 782081 XBDP chlorobis(p-chlorophenyl)methane 2.598
5216535 XBDP benzene, 1,1'-(2,2-dichloroethylidene)bis[4-bromo- (9CI) 2.416
5736492 XBDP benzene, 1,1'-(bromomethylene)bis[2,3,4,5,6-pentafluoro- (9CI) yes 5.204
87581 XH 2,3,4,5-tetraiodofuran 3.495 2176627 XH Pyridine, pentachloro- 3.693 2338105 XH 4,5,6,7-tetrachlorobenzotriazole 4.040 5599246 XH 2,4-dichloro-6-pentachlorophenoxy-S-triazine 4.162 7682340 XH benzimidazole-2,4,5,6,7-pentaCL yes 4.516 10202467 XH 2,4-dichloro-6-(P-chlorophenyl)-S-triazine 3.251 10351061 XH 4-Pyridinethiol, 2,3,5,6-tetrachloro- (7CI,8CI,9CI) 3.008 17717167 XH pyridine, 2,3,4,5-tetrachloro-6-fluoro- 3.693 18247773 XH 2,4-dichloro-6-(2,4-dichlorophenoxy)-S-triazine 3.058 22963628 XH Pyridine, 2,3,5,6-tetrachloro-4-(methylthio)- 2.962 30886261 XH 2,4-dichloro-6-(P-chlorophenoxy)-S-triazine 2.598 30886272 XH 2,4-dichloro-6-(2,4,5-trichlorophenoxy)-S-triazine 3.467 30894634 XH 2,4-dichloro-6-((P-chlorophenyl)thio)-S-triazine 2.598 22039389 XNOR 1,2,3,4,7,7-hexachloro-2-norbornene yes 5.289 7090417 XPAH biphenylene, 2,3,6,7-tetrachloro- yes 4.514 90077795 XPAH tetrachloroacenaphylene yes 4.514
90175 XUC benzenemethanol, .alpha.-(trichloromethyl)-, acetate 2.470 117088 XUC 1,3-Isobenzofurandione, 4,5,6,7-tetrachloro- 4.093 127902 XUC 1,1'-Oxybis[2,3,3,3-tetrachloropropane] 3.303
307711 XUC
Undecanoic acid, 2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11-eicosafluoro-, potassium salt yes 5.013
358236 XUC methanesulfonic acid, trifluoro-, anhydride 2.846 376501 XUC hexanedioic acid, octafluoro-, diethyl ester 2.438 398992 XUC benzothiazole, 2-methyl-5-(trifluoromethyl)- 3.589 402380 XUC 3,3-diMe-1-(3-CF3 Ph)triazene 2.470 605367 XUC naphthalene, 1,2,3,4-tetrachloro-1,2,3,4-tetrahydro- 4.033 632791 XUC 1,3-Isobenzofurandione, 4,5,6,7-tetrabromo- 4.093
711626 XUC carbonic acid, CYCLIC 3,4,5,6-tetrachloro-O-phenylene ester 4.033
63
725893 XUC benzoic acid, 3,5-bis(trifluoromethyl)- (7CI,8CI,9CI) 4.071 791504 XUC 1,3-diEHIETANE, 2,2,4,4-tetraKIS(trifluoromethyl yes 5.529 853394 XUC methanone, bis(pentafluorophenyl)- (9CI) 4.287
1043501 XUC benzene, 1,2,3,4,5-pentafluoro-6-[(2,3,4,5,6-pentafluorophenyl)thio]- yes 4.978
1203867 XUC ethanone, 2,2-dichloro-1-(2,4,5-trichlorophenyl)- 3.475 1536238 XUC methanone, (2,3,4,5,6-pentafluorophenyl)phenyl- 2.810 1717506 XUC 1,3-dithietane, 2,2,4,4-tetrafluoro- (8CI,9CI) 4.085 1825236 XUC benzoyl chloride, pentachloro- (7CI,8CI,9CI) 3.879 1928376 XUC acetic acid, (2,4,5-trichlorophenoxy)-, methyl ester 2.470 2200706 XUC 4,4'-dihydroxyoctafluorobiphenyl 3.777 2207274 XUC 1,2,3,4-tetrachloro-5,5-dimethoxy-1,3-cyclopentadiene 2.999 2338274 XUC 1H-benzimidazole, 4,5,6-trichloro-2-(trifluoromethyl)- yes 4.893 2338296 XUC 1H-benzimidazole, 4,5,6,7-tetrachloro-2-(trifluoromethyl)- yes 5.247 2341868 XUC benzenamine, N-[(pentafluorophenyl)methylenE]- 2.810 2666708 XUC Carbonimidic dichloride, (2,4-dichlorophenyl)- (9CI) 2.962 2694066 XUC benzoic acid, 2,3,6-trichloro-, methyl ester 2.567 2948201 XUC chloroacetic acid, pentachlorophenyl ester 4.071 3141251 XUC thiophene, 2,3,4-tribromo- (6CI,7CI,8CI,9CI) 2.827 3584665 XUC 1H-benzimidazole, 5-chloro-2-(trichloromethyl)- (9CI) 4.033 3760665 XUC 1-Butanone, 2-bromo-4-chloro-1-(4-chlorophenyl)- (9CI) 2.737 3958030 XUC thiophene, tetrabromo- 3.495
4089581 XUC
propanoyl fluoride, 2,3,3,3-tetrafluoro-2- 1,1,2,3,3,3-hexafluoro-2- 1,1,2,2-tetrafluoro-2-(fluorosulfonyl)ethoxy prop 3.750
4400060 XUC [1,1'-biphenyl]-4-OL, 3,4',5-trichloro- 2.598 5358065 XUC benzoic acid, 2,3,4,5-tetrachloro-6-cyano-, methyl ester 2.952 5634377 XUC ethanol, 2,2,2-trichloro-, carbonate (2:1) (9CI) 2.707
5902692 XUC Acetic acid, chloro-, 2,4,5-trichlorophenyl ester (6CI,8CI,9CI) 3.096
6022339 XUC Carbamimidic chloride, N'-(3,4-dichlorophenyl)-N,N-dimethyl- (9CI) 2.737
13252147 XUC propanoic acid, 2,3,3,3-tetrafluoro-2-(1,1,2,3,* yes 4.805 14047097 XUC 3,3',4,4'-tetrachloroazobenzene 2.416 14815873 XUC benzoic acid, 4-(trichloromethyl)-, methyl ester (9CI) 2.737
16090145 XUC ethanesulfonyl fluoride, 2- 1- difluoro (trifluoroethenyl)oxy methyl -1,2,2,2-tetrafluoroethoxy -1,1,2,2-tetrafluoro- yes 4.577
16143843 XUC 3,5-diCF3 acetanilide 3.754 17025477 XUC benzene, [(tribromomethyl)sulfonyl]- 2.567 17587223 XUC 3,5-Octanedione, 6,6,7,7,8,8,8-heptafluoro-2,2-dimethyl- 3.310 18237940 XUC 1H-benzimidazole, 4,5,6,7-tetrachloro-2-methyl- 4.033 19259111 XUC thiophene, tetraiodo- (6CI,7CI,8CI,9CI) 3.495 20405305 XUC ethanol, 2,2,2-trichloro-, phosphate (3:1) (8CI,9CI) 4.153 21132309 XUC Urea, N,N'-bis(pentafluorophenyl)- (9CI) 2.985 21757824 XUC Acetic acid, 1-(3,4-dichlorophenyl)-2,2,2-trichloroethyl ester 3.376 23383962 XUC 2,4-diCL CF3-methanesulfonanilide 3.682
24169322 XUC 1,3-dioxolane, 2-(bromomethyl)-2-(2,3,4-trichlorophenyl)- (8CI,9CI) 3.733
25108694 XUC benzene, 1-chloro-4-(2,2-dichloro-1-methylcyclopropyl)- (8CI,9CI) 3.505
25711777 XUC propanoyl fluoride, 2,3,3,3-tetrafluoro-2- 1,1,2,3,3,3-hexafluoro-2-(pentafluorophenoxy)propoxy - 4.117
26654977 XUC ethanesulfonyl fluoride,2-[1-[difluoro[(triFLUO* 4.051
27619972 XUC 1-Octanesulfonic acid, 3,3,4,4,5,5,6,6,7,7,8,8,8-tridecafluoro- yes 4.769
29091096 XUC benzene, 2,4-dichloro-1,3-dinitro-5-(trifluoromethyl)- 2.746 29684262 XUC benzenediazonium, 3,5-bis(trifluoromethyl)- 3.956 32919043 XUC Acetic acid, tribromo-, phenylmethyl ester (9CI) 2.737
39108344 XUC 1-Decanesulfonic acid, 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluoro- (9CI) 4.216
39234861 XUC benzenesulfonyl chloride, 3,5-bis(trifluoromethyl)- 3.883 42016911 XUC benzoyl chloride, 2-hydroxy-3,5-diiodo- (9CI) 2.443 50274834 XUC benzoyl chloride, 3,4,5-trichloro-2-hydroxy- (9CI) 2.999 51366257 XUC 2,2,2-trichloro-1-(3,4-dichlorophenyl)ethyl acetate 3.376 51735901 XUC 1-butanesulfinyl isoCYANATE, 1,1,2,2,3,3,4,4,4-Nonafluoro- 3.808 55759881 XUC benzene, 2,4-dichloro-1-[(4-chlorophenyl)thio]- 2.598 55773905 XUC ethylpentachlorophenoxyacetate 2.746 56046083 XUC 3-methoxypentachlorocyclohexane 3.456 56400114 XUC 3(CF3-MeO)pentachlorocyclohexane yes 4.566 56421320 XUC 2-ethoxypentachlorocyclohexane 3.475
64
56421353 XUC 3-ethoxypentachlorocyclohexane 3.475 56421364 XUC 1(CF3-MeO)pentachlorocyclohexane yes 4.566 56961912 XUC 1,4-naphthalenedione, 5-bromo-2,3-dichloro- (9CI) 3.251 57846067 XUC thiophene, 2,5-dibromo-3-(dibromomethyl)- (9CI) 3.168
59819522 XUC phosphorochloridic acid, 2-chlorophenyl 2,2,2-trichloroethyl ester (9CI) 2.810
59866727 XUC propanoic acid, 2,3-dibromo-, 2,2,2-tribromo-1-chloroethyl ester (9CI) 2.612
60066844 XUC Cyclopropanecarboxylic acid, 2,2-dimethyl-3-(2,2,2-trichloroethyl)-, ethyl ester 2.470
60207309 XUC 1,3-dioxolane, 2-(bromomethyl)-2-(2,4-dichlorophenyl)- (9CI) 3.251
60254140 XUC Cyclopropanecarboxylic acid, 2,2-dimethyl-3-(2,2,2-trichloroethyl)-, ethyl ester, trans- 2.470
62601176 XUC benzene, pentachloro[(chloromethyl)thio]- (9CI) 3.879 63980121 XUC benzoic acid, 4-(trichloromethyl)-, 2-chloroethyl ester (9CI) 2.952 64630644 XUC 2',3,4,4',5-pentachloro-2-aminodiphenyl ether 2.810 64694855 XUC Acetamide, N-(3,4-dichlorophenyl)-2,2,2-trifluoro- 3.554
65072482 XUC benzenemethanol, 3,4-dichloro-.alpha.,.alpha.-bis(trifluoromethyl)- yes 4.624
65601403 XUC 2,2,4,4-tetrachloro-2,2,4,4,6,6-hexahydro-6,6-B* 2.569 68413718 XUC benzene, 1,3,5-tribromo-2-(2-bromoethoxy)- 2.999 68516541 XUC benzenesulfonyl chloride, 4-(1,1,2,2-tetrafluoroethoxy)- 2.810
69087463 XUC propanoic acid, 3- 1- difluoro (trifluoroethenyl)oxy methyl -1,2,2,2-tetrafluoroethoxy -2,2,3,3-tetrafluoro- yes 4.769
69116730 XUC
propanoic acid, 3- 2- 1,2-difluoro-2-oxo-1-(trifluoromethyl)ethoxy -1,2,2-trifluoro-1-(trifluoromethyl)ethoxy -2,2,3,3 3.337
74298694 XUC naphthalene, 2,4,5,8-tetrachloro-1-methoxy- 3.733 74664305 XUC benzenediazonium, 3,5-bis(trifluoromethyl)- 3.956 75147205 XUC 3-Butenoic acid, 2,2,3,4,4-pentachloro-, butyl ester 2.617 76253606 XUC tetrachlorobenzyltoluene 2.416
95998702 XUC alpha,alpha-difluorodichlorotrifluoromethyl diphenylmethane* 3.485
99308244 XUC trichloro-1-(dichloropropoxy)hexane 2.617 99308255 XUC tetrachloro-1-(dichloropropoxy)hexane 3.018
121107435 XUC 2,2'-4,4'-tetrachloro-3-methyl diphenylmethane 2.416 121107479 XUC 2,2'-4,6'-tetrachloro-3-methyl diphenylmethane 2.416 121107480 XUC 2,2'-4,6'-tetrachloro-5-methyl diphenylmethane 2.416 121107833 XUC 2',3-4,6'-tetrachloro-6-methyl diphenylmethane 2.416
65
Table 4 - CAS number, chemical class, name and POP score of chemicals meeting the
elevated AC-BAP and atmospheric oxidation half-life criteria, but not meeting the POP
score criteria.
CAS Class NAME SCORE
80068 CUP chlorfenethol 0.91 84548 CUP 9,10-Anthracenedione, 2-methyl- -0.65 84628 CUP 1,2-benzenedicarboxylic acid, diphenyl ester -4.02 88062 CUP phenol, 2,4,6-trichloro- 2.40 88857 CUP phenol, 2-(1-methylpropyl)-4,6-dinitro- -1.75 92933 CUP 1,1'-biphenyl, 4-nitro- -0.23 95954 CUP phenol, 2,4,5-trichloro- 2.40 97165 CUP genite 0.35
103333 CUP diazene, diphenyl- 0.28 115322 CUP dicofol 2.18 116290 CUP tetradifon 1.80 117260 CUP bulan -1.22 117271 CUP 2-nitro-1,1-bis(p-chlorophenyl)propane -0.71 510156 CUP chlorobenzilate -1.22 513928 CUP ethene, tetraiodo- 2.52
1420071 CUP dinoterb -1.75 1836755 CUP benzene, 2,4-dichloro-1-(4-nitrophenoxy)- 0.35 1836777 CUP chlornitrofen 0.80 1861321 CUP DCPA 1.75 2439012 CUP oxythioquinox 0.40 2487016 CUP medinoterb acetate -3.51 2759719 CUP cypromide 2.70 3765579 CUP glenbar 1.75 3861414 CUP bromoxynil butyrate 0.81 3996596 CUP medinoterb -2.19 5335240 CUP [1,1'-biphenyl]-2-ol, 3,5-dichloro- 2.08 8027007 CUP dilan -0.71 13738631 CUP 1,5-dichloro-3-fluoro-2-(4-nitrophenoxy)benzene 0.80 14255880 CUP Fenazaflor 0.36 15457053 CUP benzene, 2-nitro-1-(4-nitrophenoxy)-4-(trifluoromethyl)- -1.35 27253298 CUP Neodecanoic acid, zinc salt -5.19 28159980 CUP IRGAROL 1051 -1.20 42576023 CUP bifenox -1.73 42874011 CUP nitrofluorfen 0.63 50594677 CUP acifluorfen-Me -1.52 60168889 CUP fenarimol -1.26 61444620 CUP N-[2-amino-3-nitro-5-(trifluoromethyl)phenyl]-2* 1.63 79622596 CUP fluazinam -0.41
83164334 CUP 3-pyridinecarboxamide, N-(2,4-difluorophenyl)-2-[3-(trifluoromethyl)phenoxy]- -1.84
88283414 CUP ethanol, 1-(2,4-dichlorophenyl)-2-(3-pyridinyl)-, O-methyloxime -0.19 106917526 CUP flusulfamide -0.68 131086425 CUP HC-252 Herbicide -2.75 3740929 CUP fenchlorim 2.70 101553 PBDE benzene, 1-bromo-4-phenoxy- 1.48
2050477 PBDE benzene, 1,1'-oxybis 4-bromo- 2.08 36563470 PBDE bromophenoxybenzene 1.48 2050671 PCB biphenyl, 3,3'-dichloro- 2.70 2050682 PCB 4,4'-dichlorobiphenyl 2.70 2051607 PCB 2-chlorobiphenyl 2.07 2051618 PCB 3-chlorobiphenyl 2.07 2051629 PCB 4-chlorobiphenyl 2.07 2974905 PCB biphenyl, 3,4'-dichloro- 2.70 2974927 PCB biphenyl, 3,4-dichloro- 2.70 11104282 PCB Aroclor 1221 2.07 11141165 PCB Aroclor 1232 2.07 13029088 PCB 2,2'-dichlorophenyl 2.70 16605917 PCB 2,3-dichlorobiphenyl 2.70 25569806 PCB 2,3'-PCB 2.70 27323188 PCB chlorobiphenyl 2.07 33039815 PCB dichlorobiphenyl 2.70
66
33146451 PCB 2,6-dichlorobiphenyl 2.70 33284503 PCB 2,4-dichlorobiphenyl 2.70 34883391 PCB 2,5-dichlorobiphenyl 2.70 34883415 PCB biphenyl, 3,5-dichloro- 2.70 34883437 PCB 2,4'-PCB 2.70
132567227 PCB 4,4'-dichlorobiphenyl 2.70 87401 PCBD benzene, 1,3,5-trichloro-2-methoxy- 2.38 88493 PCBD benzenesulfonyl chloride, 4-chloro-2,5-dimethyl- 1.41 98317 PCBD benzenesulfonyl chloride, 3,4-dichloro- 1.64 98602 PCBD benzenesulfonyl chloride, 4-chloro- 0.91
102363 PCBD benzene, 1,2-dichloro-4-isocyanato- 1.79 320729 PCBD benzoic acid, 3,5-dichloro-2-hydroxy- 1.95 551768 PCBD 3-methyl-2,4,6-trichlorophenol 2.38 585477 PCBD 1,3-benzenedisulfonyl dichloride 0.55 609198 PCBD 3,4,5-trichlorophenol 2.40 609609 PCBD benzenesulfonyl chloride, 2,4-dimethyl- 0.67 634935 PCBD benzenamine, 2,4,6-trichloro- 2.40 882144 PCBD propanamide, N-(3,4-dichlorophenyl)- 1.92 933755 PCBD phenol, 2,3,6-trichloro- 2.40 933788 PCBD 2,3,5-trichlorophenol 2.40
1196130 PCBD benzene, 1,3,5-trichloro-2-nitroso- 2.38 1570678 PCBD 2,6-dichlor-3,4-dimethylphenol 1.79 1965259 PCBD benzene, 1-azido-2,4-dichloro- 1.79 2057649 PCBD 2,4-dichlor-3,6-dimethylphenol 1.79 2131557 PCBD benzene, 1-chloro-4-isothiocyanato- 1.00 2213812 PCBD benzene, 2-chloro-5-(1,1-dimethylethyl)-1,3-dinitro- -0.74 2612579 PCBD benzene, 2,4-dichloro-1-isocyanato- 1.79 2888064 PCBD M-chlorobenzenesulphonyl chloride 0.91 2891170 PCBD 1,3-benzenedisulfonyl dichloride, 4-chloro- 1.19 2905239 PCBD 2-chlorobenzenesulfonyl chloride 0.91 3296057 PCBD benzene, 1-azido-4-chloro- 1.00 3426899 PCBD phosphorodichloridous acid, phenyl ester -0.25 3773146 PCBD 2,4,5-trichlorobenzenethiol 2.40 3780856 PCBD carbonylcyanide26dichlorophenylhydrazone 1.36 4163784 PCBD benzene, 1,2,4-trichloro-5-(methylthio)- 2.38 5392825 PCBD benzene, 1,4-dichloro-2-isocyanato- 1.79 5402733 PCBD benzenesulfonyl chloride, 2,5-dichloro- 1.64 6130752 PCBD benzene, 1,2,4-trichloro-5-methoxy- 2.38 6590949 PCBD benzene, 1,2-dichloro-4-isothiocyanato- 1.79 6590961 PCBD benzene, 2,4-dichloro-1-isothiocyanato- 1.79 6639301 PCBD benzene, 1,2,4-trichloro-5-methyl- 2.40 6752381 PCBD benzenesulfonyl chloride, 4-isocyanato- 0.55 7160227 PCBD propanamide, N-(3,4-dichlorophenyl)-2,2-dimethyl- 1.36 7359720 PCBD 2,3,4-trichlorotoluene 2.40 7606873 PCBD benzoic acid, 3,5-dichloro-2-hydroxy-, methyl ester 2.16 7669547 PCBD benzenesulfenyl chloride, 2-nitro- 0.35 13334731 PCBD benzene, 1,5-dichloro-2-methoxy-3-methyl- 1.79 13790142 PCBD benzene, 1,3-dichloro-4-methyl-2-nitro- 1.95 14593283 PCBD 4-(tert-butyl)-2-chloro-6-nitrophenol 0.25 14611702 PCBD thiocyanic acid, 3,5-dichloro-4-hydroxyphenyl ester 1.95 15196835 PCBD benzoic acid, 5-chloro-2-hydroxy-, ethyl ester 1.29 15945070 PCBD benzenesulfonyl chloride, 2,4,5-trichloro- 2.26 15950660 PCBD 2,3,4-trichlorophenol 2.40 17199212 PCBD 2-chloro-6-nitro-4-(1,1,3,3-tetramethylbutyl)-P* -1.69 17302828 PCBD benzoic acid, 3,5-dichloro-4-hydroxy-, ethyl ester 1.92 19040621 PCBD benzenesulfonyl chloride, 2,5-dimethyl- 0.67 21472866 PCBD benzene, 1,2,3-trichloro-5-methyl- 2.40 23165539 PCBD benzene, 4-chloro-1-isothiocyanato-2-methyl- 1.22 23378883 PCBD benzenesulfonyl chloride, 3,5-dichloro-2-hydroxy- 1.83 25097443 PCBD phosphinic chloride, (1,1-dimethylethyl)[4-(1,1-dimethylethyl)phenyl]- (9CI) -1.30 25167822 PCBD phenol, trichloro- 2.40 25167844 PCBD trichloro-1,2-benzenediol 2.38 27165226 PCBD benzenediazonium, 4-chloro-2-nitro- 1.48 28225889 PCBD benzenethiol, 2,4-dichloro-5-methyl- 1.68 28286864 PCBD benzenesulfonyl chloride, 2,4-dichloro-5-methyl- 1.83
28343284 PCBD pentanenitrile, 4,4-dimethyl-3-oxo-2-[(2,4,5-trichlorophenyl)hydrazono]- (9CI) -0.31
29682460 PCBD benzene, 1,3-dichloro-2-methyl-4-nitro- 1.95 30928246 PCBD benzenediazonium, 2,6-dichloro-4-nitro- 2.16 32139723 PCBD 1,2-benzenediol, 3,4,6-trichloro- 2.38
67
34123507 PCBD benzene, 2-chloro-4-isocyanato-1-(1-methylethyl)- 1.29 34732097 PCBD 2,3,4-trichlorobenzenesulphonyl chloride 2.26 34893920 PCBD benzene, 1,3-dichloro-5-isocyanato- 1.79 34981389 PCBD benzenesulfonyl chloride, 5-chloro-2-methyl- 1.15 35282838 PCBD benzene, 1,2,4-trichloro-3,6-dimethoxy-5-nitro- 1.88 36663759 PCBD benzene, 2-chloro-5-(1-methylethyl)-1,3-dinitro- -0.25 37693155 PCBD phenol, 2,6-dichloro-3-methyl-4-nitro- 2.16 38815393 PCBD phosphorodichloridic acid, (1,1-dimethylethyl)phenyl ester 0.07 39549274 PCBD phenol, 2,4-dichloro-3-methyl-6-nitro- 2.16 39920371 PCBD benzene, 1,3-dichloro-2-isocyanato- 1.79 41195908 PCBD benzene, 1,2-dichloro-3-isocyanato- 1.79 54135807 PCBD benzene, 1,2,3-trichloro-4-methoxy- 2.38 54135818 PCBD benzene, 1,2,5-trichloro-3-methoxy- 2.38 54135829 PCBD benzene, 1,2,3-trichloro-5-methoxy- 2.38 54932842 PCBD benzene, 1,4-dichloro-2,5-diethyl- 1.95 55191167 PCBD benzene, 1,3-dichloro-5-(1,1-dimethylethyl)-2-methoxy- 1.92 56157927 PCBD 4-chloro-2-methylbenzenesulphonyl chloride 1.15 56961207 PCBD 1,2-benzenediol, 3,4,5-trichloro- 2.38 56961321 PCBD benzoic acid, 3-chloro-2-hydroxy-, ethyl ester 1.29 56961456 PCBD phenol, 2,4,5-trichloro- 2.40 56962095 PCBD 1-propanone, 1-(4-chlorophenyl)-, oxime 1.48 57135689 PCBD benzene, 1-chloro-2-isothiocyanato-4-nitro- 1.29 60741517 PCBD phenol, 2,4-dichloro-6-methyl-3-(1-methylethyl)- 2.16 64070811 PCBD arsinic chloride, 1,2-phenylenebis- (9CI) -1.42 65724121 PCBD dichloro-p-cymene 1.95 68631049 PCBD benzenesulfonyl chloride, 2,3-dimethyl-5-nitro- -0.39 69158265 PCBD 2,4,5,6-tetrachloro-1,3-benzenediol diacetate 1.75 71463597 PCBD pentanoic acid, 2,6-dichlorophenyl ester 1.36 74098296 PCBD phenol, 6-chloro-3,4-dimethyl-2-nitro- 1.29 74220158 PCBD phenol, 2,4,6-trichloro-, calcium salt (9CI) 2.40 74664338 PCBD benzenediazonium, 4-chloro-2-nitro-, trichlorozincate(1-) (9CI) 1.48 81686411 PCBD 2,5-dichloro-p-cymene 1.95 81686444 PCBD 2,3-dichloro-p-cymene 1.95 81686466 PCBD 2,6-dichloro-p-cymene 1.95 92366353 PCBD dichloro-p-cymen-8-ol 2.16 93942611 PCBD ethanone, 1-(2,3-dichlorophenyl)-, oxime 1.95 94213226 PCBD benzeneacetaldehyde, 2,3-dichloro-a-oxo-, aldoxime (9CI) 2.16 99817364 PCBD 2-nitro-4,6-dichloro-5-ethylphenol 1.92 29446159 PCDD 2,3-dichlorodibenzo-p-dioxin 2.15 33857260 PCDD 2,7-dichlorodibenzo-p-dioxin 2.15 33857282 PCDD 2,3,7-trichlorodibenzo-p-dioxin 2.64 38178380 PCDD 1,6-dichlorodibenzo-p-dioxin 2.15 38964226 PCDD 2,8-dichlorodibenzo-p-dioxin 2.15 39227582 PCDD 1,2,4-trichlorodibenzo-p-dioxin 2.64 50585392 PCDD 1,3-dichlorodibenzo-p-dioxin 2.15 54536173 PCDD 1,2,3-trichlorodibenzo-p-dioxin 2.64 54536184 PCDD 1,2-dichlorodibenzodioxin 2.15 54536195 PCDD 1,4-dichlorodibenzodioxin 2.15 64501004 PCDD dichlorodibenzodioxin 2.15 64560130 PCDD 4,6-dichlorodibenzodioxin 2.15 67028175 PCDD 1,3,7-trichlorodibenzodioxin 2.64 69760969 PCDD trichlorodibenzodioxin 2.64 82291336 PCDD 1,3,6-trichlorodibenzo-p-dioxin 2.64 82306614 PCDD 1,3,8-trichlorodibenzo-p-dioxin 2.64 82306625 PCDD 1,3,9-trichlorodibenzo-p-dioxin 2.64 2444895 PCDE benzene, 1,1'-oxybis[4-chloro- 2.08 2689078 PCDE benzene, 1-chloro-2-phenoxy- 1.48 6842622 PCDE benzene, 1-chloro-3-(4-chlorophenoxy)- 2.08 6903657 PCDE 2,4'-dichlorodiphenyl ether 2.08 7005723 PCDE benzene, 1-chloro-4-phenoxy- 1.48 24910687 PCDE 3,5-dichlorodiphenyl ether 2.08 24910698 PCDE benzene, 1,4-dichloro-2-phenoxy- 2.08 28419694 PCDE 2,6-dichlorodiphenyl ether 2.08 28675083 PCDE dichlorophenyl ether 2.08 51892263 PCDE 2,4-dichloro-1-phenoxybenzene 2.08 55398862 PCDE monochloro diphenyl oxide 1.48 55538697 PCDE benzene, 1,2-dichloro-4-phenoxy- 2.08 5409836 PCDF 2,8-dichlorodibenzofuran 2.80 24478748 PCDF 2,4-dichlorodibenzofuran 2.80 25074673 PCDF 3-chlorodibenzofuran 2.20
68
43047990 PCDF dichlorodibenzofuran 2.80 51230490 PCDF 2-chlorodibenzofuran 2.20 58802214 PCDF 3,7-dichlorodibenzofuran 2.80 60390274 PCDF 2,6-dichlorodibenzofuran 2.80 64126858 PCDF 1,2-dichlorodibenzofuran 2.80 64126869 PCDF 2,3-dichlorodibenzofuran 2.80 70648145 PCDF 1,9-dichlorodibenzofuran 2.80 74918404 PCDF 3,6-dichlorodibenzofuran 2.80 74992964 PCDF 4-chlorodibenzofuran 2.20 74992975 PCDF 1,6-dichlorodibenzofuran 2.80 74992986 PCDF 2,7-dichlorodibenzofuran 2.80 81638371 PCDF 1,8-dichlorodibenzofuran 2.80 84761864 PCDF 1-chlorodibenzofuran 2.20 94538008 PCDF 1,3-dichlorodibenzofuran 2.80 94538019 PCDF 1,4-dichlorodibenzofuran 2.80 94538020 PCDF 1,7-dichlorodibenzofuran 2.80 94570839 PCDF 3,4-dichlorodibenzofuran 2.80 1825305 PCN naphthalene, 1,5-dichloro- 2.89 1825316 PCN naphthalene, 1,4-dichloro- 2.89 2050693 PCN naphthalene, 1,2-dichloro- 2.89 2050728 PCN naphthalene, 1,6-dichloro- 2.89 2050739 PCN naphthalene, 1,7-dichloro- 2.89 2050740 PCN naphthalene, 1,8-dichloro- 2.89 2050751 PCN naphthalene, 2,3-dichloro- 2.89 2065705 PCN naphthalene, 2,6-dichloro- 2.89 2198756 PCN naphthalene, 1,3-dichloro- 2.89 2198778 PCN naphthalene, 2,7-dichloro- 2.89 28699889 PCN naphthalene, dichloro- 2.89
98588 PXBD benzenesulfonyl chloride, 4-bromo- 0.91 98613 PXBD benzenesulfonyl chloride, 4-iodo- 0.91 99285 PXBD phenol, 2,6-dibromo-4-nitro- 1.95
106376 PXBD benzene, 1,4-dibromo- 1.72 106536 PXBD benzenethiol, 4-bromo- 0.71 118796 PXBD phenol, 2,4,6-tribromo- 2.40 147820 PXBD benzenamine, 2,4,6-tribromo- 2.40 305851 PXBD phenol, 2,6-diiodo-4-nitro- 1.95 455209 PXBD benzenesulfonyl fluoride, 4-ethyl- 0.67 583551 PXBD benzene, 1-bromo-2-iodo- 1.72 589877 PXBD benzene, 1-bromo-4-iodo- 1.72 591184 PXBD benzene, 1-bromo-3-iodo- 1.72 601843 PXBD benzoic acid, 2,6-dibromo- 1.79 607998 PXBD benzene, 1,3,5-tribromo-2-methoxy- 2.38 609223 PXBD phenol, 2,4-dibromo-6-methyl- 1.68 609234 PXBD phenol, 2,4,6-triiodo- 2.40 615429 PXBD benzene, 1,2-diiodo- 1.72 624384 PXBD benzene, 1,4-diiodo- 1.72 626006 PXBD benzene, 1,3-diiodo- 1.72 827236 PXBD 2,4-dibrom-6-nitro-anilin 1.95
1074244 PXBD benzene, 1,4-dibromo-2,5-dimethyl- 1.68 1139522 PXBD phenol, 4-bromo-2,6-bis(1,1-dimethylethyl)- -0.25 1646544 PXBD benzene, 1,4-dibromo-2,3,5,6-tetramethyl- 1.95 1940427 PXBD phenol, 4-bromo-2,5-dichloro- 2.40 1985122 PXBD benzene, 1-bromo-4-isothiocyanato- 1.00 2059769 PXBD benzene, 1-iodo-4-isothiocyanato- 1.00 2101884 PXBD benzene, 1-azido-4-bromo- 1.00 2101895 PXBD benzene, 1-azido-3-bromo- 1.00 2109128 PXBD phenol, 2,4,6-triiodo-3-methyl- 2.38 2131591 PXBD benzene, 1-bromo-3-isothiocyanato- 1.00 2432146 PXBD phenol, 2,6-dibromo-4-methyl- 1.68 2437492 PXBD 1,3-benzenediol, 2,4,6-tribromo- 2.38 2489523 PXBD benzenesulfonyl fluoride, 3-(chlorosulfonyl)- 0.55 2747173 PXBD 3,4,5-tribromocatechol 2.38 2905240 PXBD M-bromobenzenesulphonyl chloride 0.91 3217150 PXBD 4-bromo-2,6-dichlorophenol 2.40 3302394 PXBD benzene, 1-azido-2-bromo- 1.00 3336263 PXBD benzonitrile, 3-bromo-4-hydroxy-5-iodo- 1.79 3460182 PXBD benzene, 1,4-dibromo-2-nitro- 1.79 4068751 PXBD benzoic acid, 2-hydroxy-5-iodo-, methyl ester 1.48 4186521 PXBD phenol, 2,4-diiodo-6-methyl- 1.68 4524770 PXBD 6-bromo-2,4-dichlorophenol 2.40
69
4526561 PXBD 4,6-dibromo-2-chlorophenol 2.40 4619743 PXBD phenol, 2,4,6-tribromo-3-methyl- 2.38 5324130 PXBD 2,6-dibromo-4-chlorophenol 2.40 5326385 PXBD benzene, 1-iodo-2-methyl-4-nitro- 1.22 5398276 PXBD benzenamine, 2,6-diiodo-4-nitro- 1.95 5469192 PXBD benzene, 1-bromo-2,4,5-trimethyl- 1.00 6320407 PXBD benzene, 1,3,5-tribromo-2-methyl- 2.40 6380343 PXBD phenyl arsenic diiodide -0.68 6942990 PXBD benzene, 2,4-dibromo-1,3,5-trimethyl- 1.79 7191460 PXBD ethanone, 1-(2-hydroxy-3,5-diiodophenyl)- 1.95 7530270 PXBD phenol, 4-bromo-2-chloro-6-methyl- 1.68 7587157 PXBD phenol, 2,5-dichloro-4-iodo- 2.40 7745928 PXBD benzene, 2-iodo-1-methyl-4-nitro- 1.22 15046841 PXBD 2,4-dibromoanisole 1.68 17199223 PXBD 2-bromo-6-nitro-4-(1,1,3,3-tetramethylbutyl)-PH* -1.69 17199234 PXBD 2-bromo-4-(tert-butyl)-6-nitrophenol 0.25 19241384 PXBD benzene, 4-bromo-1-isothiocyanato-2-methyl- 1.22 21702841 PXBD benzene, 2,4-dibromo-1-methoxy- 1.68 22621438 PXBD salicylic acid, 5-iodo-3-nitro- 0.76 22802400 PXBD phenol, 2,6-dibromo-3,4-dimethyl- 1.79 23456931 PXBD 3,3-dimethyl-1-(4-jodphenyl)-triazen 1.48 26249127 PXBD benzene, dibromo- 1.72 27476228 PXBD tribromotoluene 2.40 28166065 PXBD benzene, 4-azido-1-fluoro-2-nitro- 1.29 28805905 PXBD dibromodimethylbenzene 1.68 29052071 PXBD butyric acid, M-iodophenyl ester 1.29 29052082 PXBD butyric acid, P-iodophenyl ester 1.29 30672709 PXBD benzenesulfonyl fluoride, 2-(chlorosulfonyl)- (9CI) 0.55 30672721 PXBD benzenesulfonyl fluoride, 4-(chlorosulfonyl)- 0.55 30812874 PXBD benzene, dibromoethyl- 1.68 34586497 PXBD benzenesulfonyl fluoride, 2-ethyl- 0.67 34586500 PXBD benzenesulfonyl fluoride, 3-ethyl- 0.67 40371640 PXBD benzene, 1-bromo-5-chloro-2-methyl-4-nitro- 1.95 41252964 PXBD benzene, 2-chloro-1-iodo-4-nitro- 1.79 41252975 PXBD benzene, 4-iodo-1-methyl-2-nitro- 1.22 41727473 PXBD benzoic acid, 3,5-dibromo-4-hydroxy-, methyl ester 2.16 42228660 PXBD benzenediazonium, 2-bromo-4,6-dinitro- 0.25 50638476 PXBD benzene, 4-bromo-2-chloro-1-methoxy- 1.68 50702380 PXBD M-iodobenzenesulphonyl chloride 0.91 51699899 PXBD benzene, 1,3-dibromo-2-methoxy-5-methyl- 1.79 52628372 PXBD tribromoaniline 2.40 54852685 PXBD phenol, 2-bromo-4-chloro-6-methyl- 1.68 55771818 PXBD benzoic acid, 3,5-dibromo-4-hydroxy-, ethyl ester 1.92 56701305 PXBD benzene, 2-iodo-1,3,5-trimethyl-4-nitro- 1.29 57018129 PXBD phenol, 2,6-dibromo-4-ethyl- 1.79 62265990 PXBD benzene, 1,3-dibromo-2-methoxy-4-methyl-5-nitro- 1.92 62778181 PXBD phenol, 2,6-diiodo-3,4-dimethyl- 1.79 63059290 PXBD benzenesulfonyl chloride, 2-iodo- 0.91 64046602 PXBD 1,3-benzenediol, 2,4,6-tribromo-5-methyl-, diacetate (9CI) 0.75 65436875 PXBD tribromomethylphenol 2.38 66974581 PXBD 1-(o-I Ph)-3,3-dimethyltriazene 1.48 68084300 PXBD phenol, 2,4,6-tribromo-, aluminum salt 2.13 71412254 PXBD Butanoic acid, 4-cyano-2,6-diiodophenyl ester 0.81 71672883 PXBD benzene, 1-bromo-4-isothiocyanato-2-methyl- 1.22 86006442 PXBD dibromochloromethylphenol 2.38
92660 PXBP 1,1'-biphenyl, 4-bromo- 2.07 92864 PXBP 1,1'-biphenyl, 4,4'-dibromo- 2.70
321608 PXBP 1,1'-biphenyl, 2-fluoro- 2.07 324743 PXBP 1,1'-biphenyl, 4-fluoro- 2.07 398232 PXBP 4,4'-difluorobiphenyl 2.70
1591317 PXBP 1,1'-biphenyl, 4-iodo- 2.07 2052075 PXBP 1,1'-biphenyl, 2-bromo- 2.07 2113511 PXBP 1,1'-biphenyl, 2-iodo- 2.07 2113577 PXBP 1,1'-biphenyl, 3-bromo- 2.07 3001158 PXBP 1,1'-biphenyl, 4,4'-diiodo- 2.70 13029099 PXBP 2,2'-dibromo-1,1'-biphenyl 2.70 20442799 PXBP 1,1'-biphenyl, 3-iodo- 2.07 27479658 PXBP dibromo-1,1'-biphenyl 2.70 37847522 PXBP 1,1'-biphenyl, 2,4-difluoro- 2.70 41604197 PXBP 1,1'-biphenyl, 4-bromo-2-fluoro- 2.70
70
53591983 PXBP 1,1'-biphenyl, 2-bromo-4-fluoro- 2.70 39073079 PXDD 2,7-dibromodibenzo-p-dioxin 2.15 50585370 PXDD 2,3-dibromodibenzo-p-dioxin 2.15 50585381 PXDD 2,3-difluorodibenzo-p-dioxin 2.15 91371141 PXDD 1,6-dibromodibenzo-p-dioxin 2.15
105836962 PXDD 2,8-dibromodibenzo-p-dioxin 2.15 86760 PXDF 2-bromodibenzofuran 2.20
10016521 PXDF 2,8-dibromodibenzofuran 2.80 65489807 PXDF 2,7-dibromodibenzofuran 2.80
83534 PXN naphthalene, 1,4-dibromo- 2.89 36316888 PXN 2,6-diiodonaphthalene 2.89
80104 silane silane, dichlorodiphenyl- 0.79 144796 silane silane, chloromethyldiphenyl- 0.28 312403 silane silane, difluorodiphenyl- 0.79 775122 silane silane, diphenyl- 0.79 776761 silane silane, methyldiphenyl- 0.28 825945 silane silane, trichloro(4-chlorophenyl)- 0.38 830466 silane benzene, 1,4-bis(trichlorosilyl)- -0.25
1000700 silane silamine, N,N'-methanetetraYLbis[1,1,1-trimethyl- -0.18 1078962 silane butanenitrile, 4-(dichlorophenylsilyl)- 0.55 1078973 silane silane, 1,4-phenylenebis[chlorodimethyl- 0.55 1631830 silane chloro-diphenyl silane 0.79 2602553 silane silane, dichlorobis(trichloromethyl)- 1.77 3401261 silane silane, dichloro(3-chloropropyl)phenyl- 1.09 3449261 silane silanamine, N-(dimethylphenylsilyl)-1,1-dimethyl-1-phenyl- -2.14 4774758 silane diphenyl methyl silicon azide -1.20 5599326 silane triethyl silicon azide -0.59 13465844 silane tetraiodosilane -4.31 13688909 silane silane, trichloro[4-(chloromethyl)phenyl]- 0.56 17887411 silane silane, dichloro(1,1-dimethylethyl)phenyl- 0.67 18081124 silane silane, chloro(4-chlorophenyl)methylphenyl- 0.88 27137855 silane silane, trichloro(dichlorophenyl)- 1.11 30540342 silane silane, diazidomethylphenyl- -0.39 33317656 silane silane, dichlorobis(3-chloropropyl)- -0.08 33434638 silane silane, trichloro(tetrachlorophenyl)- 2.26 58479611 silane silane, chloro(1,1-dimethylethyl)diphenyl- -1.20 64426364 silane silanediamine, N-(chlorodimethylsilyl)-N'-(chlorodiphenylsilyl)-1,1-dimethyl- -3.06 79793003 silane benzenesulfonyl chloride, 4-[2-(trichlorosilyl)ethyl]- 0.67
56337 Silox disiloxane, 1,1,3,3-tetramethyl-1,3-diphenyl- -2.14 540976 Silox cyclohexasiloxane, dodecamethyl- -4.80 556683 Silox hexadecamethylcyclooctasiloxane -8.21
1014660 Silox silane, trimethyl(4-nitrophenoxy)- -0.39 2116849 Silox trisiloxane, 1,1,1,5,5,5-hexamethyl-3-phenyl-3-[(trimethylsilyl)oxy]- -3.94 2351135 Silox disiloxane, 1,3-bis(bromomethyl)-1,1,3,3-tetramethyl- 0.75 2565073 Silox silanediol, diphenyl-, diacetate -3.06 2943706 Silox disiloxane, 1,3-bis(dichloromethyl)-1,1,3,3-tetramethyl- 1.96 3582727 Silox disiloxane, 1,3-dichloro-1,3-dimethyl-1,3-diphenyl- -1.20 4353779 Silox silanol, 1,1,1-trimethyl-, 1-sulfuryl chloridoyl chloride -1.42 5360043 Silox methanol, (1,1,3,3-tetramethyl-1,3-disiloxanediyl)bis-, diacetate -2.30 7288280 Silox pyrimidine, 5-methyl-2,4-bis[(trimethylsilyl)oxy]- -1.75 10448096 Silox heptamethylphenylcyclotetrasiloxane -3.06 10497059 Silox silanol, trimethyl-, phosphate (3:1) -2.30 13176697 Silox 1-tetrasiloxanol, 1,1,3,3,5,5,7,7,7-nonamethyl- (9CI) -2.30 14920924 Silox pentamethylphenyldisiloxane -0.39 17875557 Silox trisiloxane, 1,1,5,5-tetramethyl-3,3-diphenyl- -3.06 17882063 Silox silanol, trimethyl-, benzenesulfonate (7CI,9CI) -0.39 17906091 Silox 1,1,1,3,3,5,7,7,7-Nonamethyl-5-phenyltetrasilo* -3.94 17962344 Silox trisiloxane, 1,1,3,5,5-pentamethyl-3-phenyl- -1.30 18027457 Silox trisiloxane, 3-[(dimethylsilyl)oxy]-1,1,5,5-tetramethyl-3-phenyl- -2.63 18105312 Silox hexanedioic acid, bis(trimethylsilyl) ester -2.71 18132724 Silox disiloxane, 1,3-bis(3-chloropropyl)-1,1,3,3-tetramethyl- 0.63 18306291 Silox silanol, trimethyl-, 1,1'-sulfate -0.65 18407164 Silox 1,1,1,3,3,5,5-Heptamethyl-5-phenyltrisiloxane -2.19 37843268 Silox 2-Butanone, O,O'-(dimethylsilylene)dioxime -1.48 60587102 Silox 1,1,1,3,3,5,7,7,9,9,9-UNDecamethyl-5-phenyipen* -5.66
78637 UC peroxide, (1,1,4,4-tetramethyl-1,4-butanediyl)bis (1,1-dimethylethyl) -3.54 81141 UC ethanone, 1- 4-(1,1-dimethylethyl)-2,6-dimethyl-3,5-dinitrophenyl - -3.51 81152 UC benzene, 1-(1,1-dimethylethyl)-3,5-dimethyl-2,4,6-trinitro- -3.51 83669 UC benzene, 1-(1,1-dimethylethyl)-2-methoxy-4-methyl-3,5-dinitro- -2.63 84479 UC 9,10-anthracenedione, 2-(1,1-dimethylethyl)- -2.14
71
84515 UC 9,10-anthracenedione, 2-ethyl- -1.15 86000 UC 2-nitrobiphenyl -0.23 92922 UC 1,1'-biphenyl -4-carboxylic acid -0.23 93992 UC benzoic acid, phenyl ester -0.23 98737 UC benzoic acid, 4-(1,1-dimethylethyl)- 0.07
101633 UC benzene, 1,1'-oxybis 4-nitro- -2.14 116665 UC 1H-indene, 2,3-dihydro-1,1,3,3,5-pentamethyl-4,6-dinitro- -2.60 118274 UC benzoic acid, 2-[[[2-(ethoxycarbonyl)phenoxy]carbonyl]oxy]-, ethyl ester -5.30 119517 UC 1,2-propanedione, 1-phenyl-, 2-oxime 0.55 136607 UC benzoic acid, butyl ester 0.07 145391 UC benzene, 1-(1,1-dimethylethyl)-3,4,5-trimethyl-2,6-dinitro- -2.63 247165 UC thieno[2,3-B][1]benzothiophene 2.11 247529 UC thieno[3,2-B][1]benzothiophene 2.11 256917 UC 11H-dibenzo[C,F][1,2]diazepine 0.40 303219 UC Phenol, 3-methyl-6-(1-methylethyl)-2,4-dinitro- -1.75 311568 UC 5,6,11,12-tetraoxadispiro[3.2.3.2]dodecane (7CI,8CI,9CI) 2.11 464880 UC cyclopentanecarboxylic acid, 1,2,2,3-tetramethyl- 0.55 501600 UC bis(4-methylphenyl)diazene -0.72 527311 UC 1,2,3,4,5,6-cyclohexanehexone 0.55 535466 UC benzenestibonic acid -1.00 584907 UC diazene, bis(2-methylphenyl)- -0.72 585488 UC 2,6-di-t-butylpyridine -0.39 588045 UC bis(3-methylphenyl)diazene -0.72 599666 UC benzene, 1,1'-sulfonylbis 4-methyl- -1.20 602608 UC anthracene, 9-nitro- -0.65 607012 UC phosphine, ethyldiphenyl- -0.23 620553 UC benzene, 1-nitro-3-phenoxy- -0.72 620882 UC 4-nitro diphenyl ether -0.72 621307 UC benzene, 1-isothiocyanato-3-methyl- 0.35 643652 UC methanone, (3-methylphenyl)phenyl- -0.23 643947 UC 1,2-benzenediol, dibenzoate (9CI) -4.02 691247 UC 2-propanamine, N,N'-methanetetraylbis[2-methyl- -0.18 716767 UC [1,1'-biphenyl]-3-carboxylic acid -0.23 720752 UC 1,1'-biphenyl -4-carboxylic acid, methyl ester -0.72 728405 UC 2,6-ditert.butyl-4-nitro-phenol -2.19 736301 UC benzene, 1,1'-(1,2-ethanediyl)bis 4-nitro- -2.60 829834 UC diphenylarsine 0.55 880933 UC 1-methyl-4-nitronaphthalene 0.28 881038 UC naphthalene, 2-methyl-1-nitro- 0.28 892217 UC 3-nitro-fluoranthen -1.15 943157 UC benzene, 1-methyl-4-(1-methylethyl)-2-nitro- 0.07 943271 UC ethanone, 1- 4-(1,1-dimethylethyl)phenyl - 0.07 945482 UC arsine, methyldiphenyl- (6CI,7CI,8CI,9CI) -0.65 949871 UC diazene, (4-methylphenyl)phenyl- -0.23 954074 UC 9,10-anthracenedione, 1-methyl- -0.65 954461 UC 9-nitrophenanthrene -0.65
1131471 UC 3H-1,5-benzodiazepine, 2,4-dimethyl- 0.79 1195706 UC 1,1-cyclopropanedicarbonitrile, 2,2,3,3-tetramethyl- 0.67 1202364 UC benzene, 1,1'-tellurobis- 0.55 1205067 UC benzoic acid, 4-isothiocyanato-, ethyl ester -0.39 1271541 UC chromium, bis(h6-benzene)- (9CI) 0.55 1320167 UC benzoic acid, (1,1-dimethylethyl)- 0.07 1344305 UC phenol, 2-(1-methylpropyl)dinitro- -1.75 1421494 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy- -2.19 1486288 UC phosphine, methyldiphenyl- 0.28 1519364 UC 9,10-Anthracenedione, 1,4-dimethyl- -1.15 1539044 UC 1,4-benzenedicarboxylic acid, diphenyl ester -4.02 1562932 UC benzoic acid, 4-(phenylazo)- -1.20 1605534 UC diethylphenylphosphine 0.67 1620640 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, ethyl ester -3.07 1636824 UC 2-pentanone, (2,4-dinitrophenyl)hydrazone -2.63 1636835 UC 3-pentanone, (2,4-dinitrophenyl)hydrazone -2.63 1640397 UC 3H-Indole, 2,3,3-trimethyl- 1.33 1719217 UC phenol, 2,4,6-trimethyl-3-nitro- 0.07 1816962 UC benzene, 1,4-dimethyl-2- (4-methylphenyl)sulfonyl - -1.67 1817749 UC benzene, 1,1'-methylenebis 4-nitro- -2.14 1886573 UC benzene, 1-(1,1-dimethylethyl)-2-nitro- 0.07 1942616 UC diazene, (4-isocyanatophenyl)phenyl- -1.20 2033241 UC 1,3-dioxane-4,6-dione, 2,2-dimethyl- 0.35 2113588 UC 3-nitrobiphenyl -0.23
72
2131615 UC benzene, 1-isothiocyanato-4-nitro- 0.55 2131626 UC benzoic acid, 4-isothiocyanoato- 0.55 2155717 UC 1,2-benzenedicarboperoxoic acid, bis(1,1-dimethylethyl) ester -3.94 2167239 UC peroxide, (1-methylpropylidene)bis (1,1-dimethylethyl) -1.89 2273510 UC stannane, oxodiphenyl- -0.65 2348176 UC 2-hexanone, (2,4-dinitrophenyl)hydrazone -3.07 2359093 UC Isophthalic acid, 5-tert-butyl- -1.30 2444293 UC p-nitrophenyl o-tolyl ether -1.20 2491523 UC 4-nitroazobenzene -1.20 2510556 UC 9-phenanthrenecarbonitrile -0.13 2511220 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, methyl ester -2.63 2523480 UC 9H-fluorene-2-carbonitrile 0.40 2622891 UC borinic acid, diphenyl- 0.28 2852688 UC methanone, bis(3-methylphenyl)- -0.72 3056642 UC 4-tert-butylphenol, Acetate -0.39 3077972 UC 2-butanone, 3-methyl-, (2,4-dinitrophenyl)hydrazone -2.63 3282562 UC benzene, 1-(1,1-dimethylethyl)-4-nitro- 0.07 3285981 UC 9,10-anthracenedione, 1,2-dimethyl- -1.15 3382567 UC 4-t-butylnitrobenzene 0.07 3402742 UC benzene, 1-methyl-4-(4-nitrophenoxy)- -1.20 3457612 UC peroxide, 1,1-dimethylethyl 1-methyl-1-phenylethyl -0.85 3529826 UC benzene, 1-isothiocyanato-3-nitro- 0.55 3581702 UC 1-propanol, 2,2-dimethyl-, benzoate -0.39 3581735 UC 1-propanol, 2,2-dimethyl-, 4-nitrobenzoate -1.75 3646579 UC diazene, bis(4-nitrophenyl)- -2.60 3682562 UC benzoic acid, 2-(phenylazo)- -1.20 3709180 UC 1,3-dioxane-4,6-dione, 2,2,5-trimethyl- 0.67 4044659 UC benzene, 1,4-diisothiocyanato- 0.55 4067145 UC benzoic acid, 4-(1,1-dimethylethyl)-, aluminum salt 0.07 4097498 UC 4-tert-butyl-2,6-dinitrophenol -1.75 4171339 UC diazene, bis(2-nitrophenyl)- -2.60 4196898 UC 1,3-propanediol, 2,2-dimethyl-, dibenzoate -3.97 4237405 UC benzene, 1-(1-methylpropyl)-4-nitro- 0.07 4262617 UC peroxide, 1,1'-(1-methylethylidene)bis[2-(1,1-dimethylethyl) -1.48 4507986 UC 1,2-dioxolane, 3,3'-dioxybis[3,5,5-trimethyl- (8CI,9CI) -1.67 4519328 UC diarsene, diphenyl- (9CI) 0.55 4731366 UC tri(sec-butyl) aluminum -0.24 5097121 UC benzene, 1,1'-sulfonylbis[2-methyl- -1.20 5180596 UC 2,6-bis(1,1-dimethylethyl)-4-nitrobenzenamine -2.19 5393497 UC benzenamine, N,N'-(1,2-dimethyl-1,2-ethanediylidene)bis- -1.67 5406575 UC benzoic acid, 4-(1,1-dimethylethyl)-, ethyl ester -0.85 5410979 UC 3-nitrodibenzofuran -0.13 5465134 UC benzene, 1,2,4,5-tetramethyl-3,6-dinitro- -1.30 5617709 UC 5,7-dioxaspiro[2.5]octane-4,8-dione, 6,6-dimethyl- 1.33 5689190 UC benzenemethanol, a,a-dicyclopropyl- 0.95 5737133 UC 4H-cyclopenta(DEF)phenanthren-4-one 0.45 5892091 UC bismuthine, (benzoyloxy)oxo- -0.59 5950834 UC benzene, 1-nitro-2-(4-nitrophenoxy)- -2.14 6002342 UC phosphine, (1,1-dimethylethyl)diphenyl- -1.20 6099805 UC benzene, 1-(1,1-dimethylethyl)-2-methoxy-3,5-dinitro- -2.19 6338096 UC 9,10-anthracenedione, 1-mercapto- -0.65 6531357 UC 9,10-anthracenedione, 2,3-dimethyl- -1.15 6575140 UC Benzonitrile, 2,6-diethyl- 0.55 6676900 UC (2-methylphenyl)phenyldiazene -0.23 6720281 UC diazene, (4-methylphenyl)phenyl-, (Z)- -0.23 6720394 UC diazene, (4-methylphenyl)phenyl-, (E)- -0.23 7030184 UC 2-methyl-4'-nitroazobenzene -1.67 7137555 UC benzene, 1-butyl-2-nitro- 0.07 7373264 UC benzene, 1,3,5-triisocyanato-2-methyl- -1.30 7470146 UC 3-phenanthrenecarboxylic acid -0.65 7507484 UC 1,4-benzenediol, 2-(1,1-dimethylethyl)-, diacetate (9CI) -2.19 7508681 UC benzoic acid, 4-(phenylazo)-, ethyl ester -2.14 7599237 UC 1,1'-biphenyl, 2-azido- -0.23 7612966 UC 4-Isothiocyanoazobenzene -1.20 7621161 UC phosphine, (1,1-dimethylethyl)methylphenyl- 0.55 10342606 UC benzene, 1-(2-methylpropyl)-4-nitro- 0.07 10605240 UC 1,3-propanediol, 2-methyl-2-propyl-, dinitrate (7CI,8CI,9CI) -1.48 13009911 UC 1,3,5-triazine, 2,4,6-tris(2-methyl-1-aziridinyl)- -0.63
13102324 UC propaneperoxoic acid, 2,2-dimethyl-, 1,1,4,4-tetramethyl-1,4-butanediyl ester (9CI) -5.19
73
13177281 UC nitrofluoranthene -1.15 13246327 UC arsine, As,As'-1,2-phenylenebis[As,As-dimethyl- -0.59 13471697 UC benzene, 4-isocyanato-1-methyl-2-nitro- 0.07 13472087 UC Butanenitrile, 2,2'-azobis[2-methyl- -1.06 13509336 UC carbonothioIic acid, O,S-diphenyl ester -0.72 13509369 UC carbonodithioIic acid, S,S-diphenyl ester -0.72 13615388 UC 2-methyl-8-nitronaphthalene 0.28 13653628 UC Peroxide, (1-methylpropylidene)bis (1,1-dimethylpropyl) -2.71 14213015 UC 1,1'-biphenyl, 3-azido- -0.23 14284936 UC ruthenium, tris(2,4-pentanedionato-o,o )-, (oc-6-11)- -3.95 14387305 UC benzoic acid, 4,4'-oxybis-, 1,1'-dimethyl ester -3.06 15085208 UC 4-Indancarbonitrile, 7-methyl- 1.33 15128504 UC dibenz[B,E]oxepin-6,11-dione -0.65 15231784 UC 1,3-dioxane-4,6-dione, 2,2-dimethyl-5-phenyl- -0.72 15815391 UC 9,10-Anthracenedione, 1,5-dimethyl- -1.15 15986922 UC pyrazine, 2,3-dihydro-5,6-dimethyl- -0.25 15986933 UC pyrazine, 2-ethyl-5,6-dihydro-3-methyl- 0.04 16225266 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)- -1.75 16268987 UC 1,3,5-triazine-2,4,6-triamine, N,N',N''-tris(1,1-dimethylethyl)- (9CI) -3.51 16424661 UC 4-HeptaNone, 2,2,3,3,5,5,6,6-Octamethyl- -1.89 16714184 UC benzene, 2-azido-1-methyl-3-nitro- 0.07 16738888 UC 1,1-cyclopropanedicarbonitrile, 2-methyl-2-propyl- 0.67 16738899 UC 1,1-cyclopropanedicarbonitrile, 2-butyl-2-methyl- 0.55 16888758 UC 1,2-ethanediamine, N,N'-bis(1-methylethylidene)- -0.59 16968197 UC benzene, 1,1'-(1,2-ethanediyl)bis[2-nitro- -2.60 17024189 UC 2-nitrophenanthrene -0.65 17082121 UC diazene, diphenyl-, (E)- 0.28 17484285 UC benzoic acid, 3-(1,1-dimethylethyl)-5-methyl- -0.39 18213735 UC 2,4-dimethoxy-6-phenyl-S-triazine -0.72 18264954 UC 2-phenanthrylamine, 9,10-dihydro-1,3-dinitro- -2.62 18277913 UC benzoic acid, 2-(phenylazo)-, ethyl ester -2.14 18709518 UC pyrazine, 2,5-bis(1,1-dimethylethyl)- -0.39 18794393 UC pyridine, 4-(1,1-dimethylethyl)-3-[(1,1-dimethylethyl)thio]- -0.85 18794462 UC 3-picoline, 6-(tert-butylthio)- 0.55 18794484 UC pyridine, 2-[(1,1-dimethylethyl)thio]-3,5-dimethyl- 0.07 18833879 UC 3- picoline, 2-(tert-butylthio)- 0.55 19370344 UC benzene, 1-(1-methylpropyl)-2-nitro- 0.07 19377958 UC BIcyclo[3.1.0]hexaN-2-one, 1,5-bis(1,1-dimethylethyl)-3,3-dimethyl- -1.20 19694123 UC diazene, bis(1,1-dimethylpropyl)- -0.24 19715196 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-2-hydroxy- -2.19 19881186 UC benzene, 1-isothiocyanato-4-(4-nitrophenoxy)- -2.14 20332109 UC stannane, diphenylthioxo- (8CI,9CI) -0.65 20615752 UC benzene, 1-azido-4-methyl-2-nitro- 0.07 20651756 UC benzene, 1-butyl-4-nitro- 0.07 20651767 UC benzene, 1-butyl-3-nitro- 0.07 20927951 UC 2-nitrodibenzofuran -0.13 21578949 UC 1,4-benzenedicarboperoxoic acid, bis(1,1-dimethylethyl) ester (9CI) -3.94 21850408 UC hexanediperoxoic acid, 2,4,4-trimethyl-, bis(1,1-dimethylethyl) ester (9CI) -4.78 22397337 UC 1,2,4,5-Tetroxonane, 3,3,6,6,9,9-hexamethyl- -0.85 23132527 UC benzene, 1-(1,1-dimethylethyl)-3-nitro- 0.07 23383597 UC propaneperoxoic acid, 2,2-dimethyl-, 1-methyl-1-phenylethyl ester -1.75 24582545 UC methyl diphenylarsinite -1.06 24624291 UC 9,10-anthracenedione, 1-ethyl- -1.15 25005952 UC pyrazine, 2,3-dihydro-5,6-di-2-pyridyl- -1.15 25149466 UC propanenitrile, 2- (1,1-dimethylethyl)azo -2-methyl- -0.18 25722661 UC cyanic acid, (1-methylethylIdene)di-4,1-phenylE* -2.14 26328596 UC 4-Isothiocyanobenzophenone -1.20 26444202 UC AR,AR'-azotoluene -0.72 27253334 UC neodecanoic acid, calcium salt -5.19 27936341 UC methylanthracene-9,10-dione -0.65 28122841 UC benzenecarbothioIC acid, 4-nitro-, S-(4-methylphenyl) ester -2.14 28361435 UC benzene, 1,1'-sulfonylbis 3,4-dimethyl- -2.14 28842059 UC diazene, (2-methylphenyl)(3-methylphenyl)- (9CI) -0.72 28984852 UC nitrobiphenyl -0.23 29418222 UC o,p'-azotoluene -0.72 29418573 UC 2-methyl-4-nitroazobenzene -1.67 29418584 UC (4-methylphenyl)(4-nitrophenyl)diazene -1.67 30026927 UC hydroperoxide, 1-[(1,1-dimethylethyl)phenyl]-1-methylethyl (9CI) -0.85 30448432 UC 1,2-benzenedicarboxylic acid, 1,2-bis(1,1-dimethylethyl) ester -3.07 30863651 UC 2-(diazomethyl)-4,6-bis(methylthio)-S-triazine 0.07
74
31656914 UC 1,1'-biphenyl, 4-azido- -0.23 32294603 UC ditelluride, diphenyl 0.55 32366260 UC benzene, (1-azido-1-methylethyl)- 0.55 32588548 UC 9,10-anthracenedione, 2-(1,1-dimethylpropyl)- -2.62 32673259 UC phosphine, bis(1,1-dimethylethyl)phenyl- -0.85 32703790 UC 1,3-isobenzofurandione, 5-(1,1-dimethylethyl)- -0.23 33720740 UC azete, 2,3-dihydro-4-phenyl- 1.28
34296554 UC spiro[cyclopropane-1,9'-[9H]fluorene]-2,3-dicarboxylic acid, dimethyl ester, cis- (9CI) -3.15
34372093 UC pentanenitrile, 2,4-dimethyl-2-(phenylazo)- -1.30 34874878 UC benzene, 2-methoxy-1,3,5-trimethyl-4-nitro- -0.39 35147456 UC 1,2-dioxolane, 3,5-bis[(1,1-dimethylethyl)dioxy]-3,5-dimethyl- -2.63 36050927 UC phosphine, methyl(1-methylethyl)phenyl- 0.67 36210716 UC methanone, (4-azidophenyl)phenyl- -1.20 36760437 UC carbonothioic acid, O-(1,1-dimethylethyl) S-phenyl ester -0.39 37467655 UC ethanone, 1-(3,5-diethyl-2,6-dihydroxyphenyl)- -0.85 38675795 UC methanone, cyclopropyl[4-(1,1-dimethylethyl)phenyl]- -0.23 39049053 UC neodecanoic acid, titanium salt -0.24 39950931 UC 3,4-hexanedione, 2,2,5,5-tetramethyl-, monooxime -0.65 40515182 UC benzene, 4-azido-1-methyl-2-nitro- 0.07 40515193 UC benzene, 2-azido-1-methyl-4-nitro- 0.07 40888979 UC 2-propanol, 2,2'-azobis-, diacetate (ester) (9CI) -1.89 41037132 UC 1-methyl-3-nitronaphthalene 0.28 41825289 UC plumbane, bis(nitrooxy)diphenyl- (9CI) -1.89 41902425 UC 3-pentanol, 3-(1,1-dimethylethyl)-2,2,4,4-tetramethyl- -1.06 41981708 UC thiazole, 2,4-dimethyl-5-propyl- 0.35 41981719 UC thiazole, 2,5-diethyl-4-methyl- 0.35 41981720 UC thiazole, 4,5-dimethyl-2-propyl- 0.35 41981753 UC thiazole, 4-ethyl-2-methyl-5-propyl- -0.18 43052397 UC pentanedioic acid, bis(1,1-dimethylethyl) ester (9CI) -2.30 43126836 UC tert-dodecanethiol, silver(1+) salt -0.65 43133944 UC ethanone, 1-[(1,1-dimethylethyl)phenyl]- 0.07 49540854 UC 4-isothiocyanophenyl benzoate -1.67 49556163 UC neodecanoic acid, tin(2+) salt -0.24 50670503 UC 1,1'-biphenyl -4-carbonitrile, 4'-methyl- -0.23 51677419 UC butane, 2-azido-2,3,3-trimethyl- -0.59 51818554 UC neodecanoic acid, iron salt -0.24 51818565 UC neodecanoic acid, nickel salt -5.19 52235208 UC butanenitrile, 2- (1,1-dimethylethyl)azo -2-methyl- -0.24 52270447 UC neodecanoic acid, cobalt(2+) salt -0.24 52414898 UC thiazole, 2,4-diethyl-5-methyl- 0.35 52625259 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, nickel(2+) salt (2:1) -2.19 52909883 UC benzenecarbothioIC acid, 2-nitro-, S-(4-methylphenyl) ester -2.14 52909894 UC benzenecarbothioIC acid, 3-nitro-, S-(4-methylphenyl) ester -2.14 53044487 UC benzene, 1-(2-ethylphenoxy)-2,4-dinitro- -3.06 53274194 UC benzeneacetic acid, 4-nitro-, 4-methylphenyl ester -2.60 53498309 UC 2-isopropyl-4,5-dimethylthiazole -0.59 53498310 UC thiazole, 5-ethyl-4-methyl-2-(1-methylethyl)- 0.67 54106400 UC benzene, 1-methyl-2-(2-nitrophenoxy)- -1.20 54187198 UC methanone, bis(2-methylcyclopropyl)-, oxime 1.53 54357083 UC 2-methyl-6-nitronaphthalene 0.28 54518091 UC benzo[B]thiophen-2(3H)-one, 3-diazo- 1.33 54644416 UC propanoic acid, 2,2-dimethyl-, 2-(1,1-dimethylethyl)phenyl ester -1.75 54738931 UC nitroanthracene -0.65 54755203 UC 2-methyl-5-nitronaphthalene 0.28 54755214 UC 2-methyl-8-nitronaphthalene 0.28 55044525 UC benzoic acid, 4-methyl-, 4-(methoxycarbonyl)phenyl methyl ester -3.06 55044694 UC 2-butanol, 2-methyl-3-(nitroimino)-, benzoate (ester) -2.19 55076363 UC benzoic acid, 2-hydroxy-6-nitro-, phenyl ester -2.14 55172980 UC neodecanoic acid, barium salt -5.19 55203593 UC 3H-Indole, 2,3,3-trimethyl-5-(phenylsulfonyl)- -2.62 55591112 UC 1H-Indene-4-carboxylic acid, 2,3-dihydro-1,1-dimethyl-, methyl ester -0.23 55712384 UC 1H-Indene-4-carboxylic acid, 2,3-dihydro-1,1-dimethyl- 0.28 55794202 UC butanoic acid, 3,3-bis[(1,1-dimethylethyl)dioxy]-, ethyl ester -3.54 55912180 UC pentanenitrile, 2- (1,1-dimethylethyl)azo -2,4-dimethyl- -1.06 56623337 UC 2-propanol, 2,2'-azobis-, dipropanoate (ester) (9CI) -2.71 57908482 UC 2-butanol, 2,2'-azobis-, diacetate (ester) (9CI) -2.71 57910799 UC 2-butanol, 2- (1,1-dimethylethyl)azo - -0.18 58743752 UC 1,1'-biphenyl -4-carbonitrile, 4'-ethyl- -0.72 60263447 UC pyridine, 2-(3,3-dimethylbutyl)- 0.55
75
60451072 UC neodecanoic acid, vanadium salt -0.24 60633218 UC phosphine, bis(1,1-dimethylethyl)(2,2-dimethylpropyl)- -1.06 61142516 UC phosphine, (2,2-dimethylcyclopropyl)diphenyl- -1.15 61951960 UC neodecanoic acid, cadmium salt -5.19 62245470 UC 2-amino-7-nitrophenanthrene -1.15 62337911 UC phosphine, bis(2,2-dimethylcyclopropyl)phenyl- -0.65 62637983 UC 1-naphthalenol, 2-(phenylazo)-, acetate (ester) -3.09 63017878 UC 1-methylnitronaphthalene 0.28 63042080 UC carbonic acid, methyl 2-methyl-4-[(2-methylphenyl)azo]phenyl ester (9CI) -3.06 63084139 UC 1H-Indene-5-carbonitrile, 2,3-dihydro-1,1,2,3,3,6-hexamethyl- -1.20 63906558 UC 3,6-diethyl-3,6-dimethyl-1,2,4,5-tetroxane 0.55
64047274 UC Iron, [[3,3'-(1,2-ethanediyldinitrilo)bis[2-butanone] dioximato](2-)-N,N',N'',N''']- (9CI) -1.89
67801530 UC 1,4-benzenedicarboxylic acid, 2-(4-methylphenyl)-, dimethyl ester -3.06 67801563 UC 1,1'-biphenyl -2,4'-dicarboxylic acid, 5-methyl-, dimethyl ester -3.06 68132956 UC borate(1-), dimethyldiphenyl-, lithium, (T-4)- -0.23 68425592 UC benzene, (1-methylpropyl)nitro- 0.07 68455925 UC nitrophenanthrene -0.65 68459814 UC benzene, 1-(1-methylpropyl)-3-nitro- 0.07 68683181 UC neodecanoic acid, silver(1+) salt -0.24 68958521 UC neodecanoic acid, iron(3+) salt -0.24 71735198 UC phenol, 2-(1-methylpropyl)-4,6-dinitro- -1.75 71850776 UC benzene, 1-(1,1-dimethylethyl)-3,5-dimethyl-2,4-dinitro- -2.19 74630005 UC benzene, 2-(1,1-dimethylethyl)-4-methyl-1-(1-methylethyl)-3,5-dinitro- -3.07
74810392 UC tricyclo[5.1.0.02,4]octane-5-carboxylic acid, 3,3,8,8-tetramethyl-, methyl ester (9CI) -0.13
75790828 UC benzene, 1-methyl-2-nitro-4-[(4-nitrophenyl)methyl]- -2.60 77468361 UC nitrofluoranthene -1.15 80182270 UC dimethylnitrophenanthrene -1.65 80191442 UC methylnitrophenanthrene -1.15 85895814 UC pyridine, 2-(1-ethyl-1-methylpropyl)- 0.55 91137278 UC 1-methyl-5-nitronaphthalene 0.28 91137289 UC 2-methyl-7-nitronaphthalene 0.28 94158062 UC [1,1'-biphenyl]-2-ol, 2'-azido-, acetate (ester) (9CI) -2.14
105752678 UC 1-methyl-6-nitronaphthalene 0.28 116211951 UC naphthalenetetrone, methyl -0.23 116211984 UC hydroxynitrophenanthrene -1.15 116530075 UC 1-methyl-7-nitronaphthalene 0.28
91134 XAB benzene, 1,2-bis(bromomethyl)- 1.68 93527 XAB benzene, (1,2-dibromoethyl)- 1.68 94995 XAB benzene, 2,4-dichloro-1-(chloromethyl)- 2.40
102476 XAB benzene, 1,2-dichloro-4-(chloromethyl)- 2.40 589151 XAB benzene, 1-bromo-4-(bromomethyl)- 1.65 589173 XAB benzene, 1-bromo-4-(chloromethyl)- 1.65 612124 XAB benzene, 1,2-bis(chloromethyl)- 1.68 623245 XAB benzene, 1,4-bis(bromomethyl)- 1.68 623256 XAB benzene, 1,4-bis(chloromethyl)- 1.68 626153 XAB benzene, 1,3-bis(bromomethyl)- 1.68 823789 XAB benzene, 1-bromo-3-(bromomethyl)- 1.65
2014837 XAB benzene, 1,3-dichloro-2-(chloromethyl)- 2.40 2387180 XAB benzene, 1,4-bis(chloromethyl)-2-methyl- 1.79 2735059 XAB benzene, 2,4-bis(chloromethyl)-1-methyl- 1.79 2745495 XAB benzene, 1,4-dichloro-2-(chloromethyl)- 2.40 3290015 XAB benzene, 1,2-dichloro-3-(chloromethyl)- 2.40 3290060 XAB benzene, 1,3-dichloro-5-(chloromethyl)- 2.40 3335339 XAB benzene, 1-methyl-2-(trichloromethyl)- 2.38 3335340 XAB M-xylene, .alpha.,.alpha.,.alpha.-trichloro- 2.38 3335351 XAB benzene, 1-methyl-4-(trichloromethyl)- 2.38 3433805 XAB benzene, 1-bromo-2-(bromomethyl)- 1.65 3883134 XAB benzene, (2,2,2-trichloroethyl)- 2.38 6298722 XAB benzene, 1,4-bis(chloromethyl)-2,5-dimethyl- 1.95 6529539 XAB benzene, 1-(2-bromoethyl)-4-chloro- 1.68 7398449 XAB benzene, 1-(chloromethyl)-4-(dichloromethyl)- 2.38 16799056 XAB benzene, 1-(2-bromoethyl)-3-chloro- 1.68 18880007 XAB benzene, 1-(bromomethyl)-4-(1,1-dimethylethyl)- 1.48 20443985 XAB benzene, 2-(bromomethyl)-1,3-dichloro- 2.40 28347139 XAB benzene, bis(chloromethyl)- 1.68 31195172 XAB benzene, tribromoethyl- 2.38 38721710 XAB benzene, dichloro(chloromethyl)- 2.40 53459408 XAB benzene, 1-(2-chloroethyl)-4-(chloromethyl)- 1.79
76
54789309 XAB benzene, 1-(1-chloroethyl)-4-(chloromethyl)- 1.79 54789310 XAB benzene, 1-(1,2-dichloroethyl)-4-methyl- 1.79 54965003 XAB benzene, 1,4-dichloro-2-(1-chloroethyl)- 2.38 54965014 XAB benzene, 1,2-dichloro-4-(1-chloroethyl)- 2.38 54965025 XAB benzene, 1,4-dichloro-2-(2-chloroethyl)- 2.38 59216170 XAB benzene, 2,4-dibromo-1-(2-bromoethyl)- 2.38 59473459 XAB benzene, 1-(chloromethyl)-2-iodo- 1.65 74298945 XAB benzene, 1-chloro-4-(1,2-dichloroethyl)- 2.38 1667103 XABP 1,1'-biphenyl, 4,4'-bis(chloromethyl)- 1.48 1940574 XABP 9H-fluorene, 9-bromo- 2.20 24802991 XAH 2-methylamino-4-phenyl-6-trichloromethyl-S-triA* 1.38 24848484 XAH 3-(sec-butyl(4-methyl-6-(trichloromethyl)-S-*)) 0.22 24848519 XAH 3-(methyl(4-propyl-6-(trichloromethyl)-S-tri*)) 0.22 30339743 XAH 1,3,5-triazin-2-amine, 4-pentyl-6-(trichloromethyl)- 1.88 30357392 XAH 2-(tert-butylamino)-4-methyl-6-trichloromethyl* 1.88 30357427 XAH 2-methyl-4-((1,1,3,3-tetramethylbutyl)amino)-6-* -0.31 30362335 XAH 2-methyl-4-phenyl-6-(trichloromethyl)-S-triAZI* 1.98 30362471 XAH alpha,alpha-dimethyl-4,6-bis(trichloromethyl)-*- 1.90 30369165 XAH 2-amino-4-(P-tolyl)-6-(trichloromethyl)-S-triAZ* 1.38 30369574 XAH 2-amino-4-((P-chlorophenyl)thio)-6-(trichloro*) 1.80 30369585 XAH 2-((P-chlorophenyl)thio)-4-(methylamino)-6-(T*) 1.21 30576319 XAH 2-amino-4-(2,4-dichlorophenyl)thio-6-trichloro* 2.18 30863606 XAH 2-methoxy-4-phenyl-6-trichloromethyl-S-triazine 1.38 30863628 XAH 2-(P-chlorophenyl)-4-methoxy-6-trichloromethyl* 1.80
103112352 XAH ethyl-1-(2,4-dichlorophenyl)-5-trichloromethyl-* 0.99 75956 XAlka pentabromoethane 2.98
354472 XAlka ethane, pentabromofluoro- 3.32 422264 XAlka 1,1,1,2,2,3-hexachloro-3-fluoropropane 3.25 422286 XAlka 1,1,2,2,3,3-hexachloro-1-fluoropropane 3.25 422402 XAlka 1,1,1,2,3,3-hexachloro-2-fluoropropane 3.25 422786 XAlka 1,1,1,2,2,3,3-heptachloro-3-fluoropropane 3.52 422811 XAlka 1,1,1,2,3,3,3-heptachloro-2-fluoropropane 3.52 431798 XAlka 1,1,1,2,3,3-hexachloro-3-fluoropropane 3.25 507255 XAlka methane, tetraiodo- 3.39 558167 XAlka bromotriiodomethane 3.39 594218 XAlka 1,1,1-triiodoethane 1.88 594672 XAlka pentaiodoethane 2.98 594730 XAlka hexabromoethane 3.32 594898 XAlka 1,1,1,2,2,3,3-heptachloropropane 3.25 594901 XAlka octachloropropane 3.52 629094 XAlka hexane, 1,6-diiodo- 0.16 630251 XAlka ethane, 1,2-dibromo-1,1,2,2-tetrachloro- 3.32
1522889 XAlka 2,2-bis(jodmethyl)-1,3-dijodpropan 1.62 1529686 XAlka Butane, 1,2,3,4-tetrabromo- 1.74 2657672 XAlka 1,2,3,4-tetrabromobutane 1.74 3228997 XAlka propane, 1,3-dichloro-2,2-bis(chloromethyl)- 1.62 3229003 XAlka propane, 1,3-dibromo-2,2-bis(bromomethyl)- 1.62 3607781 XAlka 1,1,1,3,3,3-hexachloropropane 2.93 3849330 XAlka propane, 1,1,1,2,3,3,3-heptachloro- 3.25 14059906 XAlka dibromodiiodomethane 3.39 14349827 XAlka chlorotriiodomethane 3.39 21981339 XAlka Butane, 1,1,1,3,3-pentachloro- 2.28 24425976 XAlka 1,1,1,2,2,3-hexachloropropane 2.93 54268029 XAlka 1,2,2,3-tetrabromopropane 2.02 58443860 XAlka hexane, 1,2,5,6-tetrabromo- 1.61 79504022 XAlka bromopentachloroethane 3.32 83293827 XAlka 1,2,2,3,3-pentachlorobutane 2.28
90993 XBDP benzene, 1,1'-(chloromethylene)bis- 1.48 101768 XBDP benzene, 1,1'-methylenebis 4-chloro- 2.08 134838 XBDP benzene, 1-chloro-4-(chlorophenylmethyl)- 2.08 776749 XBDP benzene, 1,1'-(bromomethylene)bis- 1.48
2642800 XBDP 1,1-bis(p-chlorophenyl)-2-chloroethane 1.98 5789300 XBDP benzene, 1,1'-(1,2-dibromo-1,2-ethanediyl)bis- 1.48 5963495 XBDP 1,2-dichloro-1,2-diphenylethane 1.48 12110393 XBDP methylium, chlorodiphenyl-, (OC-6-11)-hexachloroantimonate(1-) 1.48 13391394 XBDP benzene, 1-chloro-3-(chlorophenylmethyl)- 2.08 24161146 XBDP methylium, chlorodiphenyl- (8CI,9CI) 1.48 25249392 XBDP bis(chlorophenyl)methane 2.08 56961478 XBDP benzene, 1-chloro-2-(chlorophenylmethyl)- 2.08 59485346 XBDP benzene, 1,1'-(1,2-ethanediyl)bis[2-bromo- 1.48
77
86986 XH quinoline, 4,7-dichloro- 2.89 3141240 XH thiophene, 2,3,5-tribromo- 0.95 3842555 XH 2-chloro-4,6-diphenyl-S-triazine -0.65 4203183 XH quinoline, 4,6-dichloro- 2.89 4350418 XH pyridine, 2-[(4-chlorophenyl)methyl]- 1.48 4409114 XH pyridine, 4-[(4-chlorophenyl)methyl]- 1.48 6484259 XH quinazoline, 4-chloro-2-phenyl- 0.43 13838341 XH 2,4-dichloro-6-(P-tolyloxy)-S-triazine 1.48 20815547 XH pyridine, 4-[(P-bromophenyl)azo]- 0.92 21039781 XH cinnoline, 6-chloro-4-phenyl- 0.43 25117504 XH pyridine, 2-[(P-bromophenyl)azo]- 0.92 25117515 XH quinoline, 2-[(P-bromophenyl)azo]- -0.65 25117526 XH quinoline, 7-[(P-bromophenyl)azo]- -0.65 25187193 XH quinoxaline, 6-chloro-2-phenyl- 0.43 30358168 XH 2-amino-4,6-bis(P-chlorophenoxy)-S-triazine -1.78 30369212 XH 2-amino-4,6-bis(P-chlorophenyl)-S-triazine -0.73 30894770 XH 2-chloro-4,6-bis(O-chlorophenoxy)-S-triazine -0.88 30894929 XH 2-chloro-4-(diazomethyl)-6-phenyl-S-triazine 0.38 30894930 XH 2-chloro-4-(P-chlorophenyl)-6-phenyl-S-triazine -0.19 30894963 XH 2,4-dibromo-6-phenoxy-S-triazine 2.08 31721923 XH 5H-benzo[4,5]cyclohepta[1,2-b]pyridin-5-one, 7-chloro-10,11-dihydro- 0.43 59666169 XH quinoline, 2,7-dichloro-4-methyl- 3.01 63905964 XH 4-pyridinethiol, 3,5-diiodo- 1.65 74129150 XH pyridine, 2-[(4-bromophenyl)methyl]- 1.48 464415 XNOR bicyclo[2.2.1]Heptane, 2-chloro-1,7,7-trimethyl-, endo- 2.16
22819701 XNOR tricyclo[5.1.0.03,5]octane, 8,8-dichloro-1,4,4-trimethyl-, (1a,3b,5b,7a)- (9CI) 3.86
30462534 XNOR bicyclo[2.2.1]heptane, 2-chloro-1,7,7-trimethyl-, (1R,2S,4R)- 2.16 259790 XPAH biphenylene 2.11
17219942 XPAH 9,10-dichlorophenanthrene 2.15 59116880 XPAH dichlorophenanthrene, isomer 2.15 86329604 XPAH dichlorofluoranthene, isomer 1.52
80079 XUC benzene, 1,1'-sulfonylbis 4-chloro- 0.91 81867 XUC 1H,3H-naphtho 1,8-cd pyran-1,3-dione, 6-bromo- 1.00 82439 XUC 9,10-anthracenedione, 1,8-dichloro- 0.94 82440 XUC 9,10-anthracenedione, 1-chloro- 0.43 82462 XUC 9,10-anthracenedione, 1,5-dichloro- 0.94 83056 XUC 2,2-bis(p-chloro-phenyl)acetic acid 0.35 84457 XUC 9,10-anthracenedione, 2,3-dichloro- 0.94 85290 XUC methanone, (2-chlorophenyl)(4-chlorophenyl)- 1.48 85461 XUC 1-naphthalenesulfonyl chloride 0.79 90904 XUC methanone, (4-bromophenyl)phenyl- 0.92 90982 XUC methanone, bis(4-chlorophenyl)- 1.48 93118 XUC 2-naphthalenesulfonyl chloride 0.79 94177 XUC peroxide, bis(4-chlorobenzoyl) -0.71
104245 XUC benzoyl chloride, 4-(phenylazo)- -0.15 117237 XUC 3H-pyrazol-3-one, 2-(2-chloro-6-methylphenyl)-2,4-dihydro-5-methyl- 0.92 129351 XUC 9,10-anthracenedione, 1-chloro-2-methyl- -0.11 133142 XUC peroxide, bis(2,4-dichlorobenzoyl) 0.08 134850 XUC methanone, (4-chlorophenyl)phenyl- 0.92 135126 XUC benzene, 4-chloro-1-(4-chlorophenoxy)-2-nitro- 0.35 322758 XUC benzene, 1-(4-chlorophenoxy)-2-nitro-4-(trifluoromethyl)- 0.63 330938 XUC di-(P-fluorphenyl)-ether 2.08 340578 XUC mecloqualone -0.65 342234 XUC methanone, bis(2-fluorophenyl)- 1.48 342256 XUC methanone, (2-fluorophenyl)(4-fluorophenyl)- 1.48 345551 XUC ethanone, 1-(2'-fluoro[1,1'-biphenyl]-4-YL)- 0.38 345926 XUC 4,4'-difluorobenzophenone 1.48 447314 XUC ethanone, 2-chloro-1,2-diphenyl- 0.38 569062 XUC 9,10-anthracenedione, 1-fluoro- 0.43 602211 XUC 1,2-benzenedicarboxylic acid, 3,4,5,6-tetrachloro-, monoethyl ester (9CI) 1.75 602255 XUC 9,10-anthracenedione, 1,4-dichloro- 0.94 602379 XUC naphthalene, 1-chloro-8-nitro- 1.48 602733 XUC 9,10-anthracenedione, 1,3-dichloro- 0.94 605403 XUC 9,10-anthracenedione, 2,6-dichloro- 0.94 605436 XUC 9,10-anthracenedione, 2,7-dichloro- 0.94 605618 XUC naphthalene, 1-chloro-4-nitro- 1.48 605630 XUC naphthalene, 1-chloro-5-nitro- 1.48 606008 XUC benzoic acid, 2-amino-3,5-dibromo-, methyl ester 2.16 607227 XUC naphthalene, 1-chloro-2-nitro- 1.48
78
607374 XUC naphthalene, 7-chloro-1-nitro- 1.48 625887 XUC thiophene, 2,5-diiodo- 1.97 631936 XUC benzenesulfonamide, 4-chloro-N-(4-chlorophenyl)-N-methyl- -0.19 632837 XUC 1-bromoantrhraquinone 0.43 677678 XUC acetyl fluoride, difluoro(fluorosulfonyl)- 0.64 712481 XUC diphenylchloroarsine 0.55 720741 XUC ethanone, 1-(4'-fluoro[1,1'-biphenyl]-4-YL)- 0.38 727515 XUC 9,10-anthracenedione, 1,6-dichloro- 0.94 727991 XUC methanone, phenyl[2-(trifluoromethyl)phenyl]- 1.38 728814 XUC methanone, phenyl[3-(trifluoromethyl)phenyl]- 1.38 728869 XUC methanone, phenyl[4-(trifluoromethyl)phenyl]- 1.38 734598 XUC phosphine, (4-bromophenyl)diphenyl- -1.17 790410 XUC benzoic acid, 4-chloro-, anhydride with 4-chlorobenzoic acid -0.19 958009 XUC 11H-dibenzo[C,F][1,2]diazepin-11-one, 3,8-dichloro- 0.94
1016779 XUC methanone, (3-bromophenyl)phenyl- 0.92 1016780 XUC methanone, (3-chlorophenyl)phenyl- 0.92 1079669 XUC phosphinous chloride, diphenyl- 0.79 1084986 XUC 11H-dibenzo[C,F][1,2]diazepine, 3,8-dichloro- 1.53
1107002 XUC 1,3-Isobenzofurandione, 5,5'- 2,2,2-trifluoro-1-(trifluoromethyl)ethylidene bis- -2.25
1111917 XUC germane, triiodomethyl- (7CI,8CI,9CI) -4.31 1137593 XUC [1,1'-biphenyl]-4-ol, 3,5-dichloro- 2.08 1185097 XUC ethanesulfenyl chloride, 1,1,2,2-tetrachloro- 1.33 1483723 XUC iodonium, diphenyl-, chloride 1.92 1483734 XUC iodonium, diphenyl-, bromide 1.92 1484500 XUC ethanone, 2-bromo-1,2-diphenyl- 0.38 1488422 XUC iodonium, (2-carboxyphenyl)phenyl-, hydroxide, inner salt 0.35 1545831 XUC diazene, (4-fluorophenyl)phenyl- 0.92 1594690 XUC 9,10-Anthracenedione, 1,7-dichloro- 0.94 1596367 XUC benzenesulfonamide, N-(4,5-dichloro-2-nitrophenyl)- -1.22 1602002 XUC P,P'-dichloroazobenzene 1.48 1613667 XUC germane, dichlorodiphenyl- 0.55 1623934 XUC [1,1'-biphenyl]-4-sulfonyl chloride -0.23 1714507 XUC 4,4'-dichlorobenzophenone, oxime 0.91 1729993 XUC 1-naphthalenecarboxylic acid, 8-bromo- 1.48 1806231 XUC methanone, (2-chlorophenyl)(4-fluorophenyl)- 1.48 1836733 XUC benzene, 1-chloro-3-methyl-2-(4-nitrophenoxy)- -0.66 1836744 XUC benzene, 1-chloro-4-(4-nitrophenoxy)- -0.15 1928398 XUC Acetic acid, (2,4,5-trichlorophenoxy)-, ethyl ester (6CI,7CI,8CI,9CI) 1.88 2005085 XUC benzoic acid, 4-chlorophenyl ester -0.15 2069412 XUC methanone, (4-bromophenyl)(4-fluorophenyl)- 1.48 2091614 XUC benzene, 1-chloro-2-(4-nitrophenoxy)- -0.15 2117693 XUC Plumbane, dichlorodiphenyl- 0.55 2181422 XUC trimethylsulfonium iodide -3.07 2303233 XUC benzene, 1-chloro-3-(4-nitrophenoxy)- -0.15 2392485 XUC benzene, 4-chloro-1-(2,4-dichlorophenoxy)-2-nitro- 0.80 2547617 XUC methanesulfonyl chloride, trichloro- 0.54 2629472 XUC stibine, chlorodiphenyl- 0.55 2712836 XUC butyranilide, 2,2,3,3,4,4,4-heptafluoro-2'-hydroxy-4'-nitro- 1.63 2729115 XUC phosphine oxide, triS(pentafluorophenyl)- 1.99 2750814 XUC naphthalene, 1,4-dichloro-5-nitro- 2.08 2790161 XUC cyclopropanecarboxamide, N-(3,4-dichlorophenyl)-1-methyl- 2.08 2892617 XUC 3-butenoic acid, 2,3,4,4-tetrachloro-, butyl ester 2.16 2899027 XUC urea, N,N'-dichloro-N,N'-bis(2,4,6-trichlorophenyl)- 1.69
2915017 XUC 4(3H)-Quinazolinone, 2-methyl-3-(a,a,a-trifluoro-o-tolyl)-, monohydrochloride (8CI) -0.34
2965777 XUC 4(3H)-Quinazolinone, 2-methyl-3-(a,a,a-trifluoro-m-tolyl)-, monohydrochloride (8CI) -0.34
3033736 XUC Peroxide, bis(2-chlorobenzoyl) -0.71 3096477 XUC 2-chloro-9H-fluoren-9-one 1.59 3236973 XUC stannane, triiodomethyl- (8CI,9CI) 1.46 3457463 XUC ethanedione, bis(4-chlorophenyl)- 0.35 3622400 XUC benzothiazole, 2-bromo-4-chloro- 2.83 3808897 XUC ethanone, 1-(2'-chloro[1,1'-biphenyl]-4-YL)- 0.38 3905962 XUC 4,4'-dinitro-octafluorobiphenyl 1.29 3988032 XUC methanone, bis(4-bromophenyl)- 1.48 4053081 XUC 1H,3H-naphtho[1,8-cd]pyran-1,3-dione, 6-chloro- 1.00 4185631 XUC naphthalene, 2-chloro-1-nitro- 1.48 4189100 XUC 2-ethylbuttersaeure-2,2,2-trichlorethylester 1.63 4340776 XUC diazene, (4-chlorophenyl)phenyl- 0.92
79
4418842 XUC diazene, (4-bromophenyl)phenyl- 0.92 4537002 XUC 1-naphthalenecarboxylic acid, 8-chloro- 1.48 4559960 XUC 1-butanone, 1-(4-bromophenyl)-4-chloro- 2.16 4693009 XUC 2H-3,1-benzoxazine-2,4(1H)-dione, 6,8-dichloro- 2.70 4841207 XUC propanoic acid, 2-(2,4,5-trichlorophenoxy)-, methyl ester 1.88 4889707 XUC benzoic acid, 2-[(4-chlorophenyl)methyl]- -0.15 5074715 XUC phosphine, bis(2,3,4,5,6-pentafluorophenyl)phenyl- 1.67 5162038 XUC methanone, (2-chlorophenyl)phenyl- 0.92 5317668 XUC benzeneacetic acid, a,a-dichloro-, ethyl ester 1.92 5323671 XUC benzenemethanol, 2,4-dichloro-, carbonate (2:1) (9CI) 0.08 5659416 XUC 2,3,4,5,5-pentachloro-2,4-pentadienoic acid 2.18 5731011 XUC ethanone, 1-(4'-bromo[1,1'-biphenyl]-4-yl)- 0.38 5774061 XUC 1-naphthalenecarboxylic acid, 3-chloro- 1.48 5858195 XUC benzo[b]thiophen-3(2H)-one, 5,7-dichloro-4-methyl- 3.01 6141953 XUC diazene, (4-chlorophenyl)phenyl-, (E)- 0.92 6240557 XUC naphthalene, 1,2-dichloro-3-nitro- 2.08 6284793 XUC methanone, (3,4-dichlorophenyl)phenyl- 1.48 6297116 XUC 9H-fluoren-9-one, 2,7-dichloro- 2.15 6530978 XUC diazene, (4-chlorophenyl)phenyl-, (Z)- 0.92 6639276 XUC diazene, (4-iodophenyl)phenyl- 0.92 6657052 XUC 2-chloro-4-phenylazophenol 0.38 6723406 XUC 1H-indene-1,3(2H)-dione, 2-[4-(trifluoromethyl)phenyl]- 0.22 6740869 XUC methanone, (1-bromocyclopentyl)(2-chlorophenyl)- 2.08 6915577 XUC 1,3,2-benzodioxabismole, 4,5,6,7-tetrabromo-2-hydroxy- 1.75 7012160 XUC 9H-fluorene, 2,7-dichloro- 2.80 7334330 XUC diazene, bis(2-chlorophenyl)- 1.48 7335833 XUC benzoic acid, 2-benzoyl-3,4,5,6-tetrachloro-, methyl ester -0.30 7791277 XUC pyrosulfuryl chloride -1.83 8072206 XUC [(4-chlorophenyl)thio](2,4,5-trichlorophenyl)diazene (9CI) 0.91 10182840 XUC iodonium, diphenyl- 2.03 13036754 XUC disulfuryl fluoride -1.83 13102346 XUC 3H-Pyrazol-3-one, 2-(2,5-dichlorophenyl)-2,4-dihydro-3-methyl- 2.08 13124179 XUC 3H-Pyrazol-3-one, 2-(3,4-dichlorophenyl)-2,4-dihydro-5-methyl- 2.08 13191361 XUC thiophene, 2,5-dibromo-3-methyl- 1.72 13304240 XUC benzoic acid, 2-[(4-chlorophenyl)azo]- (9CI) -0.66 13455000 XUC hypodiphosphorous tetraiodide -3.90 13779925 XUC niobium pentaiodide -4.72 14529125 XUC 2H-3,1-benzoxazine-2,4(1H)-dione, 6-chloro-1-methyl- 1.48 15426149 XUC diazene, 1,2-bis(3-chlorophenyl)- 1.48
15608374 XUC 1,3,5,2,4,6-triazatriphosphorine, 2,2,4,4,6-pentabromo-6-chloro-2,2,4,4,6,6-hexahydro- (8CI) -0.68
15945285 XUC benzoic acid, 3,5-dichloro-4-methoxy-, ethyl ester 1.36
15965001 XUC 1,3,5,2,4,6-triazatriphosphorine, 2,2,4,6-tetrabromo-4,6-dichloro-2,2,4,4,6,6-hexahydro- (8CI) -0.68
16032523 XUC 1,3,5,2,4,6-triazatriphosphorine, 2,4,6-tribromo-2,4,6-trichloro-2,2,4,4,6,6-hexahydro- (8CI) -0.68
16187943 XUC isothiazole, 4-bromo-3-phenyl- 2.33 16442107 XUC benzoic acid, 2,6-dichloro-, anhydride (8CI,9CI) 0.63 16574520 XUC methanone, (4-fluorophenyl)[4-(trifluoromethyl)phenyl]- 1.80 16650525 XUC 1-naphthalenecarboxylic acid, 5-chloro- 1.48 16965085 XUC carbonic acid, 1,1-dimethylethyl 2,4,5-trichlorophenyl ester 1.31 17175132 XUC carbonic acid, P-bromophenyl phenyl ester -0.15 17318706 XUC benzenesulfonyl chloride, 4-benzoyl- -1.20 17640083 XUC propionic acid, 2,2-dichloro-, pentyl ester 1.01 17969221 XUC thiazole, 4-(chloromethyl)-2-(4-chlorophenyl)- 2.70 18018092 XUC 9,10-anthracenedione, 1,3-dichloro-2-methyl- 0.37 18018332 XUC benzimidazole, 1-ethyl-2-methyl-6-nitro-5-(trifluoromethyl)- 0.80 18189190 XUC 1,2-ethanedione, 1-[4-(bromomethyl)phenyl]-2-phenyl- -0.66 18264841 XUC phenanthrene, 2-bromo-9,10-dihydro-3-nitro- -0.11
18812812 XUC 2-propanamine, N,N'-[4-(chloromethyl)-4-methyl-2,3-oxetanediylidene]bis[2-methyl- (9CI) -0.74
19055700 XUC 1H-indene-1,3(2H)-dione, 2-[3-(trifluoromethyl)phenyl]- 0.22 19762799 XUC 4-isothiazolecarbonitrile, 3-(P-bromophenyl)-5-methyl- 0.92 19811053 XUC methanone, (2,4-dichlorophenyl)phenyl- 1.48
19982877 XUC hydrazinecarboximidamide, 2-[(4-fluorophenyl)[4-(trifluoromethyl)phenyl]methylene]-, monohydrochloride (9CI) -0.47
20357793 XUC 1H-indene, 1,2-dibromo-2,3-dihydro- 2.83 20717797 XUC 2-naphthalenecarboxylic acid, 1-bromo- 1.48 21221932 XUC methanone, [3,5-bis(trifluoromethyl)phenyl]phenyl- 1.91 21854955 XUC ethanedione, bis(2-chlorophenyl)- 0.35
80
22071245 XUC methanone, [3-(bromomethyl)phenyl]phenyl- 0.38 22532689 XUC benzene, 1,2,4-trichloro-5-(4-nitrophenoxy)- 0.80 22544021 XUC benzene, 4-chloro-1-(2-chlorophenoxy)-2-nitro- 0.35 22544076 XUC benzene, 2-chloro-1-(4-chlorophenoxy)-4-nitro- 0.35 22711235 XUC 1,2-ethanedione, 1-(4-chlorophenyl)-2-phenyl- -0.15 23038360 XUC benzene, 1-bromo-4-(phenylsulfonyl)- 0.38 25004959 XUC methanesulfinyl chloride, 1,1,1-trichloro- 0.64 25006320 XUC leptophos-O-analog 0.24 25100918 XUC thiazole, 2-(4-chlorophenyl)-4-methyl- 2.07 25187068 XUC methanone, (2-chlorophenyl)(2,5-dichlorophenyl)- 1.98 25512429 XUC dichloro-1,1'-biphenyl 2.08 25981833 XUC 3H-indole, 5-chloro-2,3,3-trimethyl- 2.07 26203816 XUC acetic acid, (2,4,6-tribromophenoxy)-, ethyl ester (8CI) 1.88 27087463 XUC carbonic acid, P-chlorophenyl phenyl ester -0.15 27701662 XUC 1,1'-biphenyl, 4-bromo-3-nitro- 0.38 28202304 XUC benzene, 2-methyl-4-nitro-1-(1,1,2,2-tetrafluoroethoxy)- 2.34
28313522 XUC acetic acid, [[2-chloro-5-(trifluoromethyl)phenyl]hydrazono]cyano-, ethyl ester (9CI) 0.08
28313577 XUC acetic acid, [[4-chloro-3-(trifluoromethyl)phenyl]hydrazono]cyano-, ethyl ester (9CI) 0.08
28313691 XUC acetic acid, cyano[[2,6-dichloro-4-(trifluoromethyl)phenyl]hydrazono]-, methyl ester (9CI) 0.99
28313771 XUC acetic acid, [[4-chloro-2-(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester (9CI) 0.62
28317568 XUC butanenitrile, 2-[[2-chloro-5-(trifluoromethyl)phenyl]hydrazono]-3-oxo- (9CI) 1.18
28317944 XUC pentanenitrile, 2-[[4-chloro-2-(trifluoromethyl)phenyl]hydrazono]-4,4-dimethyl-3-oxo- (9CI) -0.44
28322783 XUC acetic acid, cyano[(2,4,5-trichlorophenyl)hydrazono]-, ethyl ester (9CI) 0.22
28343251 XUC pentanenitrile, 2-[[2-chloro-5-(trifluoromethyl)phenyl]hydrazono]-4,4-dimethyl-3-oxo- (9CI) -0.44
28343273 XUC pentanenitrile, 2-[[3,5-bis(trifluoromethyl)phenyl]hydrazono]-4,4-dimethyl-3-oxo- (9CI) -0.32
28801914 XUC 9H-fluorene, bromo- 2.20 29170089 XUC 1,1'-biphenyl, 4-iodo-4'-nitro- 0.38 29328992 XUC 1H-benzotriazoLE, 5-chloro-1-(4-chlorophenyl)- 1.53 29421736 XUC thiophene, 3,5-dibromo-2-methyl- 1.72 29558881 XUC benzoic acid, 4-chloro-, 4-fluorophenyl ester 0.91 30070634 XUC benzoic acid, 3-chloro-, 1,1'-anhydride -0.19 31039722 XUC 1H-isoindole-1,3(2H)-dione, 4,5,6,7-tetrachloro-2-(4-chlorophenyl)- 1.54 31431160 XUC methanone, (4-chloro-3-nitrophenyl)(4-fluorophenyl)- -0.19 31643148 XUC benzenemethanol,a-(trichloromethyl)-, propanoate 1.88 32752548 XUC methanone, [4-(bromomethyl)phenyl]phenyl- 0.38 33146575 XUC methanone, (4-chlorophenyl)(2,4-dichlorophenyl)- 1.98 33184520 XUC methanone, (5-fluoro-2-methylphenyl)phenyl- 0.38 33184553 XUC (5-chloro-2-methylphenyl)phenylmethanone 0.38 33713132 XUC benzamide, N-[2-chloro-5-(trifluoromethyl)phenyl]-2-hydroxy-3-nitro- -1.00 34052374 XUC methanone, (2-chloro-5-nitrophenyl)phenyl- -0.66 34113467 XUC 2-(2-chlorophenyl)-2-(4-chlorophenyl)acetic acid 0.35 34183067 XUC methanone, (3,5-dichloro-4-hydroxyphenyl)phenyl- 0.91 34223820 XUC 9H-fluoren-9-one, 3,6-dichloro- 2.15 34593754 XUC 4-(tert-butyl)-2,6-dichlorophenol 2.16 34645846 XUC Fenclofenac -0.19 35578473 XUC ethanedione, bis(4-bromophenyl)- 0.35
36865540 XUC acetic acid, [[3,5-bis(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester, (Z)- (9CI) 0.76
36865722 XUC acetic acid, [[2-chloro-4-(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester (9CI) 0.62
36905007 XUC acetic acid, [[3,5-bis(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester, (E)- (9CI) 0.76
36905041 XUC acetic acid, cyano[(2,4,5-trichlorophenyl)hydrazono]-, methyl ester, (Z)- (9CI) 0.75
37572444 XUC isothiazole, 4-bromo-3,5-bis(methylthio)- 1.00 38396213 XUC naphthalene, 2-chloro-3-nitro- 1.48 38396291 XUC naphthalene, 6-chloro-1-nitro- 1.48 38461299 XUC benzene, 2,4-dichloro-1-(2-nitrophenoxy)- 0.35 39145476 XUC benzene, 1-(4-chlorophenoxy)-2-nitro- -0.15 39145487 XUC benzene, 1,4-dichloro-2-(4-nitrophenoxy)- 0.35 40877096 XUC 1-butanone, 4-chloro-1-(4-chlorophenyl)- 2.16 41182858 XUC benzenecarboximidoyl bromide, N-methyl- 1.00 41893703 XUC benzene, [bromo(1-bromocyclopropyl)methyl]- 2.89
81
42728621 XUC methanone, [4-(chloromethyl)phenyl]phenyl- (9CI) 0.38 42771799 XUC ethanone, 1-(2-fluoro[1,1'-biphenyl]-4-yl)- 0.38 42959999 XUC methanone, (2-bromo-3-nitrophenyl)phenyl- -0.66 42989162 XUC benzoic acid, 2,4,5-trichloro-6-cyano-3-phenoxy-, methyl ester -0.83 49619434 XUC benzoic acid, 2-chloro-, 1,1'-anhydride -0.19 50274641 XUC methanone, (2-chloro-4-nitrophenyl)(4-chlorophenyl)- -0.19 50274969 XUC benzene, 2-chloro-1-[(2,4-dichlorophenyl)methyl]-4-nitro- 0.80 50594440 XUC phenol, 5- 2-chloro-4-(trifluoromethyl)phenoxy -2-nitro-, acetate (ester) -1.52 50594779 XUC phenol, 3- 2-chloro-4-(trifluoromethyl)phenoxy -, acetate 0.08 50618923 XUC benzene, 1,1'-(1,2-ethanediyl)bis[4-fluoro-2-nitro- -1.73 50807175 XUC ethanone, 2,2-dichloro-1-[4-(1,1-dimethylethyl)phenyl]- 1.36 50817726 XUC 1H,3H-naphtho[1,8-cd]pyran-1,3-dione, 4-chloro- 1.00 51282690 XUC methyl 5-(2-chloro-4-fluorophenoxy)-2-nitrobenz* -1.73 51326375 XUC ethanone, 2,2,2-trichloro-1- 4-(1,1-dimethylethyl)phenyl - 1.88 51735876 XUC phophorimidic trichloride, [(trifluoromethyl)sulfinyl]- -0.03 53218114 XUC benzeneacetic acid, 4-nitro-, 4-chlorophenyl ester -1.66 53317058 XUC 2-thiophenecarboxylic acid, 3,4,5-tribromo- 2.38 53526297 XUC naphthalene, 1-chloro-3-nitro- 1.48 53591790 XUC ethanone, 1-(2',4'-difluoro[1,1'-biphenyl]-4-yl)- 0.91 53911685 XUC 2H-Pyran-2,6(3H)-dione, 4-(4-chlorophenyl)dihydro- 0.92 54683919 XUC benzoic acid, 2-chlorophenyl ester 0.38 54934495 XUC propanoic acid, 2,2-dichloro-, 1,2-ethanediyl ester 2.17
55521147 XUC Carbonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8-hexadecafluoro-1,8-octanediyl dimethyl ester (9CI) 1.81
55683244 XUC benzeneacetic acid, 3,4-bis(2,2,3,3,3-pentafluoro-1-oxopropoxy)-, methyl ester (9CI) -0.40
55702426 XUC [1,1'-biphenyl]-2-carboxylic acid, 4,4'-dichloro-, methyl ester 0.35 55702448 XUC 1H-benzimidazole, 2-ethyl-7-nitro-1-propyl-5-(trifluoromethyl)- -0.30 55780411 XUC cycloPROP[A]INdene, 6-bromo-1,1A,6,6A-tetrahydro- 3.10 56041759 XUC benzenemethanol, 4-bromo-a-cyclopropyl-a-methyl- 2.07 56107029 XUC methanone, (4-chloro-3-nitrophenyl)phenyl- -0.66 56454150 XUC benzenebutanoic acid, G-bromo-, ethyl ester 0.25 56882521 XUC ethyl-2,4-dichlorobenzoate 2.16 56960953 XUC benzoic acid, 4-chloro-3-nitro-, (4-chlorophenyl)methyl ester -1.22 56960964 XUC benzoic acid, 2-chloro-5-nitro-, (4-chlorophenyl)methyl ester -1.22 56961003 XUC benzenesulfonic acid, 4-chloro-, (4-chlorophenyl)methyl ester -0.19 56961365 XUC naphthalene, 1-chloro-6-nitro- 1.48 56961376 XUC naphthalene, 2-chloro-6-nitro- 1.48 56961387 XUC naphthalene, 2-chloro-7-nitro- 1.48 56961398 XUC naphthalene, 3-chloro-1-nitro- 1.48 56961898 XUC 2-naphthalenecarboxylic acid, 5-chloro- 1.48 56961989 XUC 9,10-Phenanthrenedione, 2-chloro- 0.43 56961990 XUC 9,10-Phenanthrenedione, 3-chloro- 0.43 56966699 XUC benzene, 2-chloro-4-nitro-1-phenoxy- -0.15 56978508 XUC 1,1'-biphenyl, 4,4'-dichloro-2,2'-dinitro- -0.71 57025760 XUC MC-15608 [2,4'-dicl-4-CF3-3'-CO2Me-diPh ether] 0.42 57147054 XUC 1,4-benzenedimethanol, 2,3,5,6-tetrabromo-, diacetate (9CI) 0.62 57340213 XUC diazene, 1-[4-(bromomethyl)phenyl]-2-phenyl- 0.38 57396879 XUC [1,1'-biphenyl]-4-OL, 4'-chloro-, acetate -0.15 57479604 XUC ethanone, 2-(2-chlorophenyl)-1-phenyl- 0.38 58359538 XUC benzenesulfonyl chloride, 4-(phenylazo)- -1.20 58748166 XUC bis(2,4-dichlorobenzoyl) peroxide (9CI) -0.71 59089677 XUC [1,1'-biphenyl]-4-ol, 2',4'-difluoro-, acetate (9CI) 0.35 59189514 XUC methanone, (2,6-difluorophenyl)phenyl- 1.48 59396508 XUC methanone, (4-fluoro-3-methylphenyl)(4-fluorophenyl)- 0.91 59396519 XUC methanone, (2-fluoro-5-methylphenyl)phenyl- 0.38 59584224 XUC benzoic acid, 4-[[(4-chlorophenyl)imino]methyl]-, methyl ester -1.17 59700571 XUC pentachloro(2,2,3,3-tetrafluoropropoxy)cyclotri* 2.37 60066537 XUC hexanoic acid, 4,6,6,6-tetrachloro-3,3-dimethyl-, ethyl ester 2.17 61405489 XUC RH-1460 diPh thioether -1.52 62336247 XUC phosphine, (2-bromophenyl)diphenyl- -1.17 62476577 XUC benzene, 2-chloro-1-(3-nitrophenoxy)-4-(trifluoromethyl)- 0.63 62586963 XUC benzenemethanol, a,a-dicyclopropyl-4-fluoro- 1.59 62625245 XUC ethanone, 2-bromo-1- 4-(phenylazo)phenyl - -0.66 63028273 XUC 2,5-dichloro-4'-isopropylbiphenyl 0.91 63549337 XUC 1-propanone, 3-chloro-1-(5-chloro-2-methylphenyl)- 2.16 63549348 XUC 1-propanone, 3-chloro-1-(2-chloro-5-methylphenyl)- 2.16 63734623 XUC benzoic acid, 3- 2-chloro-4-(trifluoromethyl)phenoxy - 0.63 63867072 XUC Butanoic acid, 2-ethyl-, trichloro-1-methylethyl ester (9CI) 1.88 63867083 XUC Butanoic acid, 2-bromo-2-ethyl-, 2,2,2-tribromoethyl ester 2.16
82
63867107 XUC Butanoic acid, 2-chloro-2-ethyl-, 2,2,2-trichloroethyl ester 2.16 63867174 XUC 2-ethylbuttersaeure-2,2,2-tribromethylester 1.63 63979442 XUC pentanoic acid, 2,2-dibromo-, ethyl ester 0.75 64047478 XUC butanoic acid, 2-bromo-3-methyl-, 2,2,2-tribromoethyl ester 1.96 64047489 XUC butanoic acid, 2-bromo-3-methyl-, 2,2,2-trichloroethyl ester 1.96 64048904 XUC arsonous dichloride, [3-(trifluoromethyl)phenyl]- (9CI) 1.73 64057767 XUC benzenemethanol, 3,4-dichloro-, carbonate (2:1) (9CI) 0.08 64059376 XUC butanoic acid, 2-bromo-3-methyl-, trichloro-1-methylethyl ester (9CI) 2.16 64123644 XUC benzene, 2-(chloromethyl)-1-(1-methylethyl)-4-nitro- 0.76 64399264 XUC cyclopentanecarbonitrile, 1-(4-chlorophenyl)- 1.48 64667330 XUC hexanoic acid, 4,6,6,6-tetrachloro-3,3-dimethyl-, methyl ester 2.39 65237174 XUC 1,3-Isobenzofurandione, 5,6-dibromo- 3.01 65601414 XUC 2,2,4-trichloro-2,2,4,4,6,6-hexahydro-4,6,6-tri* 1.31 68015952 XUC benzene, 2-(chloromethyl)-1-(1-methylpropyl)-4-nitro- 0.25 69943471 XUC methanone, (2-chloro-4-fluorophenyl)phenyl- 1.48 70788566 XUC methanone, 4-(dichloromethyl)phenyl (2,5-dichlorophenyl)- 1.80 71463495 XUC benzoic acid, 2,6-dichloro-, 2,6-dichlorophenyl ester 1.80 71463553 XUC cyclopropanecarbonitrile, 1-(2,4-dichlorophenyl)- 3.01 71501514 XUC [1,1'-biphenyl]-2-carboxylic acid, 3'-chloro-4'-methyl- -0.15 71849984 XUC benzoic acid, 2,6-dichloro-, phenyl ester 0.91 73806710 XUC 2,6-dimethyl-2,3,5,6-tetrabromo-4-heptanone 2.16 74129105 XUC ethanone, 1-[2-bromo-4-(phenylazo)phenyl]- (9CI) -0.66 74204049 XUC 9,10-anthracenedione, 1-bromo-4-methyl- -0.11 74298912 XUC 1,1'-biphenyl, 3,4',5-trichloro-4-methoxy- 1.98 74298923 XUC 1,1'-biphenyl, 2,3',4-trichloro-4'-methoxy- 1.98 74610507 XUC benzoic acid, 2-(acetyloxy)-3,5-dibromo-, 4-bromophenyl ester -0.83 74630925 XUC 1-propanol, 1,1-dichloro-, phosphate (3:1) (9CI) 1.83
83721464 XUC methanesulfonamide, 1-chloro-N-[4,5-dichloro-2-(2,4-dichlorophenoxy)phenyl]-, Na salt 0.33
83721475 XUC methanesulfonamide, 1-chloro-N-[2,3,4-trichloro-6-(2,4-dichlorophenoxy)phenyl]-, Na salt 0.64
84392176 XUC 4'-(trifluoromethyl)-2-biphenylcarboxlic acid 0.80 84532729 XUC 1-naphthalenecarboxylic acid, 6-methoxy-5-(trifluoromethyl)- 0.80 85262948 XUC benzenemethanol, 2,4-dichloro-, propanoate 1.92 85721080 XUC methanone, [2-chloro-5-(trifluoromethyl)phenyl](4-fluorophenyl)- 2.18 85721091 XUC methanone, [2-bromo-4-(trifluoromethyl)phenyl](4-fluorophenyl)- 2.18 85897296 XUC chlorofluorenone 1.59 86914729 XUC methanone, (5-chloro-2-hydroxy-3-methylphenyl)(4-chlorophenyl)- 0.35 87750503 XUC benzene, 1-(phenylsulfonyl)-4-(trifluoromethoxy)- 0.24 87750592 XUC methanone, (2-fluoro-5-methylphenyl)[2-(trifluoromethyl)phenyl]- 1.21 87750616 XUC methanone, (2-chloro-4-fluorophenyl)(4-fluorophenyl)- 1.98 88185222 XUC 3-[2-chloro-4-(trifluoromethyl)phenoxy]B -3.04 88927428 XUC ethyl (2,3,4,5-tetrachlorophenoxy)acetate 2.34 90077740 XUC dichloro-9H-fluoren-9-one 2.15 90077751 XUC trichloro-9H-fluoren-9-one 2.64 94070856 XUC naphthalene, 1,3-dibromo-1,2,3,4-tetrahydro- 2.89
94248267 XUC methanesulfonamide, 1-chloro-N-(2-phenoxyphenyl)-, pentachloro deriv., sodium salt 0.64
95998699 XUC dichloro(trifluoromethyl)benzophenoNE 2.18
97659353 XUC 1,3,4-oxadiazol-2(3H)-one, 3-(2-chloro-4-isocyanatophenyl)-5-(1,1-dimethylethyl)- -1.66
104668679 XUC propanamide, 3,3,3-trifluoro-2-methyl-N-[4-nitro-3-(trifluoromethyl)phenyl]-, (±)- 1.32
108548727 XUC (dichlorophenyl)methyl chlorobenzoate 0.80 121325448 XUC benzoic acid, 5-[2-chloro-5-(trifluoromethyl)phenoxy]-2-nitro-, methyl ester -1.52
83
Table 5 - Explanation of the Acronyms for Chemical Class Used in Tables S2 and S3
CUP current use pesticide (or safener)
OP Obsolete pesticide
PBDE polybrominated diphenyl ether
PCB polychlorinated biphenyl
PCBD polychlorinated benzene derivative (chlorination is on benzene ring, functional
groups are simple)
PCDD polychlorinated dibenzo-p-dioxin
PCDE polychlorinated diphenyl ether
PCDF polychlorinated dibenzofuran
PCN polychlorinated naphthalene
PCS polychlorinated styrene
PFA perfluorinated organic acid
POP persistent organic pollutants designated by the Stockholm Convention
PXBD polyhalogenated benzene derivative (same as PCBD but including at least one
fluorine, bromine or iodine atom)
PXBP poly halogenated biphenyl (same as PCB but including at least one fluorine,
bromine or iodine atom)
PXDD polyhaloginated dibenzo-p-dioxin
PXDF polyhaloginated dibenzofuran
PXN polyhaloginated naphthalene
silane contains one or more silicon atoms
Silox contains one or more silicon atoms bonded to an oxygen atom
UC non-halogenated unclassified chemical, does not fall into any of the other
classifications
XAB halogenated alkyl benzene (benzene (possibly directly halogenated) substituted
with a halogenated alkyl)
XABP halogenated alkyl biphenyl (same as XAB but biphenyl instead of benzene)
XAH halogenated alkyl heterocycle (same as XAB but a heterocycle instead of
benzene)
Xalka halogenated alkane
XalkaC halogenated cycloalkane (excluding halogenated cyclohexanes)
Xalke halogenated alkene
XalkeC halogenated cycloalkene
XAN halogenated alkyl naphthalene (same as XAB but a naphthalene instead of
benzene)
XBDP alkyl-bridged halogenated diphenyl
XH halogenated heterocycle
XNOR halogenated norbornene or norbornane
XPAH halogenated polycyclic aromatic hydrocarbon
XUC halogenated unclassified chemical, does not fall into any of the other
classifications
84
Table 6 - Chemical classifications of the chemicals which meet the elevated AC-BAP
and atmospheric oxidation half-life criteria.
Elevated AC-BAP Yes Yes
Long atmospheric oxidation half-life Yes Yes
Fulfills POPs score criterion Yes No
Chemical Class Number Identified Number Identified
PCBs 123 20
Other halogenated biphenyls 4 16
Halo-alkyl biphenyls 3 2
PCNs 19 11
Other halogenated naphthalenes 0 2
Halo-alkyl naphthalenes 1 0
PCDDs 9 17
Other halogenated dibenzo-p-dioxins 1 5
PCDFs 90 20
Other halogenated dibenzofurans 3 3
PBDEs 2 3
Polychlorinated diphenylethers (PCDEs) 60 12
PCSs 3 0
chlorinated benzene derivatives 87 101
Other halogenated benzene derivatives 57 97
Halo-alkyl benzenes 48 40
Perfluorinated organic acids 13 0
Other Stockholm POPs and derivatives 33 0
Pesticides (obsolete) 6 0
Pesticides (current use) 11 43
Halogenated PAHs 2 4
Halogenated alkanes 16 32
Halogenated cycloalkanes 23 0
Halogenated alkenes 7 0
Halogenated cycloalkenes 9 0
Halogenated diphenylmethanes 5 13
Halogenated heterocycles 13 24
Halo-alkyl heterocycles 72 15
Halogenated norbornenes/norbornanes 1 3
silane 1 27
Siloxane 0 27
Unclassified: Non-Halogenated 0 319
Unclassified: Halogenated 100 347
Total 822 1203
85
Chapter 3
Development and Exploration of an Organic Contaminant Fate
Model Using Poly-Parameter Linear Free Energy Relationships
Trevor N. Brown, Frank Wania
Environmental Science & Technology 2009 43 (17), 6676-6683
Reproduced with permission from Environmental Science and Technology
Copyright 2009 American Chemical Society
86
1 Introduction Several years ago Breivik and Wania proposed that the domain of applicability of
multimedia fate models could be expanded by using poly-parameter linear free energy
relationships (PP-LFERs) to quantify phase partitioning111
. The argument for implementing
models of this type was that single-parameter linear free energy relationships (SP-LFERs),
primarily those based on regressions with KOW, make poor predictions of phase partitioning
when applied to chemicals outside of the datasets used in their parameterization112
.
Subsequent work has shown that this approach is viable for chemicals, such as
pharmaceuticals, which may be considered to be outside the range of applicability of SP-
LFER based models113
. More recently Götz et al. implemented PP-LFER based partitioning
into atmospheric transport models and demonstrated that the model results can be
significantly different from SP-LFER based models, especially for polar chemicals114
.
The PP-LFER equations applied in these studies were the linear solvation energy
relationships developed by Abraham et al.115
, or extensions thereof116,117
. These equations
include terms that describe the interactions of the solute (solute descriptors) and of the
phases involved (phase descriptors). Specific intermolecular interactions, such as dipole
interactions and hydrogen bonding, and non-specific intermolecular interactions, such as
cavitation energy and dispersive van der Waals interactions, are accounted for separately.
Three major factors limited the implementation of a PP-LFER-based multimedia fate model
at the time of Breivik and Wania‘s publication; PP-LFER equations were missing for several
environmentally relevant phases, there was no PP-LFER based method for calculating the
temperature dependence of phase partitioning, and the availability of solute descriptors for
environmentally relevant chemicals was limited. There has since been progress on all three
of these issues, and sufficient data is now available for the full implementation of a PP-
LFER-based multimedia fate model.
Much work has been done recently to address the lack of PP-LFER equations for some
environmentally relevant phases. Goss118
has proposed a modification of Abraham‘s PP-
87
LFERs which uses a single form to describe all partitioning processes, instead of separate
forms for partitioning between two condensed phases and between a gas phase and a
condensed phase, as in the traditional Abraham equations115
. This modification has a
statistical disadvantage in that two of the parameters are highly correlated, but as Goss notes
the equations are more convenient for environmental modelling because it allows for the use
of thermodynamic cycles to indirectly derive equations for partitioning systems that have not
been experimentally measured. Goss et al. have subsequently characterized the partitioning
of a wide range of chemicals to environmentally relevant phases, including natural organic
matter57,119
and atmospheric aerosol120,121
, and provided modified PP-LFERs to describe
partitioning to these phases.
A solution to the problem of describing the effect of temperature on phase partitioning has
also been presented. Mintz et al. derived Abraham PP-LFER equations for the prediction of
enthalpies of solvation for water/air and octanol/air partitioning122
, building upon previous
work deriving equations for the enthalpy of phase change for systems of two condensed
phases123,124
. Using the enthalpies compiled by Mintz et al.122
, and provided by Niederer et
al.119
and Arp et al.121
the necessary data is now available to derive modified PP-LFERs for
the enthalpies of phase change required for multimedia fate modelling.
Constructing a PP-LFER-based multimedia fate model is only a worthwhile endeavour if
there are solute descriptors available for environmentally relevant chemicals. Such
descriptors have been published for a large number of chemicals, but due to experimental
limitations many of these chemicals are small and have a single functional group45,125
. This
limitation in the datasets used to parameterize the PP-LFERs however does not appear to
limit their applicability to environmentally relevant phases126
. Some recent additions to the
available solute descriptors are more environmentally relevant; these include a series of
commonly used pesticides and pharmaceuticals127
, and polychlorinated biphenyls (PCBs)89
.
88
We outline here the implementation of PP-LFER equations into CoZMo-POP2, a non-
equilibrium, non-steady state, fugacity based multimedia fate model128
In working towards
the goal of expanding the range of multimedia fate models to more polar chemicals, a non-
steady state model is preferred. Polar chemicals will intuitively be more sensitive to seasonal
variations in the water balance, and possibly other environmental parameters, so a non-
steady state model will be required to properly describe their environmental fate.
Additionally, if significant differences are observed with a a non-steady state simulation on a
relatively short time scale then the potential differences between SP-LFER and a PP-LFER-
based models may be environmentally relevant, because we expect larger differences in
predicted environmental fate to be caused by differences in the description of environmental
partitioning as chemicals near equilibrium.
2 Methods The details concerning the selection of PP-LFER equations from the literature, and the
derivation of additional equations required are provided here and in the Appendix. A far
more difficult task than the implementation of a PP-LFER based model though is devising a
thorough and systematic comparison of the results of the two models. Where data is
available individual chemicals can be run in both the original SP-LFER based and the new
PP-LFER based CoZMo-POP2, and the results can be compared directly. This was the
strategy of Götz et al. and the differences in the model outputs for the specific chemicals
modelled are clearly demonstrated117
. We prefer to evaluate the differences for the entire
range of chemicals that the model can be applied to, and to offer specific guidance on what
results can be expected for chemicals with diverse environmental fates. Previous studies of
this type have made use of chemical space maps which show the variation in model outputs
as a function of partitioning coefficients, usually KAW, KOA and KOW129
. This approach uses
hypothetical chemicals with various combinations of the partitioning coefficients to
systematically map the chemical space. Applying this method to a PP-LFER based model,
however, is difficult, because for each combination of partitioning coefficients there are
multiple valid combinations of solute descriptors. It is, in theory, also possible that chemicals
in close proximity to each other within the chemical space may have widely different model
89
results, but as is discussed below and in the Appendix this is generally found not to be the
case; which allows the chemical space map method of testing and displaying model
sensitivity to be modified to evaluate the PP-LFER model.
2.1 PP-LFER Equations
Equation (9) is the modified PP-LFER equation presented by Goss to describe the
partitioning coefficient K of solute i between two phases j and k118
.
log Kijk = sjkSi + ajkAi + bjkBi + ljkLi + vjkVi + cjk (9)
Lowercase letters denote phase descriptors and uppercase solute descriptors. Each term
represents how different types of interactions between the solute and two phases contribute
to the overall partitioning. Specific interactions of the solute are described by Si
(polarity/dipolarity), Ai (hydrogen bond acidity) and Bi (hydrogen bond basicity) with the
corresponding phase descriptors describing the relative affinity of the two phases for those
kinds of interactions. Non-specific interactions are described by Li (log of the hexadecane air
partition coefficient) and Vi (McGowan volume) and cjk is a system constant. An analogous
equation is used to calculate the enthalpy of phase change.
ΔHijk = sjkSi + ajkAi + bjkBi + ljkLi + vjkVi + cjk (10)
Strictly speaking this is a misapplication of the PP-LFER equation, because the solute
descriptors are intended to describe free energy changes, not enthalpy changes alone.
However previous studies have used this type of equation with good results122-124
: the reason
for this success is very likely that in many cases the free energy and the enthalpy of a
partitioning process are linearly related, as has been noted by previous studies112,124,130
, and
recently validated mathematically by Macleod et al. for the enthalpy of vaporization131
. For
the purpose of modelling chemical fate at environmental temperatures and dilute
concentrations this assumption of linearity appears to be a fair one.
90
Table 7 - Phase Descriptors for PP-LFER Equations Used in this Study.
sjk ajk bjk ljk vjk cjk reference
log KAW -2.07 -3.67 -4.87 -0.48 2.55 0.59 118
log KOA 0.66 3.49 1.42 0.91 -0.14 -0.25 (a)
log KOW -1.41 -0.18 -3.45 0.43 2.41 0.34 118
log KHA 1.01 3.18 1.86 0.75 -0.17 -0.24 (b) 57
log KHW -1.07 -0.49 -3.01 0.27 2.38 0.34 (b) 57
log KQA 1.38 3.21 0.42 0.63 0.98 -7.24 (c) 121
ΔHAW (kJ/mol) -0.86 33.63 43.79 1.52 16.63 8.59 (d)
ΔHOA (kJ/mol) 6.04 -53.66 -9.19 -9.66 1.57 -6.67 (d)
ΔHOW (kJ/mol) 5.18 -20.03 34.60 -8.14 18.19 1.92 (e)
ΔHHA (kJ/mol) -4.59 -31.87 -17.81 -8.15 -2.82 -5.90 (f)
ΔHHW (kJ/mol) -5.45 1.76 25.98 -6.63 13.81 2.69 (g)
ΔHPA (kJ/mol) -
14.03 -20.73 -0.03 -3.37 -20.22 -2.89 (h)
(a) Derived by thermodynamic cycle from log KAW and log KOW of reference 118. (b) Equations from
reference 57 for 25 °C, normalized to OC content. (c) Equation for Berlin Winter aerosol, as
recommended for generic terrestrial aerosol, from reference 121. (d) Derived by for this study from the
data in the supporting information of reference 122. (e) Derived by thermodynamic cycle from ΔHWA and
ΔHOA. (f) Derived for this study from the data in the supporting information of reference 119. (g) Derived
by thermodynamic cycle from ΔHWA and ΔHHA. (h) Arp et al. suggest using the enthalpy of vaporization
as a substitute for ΔHQA, the equation shown is for the enthalpy of vaporization and was derived for this
study from the data in the supporting information of reference 130.
Table 7 summarizes the phase descriptors for Equations (9) and (10) that are used in this
study. Phases are symbolized by the subscripted letters A (air), O (octanol), W (water), H
(Leonardite humic acid), Q (water-insoluble aerosol fraction), and P (whole aerosol). Phase
descriptors provided in Table 7 for the enthalpies of phase change could not be drawn
directly from the literature, but have instead been derived for this study using data from the
relevant papers. Details concerning the derivation of these phase descriptors are provided in
the Appendix (Section 5.1).
2.2 Implementation of PP-LFERs in CoZMo-POP2.
As a fugacity based model partitioning in CoZMo-POP2 is described by Z-values, which are
half of a partition coefficient and describe the capacity of a phase to hold the chemical
91
modelled. The exact equations used to calculate the Z-values in CoZMo-POP2 are provided
in ref.128
, and the alterations to these equations required to implement PP-LFERs are
provided in the Appendix (Section 5.2). Two important SP-LFER equations are replaced
with PP-LFER equations; these are the equation for partitioning between organic carbon and
water, and the equation for partitioning between atmospheric particles and air. All other Z
values are calculated as described in ref.128
, the only difference is that all partitioning
coefficients are calculated with equation (9) and the phase descriptors from Table 7.
Temperature correction of the partitioning coefficients also remains unaltered, but the
enthalpies are now calculated with equation (10) using the phase descriptors from Table 7,
and then converted to internal energies. One SP-LFER remains in the model; partitioning to
the forest canopy, described by a regression with KOA132
. We prefer this equation over the
available PP-LFER for partitioning to tomato cuticle133
; the reasons for this are discussed in
the Appendix (Section 5.3).
2.3 Solute Descriptors.
Experimentally measured solute descriptors have been compiled from a number of literature
sources45,125,127,134-137
. In several of these sources Li values are missing for some45
, or all of
the chemical,127,135,137
and the missing values have been filled in from other sources where
available138-140
. Values for Li are not available for some chemicals; to fill in these data gaps a
regression has been parameterized to predict the Li value from the other solute descriptors,
the details of this regression are provided in the Appendix (Section 5.4). Two additional
sources of solute descriptors are used which calculated the descriptors from literature values
for various partitioning coefficients; these are PCBs89
, and polychlorinated naphthalenes86
. In
total, solute descriptors have been obtained for 1460 individual chemicals.
Values for log KAW and log KOA have been calculated for all 1460 chemicals using equation
(9) and the phase descriptors from Table 7, and are plotted in the log KAW/log KOA chemical
space (Figure 10). The domain of reasonable model applicability is outlined in red. This
domain is defined primarily by computational costs; outside this range the time step required
to ensure stability of the model‘s numerical calculation is very short resulting in long
computation times. In addition, many of the chemicals outside of this range partition so
92
strongly to a single phase that multimedia fate modelling is a misguided exercise. Chemicals
falling outside of the range of reasonable model applicability were removed from the list
leaving 932 individual chemicals.
Figure 10 - Locations of the 1460 chemicals for which solute descriptors were obtained
in the log KOA/log KAW chemical space. Values for log KOA and log KAW are calculated
using equation (9) and the phase descriptors from Table 7. The reasonable domain of
model applicability is shown in red.
Many of the remaining 932 chemicals have very similar solute descriptors, meaning the
model results would also be very similar for these chemicals. To save computational costs
the list has been further reduced by removing chemicals with very similar solute descriptors.
93
The reduction method is pseudo-random and designed to preserve all of the variability in the
combinations of solute descriptors in the full data set. A sample set of 235 chemicals has
been selected and used to perform calculations. The list of these chemicals and the details of
the reduction procedure are provided in the Appendix (Section 5.5, Table 8).
A statistical analysis of the sample set of 235 chemicals shows that there are a number of
strong inter-correlations in the data, most notably between Li and Vi, and Li and Si (Table
10). This is because solute descriptors are not random but vary systematically within
chemical classes. This presents a problem for the evaluation of the model results. Because of
the systemic variation in solute descriptors any patterns observed in the model results may be
due to the nature of the data set, and not directly related to the properties of the PP-LFER
model. To help quantify the extent of this effect a second data set of hypothetical solute
descriptors has been created which contains no inter-correlations (Table 11) or systemic
variation and the results for this data set are also discussed. Hypothetical chemicals have
been created by taking a random number from within the range of each solute descriptor
contained in the data set of real chemicals. This second data set contains 246 hypothetical
chemicals. The solute descriptor combinations are provided in the Appendix (Table 9).
2.4 Model Parameters.
The parameterization used in this study is for an environment representing the Baltic Sea
drainage basin141
. This parameterization includes both the aquatic environment, composed of
two water and four sediment compartments, and the terrestrial environment, composed of
three soil compartments and two forest canopy compartments. A single air compartment is in
contact with all surface compartments. All environmental parameters are the same for both
the SP-LFER and PP-LFER models. The model details are provided in the Appendix
(Section 5.6, Table 12).
Aside from partitioning coefficients, the model requires the input of a chemical‘s
degradation half-lives. Such values cannot be realistically defined for the data set of
94
hypothetical chemicals, and so to ensure comparability between the data sets generic values
are used for all calculations. All chemicals are assumed to be perfectly persistent in all
compartments. This is not only required because persistence cannot be calculated for the
hypothetical chemicals but because it would have a confounding effect on the results;
chemicals with very similar solute descriptors might have widely different fates if their
susceptibility to degradation is different.
Simulated chemical fate can be widely different depending on which model compartment
receives the emissions142
. Three different emission scenarios are tested; continuous
emissions to air, agricultural soil and fresh water. Each simulation is run for 10 years and the
model outputs for the final year of the simulation are used to evaluate the results.
2.5 Model Comparison.
All partitioning coefficients required for the SP-LFER model are calculated using equation
(9) and the phase descriptors from Table 7. The advantage of this method is that there is no
need to consider experimental error in the partitioning coefficients when comparing the
results, because the only source of error will be from the solute descriptors and this will
equally affect the results of both the SP-LFER and PP-LFER models. This means that the
only difference between the model inputs is the replacement of octanol based SP-LFERs
with PP-LFERs, and any observed differences will be attributable to this alteration.
However, the disadvantage of this approach is that the full variability in the model inputs is
not captured for either model because the input data is derived from linear fits of
experimental data. This shortcoming should be considered when interpreting the results of
the model comparison.
The output parameter used to compare the models is the amount in various compartments in
moles. Using the amount of chemical as the primary model output is intuitive because it is
linearly related to environmental concentrations and inter-compartment fluxes. The amount
of chemical is summed together for like compartments to obtain five output amounts for air,
soil, forest canopy, water and sediments. Amounts are averaged for the entire final year of
the simulation. To display the model outputs as chemical space plots we first calculate the
95
log KOA and log KAW values of each chemical using equation (9) and the phase descriptors
from Table 7. As can be observed in Figure 10 the chemicals are unevenly scattered in the
chemical space, so the next step is to interpolate the model outputs being plotted in the
chemical space to a grid of evenly spaced log KOA and log KAW values. Finally, extrapolation
is used to fill in the gaps where interpolation is not possible. A custom interpolation and
extrapolation method has been devised to take into consideration the fact that moving in
different directions in the chemical space has different chemical meanings. The details of this
method are provided in the Appendix (Section 5.7). The validity of this method is primarily
demonstrated by the fact that the resultant plots are sensible and relatively smooth, but
further discussion of the validity can be also be found in the Appendix (Section 5.8).
2.6 Linking Solute Descriptors to Chemical Fate.
A PP-LFER model allows for the elucidation of the relationship between individual solute
descriptors and environmental fate. This is primarily investigated using sub-sampling
statistical analysis. The basic principle is this; a random sub-sample drawn from a data set of
chemicals should have the same mean value and standard deviation for any solute descriptor.
For example, the mean value and standard deviation of Bi in the hypothetical data set are
0.574 and 0.367 respectively. Now we take a sub-sample of this data set, all hypothetical
chemicals with Ai above 0.394, and the mean value and standard deviation of this sub-
sample are 0.550 and 0.367 respectively. The statistics are almost identical, indicating that
this was a random sub-sampling with respect to Bi. If the mean value of the sub-sample
deviates from the mean value of the entire data set then this is a biased sub-sample and if the
standard deviation is significantly reduced compared to the full data set this is a selective
sub-sample. Sub-sampling statistical analysis of chemicals based on PP-LFER model outputs
allows us to detect if environmental phases have biases or selectivity for chemicals with
certain solute descriptor values, linking molecular interactions to environmental fate.
96
3 Results and Discussion
3.1 Comparison of KOC Values.
Organic carbon/water partitioning coefficients (KOC) of all 932 chemicals within the model‘s
domain of reasonable applicability, as calculated in the two models are compared in Figure
11A. In the SP-LFER model, KOW is calculated using equation (9), and then KOC is
calculated with a regression derived by Seth et al.44
. In the PP-LFER model, the KOC is the
KHW taken directly as calculated with equation (9) and the phase descriptors in Table 7.
There is surprisingly little scatter in the plot and a very strong correlation between the SP-
LFER and PP-LFER values. This is due partly to a strong similarity in the phase descriptors
for KOW and KHW (Table 7), but is also due to systematic variations in the solute descriptors
of real chemicals which reduce the variability; the same plot for the hypothetical chemicals
(Figure 16A in the Appendix) shows far more scatter. From the regression shown in Figure
11A we expect that chemicals with low KOC values will partition more strongly to organic
carbon in the PP-LFER model, and chemicals with high KOC values will partition more
weakly to organic carbon. The SP-LFER used in CoZMo-POP2 to calculate KOC is: KOC =
0.35∙KOW44
. If the SP-LFER equation is instead changed to: KOC = 0.85 KOW then there is a
nearly perfect match with the PP-LFER results. This alternate equation is similar to an
equation initially presented by Seth et al. (equation 10 of ref.44
), which they discarded
because some of the data in the regression were suspected to be inaccurate.
97
Figure 11 - (A) Correlation between the log KOC values obtained by the SP-LFER
equation and values obtained from the PP-LFER obtained for humic acid. (B)
Correlation between the logKPA values obtained by the SP-LFER and values obtained
from the PP-LFER obtained for whole aerosol.
3.2 Comparison of KPA Values.
Figure 11B compares values for atmospheric particle/air partitioning coefficients (KPA)
calculated using SP-LFERs and PP-LFERs. The SP-LFER for sorption to atmospheric
particles in CoZMo-POP2 assumes that the organic matter in the aerosol has the same
sorption properties as octanol143
. The PP-LFER values for KPA are calculated using the dual
phase model (consisting of water-insoluble and aqueous fractions) by Arp et al.120
, with an
assumed relative humidity of 80% to calculate the aqueous volume fraction of the aerosol.
While most chemicals in Figure 11B are close to the 1:1 line, there is significantly more
scatter than in the corresponding plot for KOC. On the left side of Figure 11B (log KPA < 6)
the chemicals with the largest deviations have much higher KPA values predicted by the PP-
LFER equation. These chemicals are biased towards low Si, Li and Vi, high Ai and Bi values,
and also have low calculated KAW values (log KAW < -3.5). We conclude that absorption into
the aqueous fraction of the aerosol increases the overall KPA for these chemicals. It should be
noted that this is unlikely to affect the model results because sorption of chemicals with log
KPA < 6 to atmospheric particles will be negligible. On the right side of Figure 11B (log KPA
98
> 6) the chemicals with the largest deviations have lower KPA values predicted by the PP-
LFER equation. These chemicals are biased towards large molecules with hydrogen bond
accepting groups (high Bi, Li and Vi values) and also have low KAW values (log KAW < -3.5).
Despite partitioning strongly to water these chemicals have a significantly lower predicted
KPA using the dual phase PP-LFER equation. An inspection of the phase descriptors (Table
7) reveals the reason; KOA increases more rapidly with increasing Bi and Li than KQA, so we
conclude that partitioning into the aqueous phase of the aerosol fails to offset the decreased
sorption capacity of the water-insoluble fraction of the aerosol for these specific chemicals.
A plot similar to that shown in Figure 11B was created using the PP-LFER for partitioning to
atmospheric aerosol by Götz et al.117
(Figure 17 in the Appendix). The KPA from the dual-
phase absorption mechanism by Arp et al.121
is comparable to the KPA derived from KOA,
lying close to the 1:1 line, as opposed to the KPA from the model of Götz et al. which is
consistently higher by an order of magnitude. This is likely because the latter assumes
perfect additivity of various absorptive and adsorptive fractions, which is probably an
overestimate of the sorptive capacity of atmospheric particles. Arp et al. also have derived a
PP-LFER for other aerosol collected in southern Sweden121
, which is arguably more
appropriate for the current model parameterization. However, the mean absolute deviation in
calculated KPA values from the results of the equation used here is only 0.34 log units and the
fit is very similar, as can be seen in Figure 18, so we prefer to use the equation recommended
by Arp et al.
3.3 Comparison of Model Results.
Figure 12 shows the primary model outputs as six chemical space plots, all with log KOA on
the x-axis and log KAW on the y-axis. As there are no chemicals in the dataset with a log KOW
above approximately 11 (Figure 10), this area is blacked out in Figure 12. Anomalous
contour shapes observed in some of the plots are due to irregularly spaced chemical data.
The corresponding plots for hypothetical chemicals are provided in the Appendix (Figure
19). Figure 12A to C show the phase distribution of real chemicals in the PP-LFER model
expressed as a fraction of the total mass of chemical; the coloured areas are where more than
99
50% of a chemical is found in a single phase. No chemical is ever found to be more than
50% in the forest canopy. Figure 12D to F are plots of the absolute difference between the
SP-LFER and PP-LFER model results, which is calculated by summing the absolute
differences in model results for all five phases and then dividing by two. This represents the
fraction of chemical that shifts from one compartment to another when comparing the model
results.
In all three emission scenarios partitioning to air and water is very similar, the largest
absolute differences in the phase distribution occur at high log KOA values where chemicals
primarily partition to the three solid phases; forest canopy, soils and sediment. Two
mechanisms are responsible for most of these differences; deposition to the forest canopy,
and soil water run-off. In the region defined by log KOA > 4.5 and log KOW > 4.5 (marked as
region 1) some chemicals show enhanced partitioning to soil at the expense of partitioning to
sediments in the PP-LFER model (soil/sediment enhanced by factor of up to 2.2). This is
primarily due to decreased sorption to atmospheric particles (lower KPA in PP-LFER model)
which in turn results in a faster uptake in the canopy (gaseous deposition to the canopy is
faster than particle-bound deposition) and a greater flux of chemical to the soil with falling
foliage. In the region defined by log KAW < -1.5 and 1 < log KOW < 4.5 (marked as region 2)
chemicals show enhanced partitioning to either soil or sediments, which is due to either
enhanced or depressed sorption to particulate organic carbon in run-off water (differences in
KOC) depending on the solute descriptors of the chemical (soil/sediment enhanced or
depressed by up to a factor of 1.7).
100
Figure 12 - Chemical space plots of phase distribution of the PP-LFER based model
and the sum of absolute differences in the percent phase distribution between the SP-
LFER and PP-LFER based models for each of the three emission scenarios.
The primary effect of these mechanisms is that when emissions are to air or water, most of
the differences between the SP-LFER and PP-LFER models occur in the three solid phases
(Figure 12D and F). However, when emissions are to soil most of the largest absolute
differences between the models occur in the transition zones where chemicals are distributed
between the air or water and the solid phases (Figure 12E). One of the main effects of the
mode of emission is that the amount of chemical in the receiving compartment is elevated.
This can be seen in Figure 12A, B and C: the compartment receiving emissions makes up the
largest portion of the phase distribution in each case. Considering this effect, a general rule
can be stated as follows: Whereas the primary environmental fate of chemicals is similarly
101
predicted by both models, the largest relative differences in results are found mostly in the
environmental phases which contain only a small fraction of emitted chemical (less than
1%). This is because even small absolute differences in the phase distribution will have a
significant relative effect (up to a factor of; air: 10.9, canopy: 14.5, water: 5.29, soil: 11.8,
sediment: 1.73). This is demonstrated by plotting the fraction of chemicals in any single
phase and then overlaying this with a plot of the relative differences in model results for that
phase; the largest relative differences in model results are always located outside of the area
with the highest fraction of chemicals (Figure 20, Figure 21, and Figure 22 in the Appendix).
A good example is the environmental distribution of PCB-194 resulting from emissions to
water. An uncertainty analysis was performed for PCB-194 by calculating the 95%
confidence intervals of the SP-LFER parameters and performing additional simulations with
these values, however the same could not be done for the PP-LFER model as standard errors
were not available for the solute descriptors. In the SP-LFER model 98.03% (94.75%-
99.10%) of PCB-194 is in sediments and 0.008% (0.003%-0.025%) is in air, in the PP-LFER
model 94.44% is in sediments and 0.033% is in air. The primary environmental fate of PCB-
194 in both models is sorption to sediments; the PP-LFER model results are depressed by a
negligible factor of 1.04 (1.00-1.05) relative to the SP-LFER model results. However, if we
are specifically interested in air concentrations of PCB-194 the PP-LFER model results are
elevated by a significant factor of 4.24 (1.33-11.7) relative to the SP-LFER model results.
An additional simulation has been performed to test the possible effect of degradation. Real
chemicals are assigned uniform degradation half-lives of two days in air, two months in
water and six months in soils, sediments and forest canopy. Additionally, the Characteristic
Travel Distance (CTD) is calculated for each chemical as described in ref.72
. Chemical space
plots corresponding to Figure 12A and D are provided in the Appendix (Figure 23) along
with a plot of how CTD varies with location in the chemical space for both the SP-LFER and
PP-LFER models (Figure 24). The phase distribution shifts with the incorporation of
degradation because a smaller fraction of emissions makes their way from air to the
condensed phases. However, the absolute differences in model results follow the same
102
general pattern. Interestingly, the absolute difference in percent distribution of the SP-LFER
and PP-LFER models is smaller when degradation is included (up to a factor of 23.3).
Oxidative degradation in air is assumed to only affect chemical in the gas phase; particle
bound chemical is protected. This has the effect of making the relative amount of chemical in
the gas phase smaller and therefore chemicals are less susceptible to gas phase deposition to
forests, which has been identified above as one of the mechanisms causing differences in
model results. Another observation is that the CTD for particle-bound chemicals is enhanced
in the PP-LFER model versus the SP-LFER model (up to a factor of 2.61), due to a smaller
net loss to the surface media relative to the amount of particle-bound chemicals.
A major conclusion resulting from this work is that differences in the output from the two
models are relatively small; larger differences in model outputs have been demonstrated by
varying the environmental input parameters of a multimedia fate model129
. When using a
multimedia fate model as an evaluative or screening tool we recommend that the choice of
using a SP-LFER or PP-LFER model should be based on the quality of the available
chemical input values. The more accurate mechanistic description of partitioning in a PP-
LFER model should allow for better predictive power, but this will only become realized if
the quality of the solute descriptors is as good as, or better than, the quality of the octanol-
based partition coefficients.
3.4 Linking Solute Descriptors to Chemical Fate.
A simple graphical method of comparing how each solute descriptor affects the location of a
chemical in the log KAW/log KOA chemical space is to plot the vectors of the movement.
Because the solute descriptors are of different magnitudes plotting the vector of the phase
descriptors would be misleading, so the vectors must be normalized to the magnitudes of the
phase descriptors. Figure 13 shows the vectors in the log KAW/log KOA chemical space
corresponding to the range defined by the upper quartile and lower quartile of the possible
values of each solute descriptor. This can be interpreted as the range of movement in the
chemical space induced by varying each solute descriptor across the majority of the range of
103
its values. The plot shows that the Li descriptor has the largest effect on location in the
chemical space, which in turn controls chemical fate.
A more detailed investigation of the link between chemical fate and solute descriptors can be
achieved with sub-sampling statistical analysis. Both real and hypothetical chemicals were
divided into four sub-samples based on their simulated environmental fate; chemicals which
are found primarily in air in all three emission scenarios, chemicals found primarily in water,
chemicals found primarily in one or more of the solid phases (soil, sediment, forest canopy),
and multimedia chemicals which have different environmental fates depending on the mode
of emission. Mean values and standard deviations for the solute descriptors of the full data
sets and for all sub-samples are provided in Table 13 of the Appendix. A phase is considered
to be biased towards a solute descriptor if the mean value of the sub-sample deviates from
the value of the full data set by more than 10%, and the phase is considered to be selective if
the standard deviation of the sub-sample is more than 20% lower than the value for the full
data set.
Figure 13 - Vector plots of movement in the log KAW/log KOA chemical space that can be
related to the contributions of each phase descriptor. Vector magnitudes are the value
of the upper quartile minus the value of the lower quartile of the distribution of solute
descriptors multiplied by the corresponding phase descriptor.
S A B L V
0.45
-1.42
0.79
-0.83
0.59
-2.02
3.57
-1.88
1.89
-0.10
log KOA
log K
AW
104
Hypothetical chemicals favouring the air compartment are biased and selective towards low
values of Si and Li but show no bias or selectivity to the other solute descriptors.
Hypothetical chemicals partitioning primarily into water are biased towards high values of Si
and Bi and low values of Li and Vi, with selectivity for low Li values. Hypothetical chemicals
which are primarily sorbed to the three solid phases are biased towards low Bi values and
both biased and selective towards high Li and Vi values. Multimedia hypothetical chemicals
have no biases and are selective only to intermediate Li values. The biases and selectivity of
each phase are intuitive, for example chemicals found in the air tend to be those that only
weakly experience van der Waals dispersive interactions (Li) and dipole interactions (Si).
Chemicals sorbed to solid phases tend to be large (Vi) and experience strong van der Waals
dispersive interactions (Li). Chemicals which experience strong dipole interactions (Si) and
strong specific interactions (Bi) tend to be found in the water. Two features of this analysis
are of further interest; first there is no observable bias or selectivity in any of the sub-samples
for chemicals which are hydrogen bond donors (Ai). Additionally, the multimedia chemicals
are very close to a random sub-sampling of the full data set of hypothetical chemicals.
Statistical analysis of the sub-samples of the dataset of real chemicals reveals far more biases
and selectivity than the hypothetical data set; this is undoubtedly because inter-correlations
in solute descriptors of real chemicals cause additional biases in the phase distribution. The
first notable difference is the Ai solute descriptor; real chemicals in the water phase are
biased towards high Ai values, real chemicals in the air and solid phases are biased and
selective for low Ai values, and real multiphase chemicals are biased towards low Ai values.
To understand this, the data set of real chemicals was divided into two sub-samples;
chemicals that have Ai = 0 (n = 138) and chemicals that have Ai > 0 (n = 97). The mean and
standard deviation for the solute descriptors of these sub-samples are provided in Table 14 in
the Appendix. Chemicals with Ai > 0 are biased and selective for low values of Si and high
values of Bi, and biased towards low values of Li and Vi. This means that hydrogen bond
donors in the dataset are small chemicals and also hydrogen bond acceptors, in short there
are too many alcohols. This observation explains most of the differences between the real
and hypothetical datasets. The water phase is biased towards chemicals with high Bi, and low
Li and Vi, many of the chemicals with these properties also have high Ai values causing the
water phase to be biased towards real chemicals with high Ai values. A similar argument
105
explains the differences between the two data sets for chemicals sorbed to the solid phases.
The net effect of this is that chemicals with high Ai and Bi end up in the water phase, and
chemicals with high Si, Li and Vi end up sorbed to the solid phases. This causes real
chemicals in air to be biased and selective for chemicals with low values of all descriptors.
A general conclusion is that the PP-LFER model has no selectivity for multimedia
chemicals; they are simply those chemicals that are not strongly selected for by any single
phase. Real multimedia chemicals are biased and selective for large Li and Vi, and are biased
towards high Bi and low Ai. A closer inspection of the biases, however, shows that they are
smaller in magnitude than the biases of the air, water or solid phases, so we conclude that
these biases are the result of the data set being skewed by the selection process, as discussed
in detail in the Appendix. Multimedia chemicals tend to have intermediate solute descriptor
values because chemicals with relatively high or low values of any solute descriptor
generally experience a bias towards one or more environmental phases, as demonstrated by
the selectivity of the model for multimedia chemicals with intermediate Li values. The fact
that Li descriptor has such a strong influence on chemical fate may provide an explanation
for the strong similarity between the SP-LFER and PP-LFER model results: KOA-based SP-
LFERs likely capture the majority of the van der Waals interactions represented by the Li
descriptor and so on a large scale do a reasonable job of predicting environmental fate.
4 Acknowledgements We acknowledge funding from the Long-range Research Initiative of the European Chemical
Industry Association (CEFIC). The study benefitted greatly from discussions with Kai-Uwe
Goss, Michael S. McLachlan and Knut Breivik.
106
5 Appendix
5.1 Derivation of Equations for Enthalpies of Phase Change
5.1.1 Derivation of PP-LFER Equations for ΔHAW, ΔHOA and ΔHOW
Solute descriptors and experimental ΔHAW values have been obtained from the Supporting
Information of Mintz et al.122
. As noted by Mintz et al. the L descriptor for one chemical,
erythritol, is unavailable, but a value has been calculated for this study using the regression
described elsewhere in the Appendix. Multiple linear regression is used to parameterize a
PP-LFER equation following the form proposed by Goss118
. Equation (11) is the PP-LFER
for ΔHAW: where S, A, B, L and V are the solute descriptors and the values in brackets are
the standard error associated with the phase descriptor.
ΔHAW = -0.86(±1.28)∙S + 33.63(±1.48)∙A + 43.79(±0.95)∙B
+ 1.52(±0.58)∙L + 16.63(±2.47)∙V + 8.59(±1.01)
(11)
(n = 369, R2 = 0.943, RMSE = 4.76 kJ/mol)
The standard error for s, the phase descriptor for polar/induced dipole interactions, is larger
than the phase descriptor which means that this phase descriptor has a negligible effect on
the value of ΔHAW. Despite this, the s descriptor is included in the equation to allow for the
calculation of ΔHOW by thermodynamic cycle.
Solute descriptors and experimental ΔHOA values have also been obtained from the
Supporting Information of Mintz et al. and a PP-LFER is parameterized by multiple linear
regression. The result is shown in equation (12).
ΔHOA = 6.04(±1.10)∙S – 53.66(±2.37)∙A – 9.19(±1.12)∙B
- 9.66(±0.34)∙L + 1.57(±1.40)∙V – 6.67(±0.74)
(12)
(n = 138, R2 = 0.989, RMSE = 2.58 kJ/mol)
Summing the solute descriptors from equations (11) and (12) gives ΔHOW by thermodynamic
cycle. This PP-LFER was tested by calculating ΔHOW for individual chemicals found in both
data sets by the same thermodynamic cycle. Equation (13) shows the PP-LFER and the
statistics for the comparison.
107
ΔHOW = 5.18∙S – 20.03∙A – 34.60∙B - 8.14∙L + 18.19∙V + 1.92 (13)
(n = 96, R2 = 0.784, RMSE = 4.54 kJ/mol)
All of the ΔHOW values are relatively close to zero so another test of the predictive power of
the regression is the correct prediction of the sign of ΔHOW. The correct sign is predicted for
84 of 96 chemicals which have available ΔHOW values, and in the remaining 12 cases the
values are very close to zero, within the range of -5 to 5 kJ/mol.
5.1.2 Derivation of PP-LFER Equations for ΔHHA and ΔHHW
Solute descriptors and ΔHHA values were obtained from the Supporting Information of
Niederer et al.119
. Equation (14) shows the PP-LFER equation parameterized for ΔHHA.
ΔHHA = -4.59 (±5.98)∙S – 31.87(±4.62)∙A – 17.81(±5.25)∙B
- 8.15(±2.28)∙L – 2.82(±9.42)∙V – 5.90(±5.57)
(14)
(n = 138, R2 = 0.595, RMSE = 10.99 kJ/mol)
PP-LFER equation (15) for ΔHHW has been derived by thermodynamic cycle from the
equations for ΔHHA and ΔHAW. Values of ΔHHW for have also been derived for individual
chemicals found in both data sets and these are used to test the regression.
ΔHHW = -5.45∙S + 1.76∙A + 25.98∙B - 6.63∙L + 13.81∙V + 2.69 (15)
(n = 60, R2 = 0.464, RMSE = 8.29 kJ/mol)
Values for ΔHHW are close to zero, as are the ΔHOW values, so the ability of the regression to
properly predict the sign was tested again. The correct sign is predicted for 50 of the 60
chemicals, and of the remaining chemicals an additional 7 chemicals have ΔHHW values
close to zero, within the range of -5 to 5 kJ/mol.
5.1.3 Derivation of a PP-LFER Equation for ΔHPA
Arp et al. recommend using the enthalpy of vaporization (ΔvapH) to predict the temperature
dependence of sorption to atmospheric particles121
. Values for ΔvapH were obtained from the
supporting information of Goss and Schwarzenbach130
and solute descriptors were obtained
108
from the database of solute descriptors described in the main text of this work. A total of 161
chemicals were found which had both ΔvapH values and solute descriptors.
Goss and Schwarzenbach note a number of outliers when ΔvapH is regressed versus the
natural log of liquid vapour pressure (ln VP) and attribute this to deviations from Trouton‘s
rule because of inter-molecular interactions (such as dimerization) in the gas phase or the
liquid phase130
. Some of these same chemicals are also in the dataset of Arp et al. (e.g. acetic
acid) and show the strongest deviations between ΔvapH and ΔHPA121
. We suggest that this is
because the experimental concentrations of Arp et al. were too low to allow for these kinds
of inter-molecular interactions, causing these chemicals to essentially obey Trouton‘s Rule
and deviate from the experimental ΔvapH values. In light of this we attempt to create a PP-
LFER for ΔvapH which intentionally neglects any deviations from Trouton‘s Rule.
The predicted ΔvapH values for a number of alcohols and nitroanilines had strong negative
deviations from the experimental values (ethanol, propanol, butanol, pentanol, hexanol,
octanol, cyclohexanol, 2-nitroaniline, 3-nitroaniline, 4-nitroaniline), and because they are all
relatively strong hydrogen bond donors and acceptors they very likely do not obey Trouton‘s
rule. In this case it would be because of intermolecular interactions in the liquid phase. The
ΔvapH of two carboxylic acids and two chlorophenols had strong positive deviations from the
experimental values and are also both hydrogen bond donors and acceptors indicating
possible intermolecular interactions in the gas phase (acetic acid, butanoic acid, 3-
chlorophenol, 4-chlorophenol). All of these chemicals were removed from the dataset. Four
other chemicals were identified as outliers, most likely due to erroneous ΔvapH values, and
removed from the correlation (hexachlorobenzene, 2,3-benzofuran, fluoranthene, 1-
propylamine). The most obvious of these was hexachlorobenzene, for which the
experimental value was given as 98.61 kJ/mol. The ln VP of hexachlorobenzene provided in
Goss and Schwarzenbach was -5.93130
, but in Shen and Wania the value presented is -2.3692
.
This value when used in the correlation of Goss and Schwarzenbach gives a ΔvapH value of
73.27 kJ/mol, which is close to the prediction of the PP-LFER equation. The PP-LFER
equation derived from the remaining data points is presented in equation (16):
ΔvapHPPLFER = 14.04 (±1.32)∙S + 20.74 (±1.44)∙A + 0.02 (±1.09)∙B (16)
109
+ 3.36 (±0.52)∙L + 20.23 (±1.94)∙V + 2.89
(n = 143, R2 = 0.995, RMSE = 1.82 kJ/mol)
Comparing the experimental values of ΔvapH to the ΔHPA values of Arp et al. we find the
relation shown in equation (17):
ΔHPA = (-0.995 ± 0.135) ΔvapH (17)
(n = 38, R2 = 0.297, RMSE = 8.46 kJ/mol)
Equation (18) shows the same relationship using the ΔvapH values predicted from equation
(16):
ΔHPA = (-0.988 ± 0.100) ΔvapHPPLFER (18)
(n = 38, R2 = 0.476, RMSE = 6.42 kJ/mol)
Most of the improvement in the correlation statistics is due to a dramatically improved
prediction of the ΔHPA values for chemicals which are known to deviate from Trouton‘s rule.
5.2 Calculation of Z-values in CoZMo-POP 2
As a fugacity based model, environmental phase partitioning in CoZMo-POP 2 is not
described directly by partitioning coefficients but instead by fugacity capacities (Z, mol m-3
Pa-1
). A detailed overview of the usage and nomenclature of fugacity based environmental
models is given by Mackay15
.
Seth et al.44
derived the following KOW based SP-LFER, equation (19), for partitioning
between organic carbon and water (KOC).
KOC = 0.35•KOW (19)
Equation (20) shows how KOC is used to calculate the fugacity capacity of particulate organic
carbon (ZPOC), which is in turn used to calculate partitioning to soil and sediment organic
carbon, as well as to organic carbon suspended in the water column.
110
ZPOC = ZWATER • KOC • δOC / 1•106 kg/L (20)
In these equations KOC is in units of L/kg, δOC is the density of organic carbon in g m-3
and
1•106 kg/L is a unit correction. To replace this SP-LFER equation with a PP-LFER equation
KOC is simply replaced with KHW yielding equation (21).
ZPOC = ZWATER • KHW • δOC / 1•106 kg/L (21)
Sorption to atmospheric particles in CoZMo-POP 2 follows the SP-LFER of Finizio et al.143
,
which essentially assumes that the organic matter in atmospheric aerosol has the same
sorption properties as octanol; equation (22) shows the form of the equation.
ZP = ZAIR • KOA • VFOQ • δOC / δoctanol (22)
In this equation VFOQ is the volume fraction of organic carbon in aerosol and δoctanol is the
density of octanol in g m-3
. A modified version of the two phase PP-LFER model of aerosol
presented by Arp et al. is used to describe sorption to aerosol120
, but assuming that there is no
ionization or ―salting out‖ effect. The volume of the water-soluble phase relative to the
water-insoluble phase is calculated with the growth factor (GF), which is a function of the
relative humidity (RH). The growth factor is calculated using an exponential fit of the data
provided by Arp et al.120
, shown in equation (23).
GF = 1 + 0.0245 • e (4.033 • RH)
(23)
An equation for partitioning to whole aerosol (KPA, m3 g
-1) is obtained by substituting the
growth factor into equation (23) of reference 120, producing equation (24).
KPA = KQA + (GF – 1) / KAW • δwater (24)
In this equation KQA is the partitioning coefficient between air and the water-insoluble
fraction of aerosol in units of m3 g
-1, and δwater is the density of water in g m
-3. It should be
noted that KQA was parameterized at 15 ºC, so KAW is temperature corrected to 15 ºC as well
when applying equation (24). Equation (25) shows the PP-LFER modified calculation of ZP
where δaerosol is the density of whole aerosol in units of g m-3
.
111
ZP = ZAIR • KPA • δaerosol (25)
All other Z values are calculated as described by Wania et al128
, the only difference is that all
partitioning coefficients are calculated with equation (9) and the phase descriptors from
Table 7 of the main publication.
5.3 PP-LFER for Sorption to Forest Canopy
Two equations of type (26) were derived by Horstmann and McLachlan to estimate
partitioning between deciduous (KDEC/AIR) and coniferous (KCON/AIR) forest canopies and
air132
.
KCANOPY/AIR = M • KOAN (26)
This equation can be converted to the log-linear form shown in equation (27);
log KCANOPY/AIR = N ∙ log KOA + m (27)
Where m is equal to log M and N is the same value from equation (27). The values of the
regression coefficients from Horstmann and McLachlan were m = 1.58 and N = 0.69 for
coniferous canopy and m = 1.15 and N = 0.76 for deciduous canopy. A similar study by Su
et al. for a deciduous canopy calculated a regression with coefficients of m = 2.04 and N =
0.67144
. A study by Welke et al., from which Platts et al. derived the data for their PP-LFER,
also calculates a log linear regression with coefficients m = 0.820 and N = 0.668145
. Figure
14 shows all of these regressions, where it can clearly be seen that the regression of Welke et
al. is dissimilar from the forest canopy regressions. Tomato cuticle clearly has different
partitioning properties from forest canopy, and for this reason we prefer to use the regression
of Horstmann and McLachlan.
Two additional regressions were derived using PP-LFER equations to test the effect the
training set of chemicals has on the resulting regression with KOA. The dataset Horstmann
and McLachlan used to derive their equations included a number of PAHs, HCB, PCBs and
PCDD/PCDFs. Solute descriptors were found in the database outlined in the main text of this
work for the PAHs, HCB, and PCBs, but none could be located for PCDD/PCDFs, so PCB-
112
194 and PCB-209 were added to the dataset to extend it to less volatile chemicals. The PP-
LFER of Platts et al. for partitioning between tomato cuticle matrix and air (KMXa/AIR) was
used to calculate values for KMXa/AIR for this dataset133
, and the KOA values were calculated
from the PP-LFER of Goss118
. The regression coefficients are m = -0.41 and N = 1.19.
The second regression was performed on the dataset of Welke et al.; solute descriptors were
found for 54 of the chemicals used by Welke et al. in the database described in the main text.
These solute descriptors were then used to calculate KOA values as described above, and a
regression of the form shown in equation (27) was calculated using the experimental KMXa
values. The regression coefficients are m = -0.38 and N = 1.02. The regression of log KMXa
with log KOA is similar for both the non-polar organic dataset of Horstmann and Mclachlan
and the more diverse dataset of Welke et al., so the regression for the forest canopies should
similarly be applicable to a more diverse dataset
113
Figure 14 - Linear Regressions of Plant Sorption with log KOA
Plots of the log-linear regressions of log KCANOPY/AIR with log KOA of Horstmann and
McLachlan132
and Su et al.144
, and the regression of log KMXa/AIR with log KOA of Welke
et al.145
.
5.4 PP-LFER Equation for Hexadecane/Air Partitioning
A large number of chemicals are missing values for the L descriptor. This descriptor is
unique in that it is the only Abraham descriptor that is itself a partitioning coefficient,
implying that it is possible to derive a PP-LFER to describe the interactions involved in the
partitioning process. To our knowledge this has not been done directly, but an equation can
be drawn indirectly from the literature. Abraham et al. provide an equation for partitioning
between hexadecane and water (K16/W) using the phase descriptors sjk, ajk, bjk, ejk and vjk (ref
2
4
6
8
10
12
14
2 4 6 8 10 12 14
log KOA
H&M Dec.
H&M Con.
Su et al.
Welke et al.
1:1
log K
Pla
nt/
AIR
114
45, equation 8). A regression for partitioning between water and air (KWA) using the same
phase descriptors has also been presented125
. Assuming that these processes form a complete
thermodynamic cycle, equation (28) for L (K16/A) can be derived by adding the phase
descriptors of the two equations:
log K16/A = 0.932∙S + 0.226∙A – 0.028∙B + 1.244∙E + 3.564∙V – 0.907 (28)
The two phase descriptors describing hydrogen bonding nearly cancel out; the magnitudes of
ajk and bjk are approximately -3.6 and -4.9 respectively in the PP-LFER for K16/W and 3.8 and
4.8 respectively in the PP-LFER for KWA. This is expected, as hydrogen bonding should play
a negligible role in the partitioning between hexadecane and air. To completely eliminate
hydrogen bonding a new PP-LFER was parameterized using most of the chemicals with
available experimental L values from the database outlined in the main text of this work.
Many of the solute descriptors for PCBs and PCNs were left out of the regression because
they would have skewed the dataset; the chemicals retained were those with the most
experimental data used in the original derivation of the solute descriptors. The regression
was performed only versus the S, E and V descriptors; this equation is shown in equation
(29). E is the excess molar refraction, a measure of polarizability.
log K16/A (L) = 0.951(±0.056)∙S + 1.205(±0.041)∙E
+ 3.492(±0.045)∙V – 0.816(±0.044)
(29)
(n = 621, R2 = 0.979, RMSE = 0.329)
The phase descriptors in (29) almost exactly match those of the literature derived equation
(28). The fact that it is possible to create a regression between the solute descriptors that has
such excellent statistics reveals strong inter-correlations among the descriptors, which is not
desirable for multiple linear regressions. Using this regression to fill in the missing L values
will only increase the degree of inter-correlation in the dataset, but it does include chemical
information not found in the other equations used in this work, namely the E descriptor.
5.5 Dataset Reduction Method
Two criteria were used to determine if two chemicals were similar enough to eliminate one
of the chemicals from the dataset; the first is the distance between the chemicals in the log
115
KOA / log KAW chemical space, and the second is the differences in their solute descriptors.
The full list of 932 chemicals considered was randomly ordered using Microsoft Excel by
assigning each chemical a random number and then sorting the list by the random number.
The mean and standard deviation of each solute descriptor was calculated for the full set of
932 chemicals and then the standard deviation was used to normalize all of the solute
descriptors to allow the deviations of different descriptors to be directly compared.
Each chemical is compared sequentially to every other chemical in the list that has not
already been identified as redundant and the following determined:
• If the distance between the chemicals in the log KOA / log KAW
chemical space is below a given threshold the criteria for distance is
considered met.
• The sum of the deviations between the normalized solute descriptors
of the chemicals is calculated.
• The maximum deviation between the normalized solute descriptors of
the chemicals is calculated.
• If both the sum and the maximum deviations are below the thresholds
the criteria for similar solute descriptors is considered met.
If both the distance and the similar solute descriptor criteria are met for any other chemical in
the list the current chemical is identified as redundant and the next chemical in the list is
considered. The thresholds used are distance = 0.5 units, sum of differences = 1.2 standard
deviations, maximum difference = 0.6 standard deviations. These thresholds were set to
reduce the list to a manageable size while not eliminating too much of the variability in the
dataset. In the area defined by log KOA > 6 and log KAW > -3.5 (i.e. the area containing the
PCBs and PCNs) the chemicals are much more numerous and closely spaced, so in this area
the thresholds were divided by two in order to retain more of the chemicals.
Table 8 - Real Chemicals Dataset
CAS NAME S A B L V Ref. Ref. L
54115 Nicotine 0.88 0.00 1.09 5.85 1.37 137
a
55389 Fenthion 1.75 0.00 0.65 9.43 1.99 127
a
116
57103 Hexadecanoic acid 0.60 0.60 0.45 8.31 2.44 45
a
57114 Octadecanoic acid 0.60 0.60 0.45 9.27 2.72 45
a
59507 4-Chloro-3-methylphenol 1.02 0.65 0.23 5.29 1.04 125
125
62533 Aniline 0.96 0.26 0.41 3.93 0.82 45
138
64186 Formic acid 0.60 0.75 0.38 1.25 0.32 45
a
64197 Acetic acid 0.65 0.61 0.45 1.75 0.46 125
125
67561 Methanol 0.44 0.43 0.47 0.97 0.31 125
125
68122 N,N-Dimethylformamide 1.31 0.00 0.73 3.17 0.65 125
125
69727 Salicylic acid 0.84 0.71 0.38 4.51 0.99 137
a
72208 Endrin 1.35 0.00 0.61 9.66 2.01 127
a
74953 Dibromomethane 0.67 0.10 0.10 2.89 0.60 125
125
75116 Diiodomethane 0.69 0.05 0.23 3.85 0.77 45
139
75898 2,2,2-TrifluoroethanoI 0.60 0.57 0.25 1.22 0.50 125
125
76017 Pentachloroethane 0.66 0.17 0.06 4.27 1.00 125
125
77101 phencyclidine 0.97 0.00 0.82 9.09 2.15 137
a
78400 Triethyl phosphate 1.00 0.00 1.06 4.75 1.39 125
125
78819 Isobutylamine 0.32 0.16 0.63 2.42 0.77 45
a
78875 1,2-Dichloropropane 0.60 0.10 0.11 2.86 0.78 125
125
79118 Chloroacetic acid 1.08 0.74 0.36 2.71 0.59 45
a
79243 Nitroethane 0.95 0.02 0.33 2.41 0.56 125
125
79345 1,1,2,2-Tetrachloroethane 0.76 0.16 0.12 3.80 0.88 125
125
80466 4-tert-Pentylphenol 0.89 0.56 0.41 6.17 1.48 45
a
83329 Acenaphthene 1.04 0.00 0.20 6.47 1.26 125
125
84662 Diethyl phthalate 1.40 0.00 0.88 7.37 1.71 45
a
86737 Fluorene 1.03 0.00 0.20 6.92 1.36 125
125
88095 2-Ethylbutanoic acid 0.57 0.60 0.50 3.53 1.03 45
a
88744 2-Nitroaniline 1.37 0.30 0.36 5.63 0.99 125
125
88857 Dinoseb 1.75 0.17 0.46 7.82 1.69 127
a
89714 Methyl 2-methylbenzoate 0.87 0.00 0.43 5.05 1.21 45
138
91203 Naphthalene 0.92 0.00 0.20 5.16 1.09 125
125
91236 2-Nitroanisole 1.47 0.22 0.30 5.44 1.09 127
a
92240 Naphthacene 1.70 0.00 0.33 10.75 1.82 45
138
92671 4-Aminobiphenyl 1.48 0.26 0.48 7.45 1.42 45
a
92944 p-Terphenyl 1.48 0.00 0.30 9.69 1.93 45
138
93550 Ethyl phenyl ketone 0.95 0.00 0.51 5.09 1.16 45
a
93992 Phenylbenzoate 1.42 0.00 0.47 7.51 1.54 45
a
94257 n-Butyl 4-aminobenzoate 1.47 0.32 0.59 7.38 1.60 45
a
94257 butyl 4-aminobenzoate 1.20 0.29 0.75 7.11 1.60 137
a
95578 2-Chlorophenol 0.88 0.32 0.31 4.18 0.90 125
125
95932 1,2,4,5-Tetramethylbenzene 0.60 0.00 0.19 5.11 1.28 45
a
96480 gamma-Butyrolactone 1.74 0.00 0.45 3.51 0.64 45
a
98566 4-Chlorobenzotrifluoride 0.58 0.00 0.01 3.98 1.03 137
a
99081 3-Nitrotoluene 1.10 0.00 0.28 5.10 1.03 125
125
99661 Valproic acid 0.57 0.60 0.50 4.47 1.31 137
a
100174 4-Nitroanisole 1.32 0.20 0.24 5.29 1.09 127
a
100469 Benzylamine 0.88 0.10 0.72 4.32 0.96 45
138
117
100481 4-Cyanopyridine 1.21 0.00 0.59 4.03 0.83 125
125
100845 3-Methylanisole 0.78 0.00 0.30 4.39 1.06 45
138
101815 Diphenylmethane 1.04 0.00 0.28 6.31 1.47 45
138
101848 Diphenyl ether 1.08 0.00 0.20 6.29 1.38 45
138
101973 Ethyl phenylacetate 1.01 0.00 0.57 5.67 1.36 45
a
103297 1,2-Diphenylethane 1.03 0.00 0.28 6.76 1.61 45
138
103833 N,N-Dimethylbenzylamine 0.80 0.00 0.69 5.08 1.24 45
a
104723 n-Decylbenzene 0.47 0.00 0.15 7.75 2.13 45
a
104949 4-Methoxyaniline 1.10 0.23 0.65 4.95 1.02 45
138
106445 p-Cresol 0.87 0.57 0.31 4.20 0.92 125
125
106478 4-Chloroaniline 1.13 0.30 0.32 4.89 0.94 125
125
106934 1,2-Dibromoethane 0.76 0.10 0.17 3.39 0.74 125
a
107186 Prop-2-en-1-ol 0.46 0.38 0.48 1.94 0.55 45
a
107197 propargyl alcohol 0.57 0.38 0.59 1.97 0.50 137
a
108861 Bromobenzene 0.73 0.00 0.09 4.04 0.89 125
125
108894 4-Methylpyridine 0.82 0.00 0.55 3.64 0.82 125
125
108918 Cyclohexylamine 0.56 0.16 0.58 3.80 0.95 125
125
108930 Cyclohexanol 0.54 0.32 0.57 3.76 0.90 125
125
109024 N-Methylmorpholine 0.74 0.00 0.90 3.30 0.86 125
a
109977 Pyrrole 0.73 0.41 0.29 2.63 0.58 45
a
110565 1,4-Dichlorobutane 0.95 0.00 0.17 3.79 0.92 125
a
110587 n-Pentylamine 0.35 0.16 0.61 3.14 0.91 125
125
110623 Pentanal 0.65 0.00 0.45 2.85 0.83 125
125
110805 2-Ethoxyethanol 0.50 0.30 0.83 2.82 0.79 45
138
110894 Piperidine 0.46 0.10 0.69 3.30 0.80 45
138
110918 Morpholine 0.79 0.06 0.91 2.98 0.72 125
a
111717 Heptanal 0.65 0.00 0.45 3.87 1.11 125
125
111762 2-butoxyethanol 0.50 0.30 0.83 3.81 1.07 125
125
111842 n-Nonane 0.00 0.00 0.00 4.18 1.38 125
125
111875 Octan-1-ol 0.42 0.37 0.48 4.62 1.30 45
139
111922 Di-n-butylamine 0.30 0.08 0.68 4.35 1.34 125
125
112129 2-Undecanone 0.68 0.00 0.51 5.73 1.67 125
125
112298 1-Bromodecane 0.40 0.00 0.12 5.87 1.69 45
a
112301 Decan-1-ol 0.42 0.37 0.48 5.63 1.58 45
139
112403 n-Dodecane 0.00 0.00 0.00 5.70 1.80 45
139
112425 Undecan-1-ol 0.42 0.37 0.48 5.80 1.72 45
a
112538 Dodecan-1-ol 0.42 0.37 0.48 6.28 1.86 45
a
112709 Tridecan-1-ol 0.42 0.37 0.48 6.77 2.00 45
a
112721 Tetradecan-1-ol 0.42 0.37 0.48 7.25 2.14 45
a
118741 Hexachlorobenzene 0.99 0.00 0.00 7.62 1.45 45
138
119335 4-methyl-2-nitrophenol 1.04 0.06 0.40 5.22 1.09 135
a
120127 Anthracene 1.34 0.00 0.26 7.57 1.45 125
125
120729 Indole 1.12 0.44 0.22 5.51 0.95 45
138
121142 2,4-Dinitrotoluene 1.27 0.07 0.51 6.01 1.21 127
a
121448 Triethylamine 0.15 0.00 0.79 3.04 1.05 125
125
123320 2,5-Dimethylpyrazine 0.90 0.00 0.69 3.99 0.92 45
a
118
123864 n-Butyl acetate 0.60 0.00 0.45 3.35 1.03 125
125
126738 Tri-n-butyl phosphate 0.90 0.00 1.21 7.74 2.24 45
a
129000 Pyrene 1.71 0.00 0.29 8.83 1.58 125
125
131113 Dimethyl phthalate 1.41 0.00 0.88 6.05 1.43 45
138
132650 Dibenzothiophene 1.31 0.00 0.18 7.58 1.38 45
138
134327 1-Naphthylamine 1.26 0.20 0.57 6.49 1.19 125
125
135193 2-Naphthol 1.08 0.61 0.40 6.20 1.14 125
125
135988 1-methylpropyl-benzene 0.48 0.00 0.16 4.51 1.28 125
125
136607 n-Butyl benzoate 0.80 0.00 0.46 5.97 1.50 45
a
139662 Diphenyl sulfide 1.20 0.00 0.30 7.48 1.49 45
a
140294 Phenylacetonitrile 1.15 0.00 0.45 4.70 1.01 45
138
142289 1,3-DichIoropropane 0.74 0.00 0.17 3.09 0.78 45
a
142847 Di-n-propylamine 0.30 0.08 0.69 3.35 1.05 45
138
142961 Di-n-butyl ether 0.25 0.00 0.45 3.92 1.29 125
125
143088 Nonan-1-ol 0.42 0.37 0.48 5.12 1.44 45
139
150196 3-Methoxyphenol 1.17 0.59 0.38 4.80 0.97 125
125
198550 Perylene 1.76 0.00 0.40 12.05 1.95 45
138
206440 Fluoranthene 1.55 0.00 0.20 8.83 1.59 45
138
243196 Benzo[b]fluorene 1.57 0.00 0.20 9.52 1.73 45
138
253667 Cinnoline 1.00 0.00 0.78 5.18 1.00 45
a
253827 Quinazoline 1.15 0.00 0.65 4.90 1.00 45
a
260946 Acridine 1.33 0.00 0.58 7.64 1.41 45
138
271896 Benzofuran 0.83 0.00 0.15 4.36 0.91 45
138
275514 Azulene 1.17 0.15 0.16 5.71 1.09 45
138
288131 Pyrazole 1.00 0.54 0.45 3.15 0.54 45
138
288471 Thiazole 0.80 0.00 0.45 3.00 0.60 45
a
289805 Pyridazine 0.85 0.00 0.81 3.43 0.63 45
138
290379 Pyrazine 0.95 0.00 0.62 2.92 0.63 45
138
292648 Cyclooctane 0.10 0.00 0.00 4.33 1.13 45
138
305033 tert-butyl chlorambucil 1.49 0.03 0.89 11.90 2.83 137
a
319846 alpha-HCH 1.20 0.00 0.47 7.34 1.58 134
134
319857 beta-HCH 1.18 0.12 0.58 7.63 1.58 134
134
333415 Diazinon 1.01 0.00 1.19 9.45 2.31 127
a
334485 Decanoic acid 0.60 0.60 0.45 5.46 1.59 45
a
367124 2-Fluorophenol 0.69 0.61 0.26 3.45 0.79 125
125
371415 4-Fluorophenol 0.97 0.63 0.23 3.84 0.79 125
125
402675 3-Fluoronitrobenzene 1.09 0.00 0.35 4.10 0.89 45
a
502410 Cycloheptanol 0.54 0.32 0.58 4.41 1.05 125
125
506309 Eicosanoic acid 0.60 0.60 0.45 10.24 3.00 45
a
512561 Trimethyl phosphate 1.10 0.00 1.00 3.76 0.97 45
a
513086 Tri-n-propyl phosphate 1.00 0.00 1.15 6.42 1.82 45
a
527548 3,4,5-Trimethylphenol 0.88 0.55 0.44 5.57 1.20 45
138
533584 2-Iodophenol 1.00 0.40 0.35 4.96 1.03 125
125
536743 Phenylethyne 0.58 0.12 0.24 3.74 0.91 45
a
537462 methamphetamine 0.79 0.09 0.70 5.83 1.38 137
a
542552 Isobutyl formate 0.60 0.00 0.40 2.79 0.89 125
125
119
544252 Cyclohepta-1,3,5-triene 0.46 0.00 0.20 3.53 0.86 45
a
544638 Tetradecanoic acid 0.60 0.60 0.45 7.35 2.16 45
a
555033 3-Nitroanisole 1.20 0.14 0.25 5.18 1.09 127
a
575371 1,7-Dimethylnaphthalene 0.89 0.00 0.20 6.45 1.37 45
a
583539 1,2-Dibromobenzene 0.96 0.00 0.04 5.46 1.07 45
138
584021 Pentan-3-ol 0.36 0.33 0.56 2.83 0.87 45
a
589184 4-Methylbenzyl alcohol 0.88 0.33 0.60 4.69 1.06 45
a
591355 3,5-Dichlorophenol 1.17 0.77 0.00 5.65 1.02 45
138
601898 2-Nitroresorcinol 1.35 0.01 0.48 5.44 1.01 135
a
608935 Pentachlorobenzene 0.96 0.00 0.00 6.72 1.33 45
138
616477 N-Methylimidazole 0.95 0.00 0.83 3.81 0.68 45
138
623370 3-Hexanol 0.36 0.33 0.56 3.34 1.01 125
125
623427 Methyl butanoate 0.60 0.00 0.45 2.89 0.89 125
125
625809 Diisopropyl sulfide 0.32 0.00 0.37 3.60 1.12 125
125
626028 3-Iodophenol 1.20 0.70 0.18 5.53 1.03 45
138
628057 1-Nitropentane 0.95 0.00 0.29 3.94 0.99 125
125
628717 Hept-1-yne 0.23 0.13 0.10 3.00 1.01 125
125
628819 Ethyl n-butyl ether 0.25 0.00 0.45 2.97 1.01 45
a
629594 n-Tetradecane 0.00 0.00 0.00 6.71 2.08 45
139
630206 1,1,1,2-Tetrachloroethane 0.63 0.10 0.08 3.64 0.88 125
125
634662 1,2,3,4-TetrachIorobenzene 0.92 0.00 0.00 6.17 1.21 125
125
644086 4-Methylbiphenyl 0.98 0.00 0.23 6.61 1.47 45
138
700129 Pentamethylbenzene 0.66 0.00 0.20 5.80 1.42 45
a
758963 N,N-Dimethylpropanamide 1.30 0.00 0.78 4.07 0.93 45
a
768956 Adamantan-1-ol 1.20 0.32 0.56 5.71 1.25 45
a
779022 9-Methylanthracene 1.30 0.00 0.26 8.44 1.60 45
138
781431 9,10-Dimethylanthracene 1.30 0.00 0.26 9.28 1.74 45
138
829265 2,3,6-Trimethylnaphthalene 0.86 0.00 0.21 6.99 1.51 45
a
830137 Cyclododecanone 0.86 0.00 0.56 6.67 1.71 45
a
994058 tert-Amyl methyl ether 0.21 0.00 0.60 2.98 1.01 137
a
1122549 4-Acetylpyridine 1.13 0.00 0.84 4.66 0.97 125
125
1582098 Trifluraline 0.40 0.00 1.35 8.54 2.20 127
a
1730376 1-Methylfluorene 1.06 0.00 0.20 7.44 1.50 45
138
1731846 Methyl nonanoate 0.60 0.00 0.45 5.32 1.59 45
24
1825305 PCN-6 1.06 0.00 0.09 6.78 1.33 86
86
1836755 Nitrofen 0.82 0.00 0.85 8.40 1.80 127
a
2050682 PCB-15 1.18 0.00 0.16 7.58 1.57 89
89
2050740 PCN-9 1.12 0.00 0.11 7.15 1.33 86
86
2051243 PCB-209 2.26 0.00 0.02 11.70 2.55 89
89
2051607 PCB-1 1.07 0.00 0.20 6.34 1.45 89
89
2051629 PCB-3 1.05 0.00 0.18 6.72 1.45 89
89
2189608 n-Octyl benzene 0.48 0.00 0.15 6.78 1.84 45
a
2234131 PCN-75 1.54 0.00 0.00 12.88 2.06 86
86
2303175 Triallate 0.92 0.20 0.75 8.77 2.12 127
a
2416946 2,3,6-Trimethylphenol 0.78 0.36 0.41 5.14 1.20 135
a
2819865 Pentamethylphenol 0.90 0.35 0.43 6.29 1.48 135
a
120
2921882 Chlorpyrifos 0.34 0.00 1.13 8.75 2.15 127
a
2986198 S-methylisothiourea 0.62 0.26 0.84 3.29 0.71 137
a
3376247 Phenyl-N-tert-butylnitrone 1.03 0.01 0.49 6.41 1.54 137
a
3452093 Non-1-yne 0.23 0.12 0.10 4.09 1.29 45
a
4658280 Aziprotryn 1.12 0.11 0.87 7.90 1.63 136
136
4920778 3-methyl-2-nitrophenol 1.01 0.14 0.48 5.19 1.09 135
a
7732185 Water 0.45 0.82 0.35 0.26 0.17 125
125
13429248 hexafluoropropene 0.55 0.77 0.10 1.39 0.70 125
125
15687271 Ibuprofen 0.59 0.59 0.81 6.83 1.78 137
a
26530201 Octhilinone 1.69 0.13 0.62 8.10 1.79 127
a
32598111 PCB-70 1.46 0.00 0.13 8.69 1.81 89
89
32774166 PCB-169 1.70 0.00 0.07 10.61 2.06 89
89
33091177 PCB-197 2.00 0.00 0.06 10.10 2.30 89
89
33146451 PCB-10 1.22 0.00 0.20 6.77 1.57 89
89
34622587 Orbencarb 1.17 0.00 0.86 8.75 1.96 127
a
35256850 Tebutam 1.11 0.02 1.12 8.54 2.10 127
a
35694087 PCB-194 2.00 0.00 0.06 11.19 2.30 89
89
38379996 PCB-95 1.61 0.00 0.13 8.63 1.94 89
89
38444778 PCB-32 1.35 0.00 0.17 7.67 1.69 89
89
38444905 PCB-37 1.31 0.00 0.13 8.39 1.69 89
89
39635319 PCB-189 1.85 0.00 0.07 10.89 2.18 89
89
39635353 PCB-159 1.72 0.00 0.09 9.96 2.06 89
89
40186729 PCB-206 2.13 0.00 0.04 11.45 2.43 89
89
41411625 PCB-160 1.74 0.00 0.11 9.59 2.06 89
89
41411647 PCB-190 1.87 0.00 0.09 10.45 2.18 89
89
44604902 S-2-fluoroethylisothiourea 0.60 0.27 0.81 3.71 0.87 137
a
51570446 PCN-16 1.18 0.00 0.00 7.70 1.45 86
86
51570457 PCN-33 1.24 0.00 0.00 8.27 1.58 86
86
52663782 PCB-195 2.00 0.00 0.06 10.83 2.30 89
89
52663793 PCB-207 2.13 0.00 0.04 10.90 2.43 89
89
52744135 PCB-135 1.74 0.00 0.11 9.29 2.06 89
89
53555638 PCN-28 1.24 0.00 0.00 8.54 1.58 86
86
54910893 Fluoxetine 1.30 0.10 0.93 9.45 2.24 137
a
55720382 PCN-22 1.12 0.00 0.00 8.04 1.45 86
86
58863153 PCN-74 1.48 0.00 0.00 11.41 1.94 86
86
58877886 PCN-63 1.42 0.00 0.00 10.54 1.82 86
86
67922241 PCN-40 1.24 0.00 0.00 8.86 1.58 86
86
67922263 PCN-50 1.36 0.00 0.00 9.20 1.70 86
86
68194081 PCB-150 1.74 0.00 0.11 8.92 2.06 89
89
68194161 PCB-173 1.87 0.00 0.09 10.09 2.18 89
89
69782907 PCB-157 1.72 0.00 0.09 10.27 2.06 89
89
70362504 PCB-81 1.44 0.00 0.11 9.08 1.81 89
89
74472370 PCB-114 1.59 0.00 0.11 9.39 1.94 89
89
74472494 PCB-186 1.87 0.00 0.09 9.73 2.18 89
89
103426977 PCN-67 1.42 0.00 0.00 10.04 1.82 86
86
149864813 PCN-31 1.24 0.00 0.00 9.10 1.58 86
86
121
150224207 PCN-57 1.36 0.00 0.00 9.41 1.70 86
86
150224252 PCN-59 1.36 0.00 0.00 9.59 1.70 86
86
239103572 1-(8-fluorooctyl)-2-nitroimidazole 1.62 0.00 0.76 8.11 1.86 137
a
a. The L solute descriptor was calculated with equation (29).
122
Table 9 - Hypothetical Chemicals Dataset
ID S A B L V
1 0.62 0.48 1.04 5.76 1.95 2 1.55 0.03 0.47 2.98 0.82
3 0.50 0.19 0.22 2.45 0.51
4 0.17 0.36 0.27 8.43 0.32 5 0.34 0.65 1.33 2.69 2.95
6 1.49 0.71 0.21 2.28 0.69
7 1.73 0.81 0.86 0.47 1.97 8 0.71 0.74 0.52 0.28 2.27
9 1.93 0.48 0.25 2.39 1.74
10 0.74 0.24 1.23 2.60 2.37 11 0.35 0.23 0.61 4.19 2.18
12 2.20 0.01 0.80 9.56 2.85
13 1.04 0.32 0.92 5.19 1.83 14 0.43 0.30 0.25 10.13 2.76
15 1.57 0.46 0.46 3.65 1.69
16 0.39 0.43 0.51 2.54 2.44 17 0.78 0.45 0.44 8.41 1.96
18 0.97 0.41 1.26 3.58 2.77
19 0.20 0.26 1.31 1.35 1.06 20 0.26 0.52 0.66 1.93 1.89
21 0.32 0.27 1.21 3.97 2.30
22 1.72 0.47 0.77 0.63 2.46 23 0.32 0.03 1.18 6.97 1.39
24 0.23 0.18 0.17 8.39 2.50
25 0.80 0.27 0.39 10.12 2.74 26 0.37 0.36 0.41 11.54 1.36
27 0.07 0.48 0.08 9.92 0.39
28 0.63 0.18 0.98 5.25 2.58 29 0.46 0.05 0.40 8.16 2.46
30 1.89 0.71 0.75 7.02 2.76
31 0.18 0.54 0.89 2.33 1.81 32 1.01 0.48 0.42 2.48 0.78
33 0.01 0.01 0.50 8.54 0.95
34 1.38 0.76 0.73 0.55 2.54 35 0.80 0.13 0.45 3.79 2.12
36 0.88 0.49 0.21 5.57 1.24
37 0.75 0.67 0.44 10.37 1.93 38 1.76 0.08 0.03 6.15 0.44
39 1.62 0.02 0.16 9.09 2.44
40 1.28 0.75 0.68 2.67 1.87 41 0.26 0.41 0.25 9.21 1.76
42 0.80 0.31 0.21 9.94 0.90
43 1.00 0.60 0.27 8.17 2.63 44 0.74 0.27 0.91 2.19 2.50
45 1.63 0.02 0.44 10.52 2.06
46 0.38 0.65 0.82 5.40 1.35 47 0.06 0.44 0.07 9.43 2.31
48 0.01 0.66 0.91 10.47 2.61 49 1.50 0.44 0.30 9.93 2.91
50 0.89 0.30 0.52 3.40 2.32
51 1.81 0.44 0.79 2.54 2.59 52 1.00 0.39 0.20 7.74 1.92
53 0.33 0.54 0.77 6.48 1.70
54 0.82 0.23 0.24 5.74 1.72
55 1.75 0.77 0.22 3.23 2.01
56 0.77 0.33 0.44 4.17 2.76
57 1.00 0.51 1.00 1.69 2.67 58 0.85 0.69 0.08 5.51 2.59
59 2.21 0.78 0.17 3.35 2.23
60 0.62 0.81 1.14 1.61 3.00 61 0.63 0.61 0.60 4.11 2.22
62 0.00 0.31 1.24 3.89 1.50
63 1.82 0.12 0.88 1.12 2.33 64 0.27 0.11 0.94 4.85 2.01
65 1.66 0.64 0.66 6.74 2.80
66 0.54 0.52 1.27 9.65 2.96
ID S A B L V
67 1.28 0.42 1.10 8.84 2.91
68 1.95 0.22 0.98 3.56 2.69
69 2.17 0.50 0.40 6.67 2.39 70 0.01 0.33 0.30 9.65 1.83
71 1.36 0.37 1.22 5.64 2.67
72 1.85 0.05 0.13 4.51 0.59 73 0.41 0.45 0.11 6.45 0.39
74 0.41 0.69 0.52 4.65 1.35
75 1.19 0.72 0.03 9.98 1.76 76 1.14 0.18 0.45 3.85 2.89
77 0.62 0.82 0.60 7.42 2.78
78 0.52 0.04 0.41 4.83 1.75 79 1.47 0.10 0.49 11.88 2.03
80 0.62 0.43 1.30 1.13 1.92
81 1.60 0.06 0.40 5.05 2.15 82 0.46 0.32 0.57 4.21 1.25
83 0.68 0.42 1.09 1.80 1.06
84 0.92 0.17 0.14 10.35 1.80 85 1.47 0.48 0.16 11.66 2.22
86 0.15 0.47 0.45 3.23 2.67
87 1.19 0.39 1.14 1.82 2.13 88 0.12 0.79 0.67 0.56 1.11
89 1.77 0.64 0.56 1.02 1.20
90 0.61 0.19 0.33 9.81 1.70 91 1.30 0.18 0.63 5.28 1.59
92 2.06 0.68 0.43 2.17 1.79
93 0.11 0.65 0.01 8.04 1.43 94 0.75 0.78 0.21 4.38 1.69
95 0.54 0.12 1.26 1.44 2.57
96 2.02 0.37 0.34 1.84 1.21 97 1.30 0.11 1.32 6.88 2.46
98 0.88 0.28 0.80 3.89 1.64
99 0.09 0.30 0.09 10.20 0.54 100 0.52 0.57 0.53 7.66 2.46
101 0.06 0.06 1.16 2.35 2.46
102 0.31 0.34 1.16 0.70 1.07
103 1.51 0.57 0.65 1.88 2.01
104 1.01 0.42 0.03 7.12 1.64
105 1.37 0.45 0.24 2.31 0.78 106 1.92 0.10 0.20 2.77 0.74
107 1.38 0.19 0.47 10.02 2.32
108 2.14 0.16 0.40 6.39 1.71 109 0.25 0.53 1.08 10.76 2.57
110 0.33 0.07 0.32 12.84 2.79
111 1.30 0.17 0.47 2.47 2.40 112 0.77 0.61 1.09 5.84 2.52
113 0.31 0.20 1.10 3.18 1.01
114 0.25 0.35 1.26 4.27 2.04 115 1.04 0.74 0.02 5.83 1.69
116 0.68 0.26 0.93 10.04 2.31 117 0.39 0.03 0.45 11.24 2.56
118 0.77 0.57 0.11 2.46 1.40
119 0.38 0.60 0.56 10.66 1.99 120 1.01 0.41 0.10 3.80 2.37
121 0.78 0.17 0.07 5.34 1.83
122 0.62 0.50 0.12 5.49 1.68 123 0.17 0.78 0.81 0.72 2.93
124 1.53 0.06 0.60 5.15 0.94
125 0.09 0.37 0.56 6.63 0.87 126 0.98 0.36 0.83 3.27 1.06
127 0.49 0.46 0.25 7.77 2.38
128 1.71 0.27 0.63 5.50 2.34 129 0.78 0.20 0.03 9.99 2.46
130 1.32 0.25 0.05 4.78 2.51
131 1.86 0.56 0.55 5.52 2.72 132 1.29 0.29 0.68 0.98 0.66
123
ID S A B L V
133 1.11 0.24 0.88 8.67 2.62
134 0.11 0.22 0.33 3.00 0.18 135 2.17 0.08 0.09 1.78 1.28
136 0.64 0.25 0.28 4.35 2.65
137 1.37 0.59 0.59 8.85 2.62 138 1.32 0.71 0.21 4.59 2.41
139 1.95 0.73 0.01 9.96 2.37
140 0.74 0.27 0.88 2.67 1.77 141 0.26 0.48 0.42 1.96 1.96
142 0.73 0.24 1.02 1.44 1.41
143 1.03 0.63 0.04 8.90 2.40 144 0.43 0.73 0.09 11.46 1.30
145 1.53 0.18 1.22 1.12 2.28
146 0.04 0.70 0.74 2.47 2.91 147 1.90 0.55 0.80 5.04 2.84
150 1.04 0.56 0.45 6.29 1.27
151 1.80 0.09 0.21 11.42 2.34 152 0.78 0.25 0.16 4.21 2.63
153 0.28 0.72 0.37 4.37 2.54
154 1.24 0.14 0.51 10.90 2.32 155 1.15 0.65 0.26 2.15 2.90
156 1.07 0.49 1.02 1.62 1.67
157 1.86 0.42 0.02 8.48 2.95 158 0.18 0.18 0.66 4.14 2.91
159 0.54 0.81 0.44 5.19 2.38 160 1.34 0.49 0.62 5.22 2.05
161 1.15 0.60 0.13 7.59 1.00
162 1.21 0.53 0.13 2.64 1.63 163 0.28 0.63 0.25 9.38 2.59
164 1.48 0.38 0.72 5.55 2.36
165 0.21 0.63 1.25 7.37 2.50 166 0.11 0.18 0.92 5.55 0.87
167 0.03 0.06 0.48 5.99 2.66
168 1.49 0.24 0.64 7.87 2.85 169 0.67 0.14 0.63 7.32 2.94
170 0.05 0.32 0.56 4.43 1.41
171 1.15 0.54 0.74 1.79 1.75 172 0.33 0.35 0.46 10.80 1.18
173 0.23 0.71 0.72 0.63 1.62
174 0.83 0.05 0.90 7.31 2.73 175 0.30 0.43 0.68 8.46 2.14
176 0.34 0.74 0.20 3.05 2.47
177 0.55 0.73 0.77 8.24 2.85 178 1.55 0.04 0.74 9.54 2.64
179 0.28 0.36 0.94 1.09 1.78
180 1.01 0.21 0.97 1.42 2.42 181 1.36 0.36 0.15 1.29 0.52
182 0.06 0.76 0.58 4.76 2.52
183 1.12 0.03 0.37 2.86 0.24 184 0.20 0.34 0.80 6.06 1.51
185 0.48 0.07 0.84 4.83 0.72
186 0.58 0.13 0.05 12.51 1.77 187 1.14 0.43 0.41 7.07 2.78
188 0.17 0.68 0.96 9.20 2.18
189 0.71 0.77 0.20 0.63 2.22 190 0.99 0.57 1.03 4.92 2.67
191 0.37 0.04 0.19 6.34 0.25
192 1.45 0.14 0.72 1.40 0.94 193 0.16 0.67 1.05 5.85 1.89
194 2.08 0.28 0.21 1.64 0.57
195 0.49 0.34 0.87 4.54 2.54
ID S A B L V
196 0.29 0.27 0.93 6.48 2.30
197 0.78 0.14 0.23 5.79 0.86 198 0.09 0.10 0.30 6.73 1.49
199 0.85 0.45 0.88 2.69 1.53
200 1.18 0.40 0.72 10.47 2.60 201 0.94 0.27 0.13 7.39 1.15
202 0.27 0.56 0.35 0.99 1.37
203 2.18 0.54 0.44 7.25 2.53 204 0.56 0.44 0.09 9.52 1.59
205 1.89 0.40 0.80 10.08 2.96
206 2.24 0.64 0.19 8.67 2.20 207 2.21 0.14 0.64 0.74 1.92
208 1.12 0.37 0.07 2.97 1.74
209 1.79 0.44 0.21 10.95 2.06 210 2.17 0.23 0.35 0.88 0.72
211 1.30 0.39 0.89 2.60 2.60
212 1.69 0.25 0.12 6.39 1.36 213 0.16 0.61 1.18 9.72 2.78
214 0.43 0.15 0.86 3.28 2.21
215 0.03 0.27 0.57 10.37 2.32 216 1.97 0.11 0.33 3.74 2.08
217 0.86 0.68 0.50 8.22 2.08
218 0.00 0.62 0.43 5.63 1.41 219 0.82 0.43 0.56 5.02 2.52
220 0.26 0.68 0.32 2.57 2.02 221 0.88 0.46 1.01 3.83 1.96
222 0.81 0.69 0.87 1.83 2.86
223 0.32 0.01 0.66 10.34 2.69 224 0.25 0.55 1.26 5.30 2.79
225 0.19 0.50 0.91 6.25 2.11
226 0.32 0.27 0.76 7.65 2.07 227 1.52 0.71 0.64 2.27 2.96
228 0.10 0.57 0.44 3.04 0.64
229 0.16 0.37 0.96 8.77 1.59 230 1.00 0.07 0.88 11.34 2.78
231 1.70 0.09 1.20 5.67 2.29
232 2.16 0.14 0.56 2.69 1.53 233 1.88 0.03 1.16 6.12 2.95
234 2.14 0.73 0.47 4.49 2.89
235 0.18 0.23 0.34 11.59 2.47 236 1.42 0.33 1.16 4.85 2.91
237 0.43 0.50 1.33 2.27 2.42
238 1.51 0.57 0.57 0.32 0.96 239 0.74 0.71 0.78 4.86 1.55
240 1.61 0.39 0.01 5.73 2.56
241 0.29 0.24 1.04 1.79 2.99 242 1.83 0.20 0.12 8.27 2.52
243 2.00 0.41 0.02 8.35 2.88
244 0.65 0.27 0.63 2.32 2.93 245 0.70 0.77 1.13 6.33 2.60
246 1.34 0.19 0.04 7.10 2.53
124
Table 10 - Solute Descriptor Inter-Correlations (R2): Real Chemicals
R2 S A B L V
S - 0.08 0.05 0.48 0.19
A 0.08 - 0.00 0.12 0.07
B 0.05 0.00 - 0.04 0.00
L 0.48 0.12 0.04 - 0.82
V 0.19 0.07 0.00 0.82 -
Table 11 - Solute Descriptor Inter-Correlations (R2): Hypothetical Chemicals
R2 S A B L V
S - 0.01 0.04 0.01 0.01
A 0.01 - 0.00 0.03 0.02
B 0.04 0.00 - 0.06 0.07
L 0.01 0.03 0.06 - 0.03
V 0.01 0.02 0.07 0.03 -
5.6 Description of Model and Parameters
Model details and compartments are as described in Wania et al.128
, except where alterations
have been made for the PP-LFER model as noted in the Section 5.2. An additional alteration
is the addition of a dynamic time step to ensure model stability. The nomenclature in
Appendix A of Wania et al. has been used to summarize the model parameters listed in Table
12.
The finite difference approximation remains unaltered in the model but the length of the time
step is no longer fixed and may differ for each time step. At each time step the maximum
125
relative change in fugacity of any compartment is calculated; if the maximum change is
greater than 50% the solution does not proceed, instead the length of the time step is
decreased to one third of the current value and the calculations for the current time step are
repeated. This procedure is repeated until the maximum relative change in the fugacity of
any compartment is 50% or less, and then the program proceeds to the next time step. The
time step is not permitted to fall below a value of 0.001 hours (3.6 seconds). If the maximum
relative change in the fugacity in any compartment is less than 10% the length of the time
step is increased to four thirds of the current value and the solution proceeds to the next time
step. The time step is not permitted to rise above 24 hours. This dynamic time step ensures
stability in all compartments for chemicals with extreme properties and speeds up the
calculation time for many other chemicals.
An emission rate of 1 tonne per year is used for all three emission scenarios, and a generic
molecular weight of 200 g/mol is used for all real and hypothetical chemicals to ensure that
the same number of moles of chemical is emitted for all chemicals and scenarios. Chemical
degradation half-lives and activation energies are set to arbitrarily high values to prevent
degradation from occurring. The density of atmospheric particles is calculated from the
densities of organic matter and mineral matter and the fraction of organic matter in particles,
assuming that the remainder of the particles is composed of mineral matter. The same
density is used in the PP-LFER model to ensure that the amount of particles in the air is the
same in both models.
126
Table 12 - Summary of Model Parameters
Chemical and Emission
MW 200 Molecular weight (g/mol)
EX 1 Emission rate to the compartment receiving emissions (tonnes/year)
RfA 1 Ratio relating fugacity in inflowing air to fugacity of the outflowing air
RfC 1 Ratio relating fugacity in inflowing water to fugacity of the outflowing water
Miscellaneous Model Parameters
DNOM 1000000 Density of organic matter (g/m3)
DNMM 2400000 Density of mineral matter (g/m3)
Air
HA 6000 Height of atmospheric compartment (m)
AS 1.8E+12 Surface area of drainage basin (m2)
VFSA 4 Mass of aerosol in air (μg/m3)
VFSAIN 4 Mass of aerosol in inflowing air (μg/m3)
VFOQ 0.1 Volume fraction of organic carbon on aerosol (-)
tA 42 Atmospheric residence time (in hours)
Q 6.8E+04 Particle scavenging ratio (-)
facStability 3 Stability of the winter atmosphere relative to summer conditions (-)
U3S 57 Rain rate of drainage basin and estuary (cm/year)
RHA 80 Relative Humidity of air (%)
TA1 268.8 Air temperature in January (Kelvin)
TA2 266.8 Air temperature in February (Kelvin)
TA3 271.1 Air temperature in March (Kelvin)
TA4 278 Air temperature in April (Kelvin)
TA5 282.8 Air temperature in May (Kelvin)
TA6 286.4 Air temperature in June (Kelvin)
TA7 288.2 Air temperature in July (Kelvin)
TA8 290.5 Air temperature in August (Kelvin)
TA9 282.8 Air temperature in September (Kelvin)
TA10 280.2 Air temperature in October (Kelvin)
TA11 277.7 Air temperature in November (Kelvin)
TA12 269.9 Air temperature in December (Kelvin)
WSS1 5.1 Wind speed in January, terrestrial environment (m/sec)
WSS2 5.8 Wind speed in February, terrestrial environment (m/sec)
WSS3 4.9 Wind speed in March, terrestrial environment (m/sec)
WSS4 4.6 Wind speed in April, terrestrial environment (m/sec)
WSS5 5 Wind speed in May, terrestrial environment (m/sec)
WSS6 4.8 Wind speed in June, terrestrial environment (m/sec)
WSS7 4.9 Wind speed in July, terrestrial environment (m/sec)
WSS8 4.6 Wind speed in August, terrestrial environment (m/sec)
WSS9 5.1 Wind speed in September, terrestrial environment (m/sec)
WSS10 5.6 Wind speed in October, terrestrial environment (m/sec)
WSS11 6.4 Wind speed in November, terrestrial environment (m/sec)
WSS12 5.7 Wind speed in December, terrestrial environment (m/sec)
Fresh Water
HW 5 Average (fresh) water depth (m)
frAW 0.06 Fraction of drainage basin covered by rivers and lakes (-)
frUW 0.14 Evaporation from fresh water as fraction of input (-)
vWD-P 1.03 Dry particle deposition to fresh water (m/h)
BPW 70 Primary productivity in fresh water ( g C/m2 yr )
facOWmiw 0.9 POC mineralisation fraction in fresh water (-)
MW 0.35 Fresh water partitioning into POC: Koc = M · Kow (Factor M in L/Kg)
TS1 268 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in January (Kelvin)
TS2 266.8 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in February (Kelvin)
TS3 270.9 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in March (Kelvin)
TS4 277.9 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in April (Kelvin)
TS5 282.4 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in May (Kelvin)
TS6 287.4 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in June (Kelvin)
TS7 288 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in July (Kelvin)
TS8 289.6 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in August (Kelvin)
TS9 282.1 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in September (Kelvin)
TS10 279.8 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in October (Kelvin)
TS11 275.6 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in November (Kelvin)
TS12 269.3 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in December (Kelvin)
Fresh Water Eroding Sediments (WS1)
127
HWS1 0.05 Surficial fresh water sediment depth, compartment 1 (m)
VFSWS1 0.3 Volume fraction of solids in fresh water sediment, compartment 1 (-)
frAWS1 0.6 Sediment area, compartment 1, as fraction of fresh water area (-)
VsedWS1 0.05 POC sedimentation velocity, fresh water sediment 1 (m/h)
VburWS1 5E-09 POC burial velocity, fresh water sediment 1 (m/h)
VlatWS12 1E-11 Lateral POC transport velocity within fresh water basin (m/h)
VlatWS1TS1 1E-11 Lateral POC transport between fresh water and estuary basins (m/h)
facOWS1res 0.95 POC resuspension intensity (fraction of deposition) (-)
facOWS1mis 0.9 POC mineralisation in sediment (fraction of net deposition) (-)
BWbio1poc 1E-07 Bioturbation diffusivity for sediment solids (m2/h)
U8bblW1 0.05 Mass transfer coefficient for benthic boundary layer (m/h)
Fresh Water Accumulating Sediments (WS2)
HWS2 0.05 Surficial fresh water sediment depth, compartment 2 (m)
VFSWS2 0.3 Volume fraction of solids in fresh water sediment, compartment 2 (-)
VsedWS2 0.05 POC sedimentation velocity, fresh water sediment 2 (m/h)
VburWS2 5E-08 POC burial velocity, fresh water sediment 2 (m/h)
VlatWS21 1E-11 Lateral POC transport velocity within fresh water basin (m/h)
facOWS2res 0.5 POC resuspension intensity (fraction of deposition) (-)
facOWS2mis 0.9 POC mineralisation in sediment (fraction of net deposition) (-)
BWbio2poc 1E-07 Bioturbation diffusivity for sediment solids (m2/h)
U8bblW2 0.01 Mass transfer coefficient for benthic boundary layer (m/h)
Estuary Water
HT 52 Average water depth of estuary (m)
AT 4.77E+11 Surface area of estuary (m2)
frUT 0.19 Evaporation from estuary as fraction of input (-)
facFTC 25 Factor by which the net flux between the estuary and coastal water compartments is increased by mixing
vTD-P 1.03 Dry particle deposition to estuary water (m/h)
BPT 134 Primary productivity in estuarine water ( g C/m2 yr )
facOTmiw 0.9 POC mineralisation in estuary water (% of PP)
MT 0.35 Estuarine water partitioning into POC: Koc = M · Kow (Factor M in L/Kg)
CpocOut 0.32 POC Concentration of outside marine water (mg/L)
TT1 269.9 Estuary temperature in January (Kelvin)
TT2 267.6 Estuary temperature in February (Kelvin)
TT3 270.9 Estuary temperature in March (Kelvin)
TT4 276.5 Estuary temperature in April (Kelvin)
TT5 280 Estuary temperature in May (Kelvin)
TT6 284.7 Estuary temperature in June (Kelvin)
TT7 286.7 Estuary temperature in July (Kelvin)
TT8 289.6 Estuary temperature in August (Kelvin)
TT9 284.2 Estuary temperature in September (Kelvin)
TT10 281.6 Estuary temperature in October (Kelvin)
TT11 277.8 Estuary temperature in November (Kelvin)
TT12 272.7 Estuary temperature in December (Kelvin)
WST1 5.8 Wind speed in January, Estuary (m/sec)
WST2 6.8 Wind speed in February, Estuary (m/sec)
WST3 5.5 Wind speed in March, Estuary (m/sec)
WST4 5.3 Wind speed in April, Estuary (m/sec)
WST5 5.9 Wind speed in May, Estuary (m/sec)
WST6 5.3 Wind speed in June, Estuary (m/sec)
WST7 5.8 Wind speed in July, Estuary (m/sec)
WST8 5.3 Wind speed in August, Estuary (m/sec)
WST9 6 Wind speed in September, Estuary (m/sec)
WST10 6.7 Wind speed in October, Estuary (m/sec)
WST11 7.7 Wind speed in November, Estuary (m/sec)
WST12 6.8 Wind speed in December, Estuary (m/sec)
Estuary Eroding Sediments (TS1)
HTS1 0.01 Surficial estuary sediment depth, compartment 1 (m)
VFSTS1 0.2 Volume fraction of solids in estuary sediment, compartment 1 (-)
frATS1 0.7 Sediment area, compartment 1, as fraction of estuary area (-)
VsedTS1 1.75E-02 POC sedimentation velocity, estuary sediment 1 (m/h)
VburTS1 1E-10 POC burial velocity, estuary sediment 1 (m/h)
VlatTS12 1E-07 Lateral POC transport velocity within estuary basin (m/h)
VlatTS1CS1 1E-11 Lateral POC transport between estuary and coastal water basins (m/h)
facOTS1res 0.75 POC resuspension intensity (fraction of deposition) (-)
facOTS1mis 0.05 POC mineralisation in sediment (fraction of net deposition) (-)
BTbio1poc 1E-07 Bioturbation diffusivity for sediment solids (m2/h)
128
U8bblT1 0.05 Mass transfer coefficient for benthic boundary layer (m/h)
Estuary Accumulating Sediments (TS2)
HTS2 0.1 Surficial estuary sediment depth, compartment 2 (m)
VFSTS2 0.2 Volume fraction of solids in estuary sediment, compartment 2 (-)
VsedTS2 1.75E-02 POC sedimentation velocity, estuary sediment 2 (m/h)
VburTS2 3E-08 POC burial velocity, estuary sediment 2 (m/h)
VlatTS21 1E-12 Lateral POC transport velocity within estuary basin (m/h)
facOTS2res 0.1 POC resuspension intensity (fraction of deposition) (-)
facOTS2mis 0.85 POC mineralisation in sediment (fraction of net deposition) (-)
BWbio2poc 1E-07 Bioturbation diffusivity for sediment solids (m2/h)
U8bblT2 0.01 Mass transfer coefficient for benthic boundary layer (m/h)
Agricultural Soil (E)
HE 0.18 Soil depth, agricultural soil (m)
VFAE 0.25 Volume fraction of air in agricultural soil (-)
VFWE 0.25 Volume fraction of water in agricultural soil (-)
VFSE 5E-05 Volume fraction of solids in run-off water from agricultural soil (-)
frUE 0.43 Evaporation from other soil compartments (agr.+unc.) as fraction of input (-)
OCE 0.02 Mass fraction of organic carbon solids in agricultural soil (-)
U7Emax 2.08 Mass transfer coefficient through air boundary over agricultural soil layer (m/h)
BsolidE 2.28E-08 Solid phase diffusivity in agricultural soil (m2/h)
vED-P 1.03 Maximum dry deposition velocity to agricultural soil (m/h)
ME 0.35 Partitioning relationship for POC in agricultural soil (Koc = M·Kow), Factor M (L/Kg)
Forest soil (B)
HB 0.1 Soil depth, forest soil (m)
frAB 0.479 Fraction of drainage basin covered by forest (-)
VFAB 0.25 Volume fraction of air in forest soil (-)
VFWB 0.25 Volume fraction of water in forest soil (-)
VFSB 1E-05 Volume fraction of solids in run-off water from forest soil (-)
frUB 0.26 Evaporation from forest soil as fraction of input (-)
OCB 0.02 Mass fraction of organic carbon solids in forest soil (-)
U7Bmax 0.416 Mass transfer coefficient through air boundary over forest soil layer (m/h)
BsolidB 2.28E-08 Solid phase diffusivity in forest soil (m2/h)
vBD-P 0.206 Maximum dry deposition velocity to forest soil (m/h)
MB 0.35 Partitioning relationship for POC in agricultural soil (Koc = M·Kow), Factor M (L/Kg)
Uncultivated Soil (U)
HU 0.1 Soil depth, uncultivated soil (m)
frAU 0.06 Fraction of drainage basin covered by uncultivated soil (-)
VFAU 0.25 Volume fraction of air in uncultivated soil (-)
VFWU 0.25 Volume fraction of water in uncultivated soil (-)
VFSU 3E-05 Volume fraction of solids in run-off water from uncultivated soil (-)
OCU 0.02 Mass fraction of organic carbon solids in uncultivated soil (-)
U7Umax 2.08 Mass transfer coefficient through air boundary over uncultivated soil layer (m/h)
BsolidU 2.28E-08 Solid phase diffusivity in uncultivated soil (m2/h)
vUD-P 1.03 Maximum dry deposition velocity to uncultivated soil (m/h)
MU 0.35 Partitioning relationship for POC in agricultural soil (Koc = M·Kow), Factor M (L/Kg)
Coniferous Canopy
frACon 0.73 Fraction of forest area covered by coniferous trees (-)
VFCon 1.5E-03 Coniferous canopy volume per ground area (m3/m2)
Mcon 38 Paritioning relationship for coniferous canopies, factor M (Kcan/air = M·KowN)
Ncon 0.69 Paritioning relationship for coniferous canopies, exponent N (Kcan/air = M·KowN)
vFCD-G 42.1 Dry gaseous deposition velocity to coniferous canopy, summer average (m/h)
vFCD-P 3.4 Dry particle deposition velocity to coniferous canopy, summer average (m/h)
tFC 5 Average residence time of needles (years)
frUFC 0.4 Evaporation from coniferous canopy as fraction of input (-)
Deciduous Canopy
VFDec 1.1E-03 Deciduous canopy volume per ground area (m3/m2)
Mdec 14 Paritioning relationship for deciduous canopies, factor M (Kcan/air = M·KowN)
Ndec 0.76 Paritioning relationship for deciduous canopies, exponent N (Kcan/air = M·KowN)
vFDD-G 129.6 Dry gaseous deposition velocity to deciduous canopy, summer average (m/h)
vFDD-P 27 Dry particle deposition velocity to deciduous canopy, summer average (m/h)
facVLeaf 0.1 Fraction of leaves which stays on trees during winter
tStartGrow 113 Start of deciduous canopy development (day)
tStopGrow 143 End of deciduous canopy development (day)
tStartFall 302 Start of falling leaves (day)
tStopFall 332 End of falling leaves (day)
frUFD 0.4 Evaporation from deciduous canopy as fraction of input (-)
129
5.7 Interpolation Method
In the log KAW / log KOA chemical space moving in different directions has different
chemical meanings. Moving along the x axis corresponds to changes in octanol/air
partitioning, moving along the y axis corresponds to changes in air/water partitioning,
moving in the diagonal up to the right corresponds to changes in octanol/water partitioning
and moving in the diagonal down to the right corresponds to changes in both octanol/air and
air/water partitioning while octanol/water partitioning remains fixed. A simple distance
weighted interpolation method might possibly interpolate points from data in only one
direction effectively ignoring the other types of partitioning.
The goal of the interpolation method outlined here is to ensure that points are interpolated
from data in at least three of the directions outlined above, guaranteeing that data is included
from at least two directions corresponding to the three partitioning coefficients. The
complete procedure is as follows:
• All points with input data in four directions are interpolated.
• All points with input data in three directions are interpolated.
• All interpolated points are added to the input data.
• All points with input or interpolated data in four directions are
interpolated; these newly interpolated points are also added to the
list of input data as they are interpolated.
• All points with input or interpolated data in three directions are
interpolated; these newly interpolated points are also added to the
list of input data as they are interpolated.
The last two steps are referred to as extrapolation because they are partly, or possibly even
entirely, calculated based on previously interpolated points. Given enough data points to start
from the procedure will always run to completion and fill in the entire chemical space,
extrapolating where necessary. Interpolation for each point is performed as follows:
• The angle of all points within 2.83 log units is calculated.
130
• Points are assigned to the four directions based on their angles; any
points on the dividing line between directions are counted for both
directions.
• Any direction with three or more points available for interpolation is
flagged as having sufficient data for interpolation.
• If there is data in a sufficient number of directions the mean for each
direction is calculated, and then the mean of the directional values is
calculated and used as the interpolated value.
A maximum distance of 2.83 was selected to ensure sufficient points are available for
interpolation in all directions when interpolating with a resolution of 1 log unit or less. Using
the mean value instead of a distance weighted method has a smoothing effect on the results,
making the surfaces less prone to spikes caused by outlying values. Python 2.5.2 code for the
entire interpolation and extrapolation routine is provided below.
131
#chemical space interpolation routine #written by Trevor Brown #in python 2.5.2 from math import acos from math import asin def get_angle(x1, y1, x2, y2): #return the angle that point x2,y2 is found from point x1,y1 in degrees distance = pow(pow(x2 - x1,2) + pow(y2 - y1,2),0.5) if distance == 0: return 0 else: if y2 >= y1 and x2 > x1: return_value = asin((x2 - x1) / distance) * 180 / 3.14159265358979 elif y2 < y1 and x2 >= x1: return_value = 90 + acos((x2 - x1) / distance) * 180 / 3.14159265358979 elif y2 <= y1 and x2 < x1: return_value = 180 + asin((x1 - x2) / distance) * 180 / 3.14159265358979 elif y2 > y1 and x2 <= x1: return_value = 270 + acos((x1 - x2) / distance) * 180 / 3.14159265358979 if return_value >= 360: return_value = return_value - 360 if return_value < 0: return_value = return_value + 360 return return_value def get_mean(data_points): #calculate the mean value of data_points if len(data_points) > 0: return_value = 0 for index in range(len(data_points)): return_value = return_value + data_points[index][1] return_value = return_value / (index+1) return return_value else: return 0 #read in chemical IDs, Koa, Kaw and value to be interpolated #from space delimited input file input_file = open("pyinput.txt","r") store_var = [] count = 0 for index in input_file: #id position = index.find(" ") store_var.append({"id":index[:position]}) index = index[position+1:] #x position = index.find(" ") store_var[count]["koa"] = eval(index[:position]) index = index[position+1:] #y position = index.find(" ") store_var[count]["kaw"] = eval(index[:position]) index = index[position+1:] #z position = index.find(" ") store_var[count]["z"] = eval(index[:position]) index = index[position+1:] count = count + 1 input_file.close() #define chemical space and resolution to parse over #in this case 3 <= Koa <= 13, -6 <= Kaw <= 4 #and a resolution of 0.5 log units koa_vals = [] for index in range(21): koa_vals.append(3 + float(index) * 0.5) kaw_vals = [] for index in range(21): kaw_vals.append(-6 + float(index) * 0.5) #create variables to track output
132
points = len(koa_vals) * len(kaw_vals) point_list = [] out_put = dict() #n_req is the required number of directions that interpolation #points are available from n_req = 4 #determines if the interpolation/extrapolation loop is in #interpolation or extrapolation mode extrapolate = 0 #start interpolation/extrapolation loop while points > 0: #update keeps n_req from decreasing until no more points #can be added when in extrapolation mode update = 0 for koa in koa_vals: for kaw in kaw_vals: #if this point has not already been interpolated calculate #the number of points in range and their directions if not out_put.has_key((koa,kaw)): arc_n_s = [] arc_ne_sw = [] arc_e_w = [] arc_se_nw = [] for index in store_var: angle = get_angle(koa, kaw, index["koa"], index["kaw"]) distance = pow(pow(index["koa"] - koa,2) + pow(index["kaw"] - kaw,2),0.5) if distance <= 2.83: if angle > 337 or angle < 23: arc_n_s.append([distance,index["z"]]) if angle > 22 and angle < 68: arc_ne_sw.append([distance,index["z"]]) if angle > 67 and angle < 113: arc_e_w.append([distance,index["z"]]) if angle > 112 and angle < 158: arc_se_nw.append([distance,index["z"]]) if angle > 157 and angle < 203: arc_n_s.append([-distance,index["z"]]) if angle > 202 and angle < 248: arc_ne_sw.append([-distance,index["z"]]) if angle > 247 and angle < 293: arc_e_w.append([-distance,index["z"]]) if angle > 292 and angle < 338: arc_se_nw.append([-distance,index["z"]]) #only directions with 3 or more points contribute to interpolation n = 0 if len(arc_n_s) >= 3: n = n + 1 if len(arc_ne_sw) >= 3: n = n + 1 if len(arc_e_w) >= 3: n = n + 1 if len(arc_se_nw) >= 3: n = n + 1 #if there is data in the required number of directions interpolate this point if n >= n_req: sum_weight = 0 if len(arc_n_s) >= 3: mean = get_mean(arc_n_s) sum_weight = sum_weight + mean if len(arc_ne_sw) >= 3: mean = get_mean(arc_ne_sw) sum_weight = sum_weight + mean if len(arc_e_w) >= 3: mean = get_mean(arc_e_w) sum_weight = sum_weight + mean if len(arc_se_nw) >= 3: mean = get_mean(arc_se_nw) sum_weight = sum_weight + mean z = sum_weight / n #add this point to the output out_put[(koa,kaw)] = z points = points - 1 #if in extrapolation mode add this point to the extrapolation data
133
#and set the flag "update" not to decrease n_req if extrapolate == 1: store_var.append({"koa":koa,"kaw":kaw,"z":z}) update = 1 #decrease n_req unless "update" flag is set and in extrapolation mode if not(extrapolate == 1 and update == 1): n_req = n_req – 1 #once all possible points have been interpolated start extrapolation if n_req == 2: #append all points interpolated to the list of data for extrapolation for key,value in out_put.items(): store_var.append({"koa":key[0],"kaw":key[1],"z":value}) n_req = 4 extrapolate = 1 #create list of points and interpolated values to output print_list = [] for key,value in out_put.items(): print_list.append([key[1],key[0],value]) print_list.sort() #output list to space delimited file in a matrix with row and column headers output_file = open("pyoutput.txt","w") this_row = print_list[0][0] out_string = "0 " for index in print_list: if index[0] == this_row: out_string = out_string + str(index[1]) + " " else: break out_string = out_string + "\n" output_file.write(out_string) out_string = str(print_list[0][0]) + " " for index in print_list: if index[0] == this_row: out_string = out_string + str(index[2]) + " " else: this_row = index[0] out_string = out_string + "\n" out_string = out_string + str(index[0]) + " " out_string = out_string + str(index[2]) + " " output_file.write(out_string) output_file.close() #note: to get Excel to properly plot the matrix as a surface you must delete #the zero at position 1,1 of the matrix after pasting into Excel
134
5.8 Validity of Showing Results in the Chemical Space
We expect to find a strong link between environmental fate and the solute descriptors of real
chemicals. This expectation comes intuitively from the following observation; if we plot a
series of chemicals, for example n-alcohols, in the log KAW / log KOA chemical space they are
all located close to each other in the plot. If we now overlay the plot with the chemical fate
predicted by an environmental model we find that the n-alcohols are primarily predicted to
be in the aqueous phase. Therefore we can safely assume that chemicals with solute
descriptors similar to n-alcohols will also be found in the aqueous phase of an environmental
model. However, it is in theory possible that chemicals with combinations of solute
descriptors completely different from n-alcohols will also be located in the same region of
the chemical space. If this is true then plotting the results of the PP-LFER model in the log
KAW / log KOA chemical space would not be sensible because the environmental fate, as
plotted in the chemical space, would be largely unrelated to specific combinations of solute
descriptors.
To investigate the relationship between location in the log KAW / log KOA chemical space and
solute descriptors, chemical space plots of each solute descriptor were constructed for both
real and hypothetical chemicals using the interpolation method outlined in Section 5.7. These
plots are shown in Figure 15, the real chemicals are labelled (R) and the hypothetical
chemicals are labelled (H). In addition to these plots are plots of the standard deviation of
each interpolated point for real and hypothetical chemicals; that is the variation in the data
used to interpolate each point, these plots are labelled (RD) and (HD).
For real chemicals all five solute descriptors show clear trends in the log KAW / log KOA
chemical space and for the descriptors S, B, L and V the standard deviation is relatively
small. These plots indicate that, for the four solute descriptors mentioned, areas in the
chemical space correspond to a very limited number of combinations of solute descriptors.
The A solute descriptor shows a clear trend in the chemical space, but the standard deviation
is relatively large and follows the same trend. We interpret this to mean that in the areas
where A = 0 all chemicals in the area generally also have A = 0. However, the areas in the
135
chemical space which contain values of A greater than zero have much more variability in
the possible values of A. For hypothetical chemicals the trends shown in the log KAW / log
KOA chemical space are less defined, and the standard deviations are larger relative to the
interpolated values. Only the L descriptor shows a strong trend and a strong similarity to the
corresponding plot of real chemicals. We interpret this to mean the relationship between
areas in the chemical space and solute descriptor combinations is in general much weaker for
the hypothetical chemicals; there are more possible combinations of solute descriptors that
may give each pair of KAW/KOA values.
Another method of investigating the relationship between the log KAW / log KOA chemical
space and solute descriptors is to consider the interpolated points as a second data set, and
compare the statistics of this data set with the original. The mean values and standard
deviations of both datasets have been calculated. The statistics for the interpolated points
were then divided by the statistics for the original datasets. This procedure was also
performed for a set of randomly generated numbers to check the effect of the distribution of
real and hypothetical chemicals in the chemical space; all of the calculated ratios are shown
in Table 13.
If a solute descriptor has no relation to chemical location in the chemical space each
interpolated point will have a random sampling of the possible phase descriptor values and
the interpolated value will be near the mean of the original dataset. The result will be a much
smaller standard deviation for the interpolated dataset, as much of the variation is ―averaged
out‖ by the interpolation method. However, if the relationship between a solute descriptor
and location in the chemicals space is quite strong then like values will be localized in the
same regions of the chemical space and much of the variation in the original dataset will be
preserved in the interpolation, meaning the standard deviation of the interpolated dataset will
be comparable to the standard deviation of the original dataset. The ratio of the means will
show if the interpolated dataset is skewed towards higher of lower values.
136
All solute descriptors for both the real and the hypothetical datasets show evidence of
localization, except for the A descriptor in the hypothetical dataset which shows no more
localization than the random data. Interpolated values for hypothetical chemicals are not
skewed to any great extent and, with the exception of the L descriptor, have weaker
localization than the dataset of real chemicals. For real chemicals the S, B, L and V
descriptors have strong localization, with interpolated dataset standard deviations
approximately 70-80% of the magnitude of the original dataset. The A descriptor for real
chemicals shows a relatively weak localization, but along with the B descriptor the mean
ratios show the interpolated dataset is skewed towards lower values. The statistics for A,
along with the plots in Figure 15 suggest the following explanation; chemicals with A equal
to zero are distributed randomly throughout the chemical space, but chemicals with A greater
than zero are localized. Interpolated data for A is skewed towards lower values because
chemicals with A equal to zero are collocated with chemicals that have high A values,
causing the higher values to be partially averaged out.
From the above discussion we conclude that the use of the log KAW / log KOA chemical space
to display the results of a PP-LFER model run using real chemicals is valid. The strong
localization of solute descriptors in the chemical space indicates that areas within the
chemical space correspond not just to specific combinations of log KAW and log KOA, but also
to a small number of similar combinations of the solute descriptors S, A, B, L and V.
Plotting hypothetical chemicals in the log KAW / log KOA chemical space is less valid because
the localization is weaker for all descriptors, and non-existent for the A descriptor. The plots
are useful for comparison to the plots created with real chemicals, but should not be used for
quantitative purposes such as the chemical screening exercise of Brown and Wania146
.
137
Figure 15 - Solute Descriptor Plots
Plots of the surface interpolated from the solute descriptors (S, A, B, L, V) of real
chemicals (R) and hypothetical chemicals (H) along with plots of the surface
interpolated from the RMSE of each point from the data used in the interpolation (RD
and HD).
138
Table 13 - Statistical Ratios of Interpolated Solute Descriptors
Real Chemicals
Mean Ratio
Real Chemicals
Standard Deviation
Ratio
Hypothetical
Chemicals Mean
Ratio
Hypothetical
Chemicals Standard
Deviation Ratio
S 0.949 0.792 0.916 0.287
A 0.705 0.307 0.971 0.152
B 0.727 0.663 0.880 0.378
L 1.079 0.813 1.024 0.806
V 1.103 0.737 1.077 0.484
Random 0.972 0.135 0.992 0.130
Figure 16 - KOC and KPA Correlations of Hypothetical Chemicals
(A) Correlation between the log KOC values obtained by the SP-LFER equation and
values obtained from the PP-LFER obtained for humic acid for hypothetical chemicals.
(B) Correlation between the log KPA values obtained by the SP-LFER and values
obtained from the PP-LFER obtained for whole aerosol for hypothetical chemicals.
139
Figure 17 - Götz et al. PP-LFER KPA versus SP-LFER KPA
Correlation between the log KPA values obtained by the SP-LFER and values obtained
from the PP-LFER of Götz et al. for urban aerosol117
.
Figure 18 - Aspvreten PP-LFER KPA versus SP-LFER KPA
Correlation between the log KPA values obtained by the SP-LFER and values obtained
from the PP-LFER of Arp et al. for rural aerosol from Aspvreten, Sweden121
.
y = 0.970x + 0.970
R² = 0.9762
4
6
8
10
12
14
2 4 6 8 10 12 14
log
KP
A(G
otz
)
log KPA (spLFER)
140
Figure 19 - Chemical Space Plots for Hypothetical Chemicals
Chemical space plots of phase distribution of the PP-LFER based model for
hypothetical chemicals and the sum of absolute differences in the percent phase
distribution between the SP-LFER and PP-LFER based models for each the three
emission scenarios.
141
Figure 20 - Relative Difference in Model Outputs for Air Emissions
Chemical space plots of phase distribution of the PP-LFER based model for real
chemicals, where the coloured areas correspond to more than 50% of chemicals being
partitioned to the phase noted, and the relative absolute differences in model outputs
for each compartment between the SP-LFER and PP-LFER based models for emissions
to air.
142
Figure 21 - Relative Difference in Model Outputs for Soil Emissions
Chemical space plots of phase distribution of the PP-LFER based model for real
chemicals, where the coloured areas correspond to more than 50% of chemicals being
partitioned to the phase noted, and the relative absolute differences in model outputs
for each compartment between the SP-LFER and PP-LFER based models for emissions
to soil.
143
Figure 22 - Relative Difference in Model Outputs for Water Emissions
Chemical space plots of phase distribution of the PP-LFER based model for real
chemicals, where the coloured areas correspond to more than 50% of chemicals being
partitioned to the phase noted, and the relative absolute differences in model outputs
for each compartment between the SP-LFER and PP-LFER based models for emissions
to water.
144
Figure 23 - Chemical Space Plots for Degrading Real Chemicals
Chemical space plots of phase distribution of the PP-LFER based model for degrading
real chemicals and the sum of absolute differences in the percent phase distribution
between the SP-LFER and PP-LFER based models for emissions to air.
145
Figure 24 - Characteristic Travel Distance of Degrading Real Chemicals
Chemical space plots of the characteristic travel distance in kilometres for the SP-
LFER and PP-LFER models.
146
Table 14 - Sub-sampling Statistics
n = 246 S A B L V
Hypothetical Chemicals Mean 0.916 0.394 0.572 5.500 1.999 All Data points St.Dev 0.630 0.226 0.368 3.210 0.710
n = 235 S A B L V
Real Chemicals Mean 0.985 0.136 0.392 6.068 1.367 All Data points St.Dev 0.453 0.214 0.293 2.570 0.525
n = 24 S A B L V
Hypothetical Chemicals Mean 0.615 0.387 0.517 2.209 2.193 Primarily in Air St.Dev 0.480 0.244 0.349 1.090 0.605
n = 30 S A B L V
Real Chemicals Mean 0.518 0.038 0.212 3.764 1.018 Primarily in Air Stdev 0.254 0.058 0.166 0.692 0.237
n = 78 S A B L V
Hypothetical Chemicals Mean 1.041 0.376 0.763 3.063 1.512 Primarily in Water St.Dev 0.645 0.212 0.357 1.854 0.710
n = 87 S A B L V
Real Chemicals Mean 0.888 0.246 0.558 4.183 0.954 Primarily in Water St.Dev 0.323 0.247 0.230 1.470 0.304
n = 88 S A B L V
Hypothetical Chemicals Mean 0.835 0.389 0.417 8.813 2.256 Primarily Sorbed to Solids St.Dev 0.597 0.232 0.315 1.970 0.533
n = 68 S A B L V
Real Chemicals Mean 1.328 0.060 0.194 8.982 1.895 Primarily Sorbed to Solids Stdev 0.455 0.163 0.207 1.606 0.371
n = 56 S A B L V
Hypothetical Chemicals Mean 0.999 0.430 0.575 5.101 2.188 Multimedia Chemicals St.Dev 0.669 0.228 0.355 1.889 0.666
n = 50 S A B L V
Real Chemicals Mean 0.969 0.105 0.480 6.768 1.579 Multimedia Chemicals St.Dev 0.404 0.195 0.335 1.448 0.350
n = 117 S A B L V
Hypothetical Chemicals Mean 0.899 0.587 0.540 5.319 2.109 Chemicals with High A Stdev 0.623 0.125 0.360 3.195 0.630
n = 97 S A B L V
Real Chemicals Mean 0.817 0.329 0.462 4.898 1.175 Chemicals with High A St.Dev 0.354 0.219 0.223 2.134 0.544
n = 118 S A B L V
Hypothetical Chemicals Mean 0.913 0.201 0.599 5.751 1.847 Chemicals with Low A St.Dev 0.642 0.113 0.362 3.266 0.757
n = 138 S A B L V
Real Chemicals Mean 1.104 0.000 0.343 6.891 1.503 Chemicals with Low A St.Dev 0.479 0.000 0.325 2.537 0.468
147
Chapter 4
Investigating the Time Resolution of Water and Organic Carbon
Balances and their Effect on Contaminant Fate Modeling
Trevor N. Brown, Frank Wania
Manuscript in Preparation.
148
1 Introduction Multimedia mass-balance models describing the fate and transport of contaminants in the
environment are integral tools in both the prediction of environmental hazard metrics for
hazard indicators such as long-range transport (LRT) and environmental persistence, and in
the interpretation of observed environmental distributions.24,147
Comparison of the LRT
metric Characteristic Travel Distance (CTD) and the environmental persistence metric
Overall Persistence (POV) among different models has shown that model scale, structure and
assumptions have a stronger influence on the prediction of CTD than POV.37
The observed
differences are likely an underestimation of the variability possible in CTD predictions
because all of the models assessed made the assumption of continuous rainfall, a factor that
is well known to have a strong influence on the long range transport of many
contaminants.148
Comparing empirically derived LRT metrics with model predictions for
specific chemicals has highlighted the importance of other variable atmospheric parameters
such a variable atmospheric oxidation rates and wind conditions.149-151
Interactions of the
atmosphere with the surface are also important for contaminant fate and transport; the
cycling of carbon through the terrestrial and marine systems and other fluctuations of
environmental parameters on global and local scales can exert a strong influence on the fate
of many contaminants.19,152
Atmospheric transport and deposition of contaminants is very complex, with many variable
and interacting environmental parameters to consider. Modeling chemical transport in a well-
defined, consistent climatic scenario with explicit consideration of all the major parameters
is possible with detailed parameterization,151
but many of the tools used to screen chemicals
for LRT potential use generic, time averaged environmental parameters. There is nothing
necessarily wrong with using steady state, time averaged models to screen chemicals for
environmental hazard metrics; the data inputs are minimal which facilitates high-throughput
screening of large chemical lists. Furthermore, many chemical parameters such as
partitioning properties, degradation rates, emission amounts and modes of emission are often
very uncertain and are assumed to introduce errors into the prediction of environmental risk
that are comparable to the error introduced by assuming steady state. Steady state
149
approximations and time averaging of important environmental parameters have been
explored to some degree as a compromise between model simplicity and the accuracy of the
resulting predictions;148,153
the errors and uncertainties associated with these assumptions are
explored in greater detail in this work.
In this study the human exposure model CoZMoMAN34
is altered to explore the effects of
environmental variability and time averaging of environmental parameters. CoZMoMAN is
composed of a food chain bioaccumulation model linked with a regional scale multimedia
mass balance model.128
The model has previously been parameterized for the western Baltic
Sea drainage basin, which is left largely unaltered in the current work. In addition to a
chemical mass balance the model solves mass balance equations for water and particulate
organic carbon (POC) balances. Water and POC mass balances are important because the
pools and fluxes of these two phases act as reservoirs and carrier phases for contaminants in
the environment, controlling their fate and transport. Pools and fluxes of organic carbon are
the most important for many semi-volatile, hydrophobic chemicals such as PCBs and
chlorinated pesticides, but many new chemicals of concern are less hydrophobic and the
water balance will be equally or more important for these chemicals. The CoZMoMAN
model ensures that the water and POC balances are internally consistent by optimizing them
to environmental parameters defined by the user at the beginning of each simulation,meaning
that the magnitudes and directions of the fluxes, and the relative sizes of the pools, cannot
violate the law of mass conservation.
The aim of the current work is to introduce variability into the environmental mass balances
of CoZMoMAN and then to apply the model to a range of chemicals and determine the
effect of this variability on chemical fate and transport at different time resolutions. A
previous version of the environmental portion of CoZMoMAN included a variable water
balance,154
but had many empirical fitting parameters that needed to be manually fit and
environmental input parameters were only at a monthly resolution. There are a number of
reasons to be interested in the short term variability of contaminants; long term averages of
environmental contaminant concentrations may be within safe levels, but at short time
150
resolutions spikes related to sudden strong fluxes in the water and POC balances may exceed
safe levels. Interpretation and design of environmental sampling and monitoring also
requires the consideration of variability, to determine how representative samples collected
over a short period of time are of the average concentrations. Long range transport of
chemicals through the atmosphere is also affected by short term variability due to the
interacting effects of relevant environmental parameters. To investigate these effects the time
resolution of environmental input parameters is varied on a scale ranging from hours to
months and the effects on model outputs and the LRT metric CTD are presented.
2 Methods The environmental portion of the CoZMoMAN model has five complete mass balances;
128
these are the mass balance of the contaminant, which has remained unaltered in the new
model, and the mass balances of air, organic aerosol, water and organic carbon which for this
work have all been altered to allow for more flexibility in defining intra-annual variability.
Figure 25 shows the four environmental mass balances in the version of CoZMoMAN
developed for this work, referred to as CoZMoMAN-EV. Some details of the mass balances
shown in Figure 25 are different from CoZMoMAN, these are described in the Appendix
(Section 6.1). The environmental mass balances are distributed among four subsystems of
the model; air and organic aerosol are restricted to the atmospheric subsystem, the water
balance spans the terrestrial and the aquatic subsystems and the organic carbon balance spans
the terrestrial, aquatic and sediment subsystems. Water and air balances are independently
defined, but the organic carbon balance is strongly dependant on the water balance and the
organic aerosol balance is dependent on the air balance. The mass balance of chemical is
dependent on all four environmental mass balances and spans all subsystems of the model.
Note that in Figure 25 some fluxes (arrows) cross subsystem boundaries; these fluxes are
used to introduce variability into the mass balances and the mass balances then respond to
the variability. Production and degradation fluxes also essentially cross the system
boundaries by introducing and eliminating organic aerosol or organic carbon from the mass
balance and are also used to introduce variability to the mass balances. Variability is
introduced into the model by reading in a deterministic pattern of variability for each of four
151
major environmental parameters that act on the environmental mass balances at the system
boundaries; these are temperature, wind speed, rainfall and biological production of POC in
the aquatic subsystem. A fifth environmental parameter, atmospheric OH radical
concentration, is also used to introduce variability into the model but affects only the mass
balance of contaminant.
Figure 25 - Schematic of the Major Pools and Fluxes of the Four Environmental Mass
Balances in CoZMoMAN-EV and their Interconnectivity.
152
Most of the fluxes spanning the subsystem boundaries in Figure 25 have their pattern of
intra-annual variation determined by relating them to one of the four variable environmental
parameters. Air and organic aerosol inflows and outflows are a function of wind speed. Wet
deposition of organic aerosol is a function of rainfall, and rainfall is a direct function of the
intra-annual pattern of rainfall rate. Evaporation is a function of wind speed and temperature.
Production of POC in the aquatic system is naturally a function of the primary biological
production rate, and of temperature. Degradation of POC is also a function of temperature.
Resuspension of sediments from accumulating sediments is a function of wind speed, and
resespension of sediments from eroding sediments is an indirect function of rainfall through
water flow rates. Other fluxes vary in time, but their pattern of variation is directly related to
the size of their originating pools by the user defined parameterization, rather than being
determined by external forces. All of the environmental mass balances are complete at all
points during the year, but in addition to ensuring closure there is a desire to set the relative
magnitudes of the fluxes and the sizes of the pools to specific values.
Magnitudes and sizes are variable over the course of each year, but many parameters have a
user-defined annual average to which the fluxes and pools are fitted. This is achieved with
fitting parameters which, in the case of fluxes, scale the magnitude of each flux to match the
user defined annual average magnitude without directly affecting its pattern of intra-annual
variation. A more refined approach is required to ensure the annual average sizes of the pools
are fitted to the user defined annual average. In each pool of the water, organic aerosol and
POC balances that has a user-defined annual average size one flux in the mass balance
equation is selected as the fitting parameter (the pool of air is fixed in size). This is generally
the flux in the mass balance that is the most uncertain, or difficult to determine, and therefore
adjusting the magnitude introduces little net uncertainty into the model. This procedure of
fitting the parameters of the fluxes and pools to user defined values is called optimization.
Mass balances for air and atmospheric particles are optimized simultaneously, then the mass
balances for water and POC are optimized sequentially. The goal of the mass balance fitting
algorithm is to have a continuous pattern of intra-annual variability, where the state of each
153
mass balance matches at the beginning and the end of the year, so that the same pattern can
be used during every year of the simulation when modeling contaminant fate. The fitting
algorithm solves the mass balance equations at each hour of the year using the finite
difference approximation. When the end of the year is reached the state of the mass balances
are carried over to the next year and the pattern of variable environmental parameters is
repeated starting from the beginning. An initial guess at the fitting parameters is made using
average values defined by the user and the steady state mass balances of CoZMoMAN.
Before any of the fitting parameters are altered the algorithm runs until there are no inter-
annual variations in the mass balances. Fitting parameters are adjusted at the end of each
year; if the alteration of a fitting parameter causes inter-annual variability in the associated
mass balance then fitting parameters associated with that mass balance are not adjusted again
until the inter-annual variation subsides.
Hourly resolved meteorological data consistent with the temperature and wind speed patterns
of the western Baltic Sea drainage basin were obtained from the online Canadian Climate
Norms Database155
including air temperatures, wind speed, visibility, air pressure and
weather observations. Precipitation data were obtained from the same source at a daily
resolution. Using simplistic heat transfer, solar insolation and phytoplankton growth models
an interconnected set of hourly resolved patterns was created from the initial data for air,
terrestrial and water temperatures, fresh water and marine primary production rates, rainfall,
wind speed and atmospheric OH radical concentrations; these models are described in the
Appendix (Section 6.2). Hourly resolved data were then averaged for longer time periods
evenly divisible into 8760 hours per year; 24h (daily resolution), 146h (~weekly resolution),
730h (~monthly resolution), and 2190h (seasonal resolution). Time averaged values were
then interpolated to an hourly resolution. An example of the resulting data is shown in Figure
26, for air temperature and biological production. Similar figures for other parameters,
details regarding the selection of the dataset, and the models used to process the data are
discussed in the Appendix (Figure 34 and Section 6.2).
154
Figure 26 - Air temperature (A) and primary biological production rate in fresh water
(B) at hourly, daily, weekly, monthly, and seasonal resolutions.
155
Because the environmental model used in these calculations has a low spatial resolution the
validity of the simulations with a high time resolution is uncertain. Possible differences in
the results presented here that would arise from increasing the spatial resolution of the
environmental model are difficult to predict. Depending on the implementation details the
CTDs predicted at higher time resolutions may decrease to values closer to the steady-state
predictions, but it is also possible that they may increase and the spread of possible outcomes
would certainly increase, so the exact effects are still an open question.
Simulations were run for hypothetical contaminants with specific combinations of chemical
partitioning properties. Three linked pure-phase partitioning properties are used to
approximate contaminant fate in the environment; air-water partitioning, octanol-air
partitioning and octanol-water partitioning. These properties are calculated as the logarithm
of their respective partition coefficients and are locked in a thermodynamic cycle defined by
equation (30).
log KOW = log KAW + log KOA (30)
In combination with empirical relationships relating octanol sorption to sorption in various
organic carbon pools these chemical properties predict contaminant distribution between air,
water, soil and sediment compartments. A chemical partitioning space is a grid of points
where each point corresponds to a specific combination of the three partitioning properties,
with a corresponding set of model outputs predicting the chemical fate of a hypothetical
chemical with those properties. The chemical partitioning space used to explore the model
outputs of CoZMoMAN-EV is contained within the bounds -5 ≤ log KAW≤ 1, 4 ≤ log KOA≤
12, and simulations were run for five years of continuous emissions. Temperature
dependences for each combination of log KAW and log KOA were calculated with an empirical
equation relating the partition coefficients to the enthalpies of phase change, the derivation of
which is described in the Appendix (Section 6.3). Two simulations were run for all
hypothetical chemicals; one ―labile‖ simulation and one ―persistent‖ simulation.
Environmental half-lives roughly corresponding to the persistence screening criteria of the
Stockholm Convention on POPs5 were calculated for air, terrestrial, water and sediment
compartments at their respective annual average temperatures. These values were then
156
shifted to lower and higher values to create the labile and persistent half-life values. The
temperature dependence of the degradation rates is assumed to be uniform for air and for
surface compartments at all points in the chemical space.
The mode of emission, i.e. the fractional distribution of emissions to the compartments air,
cultivated soil and fresh water, is assumed to be different for different hypothetical chemicals
and dependent on their location in the chemical partitioning space. The rationale for this is
that the mode of emission is not independent of chemical properties. A very volatile
chemical is unlikely to be emitted directly to soil or water, as fugitive emissions from
commercial or industrial use are the most likely route of release. Low-volatility chemicals
are likely to remain in place wherever they are emitted, or wash out with flowing water if
they are water soluble. Emissions of low-volatility chemicals used in commerce may be from
direct applications to soil (i.e. pesticides), amendment to agricultural soil in sewage
treatment plant biosolids, accidental spills, or leaching from landfills. Equilibrium
distribution between fresh water, cultivated soil, and a volume of air corresponding to the
surface area of these two compartments is calculated for each hypothetical chemical at the
annual average temperatures. Emissions are then distributed among these three
compartments according to the equilibrium distribution for each hypothetical chemical. It
was further assumed that the inflowing air contained no contaminants, which causes the air
concentrations to reach pseudo-steady state very rapidly, and therefore the amount of
contaminant in air is largely determined by advection into and out of the system. This means
that the simulations essentially represent a ―source region‖ for the contaminants.
3 Results A number of test calculations were run to confirm that the CoZMoMAN-EV produced model
outputs within the range of similar calculations done in CoZMoMAN; because although the
pattern of intra-annual variability is different in each model the annual average magnitudes
of the contaminant amounts should be fairly similar. This was done using monthly resolved
environmental parameters in both models. In each case the same chemical properties and
emission fractions, derived as described above, were used for both the new model and the
157
old model. Combinations of log KOA and log KAW representative of volatile, water soluble
and lipophilic chemicals for both labile and persistent chemicals were selected for the
comparison. Factor differences were calculated for the amounts of chemical in all
compartments at each hour of the last year of the simulation and then averaged. For
persistent water soluble and lipophilic chemicals the new model is within a factor of 1.5 of
the original model in all compartments, and within a factor of 2 for persistent volatile
chemicals in all compartments except for sediments which are within a factor of 7.
Differences between the models for labile chemicals are much more pronounced in some
compartments; labile water soluble and lipophilic chemicals are within a factor of 1.5 for
aquatic compartments but only within a factor of 3 for air and terrestrial compartments, and
labile volatile chemicals are within a factor of 3 in aquatic compartments and a factor of 3.5
for air and terrestrial compartments. For both labile and persistent water soluble and
lipophilic chemicals the annual pattern of variation predicted by the new and old models in
water and soils respectively are well matched in both the magnitude of the variations and in
the timing of the maximum and minimum values. However, the pattern of variation for labile
and persistent volatile chemicals is more complex in the new model than in the old model,
showing more sensitivity to air temperature, winds speed and atmospheric OH radical
concentrations. Figures comparing the patterns of variation are provided in the Appendix
(Figure 36 and Figure 37).
Simulation results from the five different time resolutions were inspected to ensure that the
total chemical amounts in the model were stable from year to year, so that differences in
other metrics could be reliably attributed to differences in time resolution. Year to year
percent differences in total chemical mass were calculated between the fourth and fifth years
of the simulation at each hour of the year and averaged. For labile chemicals at all points in
the chemical space and at all time resolutions the average year to year differences were <<
1% indicating that a pseudo-steady state had been reached by the fifth year. Persistent
chemicals at all time resolutions had year to year variations in the range of 2.7% (hourly
resolution) to 2.8% (seasonal resolution) when averaged over the entire chemical space;
strongly lipophilic persistent chemicals were furthest from pseudo-steady state peaking at
5.1% (hourly resolution) to 5.2% (seasonal resolution) year to year variation. In the fifth year
158
of the simulation the total chemical mass present in the model environment as a fraction of
the total emissions (retention fraction) was calculated for all points in the chemical space of
each time resolution. Retention fractions for labile chemicals ranged from 0.04% to 3.38%
(average 1.83%) at hourly resolutions and ranged from 0.03% to 3.37% (average 1.82%) at
seasonal resolutions. Retention fractions for persistent chemicals ranged from 0.08% to
31.1% (average 15.7%) at hourly resolutions and ranged from 0.06% to 31.3% (average
15.8%) at seasonal resolutions. In addition to very similar year to year variations and
retention fractions, the chemical distribution at each time resolution was virtually identical.
Simulations using hourly resolved environmental parameters had the largest intra-annual
variation in all compartments of the model, with simulations using lower resolutions
capturing only a fraction of the variation. The percentage of the annual variation captured for
each compartment by each time resolution was calculated using equation (31) where M is the
mass of chemical in an arbitrary compartment at times i and resolutions j:
𝐶𝑎𝑝𝑡𝑢𝑟𝑒𝑑 𝑉𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛 = 100% × 1 − 𝑀𝑖 ,𝑗 −𝑀𝑚𝑒𝑎𝑛 ,𝑗 − 𝑀𝑖 ,ℎ𝑜𝑢𝑟 −𝑀𝑚𝑒𝑎𝑛 ,ℎ𝑜𝑢𝑟
𝑀𝑖 ,ℎ𝑜𝑢𝑟 −𝑀𝑚𝑒𝑎𝑛 ,ℎ𝑜𝑢𝑟 (31)
Figure 27 summarizes the percentage of the intra-annual variation that is captured for each
compartment and time resolution for persistent and labile chemicals. As Figure 27 shows,
daily resolved environmental parameters capture most of the variability in surface
compartments for most points in the chemical space. In a number of aquatic compartments
the simulations with averaged environmental parameters capture ~0% of the intra-annual
variation for chemicals with high log KOW values because the pattern of variation in those
compartments is out of phase with the results of the simulation run with hourly resolved
parameters, but the amount of high log KOW chemical present in these compartments is small
so this result is largely unimportant. In both of the soil compartments the pattern of intra-
annual variability is poorly captured for chemicals that are primarily in air or water, and for
multimedia chemicals. The most notable result is the poor performance of the time averaged
environmental parameters for capturing the intra-annual variation of the amount of chemical
in air; the variations in air are much larger because the amount of chemical in air responds
159
faster to changes in environmental parameters than in any other compartment and only
hourly resolution is sufficient to capture this variability.
160
Figure 27 - Percentage of Hourly Resolved Variability Captured with Daily, Weekly,
Monthly and Seasonal Resolved Data for Labile and Persistent Chemicals in all Model
Compartments. Each Cell is a Chemical Space Plot Showing the Fraction of Intra-
Annual Variability Captured at Each Point; Labile and Persistent Environmental
Distributions are Provided as a Guide.
161
CTDs were calculated using two different methods. In method one, referred to as annual
CTDs, an annual averaged atmospheric residence time was calculated by taking the average
mass of chemical in air for the last year of the simulation and dividing by the sum of the
annual average net depositional flux and the annual average degradative flux and then the
CTD was calculated with the annual average wind speed. In method two, referred to as
hourly CTDs, the CTDs were calculated using the mass, net depositional flux, degradative
flux and wind speed at each hour of the final year of the simulation, and then these hourly
CTDs were averaged. However, a problem arises when using hourly resolved environmental
parameters because the atmospheric OH radical concentrations fall to zero at night and in
most cases there is no net deposition (because the system is near steady state and emissions
are mostly to surface media), creating times when the hourly atmospheric residence time
approaches infinity. Conceptually this makes sense; it simply means that for the duration of
the periods of no net losses the chemical will travel as far as the wind will carry it, but when
averaging the hourly CTDs the result is heavily skewed to unrealistically large values by the
near-infinity CTDs. To estimate more realistic CTDs when using hourly resolved
environmental data the hourly CTDs were calculated only for daytime hours, and nighttime
hours in which the net deposition exceeded 1% of average daytime degradative losses.
Nighttime travel was approximated by assuming no net loss of chemical at night, then the
number of days it would take to deplete the chemical from the air with day time loss only
was calculated, and the amount of nighttime travel that would occur during this period was
estimated with the annual average nighttime wind speed. Using this method CTDs for
persistent chemicals were in the range of 20000 to 70000 kilometers, values corresponding to
global distribution, and CTDs for labile chemicals were in the range of 2000 to 8000
kilometers. These CTDs are not necessarily unrealistic but they are very different from those
calculated for simulations using lower resolution environmental parameters and are likely
inaccurate. Therefore, for simulations that were run using hourly resolved environmental
parameters the wind speed, mass of chemical in air, depositional flux and degradative flux
were all calculated as 24 hour running averages instead of using hourly values; which
generates CTD values much more in line with the other resolutions. Figure 28 summarizes
the CTDs calculated by the hourly methods, a numerical comparison of the values calculated
by the annual and hourly methods is provided in Table 17 in the Appendix. In a steady state
162
chemical mass balance model (a level III model for example) the two methods outlined here
would produce identical CTD values, the differences between them essentially shows the
effect of periods when high air concentrations coincide with high winds, low degradation and
low deposition to the surface (Table 17 in the Appendix).
Figure 29 shows the amounts of chemical in air for each of the hypothetical labile and
persistent chemicals calculated with hourly resolved environmental parameters. The annual
average amounts in air are determined by chemical partitioning properties (log KOA,
essentially volatility) but the pattern of variability is mostly determined by environmental
parameters and degradation. During warm, sunny periods the pattern of variation for labile
chemicals is dominated by diurnal variations in temperature and atmospheric OH radical
concentrations, whereas for the persistent chemicals wind speed and temperature are
primarily responsible for the pattern of variation. Model predictions and observations of
PCBs (semi-volatile persistent chemicals) drew similar conclusions on the causes of diurnal
variation and short term variability.156
For both labile and persistent chemicals during winter
the pattern of variation is primarily determined by the effects of weather. This is partly due
to the long length of nights in the winter versus the length of nights during the summer; long
nights and low daytime atmospheric OH radical concentrations mean there is less absolute
difference between daytime and nighttime degradation during the winter. For example the
prominent peak in late January corresponds to a period of cold temperatures, low
precipitation, low winds with high cloud cover and low visibility which reduces the predicted
concentration of atmospheric OH radicals. During this period the concentration of
atmospheric particles also peaks, which increases the capacity of bulk air to hold particle
bound chemicals. These effects combine to allow chemicals to accumulate in the atmosphere
because of an elevated storage capacity with few of the normal removal processes active. To
quantify the relative LRT effectiveness during summer and winter CTDs were calculated for
April to September and October to March of the final year of the simulation using the same
method described above to calculate annual average CTDs with the results from simulations
using hourly resolved environmental parameters. For both labile and persistent chemicals
winter CTDs were on average a factor of 6 times higher than during summer.
163
Figure 28 - Hourly Characteristic Travel Distances (CTDs) in Kilometers for Labile (A-
E) and Persistent (F-J) Chemicals using Hourly (A&F), Daily (B&G), Weekly (C&H),
Monthly (D&I), and Seasonal (E&J) Resolved Environmental Parameters.
164
Figure 29 - Moles of Chemical in Air Simulated using Hourly Resolved Environmental
Parameters for Labile (A) and Persistent (B) Chemicals, Ranging from a log KOA of 4
in Red to a log KOA of 10 in Blue, and a log KAW of -5 to 1.
4 Discussion and Conclusions From the data presented in Figure 27 we suggest that weekly to monthly resolved
environmental data is sufficient for modeling organic contaminant fate in the aquatic
environment. Simulations with monthly resolved environmental parameters capture more
than 75% of the intra-annual variability in water and sediment compartments for chemicals
which are primarily found in the aquatic environment. Chemicals found primarily in the air
also have their pattern of variability in the aquatic environment well captured by monthly
165
resolved values, but chemicals primarily found in the soil may require higher resolutions to
capture their pattern of intra-annual variation in the aquatic environment, because their high
log KOW values makes them quite sensitive to changes in POC concentrations and time
averaging tends to shift the peak POC concentration to later in the year. A similar statement
about the soil environment cannot be made with certainty because many chemicals found
primarily in soil are not at pseudo-steady state, and so although greater than 90% of the intra-
annual variability in cultivated soil is captured by monthly resolved environmental
parameters this is likely because it is a simple linearly increasing trend due to continuing
emissions. Forest canopy, and therefore forest soil as well, responds quite rapidly to changes
in air concentrations, and therefore daily resolved environmental parameters are advised
when the goal is to capture the pattern of intra-annual variability. Although the water and
POC balances play an important role in the transport and fate of chemicals in the terrestrial
and aquatic environments, environmental parameters on a resolution greater than daily, and
in some cases greater than monthly, do not appear to be required to model the patterns of
seasonal variation, at least on a regional scale, because of the relatively slow response of
chemicals in these compartments to changing environmental conditions.
If the goal is to capture the pattern of intra-annual variability in air, however, higher than
daily resolution is required; especially for labile chemicals which respond rapidly to the daily
cycle of high and low temperatures and atmospheric OH radical concentrations. Parameters
affecting the amount of chemical in air should be highly resolved; hourly resolution may not
be required; it should be sufficient to capture the diel cycles of temperature and atmospheric
OH radical concentrations. Rainfall and wind speed data should be of sufficient resolution to
capture the patterns of dry and wet spells, and calm and windy spells, which in many cases
may be on the order of daily resolution. Amounts of labile and persistent chemicals in air
simulated with hourly resolved environmental parameters were correlated against air
temperature, atmospheric OH radical concentration, wind speed, rainfall and primary
biological production to determine the effects of each parameter. For labile chemicals the
order of decreasing correlation was air temperature > atmospheric OH radical concentration
> wind speed > primary biological production > rainfall. For persistent chemicals the order
of decreasing correlation was wind speed > air temperature > atmospheric OH radical
166
concentration > primary biological production> rainfall. In all cases the correlations were
negative, which may seem counter intuitive for air temperature, but this is because peak air
temperatures correspond approximately to peak atmospheric OH radical concentrations and,
due to the temperature dependence of degradation, periods of high temperature also
correspond to periods of rapid oxidation in the atmosphere. Primary biological production is
correlated with both air temperature and atmospheric OH radical concentration so it is not
clear how much of the correlation of this parameter with air concentration is due to draw
down by increased POC concentration and how much is due to coincidental correlation.
Chemical amounts in the model respond to changing environmental parameters, but because
of kinetic limitations there is a lag between the change in an environmental parameter and
the response in chemical amounts. A separate response time will be associated with each
environmental parameter, and these response times correspond approximately to the time
resolutions of the environmental parameters required to capture the variability observed in
the chemical concentration, because any fluctuations on a smaller time scale are too rapid to
produce a response in the chemical concentration. Response times of the amount of chemical
in air to environmental parameters were estimated by comparing the environmental input
parameters to the model outputs and slowly shifting the annual patterns of each
environmental parameter to later times by one hour increments. As the environmental
parameters are shifted the correlation with the amount of chemical in air increases as each
parameter is shifted closer to being in phase with the corresponding response. The shift in
time that produces the peak correlation of each environmental parameter with the amount of
chemical in air is assumed to correspond approximately to the response time to that
parameter. Using this method the response time of labile chemicals to air temperature,
atmospheric OH radical concentration, wind speed, rain fall and primary biological
production were found to be 2 hours, 4 hours, 5 hours, 12 hours and 4 hours, respectively.
The response time of persistent chemicals to the same parameters was found to be 5 hours, 6
hours, 8 hours, 15 hours and 6 hours, respectively. Primary biological production has
response times identical to those of atmospheric OH radical concentrations so the observed
effect on air concentration is likely coincidental and due to their shared dependence on solar
radiance in the underlying models used to generate their values.
167
As can be seen from Figure 29 there are large fluctuations in air concentration possible from
day to day. To quantify the maximum possible day to day variations in air concentration the
respective average amounts of chemical in air of consecutive 24-hour periods were compared
for all points in the final year of the simulation using hourly resolved environmental
parameters. For labile chemicals the maximum day to day variation in the amount of
chemical in the air was a factor of 3.8 to 4.1 difference. For persistent chemicals a maximum
factor of 4.4 to 4.8 difference in day to day amounts of chemical in air was possible. The
annual average day to day variability for labile chemicals was in the range of 14% to 21%
and for persistent chemicals was in the range of 11% to 29%. Because the model is assumed
to be well mixed on a regional scale this is likely a relatively conservative estimate of the
day to day variability in air concentration that can be attributed to changing meteorological
conditions. This is a factor to consider when analyzing data from 24h hi-volume air samples
and attempting to attribute observed day to day variability to changes in meteorological
conditions or changes in emission strength.
Figure 28 and Table 17 in the Appendix show that as the environmental parameters are
averaged to lower resolutions the hourly CTD values approach the annual CTD values. If
this trend was continued to a hypothetical annual resolution it can be seen that these two
values would converge to a single steady state CTD of about 2500 km on average for the
persistent chemicals and about 300 km on average for the labile chemicals (Figure 38). This
means that using a steady state annual averaged resolution for environmental parameters
underestimates the CTD for persistent chemicals by a factor of 3.4 on average and for labile
chemicals by a factor of 2.8 on average. It should be noted however that the range of
differences across the chemical space is relatively small; ranging from 2.9 to 3.5 for
persistent chemicals and from 2.2 to 3.0 for labile chemicals, so although the
underestimation of CTDs by a factor of approximately three seems quite significant the
relative ranking of chemicals for LRT potential is likely affected very little. This means that
when using CTD as an LRT screening metric for the prioritization of chemicals only relative
prioritization should be performed if the CTDs are calculated by a steady state mass balance
168
model, because using a cut-off value for LRT screening will result in false negatives as the
LRT potential is underestimated.
5 Acknowledgements We acknowledge funding from the Long-range Research Initiative of the European Chemical
Industry Association (CEFIC LRI-ECO3A2-USTO-0607).
169
6 Appendix
6.1 Variable Mass Balances of Air, Organic Aerosol, Water and Particulate Organic Carbon.
6.1.1 General Description.
All compartments and sub-compartments in the CoZMoMAN have fixed volumes that were
set by the user. In CoZMoMAN-EV some of the compartments have been made variable in
time; these compartments are the volume fraction of organic aerosol in air, the volume
fraction of water in soils, the volume of water in water compartments, and the volume
fraction of particulate organic carbon (POC) in water. The user no longer sets a fixed volume
for these compartments but instead sets the desired value for the annual average volume of
these compartments. The water balance is driven by rainfall, which in the original model was
assumed to be a constant value defined by the user. The organic carbon balance, which is
only solved for the aquatic compartments, is driven by primary biological production, also
having a constant value defined by the user in CoZMoMAN. In CoZMoMAN-EV both of
these driving forces have been made variant in time and the user now defines a pattern of
variation, and the desired annual averages. Many inter-compartment fluxes have also been
altered to be directly dependent on these driving forces, linked to two other environmental
parameters, temperature and wind speed, or are now dependent on the newly variable
compartment volumes. The user now also defines these inter-compartment fluxes as annual
averages instead of fixed values. Although each of the mass balances for air, organic aerosol,
water and organic carbon are allowed to vary in time their mass balance is complete at each
hour of the year, and when the compartment volumes and fluxes are averaged over the whole
year their averages are equal to the annual average values defined by the user.
For inter-compartment fluxes this is achieved simply by fitting a general empirical scaling
factor for each variable flux. All variable compartment volumes have one of the terms in
their mass balances selected as the ―fitting term‖ which is altered to fit the annual average
volume of the compartment to that defined by the user. As an initial guess to the values of
the compartment volumes and the empirical scaling factors the steady state solution to each
of the mass balances is solved with the annual averages set by the user, and then the
170
compartment volumes are set to the annual averages and the empirical scaling factors are
calculated from the steady state solution. Starting from these initial estimates the mass
balances are calculated for an entire year with a one hour time step, then the annual averages
of the compartment volumes are calculated and compared to the user defined annual
averages. The empirical scaling factors are then altered if required and the next year is
calculated starting from the compartment volumes at the end of the previous year. This
continues until a stable, optimized solution is found. Discussion of each mass balance and
equations describing the calculation of fluxes follow; nomenclature and equations are
derived from Wania et al 2006.128
6.1.2 Air Balance.
Air movement into and out of the system is determined by a user defined residence time. The
atmospheric compartment in CoZMoMAN-EV is not spatially resolved, so wind direction
and the exact shape of the air compartment are undefined. In order to link the movement of
air to the wind speed it is assumed that air moving through the system has a path length
roughly determined by the residence time and the wind speed; conceptually this means that
the air follows a fixed path determined by the topography and prevailing weather of the
modeled environment, not necessarily in a straight line. The effective wind cross section is
then defined by the atmospheric height and the surface area of the system divided by the path
length. Air fluxes are the product of wind speed and wind cross section, and the annual
magnitude of air fluxes are fitted to the user defined average residence time by altering the
path length.
The flux of air into and out of the air compartment (aG) in m3/h is solved using equation
(32), where XSX is the wind cross-section in meters of segment X (terrestrial, estuary,
coastal, open), WSX is the wind speed in m/s of segment X (terrestrial, estuary, coastal or
open), and HTA is the height of the atmosphere in meters.
171
aG = ∑ {XSX· HTA· WSX· 3600 s/h} (32)
Total wind cross section for atmospheric compartment (XST) is determined by equation (33),
where PL is the path-length of air travelling through the atmospheric compartment in meters
and ARA is the area of the atmospheric compartment in m2; and the cross section of each
segment X is solved with equation (34).
XST = ARA / PL (33)
XSX = XST· ARX / ARA (34)
Path-length is initially estimated with equation (35), where tA is the user defined atmospheric
residence time and WSS is the wind speed of the terrestrial segment, then equation (36) is
used to calculate the annual average fitted atmospheric residence time (tA(fit)) which is fit to
match tA by altering the path-length.
PL(guess) = tA· WSS· 3600 s/h (35)
tA(fit) = HTA· ARA / aG (36)
6.1.3 Organic aerosol balance.
A mass balance for atmospheric particles was constructed with inputs from inflowing air and
a term to describe particle formation, and outputs in the outflowing air, particle settling and
precipitation scavenging of particles. The annual average volume fraction of particles in air
is fitted to the user defined volume fraction by altering the value for particle formation.
Equation (37) shows the finite difference approximation of the mass balance of organic
aerosol, where VFSA is the volume fraction of organic aerosol in air, VOA is the volume of
the air compartment in m3, qGpro is surface production of aerosol, qGin is the volume of
organic aerosol flowing into the system with inflowing air, qGout is the volume of organic
aerosol flowing out of the system with outflowing air, qGset is the volume of organic aerosol
settling out of the atmosphere and qGscv is the volume of organic aerosol scavenged by
precipitation, all in units of m3/h.
172
VFSA (t+1) = {VFSA (t) · VOA + qGpro + qGin - qGout - qGset - qGscv} / VOA (37)
Values for qGin, qGout, qGset, qGscv are calculated with equations (38) through (41) where
VFSAin is the user defined volume fraction of organic aerosol in inflowing air, DDVX is the
dry deposition velocity to surface compartment x (forest, cultivated or uncultivated soil,
coniferous or deciduous canopy, and fresh, estuary, coastal and open waters), Q is the
particle scavenging ratio and wGAx is the flux of water in rainfall to the surface compartment
x.
qGin = aG·VFSAin (38)
qGout = aG· VFSA (39)
qGset = VFSA·∑ARx·DDVx (40)
qGscv = VFSA· Q · ∑wGAx (41)
As an initial estimate of the volume fraction of organic aerosol in air the VFSA(fit) is set to
the user defined annual average, VFSA(AA), and an initial estimate of the production of
organic aerosol, qGpro(guess), is calculated by solving a steady state version of equation (37)
(dVFSA·VOA/dt = 0) for qGpro and using annual average values for aG, VFSA, DDVx and
wGAx to calculate the other parameters.
6.1.4 Water balance.
Rainfall remains the only net input of water to the aquatic system but evaporation, instead of
being defined as a fraction of rainfall, is now defined as a function of temperature and wind
speed. However, the annual magnitude of evaporation from each surface compartment is
fitted to the fractional evaporation rates defined by the user with a generic fitting parameter.
Evaporation from soil is additionally a function of the volume fraction of water in soil, the
field capacity (a newly introduced parameter) and an empirical evaporation soil stress
function. The volume fraction of water in soil is allowed to vary with time and is fitted to the
user defined annual average volume fraction by altering the field capacity of the soil. Runoff
from each soil compartment to the fresh water compartment is determined by the volume
fraction of water in soil, the field capacity and an empirical runoff soil stress function. If the
173
volume fraction of water in soil reaches the porosity of the soil then all additional net water
inputs flow directly to the fresh water compartment as surface flow. The magnitude of the
annual average runoff is fitted to the user defined runoff fraction with a generic fitting
parameter. Forest canopy now intercepts only a fraction of the rainfall on the forest soil, for
deciduous canopy this fraction is additionally dependent on the volume of the canopy. Net
rainfall interception by the forest canopy is assumed to enter the forest soil by stem flow. All
bulk water compartments are now permitted to vary in depth, and therefore volume, and are
fitted to the user defined annual average depths by altering empirical outflow fractions that
determine the downstream flow of water based on the current volume of water.
Rainfall to any surface compartment x is calculated with equation (42), where wGAx is the
water flux of rain in m3/h and U3X is the rainfall to any segment X in m/h (fresh water,
estuary, soils and forest canopy comprise the terrestrial ―S‖ segment for rainfall and so all
have the same rainfall rate), and ARx is the area of the surface compartment x.
wGAx = U3X·ARx (42)
This equation is modified for the forest soil and canopy compartments and equations (43) to
(45) are used instead, where wGAFD, wGAFC, wGAB are rainfall water fluxes to deciduous
forest canopy, coniferous forest canopy and forest soil, respectively, frUFDinter and frUFCinter
are the fraction of rainfall intercepted by the deciduous and forest canopies (set to values of
0.13 and 0.22 according to Dunne and Leopold157
) and frCovD is the seasonal forest cover
fraction of the deciduous canopy.
wGAFD = U3S·ARFD·frUFDinter·frCovD (43)
wGAFC = U3S·ARFC·frUFCinter (44)
wGAB = U3S· ARB - wGAFD - wGAFC (45)
Evaporation from any water or forest canopy compartment is calculated with equation (46),
where wGxA is the evaporation flux of water from compartment x, θevapxA is a generic
empirical scaling factor, and VPH2O is the vapor pressure of water in Pa at temperature Tx in
Kelvin, estimated with equation (47), derived for this work from the experimental values.
174
wGxA = θevapxA·WSX· VPH2O (46)
VPH2O = 2.875×10-42
· Tx18.2
(47)
For soil compartments equation (46) is modified to equation (48), where SSevap is the
evaporation soil-stress function. The evaporation soil-stress function is estimated with
equation (49), where VFWx is the volume fraction of water in soil compartment x, VFWxcap
is the field capacity of soil compartment x expressed as a volume fraction, and the values of
20 and 3 are empirical parameters. The soil-stress evaporation function was derived for this
work to capture the correct behavior without the use of a detailed hydrological model. The
soil-stress evaporation function quickly falls to essentially zero when the volume fraction of
water in soil is below the field capacity, and increases rapidly at first and then starts to levels
off above the field capacity.
wGxA = θevapxA· WSX· VPH2O·SSevap (48)
SSevap = {arctan(20 · {VFWx - VFWxcap}) + arctan(20 ·VFWxcap)}3 (49)
Empirical scaling factors, θevapxA, are initially estimated by substituting equation (50) for the
term wGxA in equations (46) and (48) and then solving for θevapA(guess), where frUx is the user
defined fraction of rainfall that evaporates for surface compartment x, and other parameters
are set to their annual average values.
wGxA = frUx·wGAx (50)
The annual average fraction of rainfall that evaporates, frUx(fit), is then optimized by altering
the value for θevapxA.
Runoff from any soil compartment x to fresh water W is calculated with equation (51),
where θrunoffxW is a generic empirical scaling factor and SSrunoff is the runoff soil-stress
function. A value for the SSrunoffis calculated with equation (52), where 10 and 2 are
empirical parameters. The soil-stress runoff function has the same general shape as the soil-
stress evaporation fraction, but falls off more slowly below the field capacity and levels off
faster above the field capacity.
175
wGxW = θrunoffxW· SSrunoff (51)
SSrunoff = {arctan(10 · {VFWx - VFWxcap}) + arctan(10 ·VFWxcap)}2 (52)
An initial estimate of the value for θrunoffxWis obtained by substituting the net water
deposition (wGAx - wGxA) for wGxW in equation (51) and solving for θrunoffxWusing annual
average values for rainfall and evaporation. The annual average value for wGxW is then
optimized to fit the user defined runoff fraction for the soil (1 - frUx) by altering the value for
θrunoffxW.
Equation (53) shows the finite difference approximation for the mass balance of water in any
soil compartment x, where VOB is the volume of the soil compartment. For forest soil stem
flows from forest canopy (wGFCB and wGFDB), defined as total net deposition to forest
canopy, are also counted as inputs in the mass balance equation.
VFWx (t+1) = {VFWx (t) · VOB + wGAx - wGxA - wGxW} / VOB (53)
As an initial estimate the volume fraction of water in soil is set to the user defined annual
average volume fraction VFWX(AA), and the field capacity is set to an initial estimate of 60%
of VFWX(AA). The field capacity is then altered to optimize the annual average volume
fraction of water in soil to match user defined annual average.
Net down-stream flow fwGxy from water compartment x to water compartment y is
calculated with equation (54), where θoutflowxy is a generic empirical scaling factor, HTx is the
depth of water compartment x and ARx is the area.
fwGxy = θoutflowxy·HTx·ARx (54)
Equation (55) is the finite difference approximation for the mass balance of water in
compartment x, where fwGzx is the net flow of water from the upstream water compartment
z, which is replaced with the sum of water runoff from soils in the case of the fresh water
compartment.
176
HTx (t+1) = {HTx (t) ·ARx + wGAx + fwGzx - wGxA - fwGxy} / ARx (55)
As an initial estimate the water depth is set to the annual average defined by the user, and an
initial estimate of θoutflowxy is made by assuming steady state conditions (dHTx/dt = 0),
substituting equation (54) into equation (55) and the solving the equation for θoutflowxy using
annual averaged values for all parameters. The water depth is then optimized to the annual
average by altering the value for θoutflowxy.
6.1.5 POC balance.
Primary biological productivity is still the main source of POC in the model, with an
additional contribution from soil runoff in the fresh water compartment as described in
Wania et al.128
Organic carbon fluxes and volume fractions in the terrestrial environment are
not variable in time other than those previously described for the forest canopy, which
remain unaltered. This is a reasonable assumption because the magnitude of intra-annual
variability is quite small in comparison to the amount of organic carbon present. Two major
alterations were made to the POC balance; first, mineralization rates are now a function of
water temperature, with the annual magnitude of mineralization fitted to the user defined
mineralization fraction with a generic fitting parameter. Second, re-suspension intensities are
now a function of water inflow from upstream or runoff, in the case of eroding sediments,
and wind speed, in the case of accumulating sediments. Annual magnitudes of re-suspension
are fitted to user defined re-suspension fractions with a generic fitting parameter.
An initial estimate of all POC fluxes, oG, concentrations of POC in water compartments x,
CPOCx in units of g POC/m3, and the volume fractions of POC in sediment compartments y,
and VFSy, are made using the steady state solution described by Wania et al.128
. Starting
from this initial condition the mass balance of POC is much simpler. For each water
compartment the finite difference approximation for the concentration of POC is shown in
equation (56), where inputs and outputs are as described in Wania et al.,128
but calculated at
every hour of the year instead of only at the beginning of the simulation and δOC is the
density of organic carbon in g/m3.
177
CPOCx (t+1) = {CPOCx (t) · HTx (t) ·ARx + δOC · (inputs – outputs)} / HTx (t+1) · ARx (56)
It was found that the volume fraction of POC in sediments was virtually constant over the
year and therefore solving the mass balance at each hour of the year was unnecessary.
Instead, the volume fraction of POC in sediments was solved using the annual average
values for all optimized parameters and the equations presented by Wania et al.128
at the end
of each year during the optimization process.
Mineralization of POC in water compartment x is described by equation (57), where oGxmiw
is the flux of organic carbon mineralized in units of m3/h, θmiwx is a generic empirical scaling
factor, and Tx and Txavg are the current and average temperatures of compartment x.
oGxmiw = θmiwx· 1.047(Tx-Txavg)
·CPOCx·VOx / δOC (57)
An initial estimate of oGxmiw is made by calculating a value for the annual average,
oGxmiw(AA), using equation (58), where oGxpro(AA) is the annual average primary production
defined by the user, and facOxmiw is the user defined fraction of the primary production that
is mineralized. An initial estimate for θmiwx is made by substituting equation (58) into
equation (57) and solving for θmiwx, using annual average values to calculate the initial
estimate.
oGxmiw(AA) = oGxpro(AA)·facOxmiw (58)
The mineralization in water, oGxmiw, is then optimized to the annual average value,
oGxpro(AA), by altering the value for θmiwx. A similar solution for mineralization in any
sediment compartment xS1 or xS2is followed using equation (59) instead of equation (57),
where VFOxS1/2is the volume fraction of organic carbon in sediment compartment xS1 or
xS2 and using equation (60) instead of equation (58), where oGxS1/2sed(AA) is the annual
average sedimentation flux to sediment compartment xS1 or xS2, and facOxS1/2mis is the user
defined fraction of the sedimentation flux that is mineralized in sediments.
178
oGxS1/2mis = θmisxS1/2· 1.047(TxS-TxSavg)
·VFOxS1/2 · VOxS1/2 (59)
oGxS1/2mis(AA) = oGxS1/2sed(AA)· facOxS1/2mis (60)
Resuspension of sediments is calculated differently for the eroding sediments, xS1, and
accumulating sediments, xS2, of water compartment x. The resuspension flux of organic
carbon from eroding sediments, oGxS1res, is calculated with equation (61), and the
resuspension flux for organic carbon from accumulating sediments, oGxS2res, is calculated
from equation (62), where θresxS1 and θresxS2are generic empirical scaling factors, wGxy is the
downstream flow from water compartment x, and WSX is the wind speed in the overlying
segment X.
oGxS1res = θresxS1· wGxy2 (61)
oGxS2res = θresxS2· WSX2 (62)
Initial estimates for oGxS1res and oGxS2res are made with equations (63) and (64), where
oGxS1sed(AA) and oGxS2sed(AA) are annual average sedimentation fluxes, and facOxS1res and
facOxS2res are the user defined fractions of the sedimentation fluxes that are resuspended.
oGxS1res = oGxS1sed(AA)· facOxS1res (63)
oGxS2res = oGxS2sed(AA)· facOxS2res (64)
The resuspension fluxes are then optimized to the annual average values by altering the
empirical scaling factors θresxS1 and θresxS2.
179
Table 15 - Mass Balance Compartments and Fluxes: their Interdependence and Fitting.
Compartment Volumes
Compartment Fitted to Fitted with
volume of air not fitted, assumed fixed in time
volume fraction of organic aerosol
in air
annual average volume fraction of
aerosol in air
surface production of organic
aerosol
volume fraction of water in soil annual average volume fraction of
water in soil field capacity of soil
depth of water annual average depth of water net downstream flow of water
volume fraction of water in
sediments not fitted, assumed fixed in time
volume fraction of organic carbon
in soils not fitted, assumed fixed in time
concentration of particulate organic
carbon in water
not fitted, allowed to reach a stable
solution consistent with inputs and
outputs
volume fraction of organic carbon
in sediments
not fitted, assumed fixed in time,
but set to a value consistent with
the annual average of inputs and
outputs
Inter-compartment Fluxes
Flux Fitted to Fitted with
inflowing and outflowing air residence time of air path-length
inflowing organic aerosol not fitted, dependant on the flux of
inflowing air
outflowing organic aerosol
not fitted, dependant on the flux of
outflowing air and the volume
fraction of organic aerosol in air
dry particle settling of organic
aerosol
not fitted, dependant on the
volume fraction of organic aerosol
in air
scavenging of organic aerosol by
precipitation
not fitted, dependant on the
volume fraction of organic aerosol
in air and the rainfall flux of water
to the surface
rainfall flux of water to the surface not fitted, user defined pattern of
variation and annual magnitude
evaporation of water from surface
user defined fraction of rainfall
flux to the surface, dependant on
volume fraction of water in soil
empirical scaling factor
stemflow of water from forest
canopy to forest soil
not fitted, dependant on rainfall
interception and evaporation from
forest canopy
runoff of water from soil to fresh
water
user defined fraction of rainfall
flux to the surface, dependant on
volume fraction of water in soil
empirical scaling factor
production flux of organic carbon
to water
not fitted, user defined pattern of
variation and magnitude
mineralization flux of organic
carbon in water
user defined fraction of production,
dependant on concentration of
organic carbon in water
empirical scaling factor
sedimentation flux of organic
carbon to sediments
not fitted, dependant on
concentration of organic carbon in
water
180
resuspension of organic carbon
from eroding sediments
user defined fraction of
sedimentation, dependant on net
downstream flow of water
empirical scaling factor
resuspension of organic carbon
from accumulating sediments
user defined fraction of
sedimentation empirical scaling factor
mineralization flux of organic
carbon in sediments
user defined fraction of
sedimentation empirical scaling factor
erosion of organic carbon from soil
to fresh water
not fitted, dependant on runoff of
water from soil to fresh water
downstream transport of sediments
not fitted, dependant on volume
fraction of organic carbon in
sediments
lateral intra-basin transport of
sediments
not fitted, dependant on volume
fraction of organic carbon in
sediments
burial rate of sediments
not fitted, dependant on volume
fraction of organic carbon in
sediments
181
6.2 Meteorological Data and Climatic Scenario Definition.
Objectives. The objectives of the climatic scenario definition were to create a data set of all
environmental parameters required at a resolution of one hour, to ensure that the parameters
were internally consistent, and that they represent realistic weather patterns compatible with
the parameterization of the model. A perfect match to the parameterized environment was
not required however, because the overall objective of the current work is to investigate the
time resolution of the environmental parameters, not necessarily to model the conditions of a
specific environment. Additionally, the aim of the current work is to investigate the effects of
environmental variability on contaminant fate, not to construct a detailed climatic model, so
the models used to process the environmental data are simplistic.
6.2.1 Data source and selection.
Meteorological data were sourced from the Canadian Climate Norms Database.155
The data
is available online for free from Environment Canada, has been quality checked and is
available at hourly resolution. It was found that the climate norms of the city of Prince
George, British Columbia, Canada compared favorably to the default meteorological
parameterization in CoZMoMAN.34
Monthly resolved meteorological data were obtained for
1943 to 2009 and temperature and precipitation were compared to monthly climate norms. It
was determined that data from the year 2000 had the best fit to the Prince George climate
norms and hourly resolved meteorological data were obtained for that year. The year 2000
was a leap year, but the CoZMoMAN model uses a 365 day year so the data for February 29
was discarded. Hourly resolved data were obtained for air temperature, wind speed,
visibility, air pressure, relative humidity and records of qualitative visual weather
observations. Precipitation amounts for the same year were obtained from the same source at
daily resolution.
6.2.2 Heat Transfer Model.
In addition to air temperatures the model requires the input of terrestrial and marine
temperatures. The temperatures in these compartments generally change more slowly than in
air, so the pattern of air temperature variation should not be used directly. A simple model
182
was used to estimate heat transfer from the air to a unit of soil and water instead. For heat
transfer to soil the approximate mass of a 1×1×1 meter cubed block of a dry soil with a
porosity of 0.5 was calculated to be 1.45×106 g, which when combined with the approximate
heat capacity (~0.8J/gK) and thermal conductivity (~1.5W/mK) of dry soil gives the
approximate model in equation (65), where ΔTsoil is the change in soil temperature per hour,
Tsoil is the current temperature of soil and Tair is the temperature of air.
ΔTsoil = 4.66×10-3
·(Tair - Tsoil) (65)
The model was applied to the collected air temperatures with a time step of one hour, at each
step the average air temperature over that time step was used to calculate the change in soil
temperature expected for the next step. An initial guess was made of the soil temperature by
setting it to the initial air temperature, and then the model was run until the final soil
temperature of each year matched the first soil temperature of the same year to within two
decimal places. An identical procedure was used to generate water temperatures with
equation (66), derived from mass, heat capacity and thermal conductivity values of 1.0×106
g, 4.18J/gK and 0.6W/mK respectively. See Figure 30 for the pattern of air, soil and water
temperatures derived this way.
ΔTwater = 5.17×10-4
· (Tair - Twater) (66)
6.2.3 Insolation Model.
A simple model of solar zenith angle was use to calculate the maximum insolation, which
was subsequently corrected for cloud cover, gas scatter and reduced visibility related to
precipitation using the meteorological data collected, following accepted methods.158
Equations (67) through (71) are used to calculate the approximate uncorrected insolation at
the surface, where JH is the Julian hour of the year starting from zero, JD is the Julian day of
the year starting from zero, fy is the year fraction in radians, D is the declination in radians,
HA is the hour angle in radians, ZA is the zenith angle in radians, LT is the latitude in
decimal degrees, and IN is the pattern of insolation at the surface as a fraction of maximum
insolation in the range 0-1.A latitude of 57.52 decimal degrees was used, corresponding
approximately to the middle latitude of the Baltic Sea drainage basin.
183
fy = 2 · π · JH / 8760 (67)
D = 0.006918-0.399912 · cos(fy)+0.070257 · sin(fy)-0.006758 · cos(2·fy)
+0.000907 · sin(2·fy)-0.002697 · cos(3·fy) + 0.00148 · sin(3·fy) (68)
HA = 1/12· π · {JH - JD · 24 - 12} (69)
ZA = 0.5·π-arcsin{sin(LT·π/180) · sin(D)+cos(LT·π/180) · cos(D) · cos(HA)} (70)
IN = cos(ZA) if > 0 else 0 (71)
Gas scattering of sunlight as it passes through the atmosphere is calculated with equations
(72) and (73), where PL is the path length of light through the atmosphere, AP is the
atmospheric pressure in kPa, and GC is the gas scattering correction to insolation, a value
between -1 and 1. All values of GC less than 0 coincide with times where the insolation IN is
set to zero.
PL = {1.002432 · cos(ZA) · cos(ZA) + 0.148386 · cos(ZA) + 0.0096467}
/ {cos(ZA) · cos(ZA) · cos(ZA) + 0.149864 · cos(ZA) · cos(ZA)
+ 0.0102963 · cos(ZA) + 0.000303978} (72)
GC = 1.021-0.084· {PL · 949· AP · 0.000001+0.051}0.5
(73)
Scattering of sunlight from atmospheric water vapor was corrected for using equations (74)
through (76), where AH is the absolute humidity in Pascals, AHavg is the annual average
absolute humidity, TA is the air temperature, RH is the relative humidity in percent, WP is
the water path, and WC is the water scattering correction to insolation.
AH = (2.875×10-42
· TA18.2
) · RH / 100 (74)
WP = 1.42· {1 + (AH - AHavg) / AHavg} (75)
WC = 1 - 0.077 · (PL · WP)0.3
(76)
Aerosol absorbance and scattering of sunlight was corrected for using equation (77), where
AC is the aerosol correction to insolation.
184
AC = 0.935PL
(77)
Weather observations from the Canadian Climate Norms database are composed of either a
cloud cover observation, or - if there is precipitation -an observation of the precipitation type
and amount. A complete cloud cover dataset was created by filling in precipitation
observations with the last cloud cover observation. Cloud cover observations are Clear,
Mainly Clear, Mostly Cloudy and Cloudy, which correspond to percent approximate
coverage fractions of 0, 0.33, 0.67 and 1. It was assumed that complete cloud cover reduced
insolation at the surface by 40%, which was used to scale the percent cloud cover, as shown
in equation (78), where CC is the cloud cover correction to insolation and FC is the fraction
of cloud cover.
CC = 1 - 0.4· FC (78)
Reduced insolation due to effect of precipitation obscuring sunlight was calculated using the
weather observations and the visibility. Whenever the weather observation recorded any
precipitation, or any other weather effect that was not cloud cover such as fog, equation (79)
was used to calculate the effect, where PC is the precipitation obscurity correction to
insolation, and VI is the visibility in kilometers, in the range of 0.1km to 80.5km with a
mean value of 38.25km.
PC = 1 / exp(1 / VI) (79)
Effective insolation is finally calculated with equation (80), and is a value between 0 and 1
representing the fraction of maximum theoretical insolation at the surface.
INeff = IN · GC · WC· AC· CC · PC (80)
Figure 31 shows the annual pattern of variation for effective insolation.
6.2.4 Precipitation Model.
Precipitation rates at hourly resolutions were obtained from the daily resolution data by using
the weather precipitation observations to assign fractions of the daily total precipitation to
specific hours of each rainfall event. Precipitation observations were consolidated into
different intensities of rain and snow, with rare observations such as ―Ice Pellets‖ and
185
―Freezing Rain‖ being binned into comparable rain or snow intensities. The relative
intensities of each type and precipitation observation were estimated by counting the number
of hours each observation had in each day and multiplying by the rainfall amount for that day
and summing these amounts for the whole year. These relative intensities were then all
divided by the lowest intensity precipitation type (Snow Showers) to get the relative
precipitation intensities (RPI) shown in Table 16. Using the weather observations the length
of each precipitation event was estimated. It was possible, likely even, that the precipitation
events included hours from two or more days; to account for this the total precipitation fall
during each event was calculated using equation (81), where Pevent is the precipitation
occurring in a single event, Pi is the total precipitation occurring during day i and ti is the
number of hours of the precipitation event that fell on day i.
Pevent = ∑Pi·ti/ 24 (81)
The precipitation that occurred during each hour j of the event Pj was then calculated with
equation (82), where RPIj is the relative precipitation intensity of each hour j during the
precipitation event.
Pj = ∑ {Pevent·RPIj} / ∑RPIj (82)
All of the daily precipitation amounts were converted into hourly amounts and then
converted into fractions of the annual precipitation that fell during each hour; these are
shown in Figure 32.
6.2.5 Primary Productivity Model.
Hourly resolved primary biological productivity was estimated using hourly resolved water
temperatures (TW) and effective insolation (INeff) as input parameters for a simple biomass
growth model. INeff was normalized using equation (83).
INeff(norm) = INeff / ∑INeff (83)
Equation (84) calculates the biomass growth rate (primary biological productivity) BP where
BM is current biomass, NU is the current available nutrients and TWavg is the annual average
water temperature. In this equation the first term represents to maximum possible biomass
186
growth rate, the second term represents the uptake efficiency of the nutrients and the third
term represents the temperature dependence of the biomass growth rate.
BP = {INeff(norm)·BM ·125} · {NU / (NU + 75)} · {1.066(TW-TWavg)
} (84)
Equation (85) shows the death rate of biomass DR, where the first term represents the annual
average death rate of biomass (31.18 is a fitted value to give the model the desired pattern
and range), and the second term represent the temperature dependence of the death rate.
DR = {BM · 31.18 / 8760} · {1.047(TWavg - TW)
} (85)
The biomass model is run at an hourly time step using the finite difference approximation in
a closed system, meaning there is no net loss or gain of either nutrients or biomass other than
conversion of one to the other. Equations (86) and (87) show the mass balance calculations
for biomass and nutrients at each time step.
BM(t+1) = BM(t) + BP - DR (86)
NU(t+1) = NU(t) - BP + DR (87)
As an initial condition the biomass was set to a value of 1 and the nutrients set to a value of
99, which constrains the system to a total biomass plus nutrient sum of 100. The simulation
was run for consecutive years until the final biomass amount at the end of the year matched
the initial biomass amount of the same year within 0.01 units. The pattern of primary
biological production was normalized by dividing by the sum of the total annual production;
this pattern is shown in Figure 33.
6.2.6 Atmospheric OH Radical Concentration Model.
An hourly resolved pattern of variation in the atmospheric OH radical concentration was
derived by assuming that it follows the same pattern as effective insolation, INeff.
187
Figure 30 - Hourly Resolved Air, Water and Soil Temperatures in Kelvin.
Figure 31 - Hourly Resolved Insolation and Corrected Insolation as Fraction of
Maximum.
188
Table 16 - Relative Precipitation Intensities.
Precipitation Type Relative Precipitation Intensity (RPI)
Moderate Rain 19.05
Moderate Rain Showers 13.02
Rain 8.47
Rain Showers 5.49
Drizzle 3.71
Moderate Snow 12.87
Snow 3.09
Moderate Snow Showers 3.07
Snow Showers 1
Figure 32 - Hourly Precipitation Amounts as a Fraction of Annual Precipitation.
189
Figure 33 - Primary Biological Production Rate as a Fraction of Annual Production.
190
Figure 34 - Time Averaging of Environmental Parameters Soil Temperature (A),
Marine Water Temperature (B), Marine Primary Biological Production Rate (C),
Atmospheric OH Radical Concentration (D), Rainfall (E) and Wind Speed (F).
191
6.3 Derivation of Empirical Relationship for Predicting Temperature Dependence.
Simulations to generate chemical space plots have previously assumed fixed temperature
dependence across the range of log KAW and log KOA, which is a reasonable approximation
but unsatisfying. To make an empirical prediction useful for application to chemical space
calculations two requirements should be met; first the enthalpies of phase change should be
predicted using only the data available for hypothetical chemicals, namely the partition
coefficients, and second the enthalpies of phase change should be thermodynamically
consistent with each other. MacLeod et al. investigated the validity of predicting the enthalpy
of vaporization from the vapor pressure of a pure substance,159
a similar method is employed
here to create predictions for other phase changes. Enthalpies of phase change for air-water
and octanol-air partitioning were taken from Mintz et al.160
and experimental values for the
partitioning coefficients were found for as many of the same chemicals as possible from the
EPI Suite experimental database and Abraham et al. 41,161
Enthalpies of phase change for
octanol-water partitioning were calculated by applying a thermodynamic cycle for chemicals
where the two datasets of Mintz et al. overlapped, and missing partition coefficients were
similarly calculated when the other two values were available.
The enthalpies of phase change were then plotted versus the logarithms of the partition
coefficients, as shown in Figure 35. Note that the slopes of these three plots are similar.
Regression statistics are quite good for octanol-air partitioning and moderate for air-water
partitioning but are poor for octanol-water partitioning, in addition to the fact that the
enthalpies of phase change for octanol-water partitioning have a narrow range of values in
comparison to the other two regressions. To make the best use of all of the data, and to meet
the requirement for thermodynamic consistency, the regressions were locked into a
thermodynamic cycle by using only one slope to fit all three datasets, and fitting only the
intercepts of the octanol-air and air-water partitioning, then calculating the intercept of
octanol-water partitioning by summing the other two intercepts. The three remaining
parameters were fitted by minimizing the error of the regressions weighted by the spread of
the enthalpy values, shown in equation (88), where RMSEXY is the root mean squared error
192
of the regression for partitioning between phase X and Y, and SDXY is the standard deviation
of the enthalpies of phase change for partitioning between phases X and Y.
weighted error = RMSEOA· SDOA + RMSEAW· SDAW +RMSEOW· SDOW (88)
Equations (89) through (91) show the regression equations and statistics used to calculate the
enthalpies of phase change in kJ/mol at each point in the chemical space of log KOA and log
KAW.
ΔHOA = -7.87 · log KOA- 11.5 (89)
(n=120, R2=0.93, RMSE = 5.92)
ΔHAW = -7.87 · log KAW + 31.0 (90)
(n=323, R2=0.66, RMSE = 10.0)
ΔHOW = -7.87 · log KOW + 19.5 (91)
(n=90, R2=0.46, RMSE = 7.93)
193
Figure 35 - Correlation of Enthalpies of Phase Change in kJ/mole with the Logarithm
of their Corresponding Partition Coefficients.
194
6.4 Definition of Environmental Half-lives.
The Stockholm Convention on POPs recommends environmental half-life cut-off values for
environmental persistence of 2 days in air, 2 months in water, and 6 months in soil and
sediment.5 Because anaerobic degradation in sediments is generally considered to be slower
than degradation in soils, for most chemicals, a ratio of 1:3:9 was preferred for
water:soil:sediment environmental half-lives instead of the 1:3:3 ratio of the Stockholm
Convention.162
Environmental half-lives were calculated similarly to the methods presented
by Arnot et al.162
Equation (92) was used to calculate the half-life in hours for air, water, soil
and sediment, where P is the phase correction with values of -2.5, -1, -0.5 and 0.0 for air,
water, soil and sediments respectively, and S is -0.5 for labile chemicals and +0.5 for
persistent chemicals.
HL = 8760 · 10(P+S)
(92)
The CoZMoMAN model assumes environmental half-lives are input at 25°C, but the half-
lives calculated with equation (92) are the desired half-lives at average environmental
temperatures. To account for this the half-lives of surface media were temperature corrected
using the assumed temperature dependence of 30kJ/mol from the average annual temperature
of each compartment to 25°C. A rate constant for reaction with atmospheric OH radicals is
required for the degradation in air; this was calculated from the half-life using the annual
average atmospheric OH radical concentration and temperature correcting the value similarly
to the HL in surface media using an assumed temperature dependence of 10kJ/mol for
degradation in air and the annual average air temperature.
195
Figure 36 - Annual Pattern of Variation in the Molar Amount of Chemical in Air,
Cultivated Soil, and Fresh Water for Hypothetical Labile Chemicals with the original
CoZMoMAN Model (Red Lines) and the New Model (Blue Lines) with Monthly
Resolved Environmental Parameters.
196
Figure 37 - Annual Pattern of Variation in the Molar Amount of Chemical in Air,
Cultivated Soil, and Fresh Water for Hypothetical Persistent Chemicals with the
original CoZMoMAN Model (Red Lines) and the New Model (Blue Lines) with
Monthly Resolved Environmental Parameters.
197
Table 17 - Average, Minimum and Maximum Characteristic Travel Distances of
Persistent and Labile Chemicals at Hourly, Daily, Weekly, Monthly and Seasonal
Environmental Parameter Resolutions, in Kilometers.
Annual CTD (km)
a Hourly CTD (km)
b
Resolution Persistent Labile Persistent Labile
Hourly 2845 (2429-3141) 492 (454-512) 8531 (7147-8875)c 845 (673-909)
c
Daily 2912 (2488-3215) 475 (435-496) 8114 (7087-8330) 794 (661-833)
Weekly 2738 (2385-3000) 437 (402-455) 7972 (7135-8135) 788 (687-814)
Monthly 2605 (2316-2823) 397 (370-411) 7184 (6059-7279) 709 (633-728)
Seasonal 2605 (2420-2746) 331 (318-338) 4052 (3882-4077) 405 (395-408)
aCTD calculated from annual averaged wind speeds, masses, depositional and
degradative fluxes. bCTD calculated at each hour and then averaged over the year.
cCTD calculated with a 24 hour running average instead of hourly values.
198
Figure 38 - Extrapolation to Steady State Characteristic Travel Distance for Labile
Chemicals (A) and Persistent Chemicals (B).
199
Chapter 5
Creating a Fragment-Based QSAR with
Iterative Fragment Selection (IFS)
Trevor N. Brown, Jon Arnot, Frank Wania
Manuscript in Preparation.
200
1 Introduction National and international regulatory programs and a global treaty mandate chemical hazard
and risk assessments to protect human and environmental health. Chemical assessments
require monitoring, exposure, toxicity and chemical property information such as the
octanol-water partition coefficient (KOW) and reaction half-lives (HLs). Compared to the
number of chemicals legislated for assessment there is a paucity of available measured data,
making it necessary to use theoretical approaches to estimate chemical properties, exposure
and toxicity.26,146,163
Chemical distribution properties and HLs can be predicted from 2D
structure using quantitative structure-property relationships (QSPRs) or quantitative
structure-activity relationships (QSARs) such as those implemented in EPI SuiteTM41
and
elsewhere. Existing fragment-based QSA(P)Rs generally require some initial expert
knowledge to identify molecular fragments that influence the chemical property of interest.
More sophisticated methods for predicting chemical properties have also been developed.
For some physical chemical properties, such as KOW, multiple prediction tools are available;
however, the variability and uncertainty between predictions can be substantial.164
For other
chemical parameters, such as HLs, the availability of high throughput screening-level tools is
very limited.41
There is thus a need to develop, evaluate and compare more QSARs to
address data gaps, identify possible sources of uncertainty in training and testing datasets and
in QSAR methodologies, and quantify and reduce uncertainty in applying these predictive
methods for chemical assessments.
Metabolic biotransformation HLs are important chemical parameters for calculating
ecological and human exposure to chemicals.31
In general, the influence of metabolic
biotransformation on human exposures is much greater than the influence of chemical
partitioning properties.25
Currently, there is only one publicly available QSAR for whole
body biotransformation HLs in vertebrates (fish).41,49
Experimental data and expert
knowledge for whole body biotransformation HLs are limited and, due to the complex and
variable nature of this process, also highly uncertain.165
Expert knowledge and ―rules of
thumb‖ of environmental (microbial) degradation mechanisms and recalcitrant chemical
groups have been built into chemical class specific models, (see review by Pavan and
Worth166
) and more general models such as the BIOWIN set of models in EPI SuiteTM.48
Mechanistic knowledge of environmental microbial degradation was initially applied in the
201
development and evaluation of a fish biotransformation HL-QSAR and the results suggest
some overlap in the importance of selected chemical descriptors (fragments) and
biotransformation mechanisms in microbes and fish.41,165
The success in extrapolating this
knowledge is encouraging; however, it is possible that other mechanisms and molecular
fragments contribute to the degree of chemical biotransformation, for which there is no
current ―expert recognition‖. A tool that automatically generates hypotheses for mechanisms
and molecular fragments related to a chemical property of interest could provide new and
deeper insights into QSA(P)Rs and expand expert knowledge.
The primary objective of this research is to develop a new automated tool for the prediction
of chemical properties. The overall method of fragment generation, dataset splitting, and
model selection is referred to as Iterative Fragment Selection (IFS). This new QSA(P)R
method does not require pre-existing expert knowledge (or ―rules-of-thumb‖) on structural
information related to a specific chemical property. 2D fragments are generated
automatically from the dataset and the method iteratively selects fragments and QSARs
through an automated process of fragment selection and model testing. The secondary
objective of this research is to apply the new method to the development and evaluation of a
new set of QSARs for predicting biotransformation HLs in fish from chemical structure. Two
new sets of biotransformation HL-QSARs are developed and evaluated and then compared
with an existing fragment-based HL-QSAR.41,49
The first new HL-QSAR (referred to as IFS-
HLN) uses the dataset splitting methods developed herein and the second new HL-QSAR
(referred to as IFS-EPI) relies on nearly identical datasets used in the development and
evaluation of the HL-QSAR included in the BCFBAF model in EPI SuiteTM.41,49
Similarities and differences between the QSARs are discussed with a focus on model
performance and fragment selection.
2 Methods
2.1 Data Processing
Data processing for the generation of descriptors, model selection, cross validation and
dataset splitting was primarily performed in python 2.6, with the mathematic and scientific
202
computing modules numpy and scipy employed for multiple linear regression (MLR) and
other mathematical functions.
2.2 Overview of the Method
The first step of the method, and one of the most important, is the generation of fragments;
because all subsequent steps rely on the fragment counts for either model selection or to
calculate of chemical similarity scores used to split the dataset into training, external
validation, and cross-validation datasets. All possible fragments contained in the dataset are
generated automatically to be used as the pool of descriptor during model selection. No pre-
selection or sorting of the pool of fragments is done except for those steps required to ensure
that the final model is stable, with no collinear fragments and minimal over-fitting. A simple
set of selection rules is defined for this purpose based strictly on fragment counts rather than
chemical structure. The model selection algorithm is essentially composed of nested
iterations of fragment selection and removal, and the overall method of fragment generation,
dataset splitting, and model selection is referred to as Iterative Fragment Selection (IFS).
2.3 Generation of Descriptors
With the exception of the molecular weight (MW), the pool of descriptors drawn from for
model selection is composed entirely of fragments created from the training dataset. All of
the chemicals in the training dataset are fragmented by breaking bonds according to the
fragmentation rule: all single and aromatic bonds are broken, with the exception of those
bonds involving hydrogen. The resulting single unit fragments lack chemical context which
in some cases would make their valence and number of attached hydrogen atoms ambiguous
if converted to standard SMILES notation. Therefore, fragments were converted to a custom
variation of SMILES notation which includes explicit hydrogen atoms and differentiates
between fused aromatic atoms bonded only to other aromatic atoms and aromatic atoms with
a single bond to hydrogen or another group. To accommodate string matching the custom
SMILES notations are made unique according to the original method defined by Weininger
et al.,168
but with atom priority based on the number and type of connections instead of
counts of hydrogen and non-hydrogen attachments.
203
Larger, more complex fragments are created by recursively combining the single unit
fragments with their neighbours. The maximum number of single unit fragments contained
in a complex fragment (the recursion depth) has been set to eight, to allow for fragments that
consist of a six membered ring with the positions of two substituents defined. Larger
recursion depths suffer from diminishing gains with regards to the total calculation time, as
the number of unimportant fragments that must be sorted through increases more rapidly
than the number of useful fragments. As a result of the recursive combination algorithm the
complex fragments are non-exclusive. For example; if a fragment containing two
neighbouring aromatic carbons with attached hydrogen atoms (HccH) is searched for in
benzene the count will be six, rather than three. This means that some fragments are not
additive, causing the fragment count to sometimes be non-linearly related to the number of
atoms that are covered by the fragment, an undesirable property for a fragment-based QSAR.
Equation (93) defines a metric measuring the degree of additivity, where A is additivity, C is
the number of atoms that are contained in exactly one copy of the fragment summed for all
chemicals in the training dataset, T is the number of atoms in a single fragment and Nsum is
the sum of fragment counts for all chemicals in the training dataset.
A = C / {T · Nsum} (93)
A adopts values between 0 and 1, where a value of 0 refers to a completely non-additive
fragment and 1 a completely additive fragment. Single unit fragments always have an A of 1.
For model selection MW is treated as a single unit fragment.
There are many fragments in the pool of descriptors that are perfectly cocorrelated;
depending on the dataset, at a recursion depth of eight approximately 90% of the complex
fragments added to the fragment pool have either identical counts or are perfectly collinear
with other fragments. All fragments that have identical counts or are perfectly collinear are
grouped together into aggregate fragments and one of the fragments is selected to represent
the entire aggregate. Selection is based on the following prioritized criteria: fragments with
fewer total counts in the training dataset are preferred, fragments with perfect additivity (i.e.
A = 1) are preferred; fragments containing fewer single units are preferred; fragments with
204
higher additivity are preferred; fragments containing fewer atoms are preferred. Fragments
are filtered by each criterion sequentially, only moving to the next criterion in the case of a
tie, until only a single fragment remains.
Fragments are divided into two classes which are treated differently in the model selection.
Fragments contained in three or more chemicals are referred to as frequent fragments.
Fragments contained in only one or two chemicals are referred to as rare fragments and are
subject to special restrictions during the model selection.
2.4 Model Selection
Model selection proceeds according to an Iterative Fragment Selection (IFS) algorithm.
Fragments are added iteratively to the MLR starting from a model containing only the
intercept (the average value of the observations). Forward selection and backward removal
of fragments are based primarily on improvements to the goodness of fit (GoF) to the
training dataset, with the fragments that produce the largest or smallest improvement in GoF
being selected or removed first. Forward selection considers all fragments in the fragment
pool that are not already in the current model. A selected fragment may replace one or more
fragments already included in the model if selection rules prohibit the fragments from being
included together. Scheme 1 shows the general algorithm for model selection.
Scheme 1: Model Selection Algorithm
1. one forward selection from the pool of single unit frequent fragments
2. iterative backwards removal of rare fragments
3. repeat steps 1-2 until no further improvement in GoF is possible
4. one forward selection from the pool of single unit rare fragments
5. iterative backwards removal of rare fragments
6. repeat steps 4-5 until no further improvement in GoF is possible
7. repeat steps 1-6 until no further improvement in GoF is possible
205
8. one forward selection with replacement from the pool of all frequent fragments
9. iterative backwards removal of rare fragments
10. iterative backwards removal of frequent fragments
11. repeat steps 8-10 until no further improvement in GoF is possible
12. one forward selection with replacement from the pool of all rare fragments
13. iterative backwards removal of rare fragments
14. iterative backwards removal of frequent fragments
15. repeat steps 12-14 until no further improvement in GoF is possible
16. repeat steps 8-15 until no further improvement in GoF is possible
Steps 1 through 7 usually proceed as a simple stepwise MLR with addition of rare fragments
occurring only rarely. The resulting model from steps 1 through 7 is referred to as the base
model. In subsequent backwards removal steps (steps 9 ,10 , 13, and 14) fragments in the
base model may be removed like any other fragment. Steps 8 through 16 add and remove
fragments from the base model; no fitting is done on residuals.
2.5 Goodness of Fit Metric
The GoF metric used for model selection is the Akaike Information Criterion, corrected for
dataset size (AICC).169
AICC penalizes model complexity and was found to limit the number
of fragments selected to a favorable fraction of the number of chemicals in the training
dataset (less than a 1:5 ratio), as suggested by Dearden et al.167
The assumptions on which
the AICC is based are well suited to fitting measured, and therefore uncertain, data with
fragments derived from two-dimensional structures; namely that a model which perfectly fits
the data does not exist and instead the goal is to find the simplest model which neglects the
least information.170
206
2.6 Fragment Selection Rules
During model selection the MLR was found to be prone to instability and collinearity
between multiple fragments, likely due to the nature of the descriptors; a sparse matrix of
integers. Fragment selection rules were created to prevent combinations of fragments that
lead to instability and collinearity from being included together during model selection. In
general fragment selection rules are intended to exclude instances of fragment coincidence;
meaning fragments which are not perfectly collinear but which are present in the same
chemicals in the training dataset. Coincident fragments create the possibility of additively
combining fragment counts to create perfect collinearity between three or more fragments,
and over-fitting and instability are also possible. Single unit fragments are of course
coincident with many complex fragments and therefore applying selection rules to single unit
fragments would prevent most other fragments from being included in the model selection,
so fragment selection rules are relaxed for single unit fragments.
Four different types of fragment coincidence are prevented by fragment selection rules.
Selection rules preventing perfect coincidence, asymmetrical coincidence and imperfect
coincidence are applied to all complex fragments. General coincidence is prevented when
one of a pair of fragments is a complex fragment and the other is a single unit fragment. No
fragment selection rules are applied if a pair of frequent, single unit fragments is considered.
Perfect coincidence occurs when two fragments are present in all of the same chemicals in
the training dataset. Asymmetrical coincidence is when one fragment of a pair of fragments
occurs only in chemicals that contain the other fragment, but not in all of the chemicals.
Imperfect coincidence is when two fragments are perfectly coincident, except that each
fragment is also present in one additional chemical. General coincidence is when there are
fewer than three chemicals which contain one of the fragments but not the other. As a
general rule then, for any pair of fragments in a fragment-based model there should be three
or more chemicals which contain one of the fragments and not the other, preferably with
both fragments each having some independent occurrences.
207
Further restrictions are placed on adding rare fragments during model selection. This is
because having many rare fragments present was found to increase the instances of model -
fitting and instability, and because the regression coefficients are very sensitive to any errors
in the observed values of the one or two chemicals in which the fragment is present. First the
GoF must be significantly reduced by including the rare fragment; this has been defined as
reducing the AICC by a value of one or more. Second the standard error of the regression
coefficient for the rare fragment must be smaller than the value of the regression coefficient.
Finally the absolute residual for the chemical containing the rare fragment, or for at least one
of the chemicals if the rare fragment is contained in two chemicals, must be greater than
0.674 times the standard error of the MLR (the student‘s t-statistic corresponding to the
upper 50% of the distribution of residuals) if the rare fragment is removed. These restrictions
must be met at all times, not just when the rare fragment is selected, and rare fragments are
removed in steps 2, 5, 9 and 13 of scheme 1 if they no longer meet any of these restrictions,
even if this causes the GoF to be poorer.
2.7 Cross Validation
During model selection variations of k-fold cross validation are applied. Internal training
datasets are all combinations of k minus m folds (k is the total number of clustered sub-
groups of chemicals and m is the number of groups used for internal cross validation) with
the internal validation datasets composed of the remaining m folds in each case. Various
values of k and m are used, but in all cases a separate model is created by MLR on each
internal training dataset using the same set of fragments and the final model presented is the
average of the models developed on the individual internal training datasets. Increasing k
means there are more fold combinations possible and more separate models are created
during the model selection. Increasing m means that each model is fitted on a smaller internal
training dataset and applied to a larger internal validation dataset. Three different value
combinations are used; k = 6 and m = 2,3,4 so the same folds are used in all three cases but
distributed so that one third, one half and two thirds of the chemicals respectively are in the
internal validation datasets.
208
As multiple models are created simultaneously an overall AICC value is calculated to test the
GoF. Each model is applied to the full training dataset and the residual sum of squares
(RSOS) is calculated. These RSOS values contain the residuals for the training dataset points
used to fit the model as well as the data points from the corresponding internal validation
dataset in some ratio defined by k and m. Increasing m has the effect of weighting the GoF
metric towards measuring the predictive power of the regressions and away from measuring
the fit on the internal training datasets. The RSOS value are all summed and then divided by
the number of models to obtain an average RSOS which is then used in the standard AICC
formula.170
Division of the full training dataset into the k folds was done by clustering similar chemicals
together into the same fold. This was found to promote the selection of fragments that are
more widely applicable across different chemical classes. Chemical similarity was
determined by the similarity of the observed values and by the similarity of fragment counts,
which were quantified separately using Tanimoto coefficients to describe the degree of
similarity,171
and the total similarity taken as the product. Fragment counts and observed
values were normalized (as detailed below) and then Tanimoto coefficients were calculated
for each pair of chemicals.171
The set of fragment counts compared for each pair of
chemicals is the set of all fragments contained in one or both of the chemicals.
Fragment counts were normalized according to equation (94) where Fij is the normalized
fragment count, Nij is the fragment count for fragment j in chemical i and Sj is the number of
single unit fragments contained in fragment j.
Fij = {Nij / Sj}0.5
+ 1 (94)
The normalization gives greater weight to smaller fragments (those containing fewer single
unit fragments), and for large fragment counts the differences between similar counts are
smaller. If a fragment is present in one of a pair of chemicals but not in the other then the
value of Fij for the missing fragment is set to Fij = 0 when calculating the Tanimoto
coefficient.
209
Observed values were normalized according to equation (95), where Vi is the normalized
observation for chemical i, Oi is the observed value for chemical i, and Omean is the mean of
all observations.
Vi = {1/3} · {Oi - Omean} / max( |Oi - Omean| ) + 0.5 (95)
This equation normalizes the observed values to between 1/6 and 5/6 with the mean value of
the dataset centered at 1/2. The Tanimoto coefficient is calculated twice, once for Vi and
once for 1-Vi, and the results are averaged so that approximately the same similarity is
calculated for the same deviation between observed values regardless of whether the values
are at the top or the bottom of the range of observed values.
Normalized fragment counts and observed values are used to calculate two separate
Tanimoto similarity coefficients and the total similarity coefficient between any pair of
chemicals is taken as the product of these. Chemicals are then divided into k folds according
to scheme 2. The aim of Scheme 2 is to seed each fold with chemicals which are the most
similar to others in the dataset and then to fill the fold with chemicals that the most similar to
the seeded chemicals.
Scheme 2: Division of Training Dataset into k Folds
1. Chemicals not yet assigned to a fold are sorted by the sum of their similarity
coefficients with each other chemical that also remain unassigned.
2. The chemical that is the most similar to other chemicals that are not yet assigned to a
fold (maximum value from step 1) is added to an empty fold.
3. Chemicals not yet assigned to a fold are sorted by their similarity to the chemicals
that have already been assigned to the fold, with priority given to chemicals with the
fewest low similarity scores.
4. The chemical with the maximum value from step 3 is added to the fold.
5. Steps 3-4 are repeated until the correct size for the fold is reached.
6. Steps 1-5 are repeated until k folds have been assigned chemicals.
210
2.8 External Validation
Division into new training and testing datasets was done by sorting all chemicals into a large
number of small folds, each containing a few similar chemicals, and then selecting some
chemicals from each fold to be placed into the training dataset. Dividing the dataset into
folds was done exactly as described in the cross validation section, with the fold size set to a
value of 6 and the number of folds set to the number of chemicals divided by 6. Small folds
are best for this procedure because it ensures the chemicals in each fold are strongly similar
and the chemical classes in the dataset will therefore be well represented in both the training
and testing datasets. A value of 6 was chosen for fold size because it is the smallest number
that satisfies three criteria; first that three or more chemicals from each fold must remain in
the training dataset to ensure any unique fragments in a chemical class are frequent
fragments in the training dataset, second that at least one chemical from each fold is added to
the training dataset, and finally because it is evenly divisible by three, so that a 2:1 training
to testing dataset size ratio is obtained. When selecting chemicals to be placed in the training
dataset the first chemical selected is that with the maximum summed similarity with other
chemicals in the same fold. The second chemical selected is that with the maximum summed
similarity with other chemicals remaining in the same fold divided by the similarity with the
first chemical selected. This method ensures that from each fold of six chemicals the most
representative chemical is chosen for the testing dataset, and the second chemical, while also
representative, is as dissimilar from the first selected chemical as possible, to ensure
diversity in the testing dataset. The distribution of chemicals into the various training and
validation datasets is summarized in Table 23.
Aggregate fragments have special restrictions applied when generating fragment counts for
external validation datasets. Individual fragments that make up an aggregate fragment are
likely collinear because they are part of a larger structure, and when an aggregate fragment is
selected by the model selection algorithm this larger structure is the information that
contributes to the GoF. Aggregate fragments are only counted in chemicals in the external
testing dataset if the larger structure is also present. This is done by applying two restrictions;
211
first all of the individual fragments in the aggregate fragment must be present in a chemical
and in the same count ratios as in the training dataset. Additionally, any fragments which are
present in all of the same chemicals in the training dataset as the aggregate fragment must
also be present, though there is no restriction on the exact counts. If either of these two
restrictions is not met for a chemical in an external testing dataset then the aggregate
fragment count is set to zero for that chemical.
2.9 Experimental Datasets
The database of fish biotransformation rates compiled by Arnot et al. was used to test the IFS
model selection algorithm.165
The biotransformation rates were averaged and converted to
the base-ten logarithm of the biotransformation half-lives (log HLN according to the
nomenclature used by Arnot et al.) as described in a subsequent publication by Arnot et al.49
In addition, the model selection algorithm was applied to log KOW values, as compiled in the
supporting information of Arnot et al. using the same division into training and testing
datasets.49
These two dataset splittings and their respective 2D fragment-based QSARs are
referred to as the IFS-HLN and the IFS-KOW datasets. Two-dimensional chemical structures
(SMILES notation172
) were retrieved from the EPI SuiteTM
software package database and
manually checked for correctness. A number of chemicals in the database differed only by
their stereochemistry, which is not captured by 2D structures; for each unique 2D structure
the log HLN and log KOW values were averaged and treated as single data points. An
additional dataset that retained the training and testing dataset division of Arnot et al.49
also
had the IFS model selection algorithm applied it, and along with the resulting QSAR is
referred to as the IFS-EPI dataset. In addition, a y-scramble was performed on the full
dataset of fish biotransformation half-lives (the log HLN values were randomly reassigned to
different chemical in the training dataset), another division into training and testing datasets
was made based on this, and another model for fish biotransformation half-life was fitted on
the y-scrambled data. This dataset and the resulting QSAR are referred to as the IFS-yscr
dataset.
212
3 Results
3.1 Fitting and External Validation
A total of six predictive models were fit for fish log HLN, three variations of IFS-EPI and
three variations of IFS-HLN. An additional three variations of the IFS-KOW dataset were also
created. Three variations of non-predictive models were also fit using the IFS-yscr dataset.
For all four datasets three 2D QSAR models were generated by applying three different
internal cross-validations, one with a 1:2 training to testing dataset split, one with 1:1 and
one with 2:1. In all cases the number of folds, k was equal to 6 and the values of m were 4, 3
and 2 respectively. In each case an averaged model was generated as the linear combination
of the three separately cross validated models. Statistics are summarized for all models in
Table 18, and fits are shown in Figure 39 and Figure 41 to Figure 43 in the Appendix.
213
Table 18 - Summary of Statistics for Fitted and External Validation Datasets.
Dataseta Fragments n ME
b MAE
c R
2 AICC
d
IFS-HLN; training dataset; 1:2 c.v. ratio 36 413 -0.024 0.419 0.794 -369.8
IFS-HLN; testing dataset; 1:2 c.v. ratio 206 -0.040 0.471 0.719
IFS-HLN; training dataset; 1:1 c.v. ratio 51 413 0.001 0.352 0.855 -458.6
IFS-HLN; testing dataset; 1:1 c.v. ratio 206 -0.013 0.465 0.724
IFS-HLN; training dataset; 2:1 c.v. ratio 67 413 -0.008 0.300 0.885 -496.8
IFS-HLN; testing dataset; 2:1 c.v. ratio 206 -0.055 0.473 0.712
IFS-HLN; training dataset; average 107 416 -0.010 0.334 0.870 -319.9
IFS-HLN; testing dataset; average 206 -0.036 0.451 0.745
IFS-EPI; training dataset; 1:2 c.v. ratio 31 417 -0.009 0.453 0.764 -340.3
IFS-EPI; testing dataset; 1:2 c.v. ratio 211 -0.009 0.497 0.691
IFS-EPI; training dataset; 1:1 c.v. ratio 46 417 -0.008 0.369 0.831 -424.1
IFS-EPI; testing dataset; 1:1 c.v. ratio 211 -0.013 0.492 0.709
IFS-EPI; training dataset; 2:1 c.v. ratio 48 417 -0.007 0.356 0.844 -449.6
IFS-EPI; testing dataset; 2:1 c.v. ratio 211 -0.025 0.486 0.700
IFS-EPI; training dataset; average 94 417 -0.008 0.365 0.845 -314.9
IFS-EPI; testing dataset; average 211 -0.016 0.457 0.743
IFS-yscr; training dataset; 1:2 c.v. ratio 11 413 -0.014 0.887 0.147 86.9
IFS-yscr; testing dataset; 1:2 c.v. ratio 206 -0.039 0.967 0.009
IFS-yscr; training dataset; 1:1 c.v. ratio 25 413 -0.020 0.796 0.288 54.0
IFS-yscr; testing dataset; 1:1 c.v. ratio 206 -0.084 0.986 0.016
IFS-yscr; training dataset; 2:1 c.v. ratio 59 413 -0.034 0.607 0.554 -28.4
IFS-yscr; testing dataset; 2:1 c.v. ratio 206 -0.143 1.072 0.002
IFS-yscr; training dataset; average 70 413 -0.023 0.736 0.456 104.7
IFS-yscr; testing dataset; average 206 -0.089 0.974 0.008
IFS-KOW; training dataset; 1:2 c.v. ratio 38 413 0.008 0.335 0.948 -504.1
IFS-KOW; testing dataset; 1:2 c.v. ratio 206 0.020 0.407 0.898
IFS-KOW; training dataset; 1:1 c.v. ratio 62 413 0.001 0.247 0.971 -657.4
IFS-KOW; testing dataset; 1:1 c.v. ratio 206 -0.001 0.354 0.923
IFS-KOW; training dataset; 2:1 c.v. ratio 69 413 -0.010 0.218 0.978 -727.8
IFS-KOW; testing dataset; 2:1 c.v. ratio 206 -0.016 0.361 0.911
IFS-KOW; training dataset; average 109 413 0.000 0.241 0.973 -526.0
IFS-KOW; testing dataset; average 206 0.001 0.346 0.925
aInternal k-fold cross-validation (c.v.) ratios are noted according to relative sizes;
internal training dataset : internal testing dataset. bMean Error.
cMean Absolute Error,
dAkaike Information Criterion, corrected for dataset size
214
.
Figure 39 - Model fit and predictions from IFS-HLN for (A,B) 1:2, (C,D) 1:1, and (E,F)
2:1 ratios between internal training and internal validation datasets during cross-
validation.
The training dataset of Arnot et al.49
contained 421 chemicals, but this has been reduced to
417 for the IFS-EPI dataset by averaging the observed values for chemicals with identical 2D
structures, the number of chemicals in the testing dataset remains unaltered (211). The
training and testing dataset splits generated in this work for the IFS-HLN, dataset contain
fewer total chemicals (413 and 207 respectively) because averaging of identical 2D
structures was done before splitting the dataset into training and testing datasets instead of
afterwards.
215
In all four datasets as the cross validation ratio of data points in the internal training and
testing datasets progresses from 1:2 to 2:1 the fit on the training dataset increases, as
demonstrated by the falling mean absolute error (MAE) and the rising correlation coefficient
(R2). The number of fragments selected in all cases increases as well, but the increase in fit
offsets the increasing complexity as shown by the falling AICC values. This is an expected
result, because at a ratio of 1:2 only a third of the training dataset is in the internal training
dataset fit by MLR at any time and so the residuals are larger overall. Each fragment added
has a direct effect on the residuals of fewer chemicals, so the point at which model
complexity begins to outweigh increases in the GoF comes sooner and fewer fragments are
added to the model. At the other extreme, a 2:1 ratio means two thirds of the training dataset
chemicals are present in each internal training dataset fit by MLR so the fit to the entire
training dataset is better. Each fragment added directly decreases the residuals of more of the
chemicals in the dataset and so more fragments are added before model complexity begins to
outweigh the GoF. The training dataset for IFS-HLN has a better GoF than IFS-EPI, but also
has more fragments in each of the three cross validation ratios. The GoF on the training
dataset for IFS-KOW is considerably better than for either IFS-HLN or IFS-EPI, but this
QSAR should not be used for predictive purposes because the full range of log KOW values is
not captured by the dataset and some of the values are likely predictions from EPI Suite,
which would lend the values to prediction by a fragment based model. The fit on the training
dataset of IFS-yscr demonstrates that the model selection algorithm will fit some amount of
random noise, but the GoF is much lower than the predictive models (Figure 42 in the
Appendix).
As the fit to the IFS-yscr training dataset increases, the prediction of the corresponding
external testing dataset becomes increasingly poor, as shown by the increasing MAE and the
wider spread of the predictions, see Figure 42. The R2 values are all essentially 0 for the IFS-
yscr external testing dataset, indicating the fragment generation, model selection, cross
validation and dataset splitting described in this work are not creating an artificially high
predictive power for the resulting 2D QSAR models.
216
For the IFS-HLN and IFS-EPI and IFS-KOW datasets the prediction of the external testing
dataset follows a different order than the y-scrambled dataset and is the same in all three
cases. The best prediction of the external training dataset is for a 1:1 ratio of internal training
to internal testing dataset chemicals during the cross validation, followed by 2:1 (more
chemicals in the internal training dataset) and 1:2 (more chemicals in the internal testing
dataset). The GoF for the IFS-HLN 2D QSAR models are marginally better than the IFS-EPI
models.49
Prediction of the external testing dataset for IFS-KOW values is better than for
either the IFS-HLN of IFS-EPI datasets, so the general pattern is that models with a better fit
for their training dataset have a better fit for the external validation dataset. However,
different models fit on the same dataset do not follow this pattern, and tend to follow the
opposite pattern as demonstrated by the results for IFS-yscr. This is due to the influence of
fitting versus over-fitting; the training dataset must be fit well in order for the model to select
all of the fragments with predictive power, but if the training dataset is over-fitted then
predictive power degrades as the noise in the data is fitted as well with extraneous fragments.
Average models were created for each dataset by taking the linear combination of the three
different cross validated models and then applying them to both the training and external
testing datasets. In the case of both the training and testing dataset for the IFS-yscr dataset
and the training datasets for the predictive models the GoF of the averaged model fell within
the range of the GoF of the individual models. For the external testing datasets for the three
predictive models however the GoF was higher for the averaged model than for any of the
individual models.
Three different log HLN values can be calculated for each chemical using the models fitted
with different cross validation ratios. This was done for the IFS-HLN and IFS-EPI datasets
and the range of predictions for each chemical in the training and external training datasets
were calculated. The average range of predictions for chemicals in the training and external
testing datasets were 0.41 and 0.48 for the IFS-EPI, and 0.35 and 0.36 for the IFS-HLN.
217
3.2 Fragment Coefficients
Summaries of the fragments in each model, along with their respective coefficients, standard
errors, and counts are given in Table 24 to Table 35. For most of the fragments the standard
error of the coefficients (SE) is less than the absolute value of the coefficient. In all three
IFS-KOW models the median relative SE (SE/|coefficient|) of the fragments in each model is
less than 0.5; corresponding to a median p value of less than 0.05. The median relative SE is
higher for the IFS-HLN, IFS-EPI and IFS-yscr datasets, in the range of approximately 0.7 for
the 1:2 cross validation ratio models decreasing to approximately 0.5 for the 2:1 cross
validation ratio models. This means that for IFS-KOW the uncertainty associated with each
fragment is smaller than for the fragments in IFS-HLN or IFS-EPI, which both have an
uncertainty comparable to the fragments IFS-yscr. Fragments in the IFS-HLN and IFS-EPI
models clearly have some predictive power, as opposed to the fragments in IFS-yscr, but
noise in the log HLN data makes the predictions more uncertain than the log KOW predictions.
Uncertainty in the coefficients of the fragments in the models can lead to large prediction
errors; as noted by Arnot et al.49
chemicals may have a number of fragments with both
negative coefficients (labile groups) and positive coefficients (recalcitrant groups) and the
MLR may over-compensate for the presence of the recalcitrant groups by making the
coefficients of the labile groups even more negative. The result is fragments with inflated
coefficients and greater uncertainty, which if applied to chemicals composed of primarily
labile or primarily recalcitrant groups, will lead to large over- or under-predictions. Similar
errors are possible if the models are applied to chemicals much larger or smaller than the
chemicals in the training datasets; the fragment counts are more likely to be outside of the
range of counts in the training datasets and the uncertainty in the coefficients is compounded
by extrapolating to larger or smaller fragment counts. A 2D QSAR should be more robust
and predictive if the absolute magnitudes of the fragment coefficients are small and
chemicals with a wide range of counts are included in the training dataset.
Median absolute coefficient values for the IFS-yscr models range from 0.59 to 0.87,
considerably higher than the median absolute coefficients for the IFS-EPI models. (0.31-
218
0.41) and the IFS-HLN models (0.28-0.37). Median absolute coefficient values for IFS-KOW
models range from 0.29 to 0.35, but the log KOW values span a much larger range than the
log HLN values, so in relative terms the coefficients tend to be smaller in magnitude. This
follows the same pattern as the model GoF for the training datasets, indicating that a good fit
to the training dataset results in fragment coefficients of smaller magnitudes.
In all of the predictive models the absolute fragment coefficients of rare fragments and
aggregate fragments tend to be higher than the absolute coefficients of the other fragments;
this pattern is not apparent in the models fit to the y-scrambled dataset. This supports the
application of the restrictions on fitting and counting the rare and aggregate fragments
outlined in this work; because these fragments seem to be more prone to over-fitting, as
indicated by their generally larger coefficients, and therefore should be used with caution. To
test the effect of aggregate fragment restrictions three different scenarios were applied to two
of the IFS-HLN models (the 1:1 and 2:1 cross validation ratios). Aggregate fragments
restrictions were applied to none of the fragments, all of the fragments and only to aggregate
fragments. In both cases the best prediction of the external testing dataset was made by
applying the restrictions only to aggregate fragments (R2 = 0.724 and 0.712 as shown in
Table 18). The worst predictions were obtained when applying no restrictions (R2 = 0.674
and 0.630) and slightly poorer predictions were made when applying restrictions to all
fragments (R2 = 0.685 and 0.664).
3.3 Fragment Selection
An inspection of the fragments selected in the models fit on the y-scrambled models should
provide an idea of which types of fragments are prone to over-fitting and therefore should be
viewed with skepticism when they are present in the predictive models. By far the most
common fragments in these models are fragments which essentially capture substitution
patterns of chlorines or bromines on aromatic rings and diphenylethers, which is not
surprising given the number of PCBs, PBDEs, dioxins and furans in the dataset. Two other
relatively common types of fragments are long alkyl chains, some branching and others not,
and large fragments containing phosphorus.
219
The IFS-HLN and IFS-EPI models were compared with the IFS-yscr models to check for
fragment overlap. None of the fragments in the IFS-HLN models were present in the IFS-yscr
models, however one to three fragments in each of the IFS-EPI models were present in the
IFS-yscr.
Overlap between models developed on the same dataset but with different cross validation
ratios was also checked. At the extremes were IFS-yscr models with only 5 fragments in
common, and IFS-KOW models with 20 fragments plus MW in common. IFS-HLN models
had 14 fragments plus MW in common, and IFS-EPI models had 8 fragments in common.
This follows the same pattern as the model GoF on the training datasets, so well fit models
consistently select the same fragments when applying different cross validation ratios.
As discussed by Arnot et al.49
log KOW and MW have an important influence on
biotransformation in fish by affecting the bioavailability and equilibrium within the fish.
Residuals for model predictions of log HLN for the chemicals in the external testing datasets
were regressed against the log KOW and the MW. In all cases no relationship between the
residuals and either the log KOW or the MW was observed, even for the two IFS-EPI models
in which the MW was not included by the model selection algorithm. This result implies that
the information about fish biotransformation half-lives represented by the log KOW and the
MW are captured by the IFS algorithm.
4 Discussion and Conclusion
4.1 Model Averaging and Model Uncertainty
One of the most interesting results of the work presented here is the increased predictive
power for the log HLN external testing datasets when the three different cross validated
models were averaged. The reasons for this can be inferred from the observed results for the
four different datasets. In each 2D QSAR model there will be some fragments which explain
the observed values and have predictive power, and some extraneous fragments which are
instead fitted to noise in the dataset. Fragments which consistently have predictive power are
220
more likely to be present in two or more of the three different cross-validation models and so
will retain their full effect after averaging the models. Extraneous and over-fitted fragments
in the models however will likely be different in each case due to different splitting for the
internal training and testing datasets, and so when the models are averaged the absolute
values of their coefficients will be diminished. The median absolute coefficient value for the
average of the IFS-HLN models is 0.11, significantly smaller than the range of median values
for the individual models (0.29-0.37), so the averaged model will be less prone to over- and
under-prediction errors when applying the model to chemicals with fragment counts that are
outside of the range of counts in the training dataset. Averaging of the models must be done
after applying the IFS algorithm in order to gain the increase in predictive power; when
multiple sizes of internal training and testing datasets were included during the IFS cross
validation step the GoF was no better than the models presented in this work (data not
shown).
For predicting values for log HLN the IFS-HLN models are recommended over the IFS-EPI
models. There are a number of reasons to recommend the IFS-HLN models, the first of which
is better GoF for the training and external testing datasets. Other reasons are more consistent
prediction of log HLN between the different models, more consistent selection of fragments,
smaller absolute values for the coefficients and the absence of any fragments which appeared
in the IFS-yscr models. The average IFS-HLN model is show in Table 19 and the fit of the
training dataset and prediction of the external testing dataset are shown in Figure 40
221
Table 19 - 2D Fragment Based IFS-HLN QSAR Averaged from Three Cross-
Validations.
Range of counts and
number of chemicals containing the fragment
Fragmentd Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
C=C 0.4685 0.1610 1-2 7 1-1 4
C(H)=CH 0.4412 0.2041 1-3 8 1-1 4
c(Cl)c-ccH 0.3153 0.1030 2-4 39 2-4 22
O=CC(H)(H)CH 0.2990 0.1393 1-2 3 1-2 3
O=CC(C(H)C(H)H)H 0.2883 0.1635 1-2 4 1-1 2
O=COH a 0.2327 0.0883 2-2 2 0-0 0
HCC(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)H a 0.2150 0.0870 1-1 2 0-0 0
c(c{c}c(H)cH)(H)cH 0.1980 0.0710 1-2 6 1-1 1
O=CC(C(H)H)H a 0.1959 0.1427 1-2 2 1-1 1
Hccc(cN(=O)=O)H 0.1870 0.0889 1-2 8 1-1 5
HC(H)H 0.1824 0.0281 1-14 205 1-12 97
HccccN(=O)=O 0.1470 0.0845 1-2 3 1-1 1
Hcc(cN(=O)=O)H 0.1276 0.0747 1-2 23 1-2 10
c(cH)Cl 0.1233 0.0309 1-7 125 1-6 70
c(H)(cH)cNH c 0.1089 0.0530 1-4 10 1-5 5
HC=C(H)H 0.1068 0.0550 1-2 9 1-1 7
c(cH)Br 0.1064 0.0434 1-8 19 1-8 8
HccOP(OC(H)(H)H)(OC(H)(H)H)=S 0.1013 0.0850 1-2 10 1-2 3
ccc(c(cC(H)H)H)H 0.0918 0.0739 1-2 3 1-2 2
c(ccc)-cccc 0.0908 0.0954 1-4 6 1-2 2
HC(C(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.0820 0.0826 1-2 17 1-2 10
c(H)c(H)c(OP(O)=S)c(H)cH 0.0719 0.0347 1-2 4 2-2 1
C(H)(CH)C(H)C 0.0684 0.0526 1-4 4 2-2 2
c(H)(cH)c-cc(H)c 0.0671 0.0267 1-4 21 1-4 10
HCC(C(C(C(H)H)(H)H)(H)H)(H)H 0.0657 0.0389 1-4 21 1-4 11
c(H)(cNH)c(H)cH 0.0621 0.0775 1-4 5 1-4 5
c(c(Cl)c(c)Cl)(c)Cl 0.0565 0.0449 1-6 18 1-6 9
c(H)(cc)c(H)cH c 0.0543 0.0168 1-6 5 4-4 1
c(H)(cH)ccc 0.0504 0.0450 2-4 35 2-4 15
CC(H)C(H)H 0.0504 0.0474 1-4 5 1-2 3
HCC(C(C(H)H)H)(H)H 0.0486 0.0189 1-12 15 2-6 6
c(H)(ccH)ccO 0.0460 0.0439 1-3 15 1-2 7
c(H)(cCcc(H)cH)c(H)c 0.0453 0.0157 2-8 4 8-8 3
Cl 0.0420 0.0164 1-12 165 1-10 90
c(H)(cH)cc-cc 0.0402 0.0198 1-8 13 1-2 6
cH 0.0367 0.0099 1-15 303 1-15 151
C(H)H 0.0367 0.0169 1-28 153 1-20 74
c(cH)c-cc 0.0323 0.0189 1-8 23 1-8 10
c(H)(c)c-ccH 0.0282 0.0465 1-6 27 1-4 11
c(H)cC(H)H 0.0241 0.0149 1-7 32 1-4 16
c(Cl)ccH c 0.0205 0.0156 1-5 97 1-6 49
c 0.0167 0.0161 1-12 303 1-12 152
{c}(cH){c}{c} -0.0129 0.0275 1-20 20 2-18 10
c(c)(cc)Br c -0.0262 0.0140 2-12 5 2-20 3
O(H)c -0.0425 0.0755 1-2 29 1-1 11
HccC(C(H)H)H -0.0429 0.0332 1-4 9 2-4 4
O=COC(C(H)H)(H)H -0.0449 0.0960 1-2 8 2-2 2
c(H)(c(H)cH)c(H)cC=O -0.0497 0.0335 1-4 16 2-2 3
BrCH c -0.0519 0.0412 3-6 3 1-1 1
HC(C(C(H)(H)H)C(H)(H)H)H c -0.0596 0.0469 1-6 12 1-8 7
c(H)(cc){c}{c}cH -0.0605 0.0597 1-4 8 1-2 3
HC(C(C(H)(H)H)(C(H)(H)H)C(H)(H)H)H -0.0635 0.1023 1-2 6 1-2 4
Hcc(c(cC(H)(H)H)H)H c -0.0662 0.0709 1-2 7 1-3 5
Hccc(c(cC(H)(H)H)H)H -0.0690 0.0524 1-2 7 1-2 5
HCC(H)(H)O -0.0733 0.0657 1-3 7 1-1 4
HC(C(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)(H)H)H -0.0748 0.0539 1-2 4 2-2 2
Br c -0.0749 0.0312 1-7 30 1-10 12
222
c(H)({c}{c})c(H)c -0.0783 0.0737 1-2 13 1-1 6
{n}{c} -0.0789 0.0349 1-2 5 1-1 1
c(H)c-c(c)c -0.0829 0.0957 1-2 18 1-2 9
c(CH)(cH)c(H)c(H)cH -0.0853 0.0729 1-4 8 1-2 3
O(CH)c -0.1120 0.1138 1-2 6 0-0 0
{c}({c}c){c}{c} -0.1373 0.0797 1-1 6 1-1 4
ccc(c(C(H)(H)H)cH)H -0.1453 0.0657 1-4 6 1-2 8
O=CN -0.1517 0.0948 1-2 10 1-1 1
OC(C(H)(H)O)H -0.2006 0.1474 1-1 3 0-0 0
HC(NC(H)(H)H)(H)H -0.2339 0.0668 1-2 5 1-2 4
HCC(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H -0.2427 0.0722 1-2 4 0-0 0
ccC(C(H)(H)H)(H)H -0.2554 0.1060 1-2 3 0-0 0
Occ(cC(H)(H)H)H -0.2608 0.0942 1-1 5 1-1 3
O=COC(H)(H)H -0.2622 0.1478 1-1 3 0-0 0
O -0.2852 0.0394 1-8 104 1-8 46
Hcc(c(C(H)(H)H)cH)C(H)(H)H -0.2863 0.1616 1-2 6 1-1 3
C(H)(CH)=CH -0.2863 0.1240 1-2 3 1-2 3
ClC(C(H)(H)Cl)H -0.3204 0.1384 1-2 3 2-2 1
s -0.3539 0.1382 1-2 5 1-1 2
N -0.4526 0.1165 1-2 18 1-2 6
N(=O)=O -0.5125 0.1518 1-2 29 1-1 17
{n}(c)cH -0.5354 0.2119 1-1 3 0-0 0
N(H)H -0.5726 0.1141 1-2 23 1-2 14
NH -0.5889 0.1368 1-2 18 1-2 7
c(c)(c(H)c)Occc -0.6073 0.2560 1-1 3 1-1 1
O=C -0.6227 0.0742 1-4 48 1-2 14
OH -0.7307 0.1022 1-3 48 1-3 19
P=O -1.1274 0.1708 1-1 9 1-1 7
ClcccO
L.C.F. : c(Cl)(cCl)c(cO)Cl b 0.9384 0.2439 1-2 5 1-1 1
HCC(H)(H)C
L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H b 0.6835 0.2080 1-2 3 2-2 1
Nc(c)cH L.C.F. : c(c)(N)c(H)c
b 0.4049 0.1691 1-1 3 0-0 0
{c}(c)({c}c)c(H)cH
L.C.F. : {c}(c)(c(H)cH){c}(c)cH b 0.2500 0.1290 1-2 4 0-0 0
{c}(cH)({c}cH)cc(H)cH L.C.F. : {c}1(cc(H)cH){c}c(H)c(H)c(H)c1H
b 0.2190 0.0672 1-2 7 0-0 0
c(H)({c}{c}o)c(H)cH
L.C.F. : {c}1(c(H)cH){c}(cH)o{c}{c}1c(H)c(H)cH b 0.1511 0.1309 1-2 3 1-1 2
cC(H)=C(H)H
L.C.F. : Hcc(c(C(=C(H)H)H)cH)H a,b 0.1407 0.0982 1-2 2 0-0 0
Hc{c}{c}c(cC(H)(H)H)H L.C.F. : Hc{c}{c}(c(cC(H)(H)H)H)cH
b 0.0905 0.0995 1-2 3 0-0 0
c(c-ccc)cc
L.C.F. : c(c-cc(Cl)cCl)(Cl)c(c)Cl b 0.0780 0.0334 1-8 6 2-6 2
c(cOc(cH)cH)(H)c L.C.F. : c(c(H)c)(Oc(cH)c(H)cH)c(H)cH
b 0.0444 0.0536 1-8 11 1-4 4
c(H)(cC)c(H)cH
L.C.F. : c(H)(c(H)cH)c(H)cC b -0.0523 0.0280 1-6 5 4-4 1
Occ(c(cCC(H)(H)H)H)H
L.C.F. : FC(C(Oc1c(c(c(c(c1H)H)CC(H)(H)H)H)H)(F)F)(H)F b -0.0584 0.0326 4-4 3 4-4 1
HC(C(H)(H)Br)H Unique Chemical : 4101682
a,b -0.0761 0.0656 2-2 1 0-0 0
HC(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)H
L.C.F. : HC(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)H b -0.0771 0.0427 1-2 5 2-2 2
cc(ccN(=O)=O)H L.C.F. : cc(cc(N(=O)=O)cH)H
b -0.0939 0.1507 1-1 6 1-1 2
{c}(cH)({n}){c}cH
L.C.F. : {c}({n})({c}cH)c(H)cH b -0.1171 0.0438 1-2 5 1-1 1
c(H)(ccH)cO
L.C.F. : c(cH)(O)c(H)ccH b -0.1319 0.1000 1-2 5 1-1 2
O=COC(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H L.C.F. : O=COC(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H
b -0.1393 0.0564 2-2 3 2-2 2
O=COC(C(H)(H)CH)(H)H
Unique Chemical : 85507795 a,b -0.1632 0.1205 1-1 1 0-0 0
223
c1(H){c}({c}(cH)c(H)cc1)cH
L.C.F. : c1(H){c}({c}(cH)c(H)cc1)cH b -0.1997 0.0809 1-2 5 1-1 1
C(H)(Br)C(H)(Cl)C(H)Br
Unique Chemical : 87843 a,b -0.2174 0.1369 1-1 1 0-0 0
c(H)(cH){n}
L.C.F. : c(H)({n})c(H)cH . c(H)c(H)c(H)cH a,b -0.2596 0.1370 1-1 2 0-0 0
MW c 0.00459 0.00043 68.1-748 413 68.1-
959.1 206
intercept -1.0727 0.0925
aRare fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s)
among the chemicals containing the aggregate fragment. Unique chemical is the CAS of
the single chemical containing the aggregate fragment. Testing range and frequency
exclude counts forbidden by aggregate fragment selection rules. cTesting range exceeds
training range. dMolecular weight (MW) or fragment in pseudo SMILES format. {}
indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms by
aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of
the regression coefficient.
224
Figure 40 - Model fit and predictions for the averaged IFS HLN QSAR for training (A)
and external validation (B) dataset.
225
4.2 Comparison to Arnot et al. QSAR
Statistics for the averaged IFS-HLN model are comparable to those reported for the Arnot et
al. model,49
hereafter referred to as the EPI-HLN model, for both the training dataset
(R2=0.82, MAE=0.38 versus R
2=0.870, MAE=0.334 for IFS-HLN) and the external testing
dataset (R2=0.73, MAE=0.45 versus R
2=0.745, MAE=0.451 for IFS-HLN). The number of
chemicals with absolute residual values greater than one log unit is 9 in the training dataset
and 19 in the external testing dataset, as opposed to the 21 and 22 reported by Arnot et al.
Table 23 shows the predicted values and residuals for chemicals in both the training and
external testing datasets. There are 11 chemicals with residuals over one log unit for both the
averaged IFS-HLN model and the EPI-HLN model, these chemicals are shown in Table 20.
Most of the errors appear to be due to the inherent shortcomings of 2D fragment based
QSARs. Tetrachloromethane, 1,1,1-trichloroethane and tetrachloroethylene are all predicted
to be more recalcitrant than they actually are, likely because they are all small, highly
chlorinated molecules and the resulting steric effects are not captured by either of the 2D
fragment-based QSARs. Dimethylphosphoric acid 3-methyl-4-nitrophenyl is an example of a
chemical which is erroneously under-predicted because of the presence of too many
fragments with large negative coefficients (P=O and N(=O)=O) without any corresponding
fragments with large positive coefficients to balance them. Other chemicals‘ log HLN are
probably erroneously predicted by both QSARs because they have unique features which are
poorly captured by 2D fragment descriptors, for example hexachlorobenzene, or they
contain groups which are poorly represented in the dataset, for example
decamethylcyclopentasiloxane.
226
Table 20 - Chemicals with log HLN Residuals Greater than 1 in IFS-HLN and EPI HLN
This Work Arnot et al
CAS Chemical Name Obs. Dataset Pred. Resid. Dataset Pred. Resid.
56235 Tetrachloromethane -1.19 train -0.20 1.00 test 0.08 1.27
71556 1,1,1-Trichloroethane -1.39 train -0.15 1.24 train -0.08 1.31
132649 Dibenzofuran 1.35 train 0.22 -1.12 train -0.08 -1.43
541026 Decamethylcyclopentasiloxane 2.74 train 1.03 -1.71 test 1.72 -1.02
6639301 2,4,5-Trichlorotoluene 2.01 train 0.71 -1.31 train 0.69 -1.32
103446 2-Ethylhexylvinylether 0.95 test -0.06 -1.01 train -0.05 -1.00
118741 Hexachlorobenzene 2.44 test 0.93 -1.51 train 1.33 -1.11
127184 Tetrachloroethylene -0.88 test 0.33 1.20 test 0.53 1.41
947728 9-Chlorophenanthrene 1.80 test 0.21 -1.58 test 0.49 -1.31
1839630 1,3,5-Trimethylcyclohexane 1.71 test 0.46 -1.25 test 0.56 -1.15
2255176 Dimethylphosphoric acid 3-
methyl-4-nitrophenyl 0.01 test -1.89 -1.90 train -1.36 -1.37
Table 21 shows the accuracy of the predicted qualitative biotransformation rates of the
external training dataset made by the averaged IFS-HLN model. Qualitative rankings are the
same as defined by Arnot et al.49
There is a centrist trend in the predictions, slowly
biotransformed chemicals generally have their rates over-predicted and rapidly
biotransformed chemicals generally have their rates under-predicted. The reasons for this are
clear from Figure 40B; the slope of the regression is less than one, so the full range of log
HLN values is not completely captured. If the qualitative predictions are categorized as very
slow or slow, moderate and very fast or fast the correctly predicted fractions are 73%, 77%
and 73%, in contrast to the values 82%, 67% and 75% reported by Arnot et al.
227
Table 21 - Qualitative Prediction Accuracy for Fish log HLN Fit on Dataset Splitting
from this Work Using Average Model from Three Different Cross-Validations.
Observed Qualitative Category
Very Slow Slow Moderate Fast Very Fast Sum
Pre
dic
ted
Qu
alit
ativ
e
Cat
ego
ry
Very Slow 69% (20/29) 5% (2/40) 0% (0/62) 0% (0/64) 0% (0/11) -
Slow 28% (8/29) 50% (20/40) 2% (1/62) 0% (0/64) 0% (0/11) 0% (0/206)
Moderate 3% (1/29) 45% (18/40) 77% (48/62) 31% (20/64) 0% (0/11) 0% (0/206)
Fast 0% (0/29) 0% (0/40) 19% (12/62) 59% (38/64) 100% (11/11) 0% (0/206)
Very Fast 0% (0/29) 0% (0/40) 2% (1/62) 9% (6/64) 0% (0/11) 17% (34/206)
Sum - 0% (0/206) 0% (0/206) 1% (2/206) 21% (44/206) 61% (126/206)
4.3 Fragment Analysis
A number of the fragments in the averaged IFS-HLN model shown in Table 19 have exactly
or partially corresponding fragments in the model presented by Arnot et al., hereafter
referred to as the EPI-HLN model. However, direct comparison of coefficients is problematic
because the EPI-HLN model includes the log KOW and MW, whereas the IFS-HLN model
only includes the MW. The inclusion of these parameters means that the fragment
coefficients are essentially relative to the effect of log KOW and MW. In the IFS-HLN model
MW has a coefficient of +0.00459, in the EPI-HLN model it is approximately -0.00256.
However the EPI-HLN model has a positive coefficient for log KOW which is partially
correlated with MW. In many cases the log KOW in the EPI-HLN model will itself be
calculated with a QSAR that has hundreds of coefficients,173
many of which are also present
in the log HLN model. Taking the coefficients from the log KOW model of Meylan et al.,173
the MW of the fragments and the coefficients from Arnot et al., the total effect of each
fragment can be estimated. The fragments of both Meylan et al. and Arnot et al. are ―atom
centered fragments‖, essentially equivalent to the single unit fragments of this work, but
generally with aliphatic or aromatic attachment differentiated. Larger fragments from this
work could be compared with some of the correction factors in the works of Meylan et al.
and Arnot et al. but the relationship is even less clear.
228
Single unit fragments that appeared in all three of the individually cross-validated models
used to create the averaged IFS-HLN model were selected for comparison with the EPI-HLN
model, along with chlorine and bromine fragments; the results are shown in Table 22. The
relationship between the fragments of the averaged IFS-HLN model and the log KOW model
of Meylan et al. along with the EPI-HLN model is not always clear, because the single unit
fragments may be included in larger fragments or in correction factors. In most cases shown
in Table 22 the signs of the total effect of each fragment are in agreement, and the total effect
predicted by the averaged model from this paper falls within the same range of effects
estimated for the model of Arnot et al. This is quite good agreement given the different
methodologies, training datasets, and fragment definitions.
229
Table 22 - Comparison of Fragment Coefficients from IFS-HLN and EPI-HLN.
IFS-HLN EPI-HLN./Meylan et al.
Fragment Coef. MW eff.a sum Coef. log KOW eff.
b MW eff.
a sum
C(H)=CHc 0.441 26 0.120 0.561 2 × =CH- 0.198 0.767 0.236 26 -0.067 0.367
HC(H)H 0.182 15 0.069 0.251 CH3-aliphatic 0.245 0.547 0.168 15 -0.038 0.375
CH3-aromatic -0.087 0.547 0.168 15 -0.038 0.043
O -0.285 16 0.073 -0.212
O-aliphatic -0.023 -1.257 -0.386 16 -0.041 -0.450
O-aromatic -0.069 0.262 0.081 16 -0.041 -0.029
aliph.-O-arom. 0.000 -0.466 -0.143 16 -0.041 -0.184
N(=O)=O -0.513 46 0.211 -0.301 NO2-aliphatic 0.000 -0.813 -0.250 46 -0.118 -0.367
NO2-aromatic -0.022 -0.182 -0.056 46 -0.118 -0.196
NH -0.589 15 0.069 -0.520 NH-aliphatic 0.407 -1.496 -0.459 15 -0.038 -0.091
NH-aromatic -0.289 -0.917 -0.282 15 -0.038 -0.609
N(H)H -0.573 16 0.074 -0.499 NH2-aliphatic 0.407 1.415 0.434 16 -0.041 0.800
NH2-aromatic -0.289 -0.917 -0.282 16 -0.041 -0.612
OHd -0.731 17 0.078 -0.653
OH-aliphatic -0.062 -1.409 -0.432 17 -0.044 -0.538
OH-aromatic -0.473 -0.480 -0.147 17 -0.044 -0.664
P=O -1.127 47 0.216 -0.912 P=O -0.603 -2.424 -0.744 47 -0.120 -1.467
O=C -0.623 28 0.129 -0.494 O=C-aliphatic 0.000 -1.559 -0.478 28 -0.072 -0.550
aliph.-C(=O)-arom. 0.000 -0.867 -0.266 28 -0.072 -0.338
Cl 0.042 35 0.163 0.205
Cl-aliphatic 0.361 0.310 0.095 35 -0.091 0.365
Cl-olefinic 0.000 0.492 0.151 35 -0.091 0.060
Cl-aromatic 0.378 0.645 0.198 35 -0.091 0.485
Br -0.075 80 0.367 0.292
Br-aliphatic 0.273 0.400 0.123 80 -0.205 0.191
Br-olefinic 0.000 0.393 0.121 80 -0.205 -0.084
Br-aromatic 0.396 0.890 0.273 80 -0.205 0.465
aEffect of MW, the MW multiplied by the appropriate coefficient.
bEffect of log KOW, the corresponding
coeffiecient for log KOW from Meylan et al. multiplied by the coefficient of Arnot et al. cEach of the Arnot et al.
and Meylan et al. coefficients are counted twice. Other olefinic hydrogens appear in the averaged model, so it is
not clear if these values are directly comparable. dThe fragment O(H)c from the averaged model essential creates
a specific effect for all OH groups with an aromatic attachment, the summed effect is -0.695.
230
Several other single unit groups not in Table 22 appear in both the averaged IFS-HLN model
and the EPI-HLN model; ‗N‘ (tertiary amines), ‗C(H)H‘ (CH2 groups), and ‗cH‘ (aromatic
hydrogens). Some larger fragments also have directly comparable fragments in the EPI-HLN
model such as ‗O=COH‘ (alcohol group) and ‗O=COC(H)(H)H‘ (ester group, specifically an
acetyl group). Other fragments have a noticeable but less clear relationship to fragments in
EPI-HLN model such as the fragment ‗s‘ (thiofuran and thiazole corrections) and ‗O=CN‘
(urea correction). This is encouraging because it suggests that the IFS method can capture
the same information as fragments selected based on expert judgment and knowledge of
biodegradation pathways. The initial set of fragments used by Arnot et al. was the same as
that used in a previously developed model for aerobic biodegradation,174
and then modified
by inspecting the residuals of the regression and manually selecting fragments that helped to
correct these residuals.49
Indeed some of the manually selected fragments have
corresponding fragments in the averaged IFS-HLN model, such as the aforementioned
thiofuran and thiazole corrections, and the fragment ―o-Chloro/mono-aromatic ether‖ which
roughly corresponds to the aggregate fragment ‗ClcccO‘ from in the IFS-HLN model. Other
than the fragmentation methods described no pre-selection of fragments was performed in
this work, so this agreement illustrates the effectiveness of the model selection algorithm.
An inspection of the remaining fragments in the averaged model should reveal additional
fragments with no analogous fragments in the EPI-HLN model which are important for
predicting fish log HLN. Based on the previous discussion the aggregate fragments are
ignored because they are too uncertain. Fragments which are similar to those found in the
IFS-yscr models are also ignored, which excludes several large fragments containing
phosphorus, several long aliphatic carbon chains, and fragments which seem to capture
specific substitution patterns on aromatic rings. Fragments with a high relative standard error
for their coefficients are also considered too uncertain. Four interesting fragments meet these
requirements, two of which appear in all three of the individual IFS-HLN cross validation
models; these are the fragments ‘c(Cl)c-ccH‘ and ‗c(cH)Cl‘. The other two fragments are
‗HC(NC(H)(H)H)(H)H‘ and ‗c(cH)Br‘. The fragments ‗c(cH)Cl‘ and ‗c(cH)Br‘ increase the
predicted fish log HLN if there is a chlorine or bromine on an aromatic ring which has no
231
vicinal group; the fragment will be counted twice if there is no vicinal group on either side of
the halogen. These fragments have a roughly inverse relationship with the fragments ‗Clcc‘
and ‗Brcc‘ (not included in the model) which capture the effect of having a vicinal group
next to an aromatically substituted chlorine or bromine. All chlorine and bromine atoms
increase the HLN; the information captured by these fragments is essentially that chlorines
and bromines with vicinal groups increase the HLN values to a smaller degree. The fragment
‘c(Cl)c-ccH‘ captures an ortho-effect for PCBs (only PCBs in the dataset have the fragment
‗c-c‘ which represents two aromatic rings joined by a single bond). Biphenyl and a PCB with
four ortho-chlorines will not have this fragment, but PCBs with one to three ortho-chlorines
will have their HLN values elevated. Finally, the fragment ‗HC(NC(H)(H)H)(H)H‘ is a
tertiary amine substituted with two methyl groups, the presence of which makes a chemical
more labile, this is additive with the fragments ‗N‘ and ‗HC(H)H‘ essentially cancelling out
the positive coefficient of the methyl groups if they are on a tertiary amine.
The IFS methods illustrated here for biotransformation can be applied to other datasets for
key chemical parameters. Future work is likely to include a new QSPRs for log KOW and
vapor pressure, in addition to a QSAR for various toxic modes of action and other
parameters relevent to chemical hazard and risk assessment.
5 Acknowledgements We acknowledge funding from the Long-range Research Initiative of the European Chemical
Industry Association (CEFIC LRI-ECO13-USTO-081212).
6 Co-Author Contributions Jon Arnot assissted in the comparison of the results of the final model with the previous
model of fish biotransformation half-lives, and contributed text for the chapter introduction.
232
7 Appendix
Table 23 - Training and Validation Datasets Summary.
CAS no. or ID + Name
Average
IFS-HLN
prediction Residual
IFS-HLN
Splitting
EPI-HLN
Splitting
IFS-yscr
Splitting
(1)
6-n-Butyl-2,3-dimethylnaphthalene 0.47 -0.23 train test train
(2)
XDE-536 methyl ester -1.10 -0.22 train test train
(4) 175 Factor J
0.96 -0.62 train test test
(20)
Octaethylene glycol monotridecyl ether -0.41 -0.20 train train test
(40)
NL-93 2.27 0.34 train train train
(60) NL-133
1.73 0.71 train train train
(70)
NL-63a 0.64 -0.28 train train test
(80)
NL-63b 0.91 -0.51 train test train
(201) XDE-537 n-butyl ester
-0.82 0.04 train train train
(205)
Spinosad Factor D -0.39 0.84 train train test
(692)
2-isopropyl decalin 0.61 0.30 train train train
(694) 1 isobutyl 2,5 dimethyl cyclohexane
0.48 0.32 train train train
(696)
1-octylpyrene 1.20 0.06 train train test
(697)a
Methylnaphthalenes 0.11 0.84 train train train
91576a - - train train train (698)b
Di-Me & Et naphthalenes 0.34 0.26 train train test
575417b - - train train test (2355)
Methylisocyanothion 0.31 0.34 train train train
(6011) 3,3',4,4'-Tetrabromodiphenyl ether (BDE 77)
1.62 0.63 train test train
(6017)
2,2',4,4',5,5'-Hexabromodiphenyl ether (BDE 153) 2.07 -0.06 train train train
53190 1.42 0.23 train test train
53703 0.56 -0.35 train train test
56235 -0.20 -1.00 train test test 59507 -0.66 -0.62 train test test
60297 -0.58 -0.12 train train test
60515 -1.22 0.21 train train train 60571c 1.42 0.22 train train test
72208c - - train test test 62442 -0.95 0.40 train train train
63252 -0.97 -0.21 train test test
67663 -0.40 -0.36 train test train 67721 0.27 0.49 train train train
71556 -0.15 -1.24 train train train
72559 1.88 -0.17 train train train 74975 -0.47 -0.27 train test test
75252 -0.29 -0.41 train train train
75354 -0.54 -0.18 train train train 76448 1.34 0.10 train test train
76835 0.54 -0.69 train train test
78400 -1.56 0.47 train train train 78422 0.37 -0.66 train test test
78637 0.90 -0.13 train train train
78795 -0.47 -0.14 train train train 78875 -0.57 -0.49 train train test
233
79016 -0.34 -0.07 train train train
79345 -0.13 -0.54 train train train 79469 -0.81 -0.39 train train test
79925 0.23 0.73 train test train
80057 -0.48 0.75 train train train 81038 0.72 -0.03 train train train
82053 -0.15 -0.54 train test train
82440 -0.48 0.25 train test train 83794 -0.52 -0.07 train train train
84151 0.67 0.19 train test test
84742 -0.93 0.52 train train train 85018 0.04 0.66 train test train
85223 0.40 -0.38 train test train
85687 -0.73 -0.30 train train test 86737 0.50 0.20 train train train
86748 -0.06 0.12 train train train
87616 0.54 0.38 train train train 87683 1.31 0.76 train train train
87821 0.80 -0.37 train train train
87843 0.47 -0.79 train train train
87865 -0.14 -0.06 train train train
88062 -0.18 -0.39 train train train
88197 -0.45 -0.44 train train test 88733 -0.26 -0.32 train test test
88744 -1.11 0.65 train train train
88982 -1.05 0.48 train train train 89690 0.06 -0.70 train train test
91225 -0.80 -0.54 train train train 91236 -0.57 -0.48 train train test
91963 -1.41 -0.09 train train train
92513 0.32 0.06 train train train 92693 -0.68 -0.13 train train test
92842 -0.22 0.25 train train train
92864 1.00 0.57 train train train 94520 -0.82 -0.33 train train train
95169 -0.82 -0.04 train train test
95487 -0.94 0.42 train train train 95501 0.26 0.24 train train test
95512 -0.58 -0.02 train train train
95578 -0.78 0.40 train train train 95647 -0.92 -0.07 train train train
95761 -0.37 -0.72 train train test
95772 -0.57 -0.35 train test train 95829 -0.27 -0.86 train train test
96184 -0.84 0.31 train train train
96297 -1.00 -0.33 train test train 96764 0.36 -0.73 train train test
96968 -1.14 0.27 train test train
97234 -0.40 -0.11 train test train 98088 -0.31 -0.17 train train train
98102 -0.72 0.01 train train train
98511 0.38 -0.07 train test train 98840 -0.88 -0.02 train train test
98953 -0.56 -0.37 train train test
99081 -0.34 -0.45 train test train 99718 -0.66 -0.18 train train train
99990 -0.48 -0.53 train train train
100016 -1.09 0.58 train train test
100185 0.58 0.02 train test train
101531 -0.49 -0.38 train train test
101611 -0.05 0.02 train test test 101848 -0.18 0.09 train train test
103504 0.12 0.51 train test train
104405 -0.05 -0.18 train test train 104723 0.77 -0.14 train train test
104881 0.04 -0.03 train train train
105055 0.26 -0.27 train test test 105066 0.10 0.40 train train test
106376 0.47 -0.16 train test train
106478 -0.59 -0.02 train test train
234
107391 -0.03 0.15 train test train
107506 1.87 -0.39 train test test 108429 -0.59 -0.12 train train test
108430 -0.79 0.10 train test test
108576 0.10 0.49 train train test 108678 0.19 0.21 train train train
108861 -0.01 -1.18 train train train
109091 -1.02 0.17 train train train 110021 -0.89 0.07 train test train
110827 -0.47 -0.19 train train train
111444 -0.47 -0.34 train train train 111842 0.30 0.67 train test test
112403 0.61 -0.34 train train test
112414 0.40 0.43 train train test 115322 1.32 0.66 train train train
117180 0.48 0.01 train train train
117793 -1.52 0.46 train test train 117817 0.04 0.40 train train train
118445 0.01 0.42 train train train
118832 -0.07 -0.30 train test train
119120 -0.51 -0.19 train train train
119471 0.19 -0.16 train train train
119562 -0.30 -0.59 train train train 119619 -0.66 -0.52 train test train
120127 0.11 0.54 train train train
120785 -0.32 -0.59 train train train 120832 -0.46 -0.48 train train train
121142 -0.67 0.45 train test test 122145 -0.48 0.52 train train test
122394 0.20 0.15 train train train
123488 0.77 -0.07 train train test 124118 0.02 0.50 train train train
127902 0.74 0.30 train train train
132649 0.22 1.12 train train train 134327 -0.30 0.05 train test test
137268 -0.61 -0.57 train train train
140669 -0.07 0.03 train train train 142961 -0.17 -0.08 train train train
144194 -1.16 -0.18 train train train
156434 -0.90 -0.34 train train train 198550 0.27 -0.33 train test train
207089 0.27 -0.34 train test train
208968 0.32 0.20 train train train 214175 0.59 -0.01 train train train
218019 0.29 0.30 train train train
226368 0.28 -0.42 train train train 243174 0.33 -0.72 train train train
294622 0.14 1.19 train train train
298000 -0.09 0.64 train train test 298044 0.35 -0.32 train train train
299843 0.71 0.24 train test train
314409 -1.49 0.16 train train train 320605 0.55 1.26 train train train
330541 -1.05 -0.40 train train train
330552 -1.02 0.05 train train test 333415 0.28 -0.22 train train train
438222 1.30 0.37 train train train
488233 0.51 0.66 train train train
500287 -0.13 -0.34 train train train
526738 0.26 -0.49 train train test
535773 0.12 -0.05 train test train 541026 1.03 1.71 train test train
541731 0.36 -0.57 train train train
542187 -0.17 0.58 train train test 544014 0.25 -0.69 train test test
554007 -0.27 -0.39 train test test
554847 -1.22 0.74 train test test 555033 -0.62 -0.42 train train train
579102 -1.05 -0.07 train test test
581408 -0.03 0.06 train test train
235
591208 -0.73 -0.22 train test train
591355 -0.36 -0.55 train train train 603112 -2.21 0.64 train train train
605027 0.76 0.96 train train train
606280 -1.29 -0.01 train train train 608935 0.75 1.12 train test test
610399 -0.84 -0.56 train train train
611063 -0.03 0.87 train train test 611212 -0.35 -0.39 train train train
612226 -0.49 -0.08 train test train
616444 -0.67 -0.61 train train train 618622 0.41 0.16 train train test
626391 0.95 0.70 train test train
629594 0.81 0.68 train test train 634673 -0.21 -0.06 train train train
634935 0.02 0.65 train test train
636282 1.00 0.34 train train train 760236 -0.59 -0.69 train train train
764136 0.16 0.33 train train train
789026 1.62 0.09 train test test
791311 0.07 0.00 train train train
877101 0.68 0.44 train train test
877112 0.80 0.60 train train train 883205 0.35 -0.24 train train test
920661 -1.03 -0.29 train test train
933120 0.84 -0.12 train train train 935955 -0.20 0.11 train test train
950378 -1.18 0.03 train test train 961115 -0.27 -0.75 train train train
1024573 1.47 0.73 train train train
1460022 1.86 0.84 train test train 1490046 -0.62 -0.57 train test train
1582098 0.31 0.31 train train train
1634044 -0.22 -0.66 train train train 1634782 -1.81 0.77 train test train
1678984 0.38 0.53 train train train
1712705 0.28 1.02 train train train 1732134 0.41 0.01 train train train
1746016 1.36 0.42 train train train
1795159 0.73 0.37 train test train 1825214 2.47 0.20 train train test
1825316 1.75 0.62 train train train
1836777 0.88 -0.13 train test test 1889674 0.94 -0.07 train train train
1897456 0.47 -0.20 train train train
1912249 -0.58 -0.47 train train train 2051243 2.90 -0.03 train train train
2104645 0.26 -0.24 train train train
2173571 0.09 0.15 train test train 2212671 -1.00 -0.08 train train train
2216695 0.04 0.35 train train test
2234131 1.25 0.37 train train train 2243621 -0.54 0.02 train test train
2381217 0.39 -0.28 train train train
2385855 1.94 -0.14 train train train 2437561 0.50 0.23 train test test
2460493 -0.55 0.40 train train train
2498660 -0.77 -0.07 train train train
2541697 0.44 -0.50 train train train
2597037 -0.47 -0.28 train train train
2631405 -0.85 -0.41 train train test 2636262 0.08 0.15 train test train
2668475 1.27 -0.11 train test train
2921882 0.62 0.16 train train train 3074713 0.57 0.70 train train train
3194556 1.92 0.15 train train train
3229003 0.56 0.20 train test train 3296900 -1.33 -0.06 train train test
3674757 0.40 -0.11 train test test
3761419 0.26 -0.63 train train train
236
3761420 0.33 -0.36 train test test
3766812 -0.79 0.27 train train train 3811492 -0.40 -0.75 train test train
3846717 0.29 0.60 train train test
3864991 0.37 0.35 train train train 3891983 1.07 0.29 train train train
4101682 0.37 -0.84 train train test
4130421 0.73 0.12 train test train 4175546 0.30 0.42 train train train
4292755 0.45 0.25 train train train
4821196 0.17 0.03 train train train 4883721 -1.28 0.38 train train train
4904614 1.22 0.52 train train train
4920950 0.48 -0.49 train train test 5323568 0.83 -0.24 train train train
5325973 0.55 0.21 train test train
5428546 -1.25 0.01 train train train 5707448 0.41 -0.35 train train train
6117971 0.80 0.18 train train train
6165511 0.42 -0.32 train test train
6639301 0.71 1.31 train train train
6975980 0.60 -0.50 train train train
7045718 0.70 0.14 train train test 10394577 0.47 -0.17 train train train
13116535 0.00 -0.28 train test train
13151343 0.51 0.14 train test train 13358117 0.12 0.18 train test train
13674845 -0.77 -0.54 train train test 13674878 -0.60 0.09 train train train
13936215 -0.22 0.42 train train train
15087248 0.28 -0.49 train test train 15254258 0.43 0.16 train test train
15972608 -0.70 -0.40 train train train
16219753 -0.35 -0.22 train test test 16606023 2.03 -0.12 train test train
16958922 0.34 -0.77 train train test
17109498 -0.44 -0.59 train train train 17700093 0.11 -0.30 train test train
18094014 0.64 0.09 train train train
18181709 0.95 -0.09 train train train 18516375 0.44 -0.27 train train train
18854018 0.41 0.64 train train test
19408743 1.44 -0.13 train test train 19666309 0.43 0.34 train train train
19780746 0.33 0.60 train train train
20020024 1.03 0.17 train test train 21609905 1.18 0.09 train train train
22907728 0.58 0.39 train train test
23342258 0.39 -0.07 train train train 25154523 0.16 -0.61 train train train
25311711 -0.37 -0.15 train test train
26444495 -0.69 0.42 train train train 26898179 1.26 0.30 train train test
28106301 0.18 -0.02 train test train
28575179 1.04 0.21 train test train 29082744 1.75 0.30 train train train
29253369 0.57 -0.44 train train train
29761215 -0.33 0.19 train train test
30171803 -0.35 -0.86 train train train
32598100 2.16 -0.09 train test train
32598133 1.58 -0.23 train train test 32598144 2.22 -0.16 train train train
32669060 -0.05 -0.02 train train train
32690930 2.26 0.30 train train test 33460025 0.46 -0.01 train test train
33576920 -0.01 0.77 train test train
33857260 0.76 -0.53 train train train 34883391 1.61 0.77 train test test
35065282 2.57 0.02 train train train
35694065 2.49 -0.14 train train train
237
35694087 2.79 0.21 train train test
35822469 1.47 -0.26 train train test 35860378 1.67 0.07 train train train
36065302 1.68 0.53 train test train
36559225 2.18 0.44 train train test 37680732 2.55 -0.09 train train train
38380017 2.55 0.18 train train train
38380028 2.39 -0.08 train test train 38380039 2.69 -0.05 train train test
38411222 2.14 0.19 train test train
38411255 2.55 -0.09 train train test 38444734 2.12 0.37 train train test
38444778 2.37 0.06 train train test
38444938 2.13 -0.14 train train train 39001020 1.30 -0.08 train train test
39227286 1.21 0.06 train train test
39635319 2.55 -0.23 train test test 40186718 2.19 0.28 train train train
40321764 1.40 0.16 train train test
40458988 0.73 -0.40 train train train
41122707 0.91 -0.06 train train test
41318756 1.20 0.05 train train test
41464431 2.14 0.32 train train test 41464475 2.28 -0.10 train train train
41464511 2.50 0.05 train train train
42240733 0.80 0.47 train train train 50876329 1.76 0.42 train train train
51207319 1.12 -0.56 train train train 51655653 0.64 0.24 train train train
51908168 2.68 -0.28 train train test
52663588 2.47 -0.12 train test train 52663680 2.63 -0.18 train train test
52663704 2.49 0.01 train train test
52663715 2.59 -0.07 train test train 52663726 2.58 -0.15 train test train
52663748 2.66 -0.15 train test train
52663759 2.62 -0.15 train train test 52663782 2.82 -0.28 train test train
52663793 2.52 0.14 train test test
52712057 2.55 -0.07 train train train 52744135 2.52 -0.19 train train train
52886358 0.67 0.30 train test test
52918635 0.58 -0.07 train test test 54135807 1.27 0.13 train train train
57018049 0.27 -0.38 train test train
57117314 1.17 0.19 train test test 57117438 0.80 0.10 train train train
57465288 1.82 0.58 train test train
59080330 1.10 0.15 train train train 59080374 2.53 0.23 train train train
59365605 -1.27 0.14 train test train
60123640 1.59 0.43 train train train 60168889 0.83 -0.26 train test train
60207901 0.69 0.01 train test train
60348609 1.85 0.11 train train train 60782583 0.67 -0.63 train test train
61949766 0.46 -0.09 train train train
62338094 1.02 -0.05 train train test
63376647 0.90 0.24 train train test
66246886 1.15 0.04 train train train
66332965 -0.43 -0.18 train train train 67562394 1.25 -0.35 train test test
68515479 0.13 -0.74 train test train
68515491 -0.12 -0.35 train test train 68526852 -0.09 0.68 train train test
68526863 -0.22 0.30 train test train
69806402 -0.85 -0.22 train train train 70362457 2.14 0.24 train train train
70362468 2.22 0.58 train test test
70362504 1.67 0.02 train test test
238
71585369 1.93 -0.18 train train test
71888896 0.03 0.10 train test test 74472347 2.24 0.19 train test train
74472381 2.55 -0.71 train train train
74472449 2.79 -0.18 train test train 74472530 3.03 -0.38 train test train
74918404 0.67 -0.09 train train train
79060609 0.91 -0.48 train train train 81406373 -0.33 -0.28 train test train
83992692 2.07 -0.10 train test train
85507795 0.37 -0.97 train train train 104294168 1.75 -0.06 train train train
106220819 1.92 0.09 train train test
106220831 2.55 0.21 train train test 107534963 0.86 -0.09 train train test
120068362 0.96 0.09 train train test
126690662 0.52 0.40 train test train 156052685 -0.23 0.39 train test train
182346210 1.68 -0.34 train train test
189084648 1.84 0.11 train train train
189084682 1.65 -0.07 train test train
207122165 1.54 0.11 train train train
(5) 2-hexyl tetralin
0.78 -0.15 test test train
(6)
Spinosad Factor A -0.76 1.17 test test train
(9)
Tri-Me naphthalenes 0.27 0.30 test test test
(10)
C-12-2-LAS 0.63 -0.29 test train test
(30) C-12-5-LAS
0.47 -0.38 test train train
(50)
NL-123 1.68 -0.02 test test test
(90)
NL-83 0.62 -0.32 test train train
(203) 175 Factor L
1.08 -0.59 test train train
(693)
1,1,1 trimethyl butyl benzene 0.40 -0.39 test train test
(6014)
2,3',4,4'-Tetrabromodiphenyl ether (BDE 66) 1.47 -0.04 test train train
50293 1.62 0.31 test train train 50328 0.29 -0.25 test train test
55389 0.18 -0.54 test train train
56382 -0.09 -0.32 test train test 56553 0.34 -0.32 test test test
57749d 1.84 0.42 test test test
5103742d - - test train test 5103719d - - test test test
12789036d - - test train test
58899e 0.51 0.98 test train test 319846e - - test test test
319857e - - test test test
319868e - - test train test 62533 -1.02 0.24 test train train
72548 1.42 1.02 test train test
75092 -0.56 0.05 test train test
75650 -0.92 -0.21 test test test
78308f -0.15 0.67 test train test
1330785f - - test test test 78591 -0.56 -0.73 test train train
79005 -0.94 -0.08 test test test
80433 0.52 -0.53 test train test 82451 -1.31 0.71 test train test
82688 0.25 -0.42 test train train
86306 -0.21 -0.70 test train test 87832 0.81 -0.94 test train train
88722 -0.42 -0.06 test train train
89612 0.12 0.44 test train train
239
89634 -0.73 -0.23 test test train
90415 -0.38 -0.68 test train train 90948 -0.75 -0.02 test train train
91156 -0.20 -0.48 test train train
91203 -0.19 0.60 test train train 91941 -0.12 -0.02 test train train
95498 0.10 -0.82 test train test
95567 -0.73 0.49 test train train 95636 -0.31 0.00 test test train
95783 -0.78 -0.22 test test test
95943 0.80 0.99 test train test 96128 -0.07 -1.07 test test train
98157 0.23 0.42 test test train
98544 -0.43 -0.45 test test test 98839 -0.15 -0.56 test test test
99092 -1.03 0.26 test test train
99547 -0.04 -0.23 test test train 99627 0.41 0.14 test test test
100005 -0.14 -0.35 test train train
100174 -0.55 -0.40 test train test
100403 0.08 -0.50 test test test
100618 -0.44 -0.43 test train test
101144 0.02 -0.17 test train train 103446 -0.06 1.01 test train train
103695 -0.34 -0.31 test test train
106434 0.02 -0.86 test train train 106467 0.36 0.74 test train test
106489 -0.79 0.37 test test train 106934 -0.29 -0.54 test train train
107051 -0.54 -0.39 test train train
108361 0.47 0.69 test train test 108601 -0.05 -1.02 test test train
108703 0.79 0.40 test train train
108872 -0.12 0.70 test test test 108907 -0.07 -0.95 test train test
109693 -0.31 -0.14 test test train
110009 -0.61 -0.23 test train test 115866 -0.96 0.20 test train train
115968 -1.40 0.74 test train train
118741 0.93 1.51 test train test 119642 0.02 0.73 test train test
120718 -0.87 -0.12 test train test
120821 0.58 0.62 test test test 121697 -0.64 -0.33 test train train
121755 -0.61 0.90 test test test
122349 -0.79 -0.15 test train test 126738 -0.96 0.20 test train test
127184 0.33 -1.20 test test test
128370 0.29 0.03 test train test 129000 0.02 0.30 test test train
131099 -0.57 0.20 test test test
132650 -0.34 1.17 test test train 135886 0.19 -0.60 test train test
141935 0.26 -0.17 test train train
206440 0.40 -0.41 test test test 217594 0.26 -0.34 test train test
309002 1.37 0.00 test test train
328847 0.45 1.07 test train test
475036 0.36 0.47 test train train
479276 -1.24 0.50 test test train
493016g 0.12 1.34 test train test 493027g - - test test test
510156 0.30 -0.39 test train test
527606 -0.82 -0.48 test train train 540976 1.45 0.89 test train train
556672 0.61 1.27 test train test
613127 0.42 -0.44 test train test 623267 -0.30 -0.51 test test train
626175 -0.30 -0.32 test train train
629505 0.71 0.32 test train train
240
629732 0.81 0.86 test train test
634662 0.62 0.08 test train test 634913 -0.19 0.47 test train test
636306 -0.05 0.04 test train train
696446 -0.46 -0.33 test train train 732263 1.14 0.67 test train train
779022 0.34 -0.11 test train train
821954 0.30 -0.21 test train train 832699 0.07 0.56 test test train
872059 0.20 0.45 test test train
879390 0.12 -0.81 test train train 947728 0.21 1.58 test test train
1163195 1.97 -0.81 test train test
1241947 -0.52 0.37 test train train 1544190 1.41 1.38 test train test
1570645 -0.56 -0.27 test test test
1839630 0.46 1.25 test test train 2027170 0.27 0.73 test train train
2042140 -1.19 0.40 test train train
2051301 0.46 0.15 test train test
2104963 0.76 0.21 test train test
2189608 0.57 -0.35 test train train
2221956 0.93 0.59 test train train 2255176 -1.89 1.90 test train train
2463845 -0.09 0.10 test train train
2497065 0.49 -1.16 test train test 2980714 0.16 -0.22 test test train
3209221 0.19 0.35 test train test 3268879 1.51 -0.35 test train train
3321504 0.52 -0.39 test train train
4316658 0.14 0.93 test train test 4390049 2.48 -0.67 test test train
5124254 0.47 -1.13 test train train
5436431 1.62 -0.03 test train test 5510996 0.19 -0.41 test test test
5566347 1.79 0.44 test test test
5617414 0.55 -0.38 test test test 6842155 0.94 -0.43 test train train
7012375 1.84 0.04 test train train
7116963 0.72 -0.68 test test train 13150817 0.66 -0.34 test test train
13475826 0.58 0.15 test train train
13540506 0.45 0.00 test test test 14816183 0.08 0.76 test test train
15258738 -0.25 -1.12 test train train
15716082 1.01 0.46 test train train 15862074 1.83 0.01 test train train
16435497 0.54 0.10 test train test
16605917 1.40 0.75 test train train 17088221 0.51 -0.27 test test train
17301234 0.83 0.19 test test test
17312446 0.73 -0.08 test train train 18708708 0.21 0.50 test train train
21564170 -0.33 -0.68 test test test
25321099 0.69 -0.12 test train train 26087478 -0.43 -0.77 test test train
26952205 -0.24 0.70 test test train
28076735 1.42 0.86 test test train
28249776 -0.12 0.78 test test test
29446159 0.71 -0.38 test test test
30746588 0.79 0.23 test test train 31508006 2.44 -0.25 test train test
31710302 1.97 -0.16 test train test
32598111 2.22 0.43 test train train 34883415 1.27 -0.17 test train train
35065271 2.73 -0.19 test train train
36335678 -0.63 0.43 test train test 37680652 2.09 -0.22 test train train
38380051 2.41 -0.18 test train train
38380073 2.46 0.18 test train test
241
38380084 2.41 0.32 test test train
38444858 1.83 0.33 test test train 38640629 0.81 -0.08 test test test
39227582 0.87 -0.44 test train train
40186729 2.96 -0.28 test train train 41464395 2.32 0.01 test train train
51630581 0.81 -0.12 test test train
52663691 2.64 -0.05 test train train 52663771 2.29 0.34 test train train
57057837 0.42 -0.31 test train test
57117449 1.20 -0.20 test train train 57653857 1.44 -0.22 test train train
59039213 1.05 0.67 test train train
59261084 2.75 -0.09 test test test 60145202 2.45 0.00 test test train
60145235 2.62 -0.12 test train train
60851345 1.08 0.04 test test train 61328458 1.22 0.71 test train test
67375308 0.41 0.29 test train train
68194047 2.36 0.11 test train train
68194058 2.38 0.23 test train test
68194149 2.50 -0.27 test test train
68515480 -0.33 -0.13 test train test 71859308 1.77 0.21 test test test
74472336 2.47 0.00 test test train
74472370 2.23 0.14 test test train 74472427 2.77 -0.19 test train test
84852153 0.05 -0.01 test train test 90411511 -0.47 -0.14 test train train
112281773 0.97 0.26 test test test
120068373 0.88 -1.10 test train train 125116236 0.81 -0.58 test test train
169107215 -0.11 0.56 test test train
182677301 1.30 0.26 test test train 204256075 0.78 -0.73 test train test
207122154 2.27 -0.23 test test test
242
Figure 41 - Model fit and predictions for fish HLN using IFS-EPI for (A,B) 1:2, (C,D)
1:1, and (E,F) 2:1 ratios between internal training and internal validation datasets
during cross-validation.
243
Figure 42 - Model fit and predictions for fish HLN using IFS-yscr for (A,B) 1:2, (C,D)
1:1, and (E,F) 2:1 ratios between internal training and internal validation datasets
during cross-validation.
244
Figure 43 - Model fit and predictions for log KOW using IFS-KOW for (A,B) 1:2, (C,D)
1:1, and (E,F) 2:1 ratios between internal training and internal validation datasets
during cross-validation.
245
Table 24 - 2D Fragment Based QSAR from IFS-HLN with a 1:2 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
O=CC(C(H)H)H a 0.588 0.428 1-2 2 1-1 1
C(H)=CH 0.297 0.387 1-3 8 1-1 4
c(Cl)c-ccH 0.282 0.216 2-4 39 2-4 22 O=CC(C(H)C(H)H)H 0.261 0.395 1-2 4 1-1 2
C(H)(CH)C(H)C 0.205 0.158 1-4 4 2-2 2
c(c(Cl)c(c)Cl)(c)Cl 0.170 0.135 1-6 18 1-6 9 c(H)(cH)ccc 0.151 0.135 2-4 35 2-4 15
c(cH)Cl 0.146 0.069 1-7 125 1-6 70
HC(H)H 0.143 0.071 1-14 205 1-12 97 c(H)(c)c-ccH 0.085 0.140 1-6 27 1-4 11
c 0.050 0.048 1-12 303 1-12 152
C(H)H 0.046 0.042 1-28 153 1-20 74
cH 0.035 0.023 1-15 303 1-15 151
c(c)(cc)Br c -0.079 0.042 2-12 5 2-20 3
c(CH)(cH)c(H)c(H)cH -0.101 0.159 1-4 8 1-2 3 O(H)c -0.128 0.227 1-2 29 1-1 11
O -0.225 0.095 1-8 104 1-8 46
{n}{c} -0.237 0.105 1-2 5 1-1 1 Hcc(c(C(H)(H)H)cH)C(H)(H)H -0.291 0.348 1-2 6 1-1 3
N(=O)=O -0.298 0.305 1-2 29 1-1 17 NH -0.331 0.302 1-2 18 1-2 7
N -0.525 0.312 1-2 18 1-2 6
c(c)(c(H)c)Occc -0.528 0.478 1-1 3 1-1 1 OC(C(H)(H)O)H -0.602 0.442 1-1 3 0-0 0
N(H)H -0.621 0.273 1-2 23 1-2 14
OH -0.676 0.257 1-3 48 1-3 19 O=C -0.709 0.153 1-4 48 1-2 14
{n}(c)cH -0.822 0.505 1-1 3 0-0 0
P=O -1.067 0.350 1-1 9 1-1 7 ClcccO
L.C.F. : c(Cl)(cCl)c(cO)Cl b 0.683 0.508 1-2 5 1-1 1
HCC(H)(H)C L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H
b 0.550 0.335 1-2 3 2-2 1
{c}(c)({c}c)c(H)cH
L.C.F. : {c}(c)(c(H)cH){c}(c)cH b 0.385 0.327 1-2 4 0-0 0
HC(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)H
L.C.F. : HC(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)H b -0.231 0.128 1-2 5 2-2 2
cc(ccN(=O)=O)H L.C.F. : cc(cc(N(=O)=O)cH)H
b -0.282 0.452 1-1 6 1-1 2
O=COC(C(H)(H)CH)(H)H
Unique Chemical : 85507795 a,b -0.490 0.362 1-1 1 0-0 0
MW c 0.00429 0.00101 68.1-
748 413
68.1-
959.1 206
intercept
-1.028 0.206
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
246
Table 25 - 2D Fragment Based QSAR from IFS-HLN with a 1:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
O=CC(H)(H)CH
0.897 0.418 1-2 3 1-2 3
O=COH a 0.698 0.265 2-2 2 0-0 0 C=C 0.595 0.388 1-2 7 1-1 4
c(c{c}c(H)cH)(H)cH 0.594 0.213 1-2 6 1-1 1
C(H)=CH 0.472 0.400 1-3 8 1-1 4 c(Cl)c-ccH 0.371 0.157 2-4 39 2-4 22
ccc(c(cC(H)H)H)H 0.275 0.222 1-2 3 1-2 2
c(ccc)-cccc 0.272 0.286 1-4 6 1-2 2 HC(H)H 0.236 0.039 1-14 205 1-12 97
c(H)c(H)c(OP(O)=S)c(H)cH 0.216 0.104 1-2 4 2-2 1
c(H)(cNH)c(H)cH 0.186 0.232 1-4 5 1-4 5
c(cH)Br 0.127 0.074 1-8 19 1-8 8
c(cH)Cl 0.127 0.049 1-7 125 1-6 70
Cl 0.126 0.049 1-12 165 1-10 90 c(H)(cH)cc-cc 0.121 0.059 1-8 13 1-2 6
cH 0.076 0.019 1-15 303 1-15 151
C(H)H 0.065 0.029 1-28 153 1-20 74 {c}(cH){c}{c} -0.039 0.083 1-20 20 2-18 10
O=COC(C(H)H)(H)H -0.135 0.288 1-2 8 2-2 2 c(CH)(cH)c(H)c(H)cH -0.155 0.150 1-4 8 1-2 3
HC(C(C(H)(H)H)C(H)(H)H)H c -0.179 0.141 1-6 12 1-8 7
c(H)(cc){c}{c}cH -0.182 0.179 1-4 8 1-2 3 ccc(c(C(H)(H)H)cH)H -0.204 0.153 1-4 6 1-2 8
Hccc(c(cC(H)(H)H)H)H -0.207 0.157 1-2 7 1-2 5
HCC(H)(H)O -0.220 0.197 1-3 7 1-1 4 HC(C(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)(H)H)H -0.224 0.162 1-2 4 2-2 2
c(H)({c}{c})c(H)c -0.235 0.221 1-2 13 1-1 6
c(H)c-c(c)c -0.249 0.287 1-2 18 1-2 9 HCC(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H -0.273 0.152 1-2 4 0-0 0
O -0.300 0.056 1-8 104 1-8 46
N(=O)=O -0.321 0.153 1-2 29 1-1 17 C(H)(CH)=CH -0.349 0.290 1-2 3 1-2 3
ccC(C(H)(H)H)(H)H -0.384 0.226 1-2 3 0-0 0
O=CN -0.455 0.284 1-2 10 1-1 1 ClC(C(H)(H)Cl)H -0.472 0.312 1-2 3 2-2 1
N(H)H -0.545 0.167 1-2 23 1-2 14
NH -0.564 0.212 1-2 18 1-2 7 c(c)(c(H)c)Occc -0.602 0.444 1-1 3 1-1 1
s -0.614 0.301 1-2 5 1-1 2
O=C -0.621 0.128 1-4 48 1-2 14 HC(NC(H)(H)H)(H)H -0.702 0.200 1-2 5 1-2 4
OH -0.773 0.137 1-3 48 1-3 19
O=COC(H)(H)H -0.787 0.443 1-1 3 0-0 0 P=O -1.231 0.302 1-1 9 1-1 7
ClcccO
L.C.F. : c(Cl)(cCl)c(cO)Cl b 0.990 0.418 1-2 5 1-1 1
HCC(H)(H)C
L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H b 0.824 0.321 1-2 3 2-2 1
cC(H)=C(H)H
L.C.F. : Hcc(c(C(=C(H)H)H)cH)H b 0.271 0.299 1-2 3 0-0 0
c(H)(cC)c(H)cH
L.C.F. : c(H)(c(H)cH)c(H)cC b -0.157 0.084 1-6 5 4-4 1
C(H)(Br)C(H)(Cl)C(H)Br
Unique Chemical : 87843 a,b -0.652 0.411 1-1 1 0-0 0
c(H)(cH){n} L.C.F. : c(H)({n})c(H)cH . c(H)c(H)c(H)cH
a,b -0.779 0.411 1-1 2 0-0 0
MW c 0.00378 0.00064 68.1
-748 413
68.1-
959.1
206
intercept -1.123 0.150
247
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
248
Table 26 - 2D Fragment Based QSAR from IFS-HLN with a 2:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
C=C
0.810 0.287 1-2 7 1-1 4
HCC(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)H a 0.645 0.261 1-1 2 0-0 0
O=CC(C(H)C(H)H)H 0.604 0.290 1-2 4 1-1 2 Hccc(cN(=O)=O)H 0.561 0.267 1-2 8 1-1 5
C(H)=CH 0.554 0.256 1-3 8 1-1 4
HccccN(=O)=O 0.441 0.254 1-2 3 1-1 1 Hcc(cN(=O)=O)H 0.383 0.224 1-2 23 1-2 10
c(H)(cH)cNH c 0.327 0.159 1-4 10 1-5 5
HC=C(H)H 0.320 0.165 1-2 9 1-1 7 HccOP(OC(H)(H)H)(OC(H)(H)H)=S 0.304 0.255 1-2 10 1-2 3
c(Cl)c-ccH 0.293 0.155 2-4 39 2-4 22
HC(C(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.246 0.248 1-2 17 1-2 10
c(H)(cH)c-cc(H)c 0.201 0.080 1-4 21 1-4 10
HCC(C(C(C(H)H)(H)H)(H)H)(H)H 0.197 0.117 1-4 21 1-4 11
c(cH)Br 0.193 0.107 1-8 19 1-8 8 HC(H)H 0.167 0.025 1-14 205 1-12 97
c(H)(cc)c(H)cH c 0.163 0.051 2-4 53 2-6 21
CC(H)C(H)H 0.151 0.142 1-4 5 1-2 3 HCC(C(C(H)H)H)(H)H 0.146 0.057 1-12 15 2-6 6
c(H)(ccH)ccO 0.138 0.132 1-3 15 1-2 7 c(H)(cCcc(H)cH)c(H)c 0.136 0.047 2-8 4 8-8 3
c(cH)Cl 0.097 0.037 1-7 125 1-6 70
c(cH)c-cc 0.097 0.057 1-8 23 1-8 10 c(H)cC(H)H 0.072 0.045 1-7 32 1-4 16
c(Cl)ccH c 0.062 0.047 1-5 97 1-6 49
HccC(C(H)H)H -0.129 0.100 1-4 9 2-4 4 c(H)(c(H)cH)c(H)cC=O -0.149 0.101 1-4 16 2-2 3
BrCH c -0.156 0.124 3-6 3 1-1 1
HC(C(C(H)(H)H)(C(H)(H)H)C(H)(H)H)H -0.191 0.307 1-2 6 1-2 4 Hcc(c(cC(H)(H)H)H)H c -0.199 0.213 1-2 7 1-3 5
Br c -0.225 0.094 1-7 30 1-10 12
ccc(c(C(H)(H)H)cH)H -0.232 0.125 1-4 6 1-2 8 O -0.331 0.043 1-8 104 1-8 46
O(CH)c -0.336 0.341 1-2 6 0-0 0
ccC(C(H)(H)H)(H)H -0.383 0.224 1-2 3 0-0 0 {c}({c}c){c}{c} -0.412 0.239 1-1 6 1-1 4
s -0.448 0.285 1-2 5 1-1 2
HCC(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H -0.455 0.154 1-2 4 0-0 0 ClC(C(H)(H)Cl)H -0.489 0.274 1-2 3 2-2 1
C(H)(CH)=CH -0.510 0.233 1-2 3 1-2 3
O=C -0.538 0.100 1-4 48 1-2 14 N(H)H -0.551 0.123 1-2 23 1-2 14
Hcc(c(C(H)(H)H)cH)C(H)(H)H -0.568 0.338 1-2 6 1-1 3
c(c)(c(H)c)Occc -0.692 0.405 1-1 3 1-1 1 OH -0.744 0.095 1-3 48 1-3 19
Occ(cC(H)(H)H)H -0.783 0.283 1-1 5 1-1 3
{n}(c)cH -0.784 0.386 1-1 3 0-0 0 N -0.833 0.159 1-2 18 1-2 6
NH -0.872 0.179 1-2 18 1-2 7
N(=O)=O -0.919 0.302 1-2 29 1-1 17
P=O -1.085 0.221 1-1 9 1-1 7
Nc(c)cH
L.C.F. : c(c)(N)c(H)c b 1.215 0.507 1-1 3 0-0 0
ClcccO
L.C.F. : c(Cl)(cCl)c(cO)Cl b 1.142 0.320 1-2 5 1-1 1
HCC(H)(H)C L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H
b 0.676 0.417 1-2 3 2-2 1
{c}(cH)({c}cH)cc(H)cH
L.C.F. : {c}1(cc(H)cH){c}c(H)c(H)c(H)c1H b 0.657 0.202 1-2 7 0-0 0
c(cOc(cH)cH)(H)c
L.C.F. : c(c(H)c)(Oc(cH)c(H)cH)c(H)cH b 0.453 0.393 1-2 3 1-1 2
c(H)({c}{c}o)c(H)cH a,b 0.422 0.295 1-2 2 0-0 0
249
L.C.F. : {c}1(c(H)cH){c}(cH)o{c}{c}1c(H)c(H)cH
{c}(c)({c}c)c(H)cH L.C.F. : {c}(c)(c(H)cH){c}(c)cH
b 0.365 0.207 1-2 4 0-0 0
Hc{c}{c}c(cC(H)(H)H)H
L.C.F. : Hc{c}{c}(c(cC(H)(H)H)H)cH b 0.234 0.100 1-8 6 2-6 2
c(c-ccc)cc
L.C.F. : c(c-cc(Cl)cCl)(Cl)c(c)Cl b 0.133 0.161 1-8 11 1-4 4
Occ(c(cCC(H)(H)H)H)H L.C.F. : FC(C(Oc1c(c(c(c(c1H)H)CC(H)(H)H)H)H)(F)F)(H)F
b -0.175 0.098 4-4 3 4-4 1
HC(C(H)(H)Br)H
Unique Chemical : 4101682 a,b -0.228 0.197 2-2 1 0-0 0
{c}(cH)({n}){c}cH
L.C.F. : {c}({n})({c}cH)c(H)cH b -0.351 0.132 1-2 5 1-1 1
c(H)(ccH)cO L.C.F. : c(cH)(O)c(H)ccH
b -0.396 0.300 1-2 5 1-1 2
O=COC(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H
L.C.F. : O=COC(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H b -0.418 0.169 2-2 3 2-2 2
c1(H){c}({c}(cH)c(H)cc1)cH
L.C.F. : c1(H){c}({c}(cH)c(H)cc1)cH b -0.599 0.243 1-2 5 1-1 1
MW c 0.00571 0.00049 68.1-
748 413
68.1-
959.1 206
intercept -1.067 0.110
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
250
Table 27 - 2D Fragment Based QSAR from IFS-EPI with a 1:2 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
c-c c 0.670 0.260 1-1 60 1-2 32
c(H)({c}{c}c(H)cCH)c(H)cH a 0.567 0.394 1-1 2 1-1 1
HC(C(C(H)(H)C)H)(H)H 0.447 0.197 1-2 5 2-2 1 C(H)=CH 0.374 0.309 1-3 6 1-1 6
HC(C(C(H)(H)H)(H)H)H 0.352 0.233 1-3 46 1-3 24
Hcc(C(C(H)(H)H)C(H)(H)H)cH c 0.312 0.305 1-3 10 1-9 4 HC(C(C(C(C(H)H)(H)H)(H)H)H)H 0.301 0.394 1-3 20 1-3 13
c(cH)(Cl)c(H)c(H)c 0.234 0.201 1-4 39 1-3 28
HCC(C(H)H)H 0.133 0.091 1-10 17 2-9 10 Cl 0.123 0.076 1-12 173 1-9 87
c(c)cH 0.111 0.049 1-8 181 1-8 93
c(cH)Br 0.039 0.029 1-8 18 2-8 9
HC(C(C(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.022 0.019 1-12 18 1-5 4
ClC(H)H -0.098 0.304 1-6 11 1-2 7
OC(H)H -0.127 0.116 1-16 39 1-6 19 Hccc(c(cC(H)(H)H)H)H -0.245 0.176 1-2 6 1-2 6
cc(cOC(H)(H)H)H -0.299 0.446 1-2 3 1-1 1
P(O)=O -0.309 0.129 1-3 13 2-3 4 NH -0.377 0.224 1-2 21 1-1 4
S -0.470 0.389 1-2 9 1-2 6 N(=O)=O -0.552 0.289 1-2 34 1-2 12
OH -0.607 0.169 1-3 42 1-3 25
O=C -0.684 0.154 1-4 37 1-2 25 N(H)H -0.686 0.279 1-2 24 1-2 13
c(H)({c}ccH)c(H)cH
L.C.F. : {c}({c})(ccH)c(H)c(H)cH b 0.662 0.250 1-2 9 1-2 5
cC(C(H)(H)H)c
L.C.F. : Hc2c(c(C(c1c(c(cc(c1H)H)H)H)C(H)(H)H)c(c(c2)H)H)H b 0.418 0.324 1-2 3 1-1 1
HC([Si]C(H)(H)H)(H)H L.C.F. : HC([Si](O[Si](O[Si](O[Si](C(H)(H)H)C(H)(H)H)
(C(H)(H)H)C(H)(H)H)(C(H)(H)H)C(H)(H)H)C(H)(H)H)(H)H
a,b 0.118 0.048 4-6 2 5-5 1
Hc{c}{c}c(cC(H)H)H L.C.F. : Hc1c(c({c}({c}c1H)c(cC(H)H)H)H)H
a,b -0.257 0.161 1-2 2 0-0 0
c(cH)(N)cH
L.C.F. : c(H)(c(N)c(H)cH)cH b -0.286 0.230 1-2 4 1-2 2
{c}({c}cH){n}cH
L.C.F. : {c}({c}cH)({n}cH)c(H)cH b -0.342 0.355 1-2 3 0-0 0
MW 0.00272 0.00078 68.1-959.1
417 74.1-748
211
intercept -0.511 0.170
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
251
Table 28 - 2D Fragment Based QSAR from IFS-EPI with a 1:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
c(c{c}c(H)cH)(H)cH 0.985 0.389 1-2 4 1-1 3
C=C 0.775 0.350 1-2 7 1-1 7
C(H)=CH 0.483 0.260 1-3 6 1-1 6 [Si] c 0.422 0.288 1-6 3 1-7 3
c(-c)(cCl)c(H)cH 0.419 0.254 1-2 21 1-1 7
O=CCH 0.397 0.500 1-2 9 1-2 8 HC(C(C(H)(H)H)(H)H)H 0.313 0.195 1-3 46 1-3 24
C 0.282 0.063 1-10 54 1-4 36
c({c}{c}c(H)cH)(H)cH 0.279 0.241 1-5 22 1-3 11 c 0.259 0.022 1-12 307 1-11 149
c(-c(cH)cH)c 0.251 0.202 1-2 14 1-2 14
CH 0.240 0.049 1-17 92 1-17 63
c(H)({c}c(H){c}c(H)cH)c(H)cH 0.198 0.158 1-4 5 1-1 1
c(H)(cH)cc-cc 0.165 0.082 1-8 15 1-4 4
c(cH)(Br)cH 0.138 0.078 1-3 10 1-3 6 c(cH)Cl 0.126 0.035 1-7 129 1-6 66
C(H)H 0.061 0.039 1-28 147 1-16 86
OC(H)H -0.077 0.079 1-16 39 1-6 19 Brc(c)c -0.082 0.033 1-10 6 2-5 3
{n} -0.207 0.203 1-3 16 1-2 8 ClC(C(H)(H)Cl)H -0.230 0.160 1-2 3 2-2 1
N#C -0.275 0.259 1-2 8 1-2 6
c(cOH)Cl -0.277 0.211 1-2 4 2-2 1 {c}({c}){c}c -0.280 0.347 1-2 7 1-1 4
C(H)(O)C(H)CH -0.287 0.159 2-7 4 2-7 3
Hccc(c(cC(H)(H)H)H)H -0.355 0.164 1-2 6 1-2 6 O=CO -0.379 0.244 1-2 18 1-2 13
Hcc(cc(C(H)(H)H)cH)C(H)(H)H a -0.429 0.263 1-1 2 0-0 0
NH -0.449 0.172 1-2 21 1-1 4 N(=O)=O -0.451 0.174 1-2 34 1-2 12
c(H)(cc){c}{c}c(H)cH -0.495 0.367 1-2 3 1-2 4
O=C -0.502 0.145 1-4 37 1-2 25 S(cc)=O a -0.519 0.345 1-2 2 0-0 0
HcccOC(H)(H)H -0.551 0.385 1-2 5 0-0 0
N -0.561 0.198 1-2 17 1-2 7 c(H)(cH)cOCH -0.602 0.314 1-2 4 2-2 1
OH -0.738 0.119 1-3 42 1-3 25
N(H)H -0.774 0.202 1-2 24 1-2 13 P=O -0.922 0.229 1-1 13 1-1 4
cc(C(H)H)c c -0.942 0.616 1-1 3 1-2 2
Nc(c)cH L.C.F. : c(c)(N)c(H)c
b 0.931 0.481 1-1 3 0-0 0
HCC(CC(H)(H)CH)(H)H
Unique Chemical : 50876329 a,b 0.815 0.367 1-1 1 0-0 0
cccOC(H)(H)H
L.C.F. : Clcc(c(cOC(H)(H)H)Cl)Cl b 0.632 0.385 1-2 3 0-0 0
cCC L.C.F. : c(H)c(H)c(CC)c(H)cH
a,b 0.456 0.278 2-2 2 0-0 0
c1(H)c(H)c(H)c(C)c(H)c1H
L.C.F. : c1(H)c(H)c(H)c(C)c(H)c1H b -0.306 0.194 1-3 4 0-0 0
ccccN(=O)=O
L.C.F. : Clcc(c(c(cN(=O)=O)Cl)Cl)Cl a,b -0.596 0.342 1-2 2 0-0 0
intercept -0.463 0.100
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
252
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
253
Table 29 - 2D Fragment Based QSAR from IFS-EPI with a 2:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
C=C 1.096 0.415 1-2 7 1-1 7
[Si] c 0.749 0.274 1-6 3 1-7 3
C(=O)C(H)CH 0.565 0.287 1-2 6 1-2 5 C(H)=CH 0.512 0.223 1-3 6 1-1 6
c(H)(cc-ccc(H)cH)cH 0.468 0.204 1-4 5 1-1 1
S=PO 0.358 0.071 2-3 21 2-3 10 HC=C(H)H 0.350 0.162 1-2 12 1-1 4
C 0.258 0.080 1-10 54 1-4 36
HC(C(C(C(C(H)H)H)(H)H)H)H 0.251 0.153 1-5 7 1-5 5 c(Cl)c-ccH 0.242 0.116 2-4 41 2-4 20
HC(C(C(H)(H)H)(H)H)H 0.241 0.127 1-3 46 1-3 24
c(H)(cH)c(H){c} c 0.231 0.058 1-6 37 1-8 23
c 0.208 0.021 1-12 307 1-11 149
CH 0.198 0.037 1-17 92 1-17 63
c(cH)Cl 0.191 0.030 1-7 129 1-6 66 c(c)(Cl)cc(H)cH 0.175 0.085 1-4 24 1-2 9
HC(C(C(H)(H)H)H)(H)H 0.172 0.111 1-3 31 1-2 17
HC(CC(H)(H)H)(H)H 0.141 0.053 1-9 30 1-9 15 c(cH)Br 0.120 0.032 1-8 18 2-8 9
C(H)H 0.073 0.025 1-28 147 1-16 86 c(Cl)(ccCl)c(c)Cl 0.063 0.030 1-12 24 1-7 11
HC(C(C(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.048 0.028 1-12 18 1-5 4
{n} -0.209 0.144 1-3 16 1-2 8 S -0.234 0.188 1-2 9 1-2 6
NH -0.249 0.131 1-2 21 1-1 4
c(H)(ccH)ccOH -0.263 0.208 1-4 9 1-2 4 O -0.267 0.046 1-8 105 1-8 47
Hccc(c(cC(H)(H)H)H)H -0.288 0.162 1-2 6 1-2 6
c1(-ccH)cc(H)c(H)c(H)c1H c -0.293 0.332 1-2 4 1-4 4 N(=O)=O -0.383 0.121 1-2 34 1-2 12
O=C(C(H)H)O -0.456 0.410 1-2 3 1-1 4
O=CN -0.470 0.258 1-2 7 1-2 4 O=C -0.475 0.091 1-4 37 1-2 25
N(H)H -0.567 0.133 1-2 24 1-2 13
OH -0.742 0.102 1-3 42 1-3 25 cc(C(H)H)c c -0.924 0.402 1-1 3 1-2 2
HCC(H)(c)c
L.C.F. : c2(H)cc(H)c(H)c(C(H)(CH)c1c(H)c(H)cc(H)c1H)c2H a,b 0.890 0.458 1-1 2 0-0 0
cccOC(H)(H)H
L.C.F. : Clcc(c(cOC(H)(H)H)Cl)Cl b 0.824 0.304 1-2 3 0-0 0
c1(H)c(H)c(Cc)c(H)c(H)c1 L.C.F. : c2(H)cc(H)c(H)c(Cc1c(H)c(H)cc(H)c1H)c2H
b 0.560 0.171 2-2 5 2-2 1
c1(H)c(H)c(H)c(-c)c(H)c1H
L.C.F. : c1(H)c(H)c(H)c(-c)c(H)c1H b 0.302 0.308 1-1 9 1-1 3
c(cH)(ccH)OccH
L.C.F. : c(cH)(cc(H)c)OccH b 0.174 0.145 1-2 11 1-2 4
c(cH)(N)cH L.C.F. : c(H)(c(N)c(H)cH)cH
b -0.316 0.201 1-2 4 1-2 2
{c}(cH)({n}){c}cH
L.C.F. : {c}({n})({c}cH)c(H)cH b -0.357 0.327 1-2 5 1-1 1
BrccC(H)(H)H
L.C.F. : Brcc(cC(H)(H)H)Br a,b -0.383 0.184 1-2 2 0-0 0
C(Oc(cH)cH)H L.C.F. : Hcc(c(OC(C(H)(H)H)H)cH)H
b -0.437 0.424 1-1 3 1-1 1
Hccc(cN(H)H)H
L.C.F. : Hcc(cc(cN(H)H)H)H b -0.497 0.383 1-1 3 1-1 2
ClC(C(=C(H)H)H)H
Unique Chemical : 760236 a,b -0.842 0.439 1-1 1 0-0 0
S=O L.C.F. : c(H)(cc)ccH
a,b -0.894 0.451 1-1 2 0-0 0
intercept -0.597 0.087
254
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
255
Table 30 - 2D Fragment Based QSAR from IFS-yscr with a 1:2 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
c(H)(c(H)c(H)cCl)cCl 1.519 0.962 1-1 5 1-1 2
c(H)(c)c(cOc(cH)cH)H 1.454 0.916 1-2 4 1-1 1
O=CC(H)c 1.094 0.783 1-1 3 0-0 0 c(H)(c-cc(H)cCl)c(H)c c 0.703 0.398 1-2 13 1-4 4
C(NC(H)H)=O 0.538 0.636 1-2 4 2-2 1
c(c)(ccH)-ccccH 0.484 0.709 1-2 4 0-0 0 HCC(C(H)(C(H)H)H)(H)H 0.206 0.199 1-4 33 1-4 18
c(Cl)(c(H)c)c(c)Cl -0.409 0.501 1-3 23 1-3 15
c(Cl)(c(H)cH)c(c)Cl -0.920 0.652 1-2 22 1-2 10 c(cH)(cc)-ccc
L.C.F. : c(Cl)(cCl)c(cH)-cc(Cl)cCl b 0.593 0.454 1-2 9 1-2 5
c(c)(-cc)c(H)c
L.C.F. : c(-ccCl)(cCl)c(H)cCl b 0.376 0.394 1-2 17 1-2 6
intercept 0.395 0.108
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
256
Table 31 - 2D Fragment Based QSAR from IFS-yscr with a 1:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
c(H)(c(H)c(H)cCl)cCl - 1.548 0.753 1-1 5 1-1 2
c(H)(c)c(cOc(cH)cH)H 1.446 0.720 1-2 4 1-1 1
c(OH)(c(H)c)c(H)cH 1.219 0.747 1-1 3 1-1 3 c(OC(H)H)c 0.977 0.546 1-2 3 0-0 0
NC(H)C(H)H 0.878 0.496 1-2 4 1-1 2
HCC(C(C(H)(H)H)(C(H)(H)H)H)(H)H 0.821 0.718 1-1 4 1-1 1 c(H)(cH){c}{c}cc 0.804 0.440 1-2 3 0-0 0
c(H)(c-cc(H)cCl)c(H)c c 0.718 0.386 1-2 13 1-4 4
C(NC(H)H)=O 0.670 0.579 1-2 4 2-2 1 Sc 0.666 0.582 1-2 3 1-1 1
c(cH)(Cl)c(H)ccOH 0.614 0.419 1-2 3 1-1 1
c(cH)C(H)CH 0.435 0.232 1-4 5 4-4 1
c(cH)(-cc(H)cH)c(H)cc 0.281 0.410 1-4 6 1-4 4
HC(C(C(C(C(H)H)(H)H)(H)H)H)H 0.237 0.257 1-3 22 1-3 10
c(cH)(C(=O)ccH)c(cH)C=O -0.311 0.488 1-4 5 1-4 2 c(Cl)(c(H)c)c(c)Cl -0.463 0.430 1-3 23 1-3 15
c(H){c}1{c}c(H)c(H)c(H)c1 -0.703 0.433 1-1 3 1-1 3
c(Cl)(c(H)cH)c(c)Cl -0.959 0.480 1-2 22 1-2 10 S=O
L.C.F. : c(H)(cc)ccH a,b 1.445 0.890 1-1 2 0-0 0
Hcc(c(C(C(H)(H)H)(H)H)c(H)cH)H
L.C.F. : Hc1c(c(C(C(H)(H)H)(H)H)c(c(c1)H)H)H a,b 1.432 0.889 1-1 2 0-0 0
c(H)(cCl)c(H){c} L.C.F. : c(H)(cCl)c(H){c}{c}o{c}
b 0.972 0.468 1-2 4 0-0 0
c(cH)(cc)-ccc
L.C.F. : c(Cl)(cCl)c(cH)-cc(Cl)cCl b 0.661 0.404 1-2 9 1-2 5
c(c)(-ccc)ccH
L.C.F. : c(cCl)(-cc(Cl)cCl)c(cH)Cl b 0.503 0.588 1-2 7 1-2 2
c(c)(-cc)c(H)c L.C.F. : c(-ccCl)(cCl)c(H)cCl
b 0.397 0.331 1-2 17 1-2 6
P(=S)Oc(c)c(H)cH
L.C.F. : P(O)(=S)Oc(ccH)c(H)cH b -1.051 0.698 1-1 4 0-0 0
intercept 0.321 0.086
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
257
Table 32 - 2D Fragment Based QSAR from IFS-yscr with a 2:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and number of
chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
c(OH)(c(H)c)c(H)cH 2.077 0.657 1-1 3 1-1 3
c(H)(c)c(cOc(cH)cH)H 1.845 0.627 1-2 4 1-1 1
c(H)(c(H)c(H)cCl)cCl 1.638 0.573 1-1 5 1-1 2 Clcc(c(c(N(=O)=O)c)H)H a 1.630 0.774 1-1 2 1-1 1
HCC(C(C(H)(H)H)(C(H)(H)H)H)(H)H 1.392 0.616 1-1 4 1-1 1
cc(cc(N(H)H)cH)H 1.381 0.646 1-2 5 1-1 1 c(OC(H)H)c 1.190 0.485 1-2 3 0-0 0
NC(H)C(H)H 1.135 0.409 1-2 4 1-1 2
Hcc(c(C(H)(H)H)cH)C(H)(H)H 1.104 0.394 1-2 6 1-2 3 HC(C(C(C(C(H)(H)H)(H)H)(C(H)(H)H)H)(H)H)H 1.048 0.425 1-1 3 1-1 1
c(H)(cH){c}{c}cc 1.041 0.387 1-2 3 0-0 0
c(cH)(Cl)c(H)ccOH 0.989 0.377 1-2 3 1-1 1
c(H)(c-cc(H)cCl)c(H)c c 0.941 0.338 1-2 13 1-4 4
HCC(C(H)(H)H)(C(H)(H)H)H c 0.905 0.811 1-1 8 2-2 1
Sc 0.893 0.414 1-2 3 1-1 1 HC(=CC(C(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.876 0.418 1-1 3 1-1 1
C(=C(H)H)Cl a 0.870 0.353 2-2 1 0-0 0
Hcc(c(C(C(H)(H)H)(H)H)cH)H c 0.810 0.276 2-2 4 4-4 1 C(H)(H)OH c 0.773 0.506 1-1 5 1-2 3
c(C(H)C=O)cH 0.740 0.593 1-2 3 0-0 0 c(H)cc(H){c}cH 0.737 0.346 1-2 3 1-1 2
C(NC(H)H)=O 0.731 0.405 1-2 4 2-2 1
c(cH)(-cc(H)cH)c(H)cc 0.638 0.379 1-4 6 1-4 4 c(cH)(cc)-c(c)cc 0.616 0.281 1-2 5 1-2 3
c(cH)(-cc(H)cH)cc(H)cH c 0.519 0.318 1-2 8 2-4 3
ClC(C(C(H)(H)Cl)H)(H)H a 0.507 0.224 1-3 2 0-0 0 {c}(cH)({c}c)c(H)cH 0.475 0.103 1-4 10 1-4 7
c(H)(cH)cC(H)CH 0.428 0.215 2-4 3 4-4 1
HC(C(C(C(C(H)H)(H)H)(H)H)H)H 0.279 0.158 1-3 22 1-3 10 C 0.141 0.085 1-10 61 1-4 26
c(cH)(C(=O)ccH)c(cH)C=O -0.266 0.315 1-4 5 1-4 2
c(H)(cOH)c(H)c -0.316 0.200 1-4 10 1-2 7 c(H)(cBr)c(H)cH -0.381 0.467 1-4 4 2-2 1
c(-c(cH)cH)c -0.423 0.278 1-2 19 1-2 9
c(c-c(cH)cH)(H)c -0.521 0.317 1-3 13 1-2 8 cc(c(c(C(H)(H)H)c)H)H -0.658 0.288 1-2 6 1-2 4
O=CC(C(H)H)H a -0.787 0.562 1-2 2 1-1 1
c(H)cOc(c(H)cH)c(H)cH c -1.026 0.568 1-2 5 1-4 4 c(H)(c)c(H){c} -1.119 0.395 1-1 5 1-1 2
c(Cl)(c(H)cH)c(Cl)cCl -1.173 0.600 1-2 10 1-2 4
HCC(C(C(CH)(CH)H)(H)H)(H)H -1.187 0.717 1-1 3 1-1 1 Hcc(c(c(H)cH)N(H)H)Cl -1.248 0.681 1-2 3 1-1 1
c(H)(cc-c(c)cH)c(H)c -1.393 0.586 1-1 7 1-1 2
S=O L.C.F. : c(H)(cc)ccH
a,b 1.734 0.744 1-1 2 0-0 0
c(H)(cCl)c(H){c}
L.C.F. : c(H)(cCl)c(H){c}{c}o{c} b 1.226 0.377 1-2 4 0-0 0
c(H)(cH)cOccCl
L.C.F. : c(Cl)(cOcc(H)c(H)ccH)c(H)c b 1.092 0.432 1-2 5 0-0 0
c(H)(ccH)cO
L.C.F. : c(cH)(O)c(H)ccH b 0.944 0.518 1-2 6 1-1 1
HC(C(OP(OC(C(H)(H)H)(H)H)=S)(H)H)(H)H
L.C.F. : HC(C(OP(OC(C(H)(H)H)(H)H)=S)(H)H)(H)H b 0.886 0.539 1-1 5 1-1 3
cc(ccC(H)H)H
L.C.F. : Hccc(ccC(C(H)H)(H)H)H b 0.868 0.543 1-2 3 2-2 1
BrccC(H)(H)H L.C.F. : Brcc(cC(H)(H)H)Br
a,b 0.773 0.554 1-2 2 0-0 0
Hcc({c}c(c({c}({c}){c})H)H)H
L.C.F. : {c}32c(H)c(H){c}(cH){c}(cH){c}3c(H)c(H){c}1c(H)c(H)c(H)c(H){c}21
a,b 0.641 0.312 1-2 2 0-0 0
c(c)(-cc)c(H)c
L.C.F. : c(-ccCl)(cCl)c(H)cCl b 0.546 0.219 1-2 17 1-2 6
258
c(H)(cc)c(H)c-cc
L.C.F. : c(H)(c(H)c-ccCl)ccCl b 0.318 0.346 1-2 11 1-2 6
c(O)(cCl)c(H)cH
L.C.F. : c(O)(cCl)c(H)c(H)c b -0.604 0.646 1-2 6 1-1 2
c(H)(cOP=S)c(H)cH L.C.F. : P(O)(=S)Occ(H)c(H)c(H)cH
b -0.606 0.297 1-2 3 0-0 0
cc(c(cC(C(H)H)(C(H)H)H)H)H
L.C.F. : Hccc(c(cC(C(C(H)H)(H)H)(C(H)H)H)H)H a,b -0.766 0.551 1-2 2 0-0 0
c(H)({c})c-c
L.C.F. : {c}21c(H)c(H)c(H)c(H){c}2c(H)cc(-cc(H)c(H)cH)c1H a,b -0.868 0.339 1-2 2 0-0 0
HccN(C(H)H)H L.C.F. : Hcc(c(cN(C(C(H)(H)H)(H)H)H)H)H
a,b -0.893 0.567 1-2 2 0-0 0
c(H)(c(H)c(H){c})c
L.C.F. : c1(H)c{c}{c}c(H)c1H b -1.343 0.360 1-2 7 1-2 4
intercept 0.124 0.074
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
259
Table 33 - 2D Fragment Based QSAR from IFS-KOW with a 1:2 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
HC=C(H)H 0.471 0.280 1-2 9 1-1 7
C(H)=CH 0.349 0.244 1-3 8 1-1 4
HC(C(C(H)(H)H)(C(H)(H)H)C(H)(H)H)H 0.348 0.234 1-2 6 1-2 4 c 0.336 0.058 1-12 303 1-12 152
HC(C(C(H)(H)H)(C(H)(H)H)H)H 0.293 0.183 1-2 14 1-1 6
HC(H)H 0.288 0.044 1-14 205 1-12 97 HCC(OC(H)H)=O 0.264 0.430 1-1 4 1-1 1
C(H)H 0.225 0.044 1-28 153 1-20 74
HC(C(C(H)(H)H)(H)H)H 0.194 0.267 1-3 43 1-3 26 c(O)(cH)c(H)c(H)cH c 0.186 0.125 1-4 8 3-6 4
{c}({c})cH 0.154 0.049 1-16 54 2-12 25
HC(C(C(C(H)(C(H)H)H)(H)CH)(H)H)H 0.141 0.300 1-2 5 1-2 2
c(OH)c(H)c 0.140 0.094 1-2 8 1-1 2
c(H)c(H)c(H)cH 0.085 0.054 1-6 108 1-6 51
HccC(H)(H)H -0.082 0.079 1-6 39 1-4 21 HC(CC(H)H)H -0.143 0.314 1-6 9 1-1 3
ClC(C(H)(H)Cl)H -0.147 0.131 1-2 3 2-2 1
C(H)(O)C(H)CH -0.168 0.207 2-7 4 3-7 2 cc -0.198 0.052 1-12 199 1-12 95
Br c -0.414 0.082 1-7 30 1-10 12 O=CC(H)H -0.519 0.308 1-4 9 1-2 5
O -0.616 0.079 1-8 104 1-8 46
O=C -0.668 0.125 1-4 48 1-2 14 HC(OC(H)H)H -0.708 0.308 1-8 7 3-3 1
O=CC(H)(H)H -0.814 0.498 1-2 3 0-0 0
HNC(H)(H)H -0.863 0.325 1-1 5 1-1 2 N(=O)=O -0.906 0.229 1-2 29 1-1 17
{n} -0.937 0.199 1-3 18 1-3 6
OH -0.972 0.158 1-3 48 1-3 19 N(H)H -1.042 0.256 1-2 23 1-2 14
N -1.220 0.221 1-2 18 1-2 6
P=O -1.567 0.365 1-1 9 1-1 7 C(H)(C)(CH)CH
L.C.F. : HCC1(C2(C(C(C1H)(C(=C2Cl)Cl)Cl)(Cl)Cl)Cl)H b 0.107 0.203 1-2 3 1-2 2
n L.C.F. : c(cH)Cl . c(cH)cH
b -0.385 0.413 1-1 4 1-1 3
HC(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)H
L.C.F. : HC(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)H b -0.662 0.292 1-2 5 2-2 2
S=O
L.C.F. : c(H)(cc)ccH a,b -0.846 0.261 1-1 1 0-0 0
S(=O)=O L.C.F. : c(H)c . O=S(c)=O
b -1.764 0.475 1-1 4 1-1 3
MW c 0.01220 0.00120 68.1-
748 413
68.1-
959.1 206
intercept 0.862 0.167
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
260
Table 34 - 2D Fragment Based QSAR from IFS-KOW with a 1:1 Internal Training to
Internal Validation Ratio During Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
C(H)=CH 0.505 0.199 1-3 8 1-1 4
HCC(C(C(C(C(C(H)H)(H)H)H)(H)H)(H)H)(H)H 0.471 0.164 1-2 3 1-1 3 HC=C(H)H 0.469 0.149 1-2 9 1-1 7
C=C(H)H 0.383 0.222 1-1 9 1-1 3
O=C(N)NH 0.329 0.367 1-1 3 0-0 0 HC(H)H 0.328 0.026 1-14 205 1-12 97
N(H)(H)cc 0.313 0.182 1-2 10 1-2 10
c1(H)c(H)c(OP=S)c(H)c(H)c1 0.293 0.163 1-1 3 1-1 1 c 0.277 0.030 1-12 303 1-12 152
c(H)(cC)cc 0.269 0.134 1-2 8 1-2 3
HC(C(C(H)(H)H)(C(H)(H)H)H)H 0.267 0.140 1-2 14 1-1 6
c(OH)c(H)c 0.259 0.085 1-2 8 1-1 2
C(H)H 0.242 0.021 1-28 153 1-20 74
HC(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.229 0.242 1-2 20 1-2 15 C(H)=C 0.223 0.173 1-2 11 1-2 4
HC(C(C(C(H)(C(H)H)H)(H)CH)(H)H)H 0.220 0.228 1-2 5 1-2 2
c(H)(cO)c(H)cH c 0.165 0.155 1-5 15 3-6 5 c(H)(c(-c)c(H)cH)cH 0.164 0.117 1-2 17 1-2 8
O(ccH)ccH 0.143 0.127 1-4 12 1-4 9 {c}({c})cH 0.137 0.030 1-16 54 2-12 25
Hcc(cC(H)(H)H)H c 0.125 0.166 1-2 20 1-3 15
ClC=C 0.080 0.036 2-6 6 2-4 4 cH 0.051 0.014 1-15 303 1-15 151
Brcc c -0.066 0.056 1-12 16 1-20 9
c(H)(c)ccc -0.074 0.055 2-4 46 2-4 19 HC(CC(H)H)H -0.152 0.164 1-6 9 1-1 3
ClC(H)H -0.173 0.197 1-6 11 1-3 7
NCH -0.185 0.255 1-1 4 1-1 2 cc(C(H)(H)H)c -0.204 0.142 1-2 5 1-1 1
Hcc(C(H)(H)H)cH -0.214 0.183 1-3 14 1-1 12
c(H)(cOcc)cccH -0.216 0.160 1-2 8 1-2 4 HC(C(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H)H -0.223 0.098 1-3 6 1-2 5
HcccC(H)(H)H -0.233 0.091 1-4 24 1-3 12
HC(C(H)(H)O)(H)H -0.251 0.113 1-3 15 1-2 7 Hcc(OC(H)(H)H)c -0.349 0.279 1-2 4 1-1 1
Hcc(OC(C(H)(H)H)H)cH -0.368 0.290 1-2 4 0-0 0
Br c -0.380 0.064 1-7 30 1-10 12 O=C -0.604 0.075 1-4 48 1-2 14
O -0.612 0.055 1-8 104 1-8 46
HC(OC(H)H)H -0.616 0.188 1-8 7 3-3 1 N(H)(C=O)cc(H)cH -0.622 0.394 1-2 5 0-0 0
O=CC(H)H -0.739 0.198 1-4 9 1-2 5
N(=O)=O -0.877 0.118 1-2 29 1-1 17 {n} -0.906 0.120 1-3 18 1-3 6
OH -1.032 0.100 1-3 48 1-3 19
N#C -1.100 0.289 1-2 6 1-2 8 P=O -1.171 0.245 1-1 9 1-1 7
HNC(H)(H)H -1.199 0.255 1-1 5 1-1 2
N(H)H -1.250 0.164 1-2 23 1-2 14
N -1.343 0.151 1-2 18 1-2 6
P(=S)Occc(H)cCl
L.C.F. : Clcc(ccOP(O)(O)=S)H b 0.611 0.295 1-1 3 0-0 0
c(Cl)(c-ccH)c(H)c
L.C.F. : c(c-ccH)(Cl)c(H)cCl b -0.068 0.052 1-2 12 1-2 7
c(H)(c)c-ccc L.C.F. : c(Cl)(c-cc(H)cCl)cCl
b -0.160 0.280 1-4 17 1-2 7
c(c)c-cc
L.C.F. : c(c-ccCl)(Cl)cCl b -0.227 0.126 1-8 25 1-8 10
c(C)c
L.C.F. : c(C)(c)cH b -0.241 0.131 1-2 12 1-2 3
c(Cl)cCH b -0.465 0.222 1-1 4 1-1 1
261
L.C.F. : c(Cl)(cH)c(CH)c(H)cH
O=CH Unique Chemical : 104881
a,b -0.532 0.237 1-1 1 0-0 0
HC(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)H
L.C.F. : HC(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)H b -0.704 0.178 1-2 5 2-2 2
C(=O)C(H)C(H)O
L.C.F. : O=CC(C(C(H)H)(H)O)H b -1.044 0.559 1-1 3 1-1 2
S=O L.C.F. : c(H)(cc)ccH
a,b -1.273 0.247 1-1 1 0-0 0
S(=O)=O
L.C.F. : c(H)c . O=S(c)=O b -1.902 0.352 1-1 4 1-1 3
MW c 0.01200 0.00070 68.1-
748 413
68.1-
959.1 206
intercept 0.724 0.112
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
262
Table 35 - 2D Fragment Based QSAR from IFS-KOW with a 2:1 Internal Training to
Internal Validation Ratio during Cross-Validation.
Range of counts and
number of chemicals containing the fragment
Fragmentd
Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.
O=C(N)NH 0.772 0.313 1-1 3 0-0 0
{n}(c)c 0.510 0.180 1-3 4 1-3 2
HC=C(H)H 0.471 0.108 1-2 9 1-1 7 C=C 0.467 0.221 1-2 7 1-1 4
C(H)=CH 0.416 0.157 1-3 8 1-1 4
C=C(H)H 0.390 0.154 1-1 9 1-1 3 c1(H)c(H)c(H)c(O)c(H)c1H c 0.379 0.253 1-2 6 1-3 4
c(OH)c(H)c 0.364 0.088 1-2 8 1-1 2
N(H)(H)cc 0.308 0.140 1-2 10 1-2 10 C(H)=C 0.287 0.132 1-2 11 1-2 4
HC(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.260 0.196 1-2 20 1-2 15
HC(H)H 0.254 0.020 1-14 205 1-12 97
C(H)H 0.246 0.017 1-28 153 1-20 74
c(H)(cH)cOcc(H)c 0.239 0.166 1-2 6 1-2 5
c 0.220 0.019 1-12 303 1-12 152 cccN(=O)=O 0.219 0.191 1-2 4 1-2 3
ccc(cC(C(H)(H)H)C(H)(H)H)H 0.192 0.083 2-6 4 6-6 1
c(cH)(ccH)C(=O)c(c)cH 0.187 0.103 1-4 4 1-4 2 HCC(C(H)(H)H)C(H)(H)H 0.175 0.134 1-3 7 2-2 1
Hcc(c(C(C(H)(H)H)C(H)(H)H)c(cH)H)H 0.172 0.063 1-3 6 2-3 3 {c} 0.159 0.031 2-8 57 2-8 26
c(H)(ccOH)cCl 0.142 0.088 1-2 4 1-1 1
HCC(H)(H)H 0.135 0.031 1-6 55 1-4 26 HC(C(C(C(H)H)(H)H)H)H c 0.111 0.116 1-3 24 1-6 15
HC(C(C(H)(H)H)C(H)(H)H)H c 0.100 0.059 1-6 12 1-8 7
c(H)(cH)c-c 0.085 0.037 1-6 46 1-4 22 cH 0.068 0.010 1-15 303 1-15 151
{c}({c})({c}cH)c(H)cH c 0.064 0.059 1-10 14 3-12 6
c(H)(c)c(cCc(cH)cH)H -0.104 0.062 2-4 4 4-4 3 BrC(H)(H)C -0.135 0.189 1-4 3 0-0 0
HC(C(H)(c)C(H)H)H -0.152 0.133 1-2 3 1-1 2
c(H)cOC(H)H -0.156 0.117 1-2 5 0-0 0 ClC(H)H -0.171 0.169 1-6 11 1-3 7
cc(c(ccC(H)(H)H)H)H -0.174 0.189 1-2 10 1-1 4
OCH -0.200 0.065 1-12 18 1-12 6 HCC(C(C(H)(H)H)C(H)H)H -0.230 0.129 1-3 3 0-0 0
Occ(cC(H)(H)H)H -0.238 0.279 1-1 5 1-1 3
NC(H)C(H)H -0.242 0.143 1-2 4 1-1 2 HC(C(H)(H)O)(H)H -0.295 0.087 1-3 15 1-2 7
Br c -0.378 0.038 1-7 30 1-10 12
HNC(H)(H)H -0.461 0.243 1-1 5 1-1 2 O -0.484 0.044 1-8 104 1-8 46
O=CC(H)H -0.489 0.147 1-4 9 1-2 5
Hcc(OC(H)(H)H)c -0.495 0.189 1-2 4 1-1 1 C(H)(H)OH -0.516 0.193 1-2 6 1-1 2
HC(OC(H)H)H -0.660 0.119 1-8 7 3-3 1
NH -0.735 0.163 1-2 18 1-2 7 O=C -0.739 0.066 1-4 48 1-2 14
OH -0.903 0.087 1-3 48 1-3 19
N(=O)=O -0.906 0.098 1-2 29 1-1 17
N#C -0.933 0.168 1-2 6 1-2 8
{n} -1.006 0.087 1-3 18 1-3 6
HC(C(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)(H)H)H -1.204 0.212 1-2 4 2-2 2 N -1.214 0.124 1-2 18 1-2 6
N(H)H -1.220 0.114 1-2 23 1-2 14
P=O -1.284 0.186 1-1 9 1-1 7 c1(H)cccc(H)c1C
L.C.F. : c1(H)cccc(H)c1C b 0.721 0.404 1-1 4 1-1 2
P(=S)Occc(H)cCl L.C.F. : Clcc(ccOP(O)(O)=S)H
b 0.495 0.271 1-1 3 0-0 0
N(=C)Nc
L.C.F. : N(ccH)(C=O)N=C . N(ccH)N=CO a,b 0.439 0.323 1-1 2 0-0 0
263
S=POcc(c(cN(=O)=O)H)H
L.C.F. : S=P(Oc(c(c(cN(=O)=O)H)H)cH)O b 0.189 0.094 1-2 4 2-2 1
{c}(cH)({c}c(H)c)o{c}cH
L.C.F. : {c}(cH)({c}c(H)cCl)o{c}c(H)cCl b 0.100 0.057 1-4 4 0-0 0
HC(cc(cC(C(H)(H)H)C(H)(H)H)H)(H)H L.C.F. : HC(cc(c(C(C(H)(H)H)C(H)(H)H)c)H)(H)H
a,b -0.110 0.046 2-6 2 0-0 0
c(c)c-cc
L.C.F. : c(c-ccCl)(Cl)cCl b -0.154 0.036 1-8 25 1-8 10
c(H)(cc)cOccBr
L.C.F. : c(Br)(c(H)cOccBr)c(cH)Br b -0.316 0.202 1-2 4 1-2 3
Hcc(ccC(F)(F)F)H L.C.F. : Hcc(cc(C(F)(F)F)cH)H
a,b -0.498 0.290 1-1 2 0-0 0
O=CH
Unique Chemical : 104881 a,b -0.700 0.252 1-1 1 0-0 0
S=O
L.C.F. : c(H)(cc)ccH a,b -1.578 0.318 1-1 1 0-0 0
S(=O)=O L.C.F. : c(H)c . O=S(c)=O
b -1.961 0.327 1-1 4 1-1 3
MW c 0.01170 0.00050 68.1-
748 413
68.1-
959.1 206
intercept 0.790 0.085
aNon-significant fragment.
bAggregate Fragment. L.C.F. is the largest common fragment(s) among the
chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical
containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate
fragment selection rules. cTesting range exceeds training range.
dFragment or molecular weight (MW) in
pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms
by aromatic bonds. All hydrogens are explicit. eRegression coefficient.
fStandard Error of the regression
coefficient.
264
Chapter 6
Conclusions
265
1 Summary of Major Conclusions Chapter 2 of this thesis presented the results from a large-scale chemical hazard assessment
which used a dualistic paradigm to screen a large list of chemicals for the potential to be
Arctic contaminants. Over 100,000 chemicals from the database of the EPI Suite software
package41
were screened first using a structural profile of known Arctic contaminants and
also with hazard metrics similar to those suggested by the Stockholm Convention.5
Chemicals were additionally screened using presence on a high-production volume list as an
additional hazard metric for human exposure potential and a list of 120 candidate Arctic
contaminants was presented. Differences in the results of the two screening paradigms were
compared and three classes of chemicals were identified that have a high structural
resemblance to POPs, but do not have the correct chemical properties to become Arctic
contaminants; these are highly volatile chemicals, highly water soluble chemicals and
particle-bound chemicals.
Chapter 3 presented the implementation of a more sophisticated and mechanistically
satisfying description of the sorption of chemicals to environmental organic matter phases in
the multimedia mass balance model CoZMo-POP2.128
Equations and methods were derived
and presented to allow for similar implementations in other mass balance models. A
sensitivity analysis using hypothetical persistent chemicals showed that differences in the
model outputs were relatively small in comparison to the model‘s sensitivity to other
environmental parameters such as rainfall rate and temperature. The relationship between
environmental fate and the various types of interactions that a chemical undergoes with
environmental media were also explored. It was found that non-specific van der Waals
interactions had the greatest controlling effect on environmental fate, and that this interaction
was highly correlated with sorption to octanol, which explains why using octanol as a
surrogate for organic matter is an adequate approximation in a regional scale multimedia
mass balance model for the purposes of calculating chemical hazard and risk assessment
metrics.
266
Chapter 4 explored the effect of increasing the time resolution of major environmental
parameters and the implementation of dynamically changing mass balances of air, organic
aerosol, water and organic carbon in the CoZMoMAN model.34
Methods for the
implementation of dynamic mass balances of environmentlly relevant phases were presented,
along with the creation of a consistent climatic scenario. A long-range transport hazard
metric, the characteristic travel distance (CTD) was used to investigate these changes for
both labile and persistent chemicals with a large variety of enviromental fates, and it was
found that using a static, steady-state description of environmental parameters can lead to a
factor of 3 under-estimation of the CTD. Although the factor difference in the results was
relatively large, they were consistent across the range of chemical persistence and
environmental distribution; meaning that using CTDs calculated with the more simplistic
static environmental parameters is acceptable so long as they are only used to judge the long-
range transport hazard of chemicals relative to each other.
Chapter 5 of this thesis presents the creation and application of a new chemical property
prediction method. This method is an integrated approach that involves creating two
dimensional structural fragments from chemicals in a dataset, a rational splitting of the
dataset into external and internal training and validation datasets, and a robust model
selection algorithm that is designed to find a good model fit for noisy datasets without
overfitting. This method takes a chemical property dataset and with no outside guidance
automatically creates a quantitative structure property (or activity) relationship (QSPR or
QSAR). As a test of the method a QSAR is created for whole body in vivo biotransformation
half-lives in fish,165
which is an important parameter in determining bioaccumulation hazard
metrics for the aquatic foodchain. Statistics and residuals of the half-life predictions and the
fragments selected for inclusion in the example QSAR compare favorably with those for a
QSAR created with expert judgement,49
suggesting the method will be a useful tool in
predicting other uncertain properties required for chemical hazard and risk assessment.
267
2 Recommendations and Future Directions Recommendations resulting from this thesis fall under two categories; recommendations
regarding the validity of simplifying assumptions in mass balance models used to calculate
chemical hazard metrics, and recommendations for the prediction and usage of chemical
properties in hazard and risk assessment. Chapter 2 demonstrated how the outputs of a
multimedia mass balance model combined with a foodweb bioaccumulation and human
exposure model could be successfully used to simultaneously screen chemicals for long-
range transport and bioaccumulation hazard metrics. There were many simplifying
assumptions in the underyling multimedia environmental model and a possible effect of this
was noted for particle-bound chemicals which appeared to have a depressed potential for
long-range transport compared to expectations. Chapter 3 and Chapter 4 followed this up by
further investigating the effects of these simplifying assumptions on model outputs for
exposure relevant environmental media and for the CTD, a long-range transport hazard
metric. In Chapter 4 it was found that the assumption of constant rainfall, along with a low
time resolution for other environmental parameters, caused the CTD to be underestimated by
up to a factor of three. The effect was relatively uniform across a wide range of chemical
persistence and partitioning behaviour though, and the differences between chemical
amounts in environmental phases directly relevant for exposure (air and water) were
similarly uniform in Chapter 3. Based on these results it is recommended that future large-
scale chemical hazard assessment excercises which use hazard metrics calculated with mass
balance models should be prioritization oriented rather than screening oriented, because
simplifying assumptions in these mass balance models have been shown here to affect the
absolute magnitude of chemical hazard metrics they are used to calculate. In many cases the
simplifying assumptions may lead to only small differences from more sophisticated models,
but in screening oriented hazard assessment the use of cut-offs will inevitably lead to false
positives and false negatives as a result. A prioritization oriented hazard assessment is likely
to be much less uncertain, because the detrimental effect of the simplifying assumptions is
experienced approximately equally for a wide range of chemicals and the hazard of the
assessed chemicals relative to each other is likely to remain largely unaltered.
268
In Chapter 2 three major data deficiencies were identified; first there was no widely
applicable QSAR available to make predictions for biodegradation in biota, and second there
was similarly no widely applicable QSAR available for toxicity. Additionally, there is little
or no publically available data for the production volumes of most chemicals or their uses,
making the estimation of chemical emissions nearly impossible. These data deficiencies have
the net result that large scale chemical screening and prioritization excercises are currently
limited to hazard based metrics, rather than risk based metrics. As discussed in Chapter 1
risk is the product of exposure and adverse effects, and at the moment both of these metrics
are unavailable on a large scale using publically available data; the lack of emission
estimates means chemical exposures cannot be unequivocally ranked and the lack of toxicity
predictions means that the potential for adverse effects is also inaccessible. Using the new
chemical property prediction method presented in this thesis a screening and prioritization
metric for assessing a chemical‘s biodegradation potential is closer to being a reality, which
would allow for a more detailed large-scale chemical hazard assessement. Future work may
also see the development of a widely applicable prediction method for toxicity, which would
bring the state of the science another step closer to fully automated risk assessment. By far
the largest uncertainty, and data need, though remains chemical emissions estimates. This is
unfortunate because, as an extensive property, production volume is independent of chemical
structure and cannot be predicted. Possibilities for further research along this avenue are
limited, the collection and distribution of production data is in the jurisdiction of chemical
regulators and chemical industry. Researchers can further the state of the science in this
direction only by acting as advocates for greater openness in chemical production and use.
269
References
270
1 CAS, Chemical Abstracts Service - Database Counter; http://www.cas.org/cgi-
bin/cas/regreport.pl; Accessed 03 14, 2011.
2 The 2007 OECDlist of high production volume chemicals; OECD; OECD Environment,
Health and Safety Publications: Environmental Directorate: Paris, 2009.
3 van Leeuwen, C.J.; Vermeire, T.G.; (eds.), Risk assessment of chemicals: an introduction.
Second edition; Kluwer Academic Publishers: Dordrecht, 2007.
4 Nielsen, E.; Ostergaard, G.; Larsen, J.C. Toxicological risk assessment of chemicals: a
practical guide; Informa Healthcare: New York, 2008.
5 Final act of the conference of plenipotentiaries on the Stockholm Convention on persistent
organic pollutants; UNEP; United Nations Environment Program: Geneva, 2001.
6 Registration, Evaluation and Authorization of Chemicals (REACH); European
Commission; COM 2003: 2003, 644 final.
7 Ecological categorization of substances on the domestic substances list (DSL);
Environment Canada; Government of Canada: Ottawa, 2004.
8 Lippmann, M.; Cohen, B.S.; Schlesinger, R.B. Environmental health science: recognition,
evaluation, and control of chemical and physical health hazards; Oxford University Press:
New York, 2003.
9 van Wijk, D.; Chenier, R.; Henry, T.; Hernando, M.D.; Schulte, C. Integrated approach to
PBT and POP prioritization and risk assessment. Integr. Environ. Assess. Manag. 2009, 5,
697-711.
10 Scheringer, M. Persistence and spatial range as endpoints of an exposure-base assessment
of organic chemicals. Environ. Sci. Technol. 1996, 30, 1652-1659.
11 Breivik, K.; Sweetman, A.; Pacyna, J.M.; Jones, K.C. Towards a global historical
emission inventory for selected PCB congeners - a mass balance approach-3. an update. Sci.
Total Environ. 2007, 377, 296-307.
12 Erickson, M.D.; Kaley II, R.G. Applications of polychlorinated biphenyls. Environ. Sci.
Pollut. Res. 2011, 18, 135-151.
271
13 Trudel, D.; Horowitz, L.; Wormuth, M.; Scheringer, M.; Cousins, I.T.; K, Hungerbuhler.
Estimating consumer exposure to PFOS and PFOA. Risk Anal. 2008, 28, 251-269.
14 Schwarzenbach, R.P.; Gschwend, P.M.; Imboden, D.M. Environmental Organic
Chemistry, 2nd Ed; Wiley: , 2002.
15 Mackay, D. Multimedia environmental models - the fugacity approach, 2nd ed.; Lewis
Publishers: Chelsea, 2001.
16 Preparation of a product register based exposure index; Fischer, S.; Almkvist, Å.; E,
Karlsson.; Åkerblom, M. ; Swedish Chemicals Agency: Sweden, 2006.
17 Cowan-Ellsberry, C.E.; S, McLachlan.M.; A, Arnot.J.; Macleod, M.; E, McKone..T.;
Wania, F. Modeling exposure to persistent chemical in hazard and risk assessment. Integr.
Environ. Assess. Manag. 2009, 5, 662-679.
18 Arnot, J.A.; Armitage, J.M.; McCarty, L.S.; Wania, F.; Cousins, I.T.; Toose-Reid, L.
Toward a consistent evaluative framework for POP risk characterization. Environ. Sci.
Technol. 2011, 45, 97-103.
19 MacLeod, M.; Riley, W.J.; McKone, T.E. Assessing the influence of climate variability
on atmospheric concentrations of polychlorinated biphenyls using a global-scale mass
balance model (BETR-Global). Environ. Sci. Technol. 2005, 39, 6749-6756.
20 Powell, A.; Mackay, D.; Webster, E.; Arnot, J.A. Modeling bioaccumulation using
characteristic times. Environ. Toxicol. Chem. 2009, 28, 272-278.
21 Boxall, A.B.A.; Sinclair, C.J.; Fenner, K.; Kolpin, D.; Maud, S.J. When synthetic
chemicals degrade in the environment. Environ. Sci. Technol. 2004, 38, 368-375.
22 Fenner, K.; Kooijman, C.; Scheringer, M.; Hungerbuhler, K. Including transformation
products into the risk assessment for chemicals: the case of nonylphenol ethoxylate usage in
Switzerland. Environ. Sci. Technol. 2002, 36, 1147-1154.
23 Czub, G.; McLachlan, M.S. Bioaccumulation potential of persistent organic chemicals in
humans. Environ. Sci. Technol. 2004, 38, 2406-2412.
272
24 MacLeod, M.; Scheringer, M.; McKone, T.E.; Hungerbuhler, K. The state of multimedia
mass-balance modeling in environmental science and decision-making. Environ. Sci.
Technol. 2010, 44, 8360-8364.
25 McLachlan, M.S.; Czub, G.; MacLeod, M.; Arnot, J.A. Bioaccumulation of organic
contaminants in humans: A multimedia perspective and the importance of biotranformation.
Environ. Sci. Technol. 2011, 45, 197-202.
26 Arnot, J.A.; D, Mackay. Policies for chemical hazard and risk priority setting: Can
persistence, bioaccumulation, toxicity and quantity information be combined?. Environ. Sci.
Technol. 2008, 42, 4648-4654.
27 Scheringer, M.; Jones, K.C.; Matthies, M.; Simonich, S.; van de Meent, D. Multimedia
partitioning, overall persistence, and long-range transport potential in the context of POPs
and PBT chemical assessments. Integr. Environ. Assess. Manag. 2009, 5, 557-576.
28 Mackay, D. Finding fugacity feasible. Environ. Sci. Technol. 1979, 13, 1218-1223.
29 Mackay, D.; Fraser, A. Bioaccumulation of persistent organic chemicals: mechanisms and
models. Environ. Pollut. 2000, 110, 375-391.
30 Daugherty, J. Assessment of chemical exposures: calculation methods for environmental
proffessionals; Lewis Publishers: Boca Raton, 1998.
31 Arnot, J.A.; Mackay, D.; Parkerton, T.F.; Zaleski, R.T.; Warren, C.S. Multimedia
modeling of human exposure to chemical substances: the roles of food web biomagnification
and biotransformation. Environ. Toxicol. Chem. 2010, 29, 45-55.
32 Arnot, J.A.; Mackay, D.; Sutcliffe, R.; Lo, B. Estimating farfield organic chemical
exposures, intake rates and intake fractions to human age classes. Environ. Model. Soft.
2010, 25, 1166-1175.
33 McKone, T.E.; Daniels, J.I. Estimating human exposure through multiple pathways from
air, water and soil. Regul. Toxicol. Pharmacol. 1991, 13, 36-61.
34 Breivik, K.; Czub, G.; McLachlan, M.S.; Wania, F. Towards an understanding of the link
between environmental emissions and human body burdens of PCBs using CoZMoMAN.
Environ. Intern. 2010, 36, 85-91.
273
35 Cahill, T.M.; Mackay, D. Complexity in multimedia mass balance models: When are
simple models adequate and when are more complex models necessary?. Environ. Toxicol.
Chem. 2003, 22, 1404-1412.
36 Beyer, A.; Mackay, D.; Matthies, M.; Wania, F.; Webster, E. Assessing long-range
transport potential of persistent organic polluntants. Environ. Sci. Technol. 2000, 34, 699-
703.
37 Fenner, K.; Scheringer, M.; MacLeod, M.; Matthies, M.; McKone, T.; Stroebe, M.;
Beyer, A.; Bonnell, M.; Le Gall, A.C.; Klasmeier, J.; Mackay, D.; Van de Meent, D.;
Pennington, D.; Scharenberg, B.; Suzuki, N.; Wania, F. Comparing estimates of persistence
and long-range transport potential among multimedia models. Environ. Sci. Technol. 2005,
39, 1932-1942.
38 Stroebe, M.; Scheringer, M.; Hungerbuhler, K. Measures of overall persistence and the
temporal remote state. Environ. Sci. Technol. 2004, 38, 5665-5673.
39 Mackay, D.; McCarty, L.S.; MacLeod, M. On the validity of classifying chemicals for
persistence, bioaccumulation, toxicity, and potential for long-range transport. Environ.
Toxicol. Chem. 2001, 20, 1491-1498.
40 Czub, G.; Wania, F.; McLachlan, M.S. Combining long range transport and
bioaccumulation considerations to identify potential Arctic contaminants. Environ. Sci.
Technol. 2008, 42, 3704–3709.
41 Exposure Assessment Tools and Models, Estimation Program Interface (EPI) Suite; U.S.
EPA; : Washington DC, 2005.
42 Wegmann, F.; Cavin, L.; MacLeod, M.; Scheringer, M.; Hungerbuhler, K. The OECD
software tool for screening chemicals for persistence and long-range transport. Environ.
Modell. Software 2009, 24, 228-237.
43 Fisk, A.T.; Norstrom, R.J.; Cymbalisty, C.D.; Muir, D.C.G. Dietary assumulation and
depuration of hydrophobic organochlorines: Bioaccumulation parameters and their
relationship with the octanol/water partition coefficient. Environ. Toxicol. Chem 1998, 17,
951-961.
274
44 Seth, R.; Mackay, D.; Muncke, J. Estimating the organic carbon partition coefficient and
its variability for hydrophobic chemicals. Environ. Sci. Technol. 1999, 33, 2390-2394.
45 Abraham, M.H.; Chadha, H.S.; Whiting, G.S.; Mitchell, R.C. Hydrogen Bonding. 32. An
analysis of water-octanol and water-alkane partitioning and the Δlog P parameter of Seiler. J.
Pharm. Sci. 1994, 83, 1085–1100.
46 Neiderer, C.; Goss, K.-U. Quantum chemical modeling of humic acid/air equilibrium
partitioning of organic vapors. Environ. Sci. Technol. 2007, 41, 3646-3652.
47 Wittekindt, C.; Goss, K.-U. Screening the partition behavior of a large number of
chemicals with a quantum-chemical software. Chemosphere 2009, 76, 460-464.
48 Biodegradation Probability Program (BIOWIN); U.S. EPA; EPI Suite: Washington DC,
2005.
49 Arnot, J.A.; Meylan, W.; Tunkel, J.; Howard, P.H.; Mackay, D.; Bonnell, M.; Boethling,
R.S. A quantitative structure-activity relationship for predicting metabolic biotransformation
rates for organic chemicals in fish. Environ. Toxicol. Chem. 2009, 28, 1168-1177.
50 Dearden, J.C. In silico prediction of ADMET properties: how far have we come?. Expert
Opin. Drug Met. 2007, 3, 635-639.
51 Introduction and Executive Summary; Bowes, M.; Brown, M.; Rivas, Y.; Butler, D.;
Zeiger, E. ; Office of Technology Assessment: Washing DC, 1995.
52 Second national report on human exposure to environmental chemicals; CDC; National
Center for Environmental Health, Centers for Disease Control and Prevention: Atlanta, 2003.
53 Lau, C.; Anitole, K.; Hodes, C.; Lai, D.; Pfahles-Hutchens, A.; Seed, J. Perfluoroalkyl
acids: A review of monitoring and toxicological findings. Toxicol. Sci. 2007, 99, 366-394.
54 Wania, F. Potential of degradable organic chemicals for absolute and relative enrichment
in the Arctic. Environ. Sci. Technol. 2006, 40, 569-577.
55 Muir, D.C.G.; Howard, P.H. Are there other persistent organic pollutants? A challenge for
environmental chemists. Environ. Sci. Technol. 2006, 40, 7157-7166.
275
56 Xiao, H.; F, Wania. Is vapor pressure or the octanol-air partition coefficient a better
descriptor of the partitioning between gas phase and organic matter?. Atmos. Environ. 2003,
37, 2867-2878.
57 Niederer, C.; Goss, K.-U.; Schwarzenbach, R.P. Sorption equilibrium of a wide spectrum
of organic vapors in Leonardite humic acid: modeling of experimental data. Environ. Sci.
Technol. 2006, 40, 5374-5379.
58 Arp, H.P.H.; Niederer, C.; Goss, K-U. Predicting the partitioning behavior of various
highly fluorinated compounds. Environ. Sci. Technol. 2006, 40, 7298-7304.
59 Atkinson, R.; Guicherit, R.; Hites, R.A.; Palm, W.U.; Seiber, J.N.; de Voogt, P.
Transformations of pesticides in the atmosphere: A state of the art. Water Air Soil Pollut.
1999, 115, 219-243.
60 High production volume (HPV) challenge program. The HPV voluntary challenge
chemical list; U.S. EPA; U.S. Environmental Protection Agency: Merrifield, 2006.
61 European chemical substances information system; ECB; European Chemical Bureau,
European Commission: Ispra, 2006.
62 The 2004 OECD list of high production volume chemicals; OECD; OECD Environment,
Health and Safety Publications: Paris, 2004.
63 Interagency testing committee under the toxic substances contral act (TSCA); U.S. EPA;
U. S. Environmental Protection Agency, Office of Pollution Prevention and Toxics: Washing
DC, 2003.
64 Pesticide product information system;
http://www.epa.gov/opppmsdl/PPISdata/index.html; Accessed December 15, 2007.
65 The WHO recommended classification of pesticides by hazard and guidelines to
classification: 2004; WHO; World Health Organization: Geneva, 2005.
66 Wania, F. Assessing the potential of persistent organic chemicals for long-range transport
and accumulation in polar regions. Environ. Sci. Technol. 2003, 37, 1344-1351.
276
67 Hilal, S.H.; Karickhoff, S.W.; Carreira, L.A. Prediction of the solubility, activity
coefficient and liquid/liquid partition coefficient of organic compounds. QSAR Comb. Sci.
2004, 23, 709-720.
68 Baird, C.; Cann, M. Environmental Chemistry; W.H. Freeman & Company: New York,
2005.
69 Kwok, E.S C.; Atkinson, R. Estimation of hydroxyl radical reaction rate constants for
gas-phase organic compounds using a structure-reactivity relationship: An update. Atmos.
Environ. 1995, 29, 1685-1695.
70 Estimation accuracy of the Atmospheric Oxidation Program; U.S. EPA; U.S.
Environmental Protection Agency: Washington DC, 2000.
71 Goss, K.-U.; Bronner, G. What is so special about the sorption behavior of highly
fluorinated compounds?. J. Phys. Chem. A 2006, 110, 9518-9522.
72 Breivik, K.; Wania, F.; Muir, D.C.G.; Alaee, M.; Backus, S.; Pacepavicius, G. Empirical
and modeling evidence of the long-range atmospheric transport of decabromodiphenyl ether.
Environ. Sci. Technol. 2006, 40, 4612-4618.
73 Ewald, G.; Larsson, P.; Linge, H.; Lennart, O.; Szarzi, N. Biotransport of organic
pollutants to an inland Alaska lake by migrating Sockeye Salmon (Oncorhynchus nerka).
Arctic 1998, 51, 40-47.
74 Koistinen, J.; Mussalo-Rauhamaab, H.; Paasivirta, J. Polychlorinated diphenyl ethers,
dibenzo-p-dioxins and dibenzofurans in Finnish human tissues compared to environmental
samples. Chemosphere 1995, 31, 4259-4271.
75 Robust summaries & test plans: Chlorinated pyridine category; The Dow Chemical
Company; U.S. Environmental Protection Agency: Merrifield, 2005.
76 Fire resistent multilayer film with at least one diffusion barrier layer and its use for
vacuum insulation panels; Kaczmarek, D.; Jacobsen, S. ; CODEN:EPXXDW EP 1607217
A1 20051221 CAN 144:52758 AN 2005:1328000 CAPLUS: Eur. Pat. Appl., 2005.
277
77 Siloxane gels for personal care products; Katsoulis, D.E. ; CODEN: USXXAM US
5486566 A 19960123 CAN 124:211487 AN 1996:121341 CAPLUS: United States of
America, 1996.
78 Braune, B.M.; Outridge, P.M.; Fisk, A.T.; Muir, D.C G.; Helm, P.A.; Hobbs, K.;
Hoekstra, P.F.; Kuzyk, Z.A.; Kwan, M.; Letcher, R.J.; Lockhart, W.L.; Norstrom, R.J.;
Stern, G.A.; Stirling, I. Persistent organic pollutants and mercury in marine biota of the
Canadian Arctic: An overview of spatial and temporal trends. Sci. Total Environ. 2005, 351-
352, 4-56.
79 Evans, M.S.; Muir, D.; Lockhart, W.L.; Stern, G.; Ryan, M.; Roach, P. Persistent organic
pollutants and metals in the freshwater biota of the Canadian Subarctic and Arctic: An
overview. Sci. Total Environ. 2005, 351-352, 94-147.
80 Gamberg, M.; Braune, B.; Davey, E.; Elkin, B.; Hoekstra, P.F.; Kennedy, D.; C,
Macdonald.; Muir, D.; Nirwal, A.; Wayland, M.; Zeeb, B. Spatial and temporal trends of
contaminants in terrestial biota from the Canadian Arctic. Sci. Total Environ. 2005, 351-352,
148-164.
81 Li, N.; Wania, F.; Lei, Y.D.; Daly, G.L. A comprehensive and critical compilation,
evaluation, and selection of physical-chemical property data for selected polychlorinated
biphenyls. J. Phys. Chem. Ref. Data 2003, 32, 1545-1590.
82 Su, Y.; Lei, Y.D.; Daly, G.L.; Wania, F. Determination of octanol-air partition coefficient
(KOA) values for chlorobenzenes and polychlorinated naphthalenes from gas
chromatographic retention times. J. Chem. Eng. Data 2002, 47, 449-455.
83 Feigel, M. Phenyl rotation in diphenylether and diphenylmethane calculated with ab initio
methods. J. Mol. Struct. (Theochem) 1996, 366, 83-88.
84 Sawunyama, P.; Bailey, G.W. Modeling the interaction of agrochemicals with
environmental surfaces: Pesticdes on rutile and organo-rutile surfaces. J. Mol. Struct.
(Theochem) 2001, 541, 119-129.
85 Lundgren, K.; Tysklind, M.; Ishaq, R.; Broman, D.; van Bavel, B. Polychlorinated
naphthalene levels, distribution, and biomagnification in a benthic food chain in the Baltic
Sea. Environ. Sci. Technol. 2002, 36, 5005-5013.
278
86 Abraham, M.H.; Al-Hussaini, A.J.M. Solvation descriptors for the polychloro-
naphthalenes: estimation of some physicochemical properties. J. Environ. Monit. 2001, 3,
377–381..
87 Corsolini, S.; Kannan, K.; Imagawa, T.; Focaradi, S.; Giesy, J.P. Polychloronaphthalenes
and other dioxin-like compounds in Arctic and Antarctic marine food webs. Environ. Sci.
Technol. 2002, 36, 3490-3496.
88 Steffen, C.; Borga, K.; Skaare, J.U.; Bustnes, J.O. The occurence of organochlorines in
marine avian top predators along a latitudinal gradient. Environ. Sci. Technol. 2006, 40,
5139-5146.
89 Abraham, M.H.; Al-Hussaini, A.J.M. Solvation parameters for the 209 PCBs: calculation
of physicochemical properties. J. Environ. Monit. 2005, 7, 295–301..
90 Kannan, K.; Yun, S.H.; Evans, T.J. Chlorinated, brominated and perfluorinated
contaminants in livers of polar bears from Alaska. Environ. Sci. Technol. 2005, 39, 9057-
9063.
91 Knudsen, L.B.; Borga, K.; Jorgensen, E.H.; van Bavel, B.; Schlabach, M.; Verreault, J.;
Gabrielsen, G.W. Halogenated organic contaminants and mercury in northern fulmars
(Fulmarus glacialis): levels, relationships to dietary descriptors and blood to liver
comparison. Environ. Pollut. 2007, 146, 25-33.
92 Shen, L.; Wania, F. Compilation, evaluation, and selection of physical-chemical property
data for organochlorine pesticides. J. Chem. Eng. Data 2005, 50, 742-768.
93 Stern, G.A.; MacDonald, C.R.; Armstrong, D.; Dunn, B.; Fuchs, C.; Harwood, L.; Muir,
D.C.G.; Rosenberg, B. Spatial trends and factors affecting variation of organochlorine
contaminants levels in Canadian Arctic beluga (delphinapterus leucas). Sci. Total Environ.
2005, 351, 344-368.
94 Sørmo, E.G.; Salmer, M.P.; Jenssen, B.M.; Hop, H.; Bæk, K.; Kovacs, K.M.; Lydersen,
C.; Falk-Petersen, S. Biomagnification of polybrominated diphenyl ether and hexa-
bromocyclododecane flame retardants in the polar bear food chain in Svalbard, Norway.
Environ. Toxicol. Chem. 2006, 25, 2502-2511.
279
95 Cetin, B.; Odabasi, M. Measurement of Henry‘s law constants of seven polybrominated
diphenyl ether (PBDE) congeners as a function of temperature. Atmos. Environ. 2005, 39,
5273–5280.
96 Harner, T.; Shoeib, M. Measurements of octanol-air partition coefficients (KOA) for
polybrominated diphenyl ethers (PBDEs): Predicting partitioning in the environment. J.
Chem. Eng. Data 2002, 47, 228-232.
97 Muir, D.C.G.; Backus, S.; Derocher, A.E.; Dietz, R.; Evans, T.J.; Gabrielsen, G.W.;
Nagy, J.; Norstrom, R.J.; Sonne, C.; Stirling, I.; Taylor, M.K.; Letcher, R.J. Brominated
flame retardants in Polar Bears (Ursus maritimus) from Alaska, the Canadian Arctic, East
Greenland, and Svalbard.. Environ. Sci. Technol. 2006, 40, 449-455.
98 Fuglei, E.; Bustnes, J.O.; Hop, H.; Mørk, T.; Bjornfoth, H.; van Bavel, B. Environmental
contaminants in arctic foxes (Alopex lagopus) in Svalbard: relationships with feeding
ecology and body condition. Environ. Pollut. 2007, 146, 128-138..
99 de Wit, C.A.; Alaee, M.; Muir, D.C.G. Levels and trends of brominated flame retardants
in the Arctic. Chemosphere 2006, 64, 209-233.
100 Paasivirta, J.; Sinkkonen, S.; Mikkelson, P.; Rantio, T.; Wania, F. Estimation of vapor
pressures, solubilities and Henry's Law constants of selected persistent organic pollutants as
functions of temperature. Chemosphere 1999, 39, 811-832..
101 Vorkamp, K.; Riget, F.; Glasius, M.; Pecselia, M.; Lebeuf, M.; Muir, D.C.G. Chloro-
benzenes, chlorinated pesticides, coplanar chlorobiphenyls and other organochlorine
compounds in Greenland biota.. Sci. Total Environ. 2004, 331, 157–175..
102 Mckinney, M.A.; de Guise, S.; Martineau, D.; Beland, P.; Lebeuf, M.; Letcher, R.J.
Organohalogen contaminants and metabolites in beluga whale (delphinapterus leucas) liver
from two Canadian populations. Environ. Toxicol. Chem. 2006, 25, 1246–1257.
103 Xiao, H.; Li, N.; Wania, F. Compilation, evaluation, and selection of physical-chemical
property data for α-, β-, γ-hexachlorocyclohexane. J. Chem. Eng. Data 2004, 49, 173-185.
104 Coelhan, M.; Reil, I.; Rimkus, G.; Parlar, H. Peak patterns of chlorostyrenes in fish and
fish oils from the North Atlantic. Environ. Sci. Technol. 2000, 34, 4695-4700.
280
105 Verreault, J.; Norstrom, R.J.; Ramsay, M.A.; Mulvihill, M.; Letcher, R.J. Composition
of chlorinated hydrocarbon contaminants among major adipose tissue depots of Polar Bears
(Ursus maritimus) from the Canadian high Arctic. Sci. Total Environ. 2006, 370, 580–587.
106 Smithwick, M.; Norstrom, R.J.; Mabury, S.A.; Solomon, K.; Evans, T.J.; Stirling, I.;
Taylor, M.K.; Muir, D.C.G. Temporal trends of perfluoroalkyl contaminants in polar bears
(ursus maritimus) from two locations in the North American Arctic, 1972-2002. Environ.
Sci. Technol. 2006, 40, 1139-1143.
107 Reth, M.; Ciric, A.; Christensen, G.N.; Heimstad, E.S.; Oehme, M. Short- and medium-
chain chlorinated paraffins in biota from the European Arctic — differences in homologue
group patterns. Sci. Total Environ. 2006, 367, 252–260.
108 Fisk, A.T.; Rosenberg, B.; Cymbalisty, C.D.; Stern, G.A.; Muir, D.C.G. Octanol/ water
partition coefficients of toxaphene congeners determined by the "slow-stirring" method.
Chemosphere 1999, 39, 2549-2562.
109 Carlson, R. Silent Spring; Houghton Mifflin Co.: Boston, 1962.
110 Martin, J.W.; Smithwick, M.M.; Braune, B.M.; Hoekstra, P.F.; Muir, D.C G.; Mabury,
S.A. Identification of long-chain perfluorinated acids in biota from the Canadian Arctic.
Enrion. Sci. Technol. 2004, 38, 373-380.
111 Breivik, K.; Wania, F. Expanding the applicability of multimedia fate models to polar
organic chemicals. Environ. Sci. Technol. 2003, 37, 4934-4943.
112 Goss, K.-U.; Schwarzenbach, R.P. Linear free energy relationships used to evaluate
equilibrium partitioning of organic compounds. Environ. Sci. Technol. 2001, 35, 1-9.
113 Żukowska, B.; Breivik, K.; Wania, F. Evaluating the environmental fate of
pharmaceuticals using a level III model based on polyparameter linear free energy
relationships. Sci. Total Environ. 2006, 359, 177–187.
114 Götz, C.W.; Scheringer, M.; MacLeod, M.; Wegmann, F.; Schenker, U.; Hungerbühler,
K. Dependence of persistence and long-range transport potential on gas-particle partitioning
in multimedia models. Environ. Sci. Technol. 2008, 42, 3690–3696.
281
115 Abraham, M.H. Scales of solute hydrogen-bonding: their construction and application to
physicochemical and biochemical processes. Chem. Soc. Rev. 1993, 22, 73-83.
116 Poole, S.K.; Poole, C.F. Chromatographic models for the sorption of neutral organic
compounds by soil from water and air. J. Chromatogr. A 1999, 845, 381–400.
117 Götz, C.W.; Scheringer, M.; MacLeod, M.; Roth, C.M.; Hungerbühler, K. Alternative
approaches for modeling gas/particle partitioning of semivolatile organic chemicals: model
development and comparison. Environ. Sci. Technol. 2007, 41, 1272–1278.
118 Goss, K.-U. Predicting the equilibrium partitioning of organic compounds using just one
linear solvation energy relationship (LSER). Fluid Phase Equilib. 2005, 233, 19–22.
119 Niederer, C.; Goss, K.-U.; Schwarzenbach, R.P. Sorption equilibrium of a wide
spectrum of organic vapors in Leonardite humic acid: experimental setup and experimental
data. Environ. Sci. Technol. 2006, 40, 5368-5373.
120 Arp, H.P.H.; Schwarzenbach, R.P.; Goss, K.-U. Ambient gas/particle partitioning. 1.
sorption mechanisms of apolar, polar, and ionizable organic compounds. Environ. Sci.
Technol. 2008, 42, 5541–5547.
121 Arp, H.P.H.; Schwarzenbach, R.P.; Goss, K.-U. Ambient gas/particle partitioning. 2: the
influence of particle source and temperature on sorption to dry terrestrial aerosols. Environ.
Sci. Technol. 2008, 42, 5951–5957.
122 Mintz, C.; Clark, M.; Acree, W.E.; Abraham, M.H. Enthalpy of solvation correlations
for gaseous solutes dissolved in water and in 1-octanol based on the Abraham model. J.
Chem. Inf. Model. 2007, 47, 115-121.
123 Abraham, M.H.; Whiting, G.S.; Fuchs, R.; Chambers, E.J. Thermodynamics of solute
transfer from water to hexadecane. J. Chem. Soc., Perkin Trans. 2 1990, , 291-300.
124 Fuchs, R.; Abraham, M.H.; Kamlet, M.J.; Taft, R.W. Solute-solvent interactions in
chemical and biological systems. IV. Correlations of ΔG, ΔH and TΔS of transfer of
aliphatic and aromatic solutes from 2,2,4-trimethylpentane to aqueous solution. J. Phys. Org.
Chem. 1989, 2, 559-564.
282
125 Abraham, M.H.; Andonian-Haftvan, J.; Whiting, G.S.; Leo, A.; Taft, R.S. Hydrogen
bonding. Part 34. The factors that influence the solubility of gases and vapours in water at
298 K, and a new method for its determination. J. Chem. Soc., Perkin Trans. 2 1994, , 1777–
1791.
126 Goss, K.-U.; Arp, H.P.H.; Bronner, G.; Niederer, C. Non-additive effects in the
partitioning behaviour of various aliphatic and aromatic molecules. Environ. Toxicol. Chem.
2009, 28, 52-60.
127 Tülp, H.C.; Goss, K.-U.; Schwarzenbach, R.P.; Fenner, K. Experimental determination
of LSER parameters for a set of 76 diverse pesticides and pharmaceuticals. Environ. Sci.
Technol. 2008, 42, 2034–2040.
128 Wania, F.; Breivik, K.; Persson, N.J.; McLachlan, M.S. CoZMo-POP 2 - A fugacity-
based dynamic multi-compartment mass balance model of the fate of persistent organic
pollutants.. Environ. Model. Software 2006, 21, 868-884.
129 Meyer, T.; Wania, F.; Breivik, K. Illustrating sensitivity and uncertainty in
environmental fate models using partitioning maps. Environ. Sci. Technol. 2005, 39, 3186-
3196.
130 Goss, K.-U.; Schwarzenbach, R.P. Empirical prediction of heats of vaporization and
heats of adsorption of organic compounds. Environ. Sci. Technol. 1999, 33, 3390-3393.
131 MacLeod, M.; Scheringer, M.; Hungerbühler, K. Estimating enthalpy of vaporization
from vapor pressure using Trouton‘s Rule. Environ. Sci. Technol. 2007, 41, 2827-2832.
132 Horstmann, M.; McLachlan, M.S. Atmospheric deposition of semi-volatile compounds
to two forest canopies. Atmos. Environ. 1998, 32, 1799-1809.
133 Platts, J.A.; Abraham, M.H. Partition of volatile organic compounds from air and from
water into plant cuticular matrix: an LFER analysis. Environ. Sci. Technol. 2000, 34, 318-
323.
134 Goss, K.-U.; Arp, H.P.H.; Bronner, G.; Niederer, C. Partition behavior of
hexachlorocyclo-hexane isomers. J. Chem. Eng. Data 2008, 53, 750–754.
283
135 Abraham, M.H.; Dearden, J.C.; Bresnen, G.M. Hydrogen bonding, steric effects and
thermodynamics of partitioning. J. Phys. Org. Chem. 2006, 19, 242–248.
136 Abraham, M.H.; Enomoto, K.; Clarke, E.D.; Rośes, M.; Ràfols, C.; Fuguet, E. Henry‘s
Law constants or air to water partition coefficients for 1,3,5-triazines by an LFER method. J.
Environ. Monit. 2007, 9, 234–239.
137 Abraham, M.H.; Ibrahim, A.; Zhao, Y.; Acree, W.E. A data base for partition of volatile
organic compounds and drugs from blood/plasma/serum to brain, and an LFER analysis of
the data. J. Pharm. Sci. 2006, 95, 2091-2100.
138 Abraham, M.H. Hydrogen bonding. Part 27. Solvation parameters for functionally
substituted aromatic compounds and heterocyclic compounds, from gas-liquid chromato-
graphic data. J. Chromatogr. 1993, 644, 95-139.
139 Abraham, M.H.; Grellier, P.L.; McGill, R.A. Determination of olive oil-gas and
hexadecane-gas partition coefficients, and calculation of the corresponding olive oil-water
and hexadecane-water partition coefficients. J. Chem. Soc. Perkin Trans. 2 1987, , 797-803.
140 Abraham, M.H.; Andonian-Haftvan, J.; My Du, C.; Osei-Owusu, J.P.; Sakellariou, P.;
Shuely, W.J.; Poole, C.F.; Poole, S.K. Comparison of uncorrected retention data on a
capillary and a packed hexadecane column with corrected retention data on a packed
squalane column. J. Chromatogr. A 1994, 688, 125-134.
141 Breivik, K.; Wania, F. Evaluating a model of the historical behavior of two hexachloro-
cyclohexanes in the Baltic Sea environment. nviron. Sci. Technol. 2002, 36, 1014-1023.
142 Webster, E.; Mackay, D.; Wania, F. Evaluating environmental persistence. Environ.
Toxicol. Chem. 1998, 17, 2148 -2158.
143 Finizio, A.; Mackay, D.; Bidleman, T.; Harner, T. Octanol-air partition coefficient as a
predictor of partitioning of semi-volatile organic chemical to aerosols. Atmos. Environ. 1997,
31, 2289-2296.
144 Su, Y.; Wania, F.; Harner, T.; Lei, Y.D. Deposition of polybrominated diphenylethers,
polychlorinated biphenyls, and polycyclic aromatic hydrocarbons to a boreal deciduous
forest. Environ. Sci. Technol. 2007, 41, 543-540.
284
145 Welke, B.; Ettlinger, K.; Riederer, M. Sorption of volatile organic chemicals in plant
surfaces. Environ. Sci. Technol. 1998, 32, 1099-1104.
146 Brown, T.N.; Wania, F. Screening chemicals for the potential to he persistent organic
pollutants: a case study of Arctic contaminants. Environ. Sci. Technol. 2008, 42, 5202-5209.
147 Klasmeier, J.; Matthies, M.; MacLeod, M.; Fenner, K.; Scheringer, M.; Stroebe, M.; Le
Gall, A.C.; McKone, T.; Van de Meent, D.; Wania, F. Applicationof multimedia models for
screening assessment of long-range transport potential and overall persistence. Environ. Sci.
Technol. 2006, 40, 53-60.
148 Jolliet, O.; Hauschild, M. Modeling the influence of intermittent rain events on long-
term fate and transport of organic pollutants. Environ. Sci. Technol. 2005, 39, 4513-4522.
149 Breivik, K.; Wania, F.; Muir, D..C G.; Alaee, M.; Backus, S.; Pacepavicius, G.
Empirical and modeling evidence of long-range atmospheric transport of decabromodiphenyl
ether. Environ. Sci. Technol. 2006, 40, 4612-4618.
150 Muir, D.C.G.; Teixeira, C.; Wania, F. Empirical and modeling evidence of regional
atmospheric transport of current use pesticides. Environ. Toxicol. Chem. 2004, 23, 2421-
2432.
151 Gasic, B.; MacLeod, M.; Scheringer, M.; Hungerbuhler, K. Assessing the impact of
weather events at mid-latitudes on the atmospheric transport of chemical pollutants using a
2-dimensional multimedia meteorological model. Atmosph. Environ. 2010, 44, 4489-4496.
152 Nizzetto, L.; MacLeod, M.; Borga, K.; Cabrerizo, A.; Dachs, J.; Di Guardo, A.;
Ghirardello, D.; Hansen, K.M.; Jarvis, A.; Lindroth, A.; Ludwig, B.; Monteith, D.; Perlinger,
J.A.; Scheringer, M.; Schwendenmann, L.; Semple, K.T.; Wick, L.Y.; Zhang, G.; Jones,
K.C. Past, present and future controls on levels of persistent organic pollutants in the global
environment. Environ. Sci. Technol. 2010, 44, 6526-6531.
153 Lammel, G. Effects of time-averaging climate parameters on predicted
multicompartmental fate of pesticides and POPs. Environ. Pollut. 2004, 128, 291-302.
154 Daly, G.L.; Wania, F. Simulating the influence of snow on the fate of organic
compounds. Environ. Sci. Technol. 2004, 38, 4176-7186.
285
155 National Climate Data and Information Archive;
http://climate.weatheroffice.gc.ca/climateData/canada_e.html; Accessed January 15, 2011.
156 MacLeod, M.; Scheringer, M.; Podey, H.; Jones, K.C.; Hungerbuhler, K. The origin and
significance of short-term variability of semi-volatile contaminants in air. Environ. Sci.
Technol. 2007, 41, 3249-3253.
157 Dunne, T.; Leopold, L.B. Water in Environmental Planning. 1978, , 88.
158 Finlayson-Pitts, B.J.; Pitts, J.N. Chemistry of the upper and lower atmosphere: theory,
experiments, and applications ; Elsevier Inc.: , 2000.
159 MacLeod, M.; Scheringer, M.; Hungerbuhler, K. Estimating enthalpy of vaporization
from vapor pressure using Trouton's rule. Environ. Sci. Technol. 2007, 41, 2827-2832.
160 Mintz, C.; Clark, M.; Acree, W.E.; Abraham, M.H. Enthalpy of solvation correlations
for gaseous solutes dissolved in water and in 1-octanol based on the Abraham model. J.
Chem. Inf. Model. 2007, 47, 115-121.
161 Abraham, M.H.; Chadha, H.S.; Whiting, G.S.; Mitchell, R.C. Hydrogen bonding 32. an
analysis of water-octanol and water-alkane partitioning and the delta log P parameter of
Seller. J. Pharm. Sci. 1994, 83, 1085-1100.
162 Development and application of models of chemical fate in Canada: practical methods
for environmental biodegradation rate estimation. Report to Environment Canada; Arnot,
J.A.; Mackay, D.; Gouin, T. ; Trent University: Peterborough, 2005.
163 Howard, P.H.; Muir, D.C.G. Identifying new persistent and bioaccumulative organics
among chemicals in commerce. Environ. Sci. Technol. 2010, 44, 2277-2285.
164 Zhang, X.M.; Brown, T.N.; Wania, F.; Heimstad, E.S.; Goss, K.-U. Assessment of
chemical screening outcomes based on different partitioning property estimation methods.
Environ. Int. 2010, 36, 514-520.
165 Arnot, J.A.; Mackay, D.; Parkerton, T.F.; Bonnell, M. A database of fish
biotransformation rates for organic chemicals. Environ. Toxicol. Chem. 2008, 27, 2263-2270.
166 Pavan, M.; Worth, A.P. Review of estimation models for biodegradation. QSAR Comb.
Sci. 2008, 27, 32-40.
286
167 Weininger, D.; Weininger, A.; Weininger, J.L. SMILES. 2. Algorithm for generation of
unique SMILES notation. J. Chem. Inf. Comput. Sci. 1989, 29, 97-101.
168 McQuarrie, A.D.R. Regression and Time Series Model Selection; World Scientific:
Singapore (New Jersey, London), 1998.
169 Dearden, J.C.; Cronin, M.T.D.; Kaiser, K.L.E. How not to develop a quantitative
structure–activity or structure–property relationship (QSAR/QSPR). SAR QSAR Environ.
Res. 2009, 20, 241-266.
170 Burnham, K.P.; Anderson, D.R. Multimodel inference: understanding AIC and BIC in
model selection. Sociol. Method. Res. 2004, 33, 261-304.
171 Willett, P.; Barnard, J.M.; Downs, G.M. Chemical similarity searching. J. Chem. Inf.
Comput. Sci. 1998, 38, 983-996.
172 Weininger, D. SMILES a chemical language and information system. 1. Introduction to
methodology and encoding rules. J. Chem. Inf. Comput. Sci. 1988, 28, 31-36.
173 Meylan, W.M.; Howard, P.H. Atom/fragment contribution method for estimating
octanol-water partition coefficients. J. Pharm. Sci. 1995, 84, 83-92.
174 Boethling, R.S.; Howard, P.H.; Meylan, W.; Stitleler, W.; Beauman, J.; Tirado, N.
Group contribution method for predicting probability and rate of aerobic biodegradation.
Environ. Sci. Technol. 1994, 28, 459-465.
175 Zhang, X.M.; Brown, T.N.; Wania, F.; Heimstad, E.S.; Goss, K.-U. Assessment of
chemical screening outcomes based on different partitioning property estimation methods.
Environ. Int. 2010, 36, 514-520.