1

URBAN STORMWATER PARTICLE AND DISINFECTION MODELING

By

JOSHUA A. DICKENSON

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2011

2

© 2011 Joshua A. Dickenson

3

To the glory of God and His Son, Jesus Christ

4

ACKNOWLEDGMENTS

I thank my advising professor, Dr. Sansalone, for his countless hours of advising

and storm-chasing and for his financial and intellectual investment in my growth and

development. I thank Dr. Delfino, Dr. Fregly, and Dr. Heaney for serving on my

committee and for their time, critiques, and refinement of my research. I thank the

undergraduate research assistants in the Urban Stormwater Lab for their tireless work

and long hours in the lab and on call. I thank the professors of my classes for sharing

their knowledge and equipping me with vision. I thank my colleagues for open ears and

creative critiques during brainstorming sessions. I thank Ruthie for keeping the

basement of Black Hall clean and for her kind words at every chance meeting. And I

thank my family for their love and support – the thought of them always makes me

smile.

5

TABLE OF CONTENTS page

ACKNOWLEDGMENTS .................................................................................................. 4

LIST OF TABLES ............................................................................................................ 8

LIST OF FIGURES .......................................................................................................... 9

LIST OF ABBREVIATIONS ........................................................................................... 11

ABSTRACT ................................................................................................................... 13

CHAPTER

1 INTRODUCTION .................................................................................................... 15 Overview ................................................................................................................. 15

Literature Review .................................................................................................... 18

2 DISCRETE PHASE MODEL REPRESENTATION OF PARTICULATE MATTER FOR SIMULATING PARTICULATE MATTER SEPARATION BY HYDRODYNAMIC UNIT OPERATIONS ................................................................. 23 Overview ................................................................................................................. 23

Objectives ............................................................................................................... 24 Methodology ........................................................................................................... 25

Selected Particle Size Distributions .................................................................. 26 Computational Fluid Dynamics (CFD) .............................................................. 27 Discrete Phase Model (DPM) ........................................................................... 28

Population Balance .......................................................................................... 30 Hydrodynamic Separators (HS) ........................................................................ 30

Experimental design for baffled HS tests ................................................... 31 Effluent sampling protocol .......................................................................... 31

Supernatant sampling protocol .................................................................. 31 Mass recovery methodology and protocol ................................................. 32 Laboratory analyses ................................................................................... 32 SSC methodology and protocol ................................................................. 33 PSD methodology and protocol ................................................................. 33

Head loss by manual measurement ........................................................... 34 Turbidity ..................................................................................................... 34 QA/QC ....................................................................................................... 35 Efficiency calculation .................................................................................. 35

CFD Model Dataset Creation ........................................................................... 36

Model Validation ............................................................................................... 38 Results and Discussion........................................................................................... 38

Baffled HS Performance ................................................................................... 38 Model Validation Results .................................................................................. 40 PSD Discretization Results ............................................................................... 40

6

Power Law Model (PLM): ................................................................................. 43 Computational Time ......................................................................................... 44

3 OVERALL RATE KINETICS MODEL OF SODIUM HYPOCHLORITE DEMAND BY THE DISSOLVED AND PARTICULATE MATTER FRACTIONS IN URBAN RAINFALL-RUNOFF............................................................................................... 58 Methodology ........................................................................................................... 60

Catchment ........................................................................................................ 60 PM Fractionation .............................................................................................. 61

Batch Reactor Framework ................................................................................ 61 Batch Reactor Setup ........................................................................................ 63

Analytical Methods ........................................................................................... 64 Parallel Second Order Demand Model for Dissolved Phase ............................ 65 Second Order Potential Driving Model for the PM Fractions ............................ 68 Model Evaluation .............................................................................................. 70

Results and Discussion........................................................................................... 70 Control Reactors .............................................................................................. 70

Kinetics Model for Dissolved Phase ................................................................. 70 PM Kinetic Model ............................................................................................. 73

4 SODIUM HYPOCHLORITE DISINFECTION OF INDICATOR ORGANISMS ASSOCIATED WITH URBAN STORMWATER PARTICLES ................................. 84 Methodology ........................................................................................................... 86

PM Fractionation .............................................................................................. 87 Microbiological Enumeration ............................................................................ 88

Batch Reactors ................................................................................................. 89 Residual Chlorine ............................................................................................. 91

Results and Discussion........................................................................................... 92

Batch Reactor Results ...................................................................................... 95 Indicator Organism Partitioning ........................................................................ 97

5 ADVANCED COMPUTATIONAL MODELING OF FREE CHLORINE DEMAND AND DISINFECTION IN UNIT OPERATIONS AND PRECESSES LOADED BY URBAN STORMWATER ...................................................................................... 107 Objectives ............................................................................................................. 108 Methodology ......................................................................................................... 109

Batch Reactor Setup and Initialization ............................................................ 115 CSBR Validation ............................................................................................. 116

Results and Discussion......................................................................................... 116

6 CONCLUSION ...................................................................................................... 127 Free Chlorine Kinetics........................................................................................... 127

Dissolved Phase Reaction Kinetics ................................................................ 127

Particulate Kinetics ......................................................................................... 128

Computational Modeling ....................................................................................... 129 PM Fate and Transport .................................................................................. 129 CFD Free Chlorine Reaction Kinetics ............................................................. 130

7

APPENDIX: ADDITIONAL FIGURES .......................................................................... 132

LIST OF REFERENCES ............................................................................................. 136

BIOGRAPHICAL SKETCH .......................................................................................... 144

8

LIST OF TABLES

Table page 2-1 Table of cumulative gamma distribution modeled gradations ............................. 45

2-2 Morsi and Alexander drag equation and coefficients for a sphere (1972) ........... 45

2-3 Hydrodynamic separator experimental run information and operational parameters ......................................................................................................... 46

2-4 Hydrodynamic separator performance ............................................................... 47

2-5 PM gradations with mean power law model parameters .................................... 47

2-6 Discrete phase model computational time for the baffled hydrodynamic separator ............................................................................................................ 48

2-7 Discrete phase model computational time for the screened hydrodynamic separator ............................................................................................................ 48

3-1 Summary of hydrologic and PM event mean concentration indices for captured events. ................................................................................................. 76

3-2 Global parallel 2nd order demand model parameters for the dissolved phase. ... 76

3-3 Hypochlorite event-based ultimate demand of urban stormwater fractions for the monitored storms. ......................................................................................... 77

4-1 Batch reactor experimental matrix of PM fractions, HOCl dose, and event date. ................................................................................................................... 99

4-2 Batch reactor particle granulometry. ................................................................. 100

4-3 Event mobilization of indicator organisms and percentage of transported organisms associated with each PM fraction. ................................................... 101

5-1 Model parameters for the dissolved parallel second order and PM potential driving force equations. .................................................................................... 120

9

LIST OF FIGURES

Figure page 2-1 Experimental validation of full scale units ........................................................... 49

2-2 Cumulative PSDs utilized in the study. ............................................................... 50

2-3 Results from the full-scale experimental testing on the baffled HS ..................... 51

2-4 Results comparing influent loading concentrations on the baffled HS ................ 52

2-5 Computational results for the screened HS ........................................................ 53

2-6 Computational results for the baffled HS ............................................................ 54

2-7 CFD per-particle size efficiency surfaces for both the screened HS (A) and the baffled HS (B) ............................................................................................... 55

2-8 CFD per-particle size efficiency differential surface for the screened HS and the baffled HS ..................................................................................................... 56

2-9 Predictive results of the power law model for RPD with increasing DN .............. 57

3-1 PSD of quintessential fractions from the batch reactors ..................................... 78

3-2 Physical representation of the parameters of the parallel 2nd order demand model ................................................................................................................ 79

3-3 Predictive fit of the dissolved fraction parallel 2nd order demand model ............. 80

3-4 Transient loading of CODd on the small urban catchment in north central Florida ................................................................................................................ 81

3-5 Maximum particle free chlorine demand ............................................................. 82

3-6 The modeling results of the second order PM chlorine demand model .............. 83

4-1 Event mean most probable number per 100 mL box-plot for twenty-five wet weather events on a small urban watershed in north central Florida ................ 102

4-2 Hypochlorite inactivation kinetics of particle associated coliform organisms on suspended, settleable, and sediment PM ......................................................... 103

4-3 Log removal of particle associated coliforms for the 04-Nov-2010 (Panel A, B) event with an initial hypochlorite dose of 45 mg/L ........................................ 104

4-4 Log removal of particle associated coliforms on sediment PM across the inoculation doses of 15, 30, and 45 mg/L ......................................................... 105

10

4-5 Partitioning of particle associated organisms to suspended, settleable, and sediment PM fractions ...................................................................................... 106

5-1 Physical batch reactor showing stirplate, aluminum foil jacket, and water quality electrodes .............................................................................................. 121

5-2 Illustration of the fluid zones within the batch reactor ....................................... 122

5-3 Histogram analysis of the computational mesh CFD free chlorine concentration .................................................................................................... 123

5-4 Comparison of the second order CFD dissolved demand model with experimental results ......................................................................................... 124

5-5 Comparison of the second order potential driving PM CFD model with experimental results. ........................................................................................ 125

5-6 Comparison of the composite dissolved and PM CFD model with batch reactor data. ..................................................................................................... 126

A-1 Continuously stirred batch reactor (CSBR) schematic. ..................................... 132

A-2 Schematic of monitored urban sub-catchment in Gainesville, FL showing contributing impervious surface. ....................................................................... 133

A-3 Control CSBRs showing hypochlorite kinetics in Nanopure DI for 8 h at 15 mg/L and 24 h at 45 mg/L ................................................................................. 134

A-4 Control CSBRs comparing autoclave sterilized and non-autoclave sterilized stormwater Matrix ............................................................................................. 135

11

LIST OF ABBREVIATIONS

ADB Azide dextrose broth

BGB Brilliant green bile broth

BHI 6.5% NaCl brain heart infusion broth

CFD Computational fluid dynamics

COD Chemical oxygen demand

CODd Dissolved chemical oxygen demand

CSBR Continuously stirred batch reactor

DN Discritization number

DOC Dissolved organic carbon

DPD N,N-diethyl-p-phenylenediamine

DPM Discrete phase model

EMC Event mean concentration

EPA The United States Environmental Protection Agency

FC Fecal coliform

FS Fecal streptococcus

HS Hydrodynamic separator

LTB Lauryl triptose broth

LTB-MUG Lauryl triptose broth amended with 4-methylumbelliferyl-β-D-glucuronide

MBE Mass balance error

MPN Most probable number

MS4 Multiple separate storm sewer system

NJCAT New Jersey Corporation for Advanced Technology

NJDEP New Jersey Department of Environmental Protection

NRMSE Normalized root mean square error

12

PCR Polymerase chain reaction

PLM Power law model

PND Particle number density

PSD Particle size distribution

PM Particulate matter

RANS Reynolds averaged Navier-Stokes

RPD Relative percent difference

RPDave The average relative percent difference

RTD Residence time distribution

SE Standard error

SSC Suspended sediment concentration

TMDL Total maximum daily load

UDF User defined function

VF Volume fraction

WWTP Waste water treatment plant

ΔEMC The change in the event mean concentration

13

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

URBAN STORMWATER PARTICLE AND DISINFECTION MODELING

By

Joshua A. Dickenson

May 2011

Chair: John Sansalone Major: Environmental Engineering Sciences

Urban stormwater is a component of the complex hydrologic water cycle whose

genesis is the result of anthropogenic modification of environmental hydrologic

pathways. This hydrologic modification results in volumetric transport of water and the

mobilization of potential particulate, chemical, biological, and nutrient contaminants.

Moving forward with sustainable development and re-development will require

identification of new technologies and novel reuse resources and an integrated design

and management approach that mitigates deleterious environmental impacts of urban

stormwater runoff and establishes potential reuse applications to alleviate over-

exploitation of current environmental freshwater sources. However, due to the inherent

differences in the nature and constituents of urban stormwater as compared to

well-studied wastewater and environmental water, research needs to characterize and

identify the necessary methods to facilitate safe and sustainable potentials for reuse.

This document proffers experimental research and modeling that explores the role of

particle separation with implications for the potential of utilizing chlorine disinfection for

urban stormwater for reuse. From the experimental perspective, findings include

significant loadings of planktonic and particle associated bacteriological indicator

14

organisms (PAOs) observed in stormwater runoff from a small urban catchment at the

University of Florida as well as high levels of chlorine demand due to dissolved and

particulate constituents of the runoff. Free chlorine disinfection applied to reactors with

differing fractions of event mobilized particulate matter (PM) found that sediment PM

(ϕ > 0.75 µm) exerts a shielding effect that protects PAOs from disinfection at the

applied doses while PAOs in the suspended PM (0.45 µm < ϕ < 25 µm) and settleable

PM (25 µm < ϕ < 75 µm) were successfully inactivated. Computational fluid dynamic

(CFD) models demonstrated that the Lagrangian discrete phase model required

discritization numbers (DN) of heterodisperse gradations and gradations of medium

dispersivity to be in the range of 8 to 16 to minimize computational error, while the

median particle size, the d50m, was sufficient for monodisperse distributions. An

Eulerian-Lagrangian CFD modeling and design methodology implementing the

observed chlorine kinetics was implemented and validated utilizing batch reactor

experimentation.

15

CHAPTER 1 INTRODUCTION

Overview

Moving into the 21st century, appropriations of water resources and water use and

reuse will become an increasingly central factor in the ongoing, sustainable

development of the developed and developing world. True sustainable development

will require a multipronged effort including policy makers, engineers, city planners,

architects and end users focusing on conservation by means of new technology and

processes in addition to habit changes by the end user. An ever increasing body of

evidence points to the detrimental impact of point and non-point source anthropogenic

pollution on the local ecologies of receiving bodies of water. Water reuse has been

identified as a means to reduce this impact as well as supplement current water

resources thereby mitigating environmental drain of source waters.

Much of the effort dedicated to reuse to date has focused on reuse of wastewater.

Reuse of wastewater has been particularly attractive due to pre-existing, centralized

infrastructure and the well documented environmental impact of wastewater nutrients on

receiving bodies of water. Implementation of total maximum daily loads (TMDLs) by

federal and state authorities has transcribed this environmental impact into an

economical impact on local utilities, who have invested in reuse research and

infrastructure. Traditionally, much of the reuse of wastewater has focused on non-

potable reuses, for example irrigation of home lawns and golf courses. However,

current and predicted water scarcity in the U.S. has made the potable reuse of

wastewater a necessity in certain regions.

16

Research into grey water reuse is beginning to bloom. The inchoate research and

technologies have focused on localized applications of captured grey water to in-home

(i.e. toilet flushing) and small outdoor irrigation projects (i.e. lawn or garden). Localized

reuse of grey water potentially could reduce loads to wastewater treatment plants as

well as source water demand. A downside to these technologies is that many of these

applications and technologies will need to be implemented by the end user on a local

scale due to lack of current grey water infrastructure and the high cost of infrastructure

retrofitting.

A potential alternative reuse source is urban stormwater. Increasing development

of impermeable infrastructure continues to increase the volumetric flow of urban

stormwater. Traditionally urban stormwater is volumetrically contained to prevent

localized flooding and subsequently returned to the water cycle via evaporation,

infiltration, or directly into surface waters. Urban stormwater has been identified as non-

point source nutrient and silting pollution and municipalities are at the beginning stages

of being required to treat urban stormwater to comply with established TMDL

regulations for impaired water bodies.

The physical-chemical constituents of urban runoff differ from the traditionally

treated environmental waters and wastewater. The particulate matter (PM) of

environmental waters and wastewater are primarily organic particles, whereas a

significant portion of the entrained particulate matter is inorganic. Some of the inorganic

particulate matter – especially in the runoff from impermeable pavement – is made up of

heavy metal debris from rubber tire wear. Fecal microbial pollution of urban

stormwaters is possible due to animal deposits, inappropriate application of non-

17

chemical fertilizers, or possible sanitary sewer overflows. However, the direct

anthropogenic fecal contaminant link is not present in raw urban stormwater as it is

naturally in untreated wastewater reducing the likelihood of certain opportunistic human

pathogens. In addition, urban stormwater PM tends to be denser, coarser, and more

hetero-disperse than wastewater.

Urban stormwater is also distinct from environmental and wastewater flows by its

stochastic and transient nature. Urban stormwater flows can vary greatly between

events and during the duration of a single event. With respect to PM, stormwater flows

tend to be front loaded – flushing out much of the PM during the initial phase of the

storm. The transient and hetero-disperse nature of urban stormwater flows will require

advanced modeling technologies and techniques to account for their time sensitive

behavior.

Many of the established technologies and processes for the treatment of

environmental waters and wastewater need to be re-evaluated for the purpose of urban

stormwater treatment due to the aforementioned inherent differences in constituents

and for stormwater‟s inherent stochastic and transient nature. Sound research is

necessary prior to the development and application of urban stormwater non-potable

reuse in and around human populations and for potential potable reuse.

Following this macroscopic motivation for urban stormwater research for reuse,

the specific focus of the proposed research is the investigation of the kinetics of free

chlorine demand and disinfection of particle laden urban stormwater and the

subsequent advanced modeling of these processes in computational fluid dynamics

with the special treatment for the hetero-disperse nature of PM in urban stormwater.

18

The dual pronged focus will elucidate fundamental biological, chemical, and physical

properties of the involved kinetic and particulate processes and couple these properties

with direct application to pertinent and realizable design and modeling methodology.

The fundamental hypothesis for the quantification of free chlorine demand is that

urban stormwater will have a unique particulate and soluble based chlorine demand that

differs from both environmental water and wastewater. Particulate matter has been

shown to shield associated organisms from microbial inactivation processes. These

studies have been primarily performed on organic particulate matter, thus a concrete

study illuminating the effect of inorganic particulate matter shielding of associated

microorganisms is warranted. In addition, the effects of gradation characteristics such

as uniformity on discrete computational particle modeling will be elucidated to ensure

accurate modeling with minimal computational cost. The primary motivation for the

development and incorporation of advanced computational modeling is the likelihood of

potentially complex and specific retrofit designs and to reduce unnecessary over design

of disinfection systems.

Literature Review

Disinfection by chlorination is the most utilized form of microbial inactivation

employed in the world today (Hrudey and Hrudey 2004). Due to its relative low cost and

current saturation of installed infrastructure, it can be expected that this trend will

continue for some time. As a primary factor in the effectiveness and cost in this

disinfection process, chlorine demand has been much studied and modeled. This

chemical concept of chlorine demand has also been correlated to disinfection byproduct

formation (Clark 1998). Frequently, chlorine demand has been modeled as first order

decay, however, second order, nth order power law, and parallel first order kinetics

19

have been employed to better fit the data under situations where additional chlorine

demand by factors other than just bulk chlorine decay (Clark 1998). Temperature‟s

significant role in reaction rates has been illustrated in the chlorine demand concept by

modifying the first order reaction rate constants by a factor of 2.5 over a range of 10°C

(Powell et al. 2000).

PM has a deleterious effect on the disinfection process. LeChevallier (1981)

documented the hindering effect of PM, by turbidity as a surrogate, on the disinfection of

environmental surface waters in Oregon. This study identified two mechanisms of

impediment. The first, inherent oxidant demand of the PM decreased the free chlorine

available for disinfection. Second, the PM shielded bacteria that were attached to its

internal pore structure from the free chlorine and potentially from detection using the

membrane filtration technique. Berman (1988) documented similar findings in the

organic rich PM in wastewaters. In addition Berman contributed observations that the

free chlorine permeated particles of diameter <7µm at a rate faster than it permeated

particles of larger diameter. Dietrich (2003) extended these findings and modeled the

permeability of PM to free chlorine with a radial diffusion model. The effect of initial

concentration of free chlorine was shown to be a primary driver in the maximum particle

size that chlorine could effectively disinfect – a size identified as the critical particle

diameter.

The effect of PM and organics during chlorine disinfection of grey water for reuse

was recently studied by Winward (2008). This investigation demonstrated that for grey

water, organic material quantified by total organic carbon only affected the disinfection

process by creating chlorine demand, whereas PM was implicated in shielding particle

20

associated coliforms from the disinfectant. Similar to Dietrich, this study documented

the effect of initial concentration on the ability of the disinfectant to penetrate particles of

increasing size.

The above mentioned previous research on the interaction of PM and disinfectant

on the microbial inactivation process has exclusively focused on organic wastewater

and environmental PM. Significant differences between wastewater and urban

stormwater PM have been documented. Kim and Sansalone (2010) compared PM from

an urban stormwater event to influent at a wastewater treatment plant (WWTP) in Baton

Rouge, Louisiana. The study found that the influent PM to the WWTP was relatively

fine (d50m = 26 µm), mono-disperse (d80m/d20m = 3.1), had a specific gravity of 1.5, and

had a ratio of volatile suspended solids to total suspended solids of 76% (a surrogate

indicator of organic content). In comparison, the urban stormwater event mean was

relatively course (d50m = 136 µm), hetero-disperse (d80m/d20m = 50.5), had a specific

gravity of 2.3, and had a ratio of volatile suspended solids to total suspended solids of

27%. These differences potentially indicate a distinction in behavior for the interaction

of urban stormwater PM with chlorination. The more hetero-disperse coarse gradation

indicates that particle shielding of particle bound microorganisms could be a possible

hindrance to the efficacy of the process (albeit the coarseness and density a boon to

gravitational treatment). In addition the primarily inorganic nature of urban stormwater

particles should indicate reduced instantaneous chlorine demand as well as a reduction

in potential disinfectant byproduct formation as compared to wastewater.

In addition to the above disparity in constituent makeup of PM between

wastewater and urban stormwater, the delivery and loading mechanisms likewise differ.

21

These differences have implications for representative characterization as well as

modeling. The delivery and volume of wastewater flows are generally predictable. The

delivery of PM in urban stormwater flows is governed by flow rate and watershed PM

loading (Sansalone and Kim 2008) and thereby not known a priori (Sansalone and

Buchberger 1997). The flow volume of a storm coupled with the available dry

deposition of the watershed dictates if the flow is mass or flow limited (Christina and

Sansalone 2003) affecting the temporal delivery of PM. These differences dictate

necessary characterization of entire events. Pollutant yields are also non-uniformly

distributed over PM partitions. More metal mass tends to be present on coarse PM,

whereas a higher concentration of metal pollutants tends to be present on finer PM

(Sansalone and Ying 2008). These pollutant properties require representative sampling

of PM concentration (Sansalone and Kim 2008) and characterization of particle size

distribution (Kim and Sansalone 2008).

The presence of microbial indicator organisms has been identified in urban

stormwater flows with a portion of the indicator organisms existing in a particle-bound

state (Charaklis et al. 2005). The presence of particle associated microorganisms has

significant implications for the importance of particle treatment technologies such as

filtration for suspended fractions (< 25 µm) or gravitational settling for sediment (> 75

µm) or settleable (< 75 µm, > 25 µm) fractions. While particle associated indicator

organisms have not shown similar frequency across organisms to centrifugal fraction

partitioning, there was no significant intra-storm variation of particle associated

partitioning for individual microbial indicators (Krometis et al. 2007). However, as

particle loading is not uniform across the duration of a storm and tends to be

22

concentrated toward the rising limb of the hydrograph, likewise particle associated

microbial loading is not uniform across the storm (Krometis et al. 2007). In addition,

research has shown that the particle partitioning of Clostridium perfringens spores

mimics that of the protozoan cysts of Giardia and oocysts of Cryptosporidium for which

it has been proposed as an indicator organism (Chizek et al. 2008).

Computational Fluid Dynamics (CFD) is a state of the art tool for modeling

complex fluid flows, particle settling phenomena, and simulating reaction kinetics

spatially and transiently. CFD has been previously used to model steady and transient

settling in urban stormwater unit operations (Pathapati and Sansalone 2009) as well as

in the optimization of a potable water settling basin (Goula et al. 2008). CFD has also

been utilized by Baawain (2006) as an optimization tool for the design of a disinfectant

storage-contact chamber. This study sought to improve the performance of the contact

chamber by geometrically improving the mixing and flow characteristics of the basin.

Baawain‟s application of CFD was an extrapolation of the U.S. Surface Water

Treatment Rule of the Ct10 concept and sought to maximize t10, but did not apply CFD to

model the oxidant demand or integrate localized Ct values across the domain.

23

CHAPTER 2 DISCRETE PHASE MODEL REPRESENTATION OF PARTICULATE MATTER FOR SIMULATING PARTICULATE MATTER SEPARATION BY HYDRODYNAMIC UNIT

OPERATIONS

Overview

Particulate matter (PM) is a common pollutant in rainfall-runoff (Heaney and Huber

1984) and surface waters (Delfino 1977). As historically practiced, gravimetric indices

such as total suspended solids (TSS) and suspended sediment concentration (SSC) do

not provide particle size distribution (PSD) representation. However, PM discretization

methods allow examination of chemical and biological interactions with PSDs, for

example with metals (Chellum and Wiesner 1997). Silt- and clay-size PM provide

habitat and a protective matrix for microbes (Lünsdorf et al. 2000) and is a mobile

substrate that provides disinfection resistance (LeChevallier et al. 1981). As a result,

PM clarification is common for treatment of stormwater (Small and DiToro 1979) and

wastewater (Tchobanoglous et al. 2003) or for water treatment optimization (Boccelli et

al. 2004). PM discretization as a function of hetero-dispersivity is important in modeling

treatment and fate of PM and PM-bound pollutants.

A common analysis for PM separation by treatment unit operations is the overflow

rate theory with the inherent assumption of Type I gravitational settling, where PM

settles discretely with negligible particle-particle interaction and impact on the fluid flow

field. This basic theory can be combined with a constitutive model of PM settling

described by Newton‟s law. Coupling overflow rate, Newton‟s Law and PSD

discretization, while less common, permits a more complete description of discrete PM

settling irrespective of the continuous fluid phase dynamics. The most basic PSD

24

discretization and separation representation is using a mass-based median size (d50m),

with increasingly accurate representation generated by higher PSD discretization.

While the coupling of a representative PSD and Newton‟s Law with overflow rate

provides a more accurate description of the discrete PM phase, a major shortcoming of

the basic overflow rate concept is the lack of a quantitative coupling with the potential

hydrodynamic complexity of the fluid phase. Coupling the dynamics of the fluid phase

can range from semi-empirical overflow rate models such as Hazen‟s law (Hazen 1904,

Fair et al. 1968) to a fundamental description with Navier-Stokes equations (Pathapati

and Sansalone 2009). Computational fluid dynamics (CFD) has significantly improved

modeling of PM separation and fate for complex geometries, non-ideal flow fields, and

transient flows (Patruno et al. 2009, Pougatch et al. 2009). However, the level of PM

discretization as a function of PSD hetero-dispersivity is required when modeling unit

operations with a Lagrangian-Eulerian CFD approach.

PM laden flows represent a challenging modeling phenomenon with discrete PM

sizes, as PSDs can be considered a continuum. Using a discrete phase model (DPM)

to simulate PM separation requires discrete PM sizes for a continuous PSD. As hetero-

dispersivity increases, the particle number exponentially increases with increasing

computational effort. Identifying PSD discretization needed for acceptable results and

computational economy is important.

Objectives

A primary study objective is the examination of PM discretization requirements in a

CFD model for selected levels of granulometric size hetero-dispersivity. Additionally,

this study illustrates the impact of gradation uniformity and overall gradation fineness on

PSD discretization requirements for CFD modeling of two different hydrodynamic

25

separators (HS) commonly utilized worldwide for treatment of urban drainage. Using

controlled physical modeling of a baffled HS and screened HS shown in Figure 2-1, to

validate the CFD model, the study hypothesized that the error in modeling PM

separation is a function of the PSD discretization.

Methodology

A granulometric attribute of PM is the PSD. Urban drainage PM in wet weather

(rainfall-runoff) or dry weather (wastewater) flows is hetero-disperse. To explore the

effect of PSD dispersivity on PSD discretization requirements a rubric is needed to

characterize the gradation uniformity. Folk and Ward (1957) proposed a sorting

coefficient (σI) for this granulometric attribute. With the sorting coefficient, gradations of

similar uniformity can be generated at a chosen d50m. Equation 2-1 presents the sorting

coefficient modified by a negative sign since percentiles are reported as % finer by

mass whereas Folk and Ward present them as % greater.

(

) (2-1)

In this expression is the phi-scale particle size gradation parameter (McManus 1966),

and is defined by Equation (2-2).

(2-2)

Where n is the percentile, n is the phi parameter of the nth finer percentile, dn is the PM

diameter of the nth finer percentile, and d0 is the unit length to non-dimensionalize the

equation.

In order to systematically study the effect of PSD discretization for size gradations

of differing σI and d50m, a methodology is needed to generate PSDs that vary only in

26

these two parameters while simultaneously generating representative urban drainage

PM gradations. Previous studies (Sansalone and Ying 2008, Lin et al. 2009) have

utilized a two parameter cumulative gamma distribution to model hetero-disperse urban

drainage PSDs. The shape and scale factor parameters in the cumulative gamma

distribution, k and λ, are physically analogous to σI and d50m. Equations 2-3 and 2-4

represent the gamma distribution (f(x)) and cumulative gamma distribution (F(x)),

respectively.

( ) ( )

( )

( )

(2-3)

( ) ∫ ( )

(2-4)

Where k is the gamma shape factor; λ is the gamma scale factor; and x is the particle

diameter.

Selected Particle Size Distributions

Table 2-1 and Figure 2-2 describe the chosen gradations for this study. Gradation

characteristics are selected to elucidate the effects of PSD dispersivity and also the

PSD d50m on the CFD simulation results for different levels of PSD discretization using a

discretization number (DN) at a fixed density (ρ = 2.65 g/cm3; silica sand). The nine

gradations are generated by identifying gamma distribution k and λ parameters that

matched the 3x3 gradation matrix of mono-disperse (σI < 0.35; in this study: σI = 0.11),

moderately-disperse (σI ≈ 1.00; in this study: σI = 1.03) and hetero-disperse (σI > 2.00;

in this study: σI = 2.64) PSDs. The gravimetric median sizes for these PSDs are coarse

(d50m = 100 µm), fine (d50m = 66.7 µm), and very fine (d50m = 33.3 µm).

27

Computational Fluid Dynamics (CFD)

CFD simulates behavior of unit operations at a variety of scales, from pilot-scale

behavior to smaller-scale phenomena, such as the internal flow field velocities and

pressure distributions as well as PM transport and fate. CFD is utilized to model

behavior of each HS as a function of granulometric attributes (PSD, d50m and σI) and

flow. CFD is based on numerical solutions to Navier-Stokes equations across a domain.

Specifically, to model the flow fields in this study, Reynolds Averaged Navier-Stokes

(RANS) equations (Ferziger and Peric 2002) are utilized with Fluent 6.3.26. The RANS

equations decompose the bulk, time independent fluid flow from the transient turbulent

fluctuations. Averaged over a time-scale much larger than the time-scale of the

fluctuations, the transient turbulent fluctuations become zero leaving only the bulk fluid

flow and stationary turbulence structures. Equations 2-5 and 2-6 succinctly describe the

steady state RANS continuity and momentum equations respectively.

( )

(2-5)

( )

( )

(2-6)

In these equations ρ is fluid density; xi is the ith direction vector; is the Reynolds

averaged velocity in the ith direction; is the Reynolds averaged pressure; and gi is the

sum of body forces in the ith direction (i.e. gravity in the negative z direction). The

decomposition of the non-linear convection in the momentum equation results in

Reynolds Stresses – . The Reynolds Stresses are unknown quantities and can be

modeled with the semi-empirical κ-ε model (Equations 2-7 through 2-8) that was

28

developed by Shih et al. (1995) which performs well for rotating homogeneous

(screened HS) and boundary-free (baffled HS) shear flows.

(

)

(2-7)

(

)

√

(2-8)

[

]

√

(2-9)

The constants: σk = 1.0, σε = 1.2, and C2 = 1.9. In Equations 2-7, 2-8, and 2-9, k is the

turbulent kinetic energy; ε is the turbulent energy dissipation rate; S is the mean strain

rate; νT is the eddy viscosity; ν is the fluid viscosity; and and are defined as

above.

The velocity and pressure flow fields in the computational domain are solved by

the coupled solution of the RANS equations and the κ-ε model utilizing second order

upwind cellular approximations on a tetrahedral meshing scheme. The geometric mesh

size of the screened HS is 1.3 million cells and 3.1 million cells for the baffled HS.

Discrete Phase Model (DPM)

To model transport of PM within HS units, a mixed mode Eulerian-Lagrangian

reference frame is utilized where fluid velocity and pressure flow fields are modeled in

an Eulerian or control volume reference frame and PM is modeled as discrete particles

in a Lagrangian or particle tracking reference frame. PM transport modeled in the

discrete phase is integrated across the fluid velocity and pressure flow fields modeled in

the Eulerian reference frame. This does not account for particle influence on velocity

and pressure flow fields and is restricted to dilute fluid flows with <10% volume fraction

(VF) (Brennen 2005). Dilute flows are governed by Type 1 settling. Even with this

29

restriction, many flow situations, including the HS units can be successfully modeled as

dilute flows (this study: VF << 10%). Furthermore, as a dilute, non-agglomerating flow,

particle collisions are negligible and the resulting CFD model is independent of

concentration.

In the discrete phase, particles are tracked by calculating particle acceleration as

a particle moves in the flow field. This is an evaluation of Newton‟s second law

enumerated in Equations 2-10 through 2-13 were i represents the three dimensional

direction (i.e. x, y, z).

( ) ( )

(2-10)

(2-11)

| |

(2-12)

(2-13)

The formulation of Equation 2-10 is particle acceleration ⁄ is equal to the

summation of the forces per unit particle mass. ( ) is the drag force per unit

particle mass; and ( ) ⁄ is buoyancy/gravitational force per unit particle mass.

Equation 2-12 is the definition of the relative Reynolds number for flow around a sphere.

In Equation 2-11 and 2-12 ρ is the fluid‟s density; ρp is the particle‟s density; dp is

particle diameter; vp_i is particle velocity in the ith direction; vi is the localized fluid

velocity in the ith direction; and µ is the dynamic viscosity. Morsi and Alexander (1972)

present Equation 1-13 as a mathematical correlation for the drag coefficient of spherical

30

particles that predicts experimentally obtained values to within 2% up to Re = 50,000.

Coefficients to Equation 2-13 are in Table 2-2.

Population Balance

The computational DPM is linked to quantification of the HS performance by the

PM population balance equation. In this study the unit operation behavior is

characterized by the change in event mean concentration (ΔEMC). For steady-state

conditions, assuming no particle coalescence, breakage, nucleation or destruction, a

population balance equation is utilized (Equation 2-14).

∑ ( )

∑ ( )

∑ ( )

(2-14)

Where fd(ξ) is a discrete particle number density (PND) function with the internal particle

diameter coordinate ξ; and xξ is the mass fraction of a particle with diameter ξ.

Therefore, the summations represent the total mass fractions of PM injected in the

influent, eluted in the effluent, or remaining within the unit operation. Note that Σ()infl =

1. From this analysis, the computationally modeled HS performance is given by ΔEMC

= 1 – (Σ()effl / Σ()infl) * 100%. An in depth presentation of the population balance equation

can be found in Jakobsen (2008).

Hydrodynamic Separators (HS)

PM separation of a baffled HS and screened HS is used to examine the role of

PSD discretization for differing flow fields with physically-validated CFD models. The

screened HS is a 1.8 m dia. unit designed to include a potential swirling mechanism and

a 2400 m screen to deflect PM from the flow stream. The baffled HS is a 1.8m dia. unit

designed to provide volumetric isolation of settled PM, trap oil, grease and floatables,

and settle PM. Units are illustrated in Figure 2-1.

31

Experimental design for baffled HS tests

The parameters selected in the experimental design include:

Particle concentration (100 and 300 mg/L), as suspended sediment conc. (SSC)

Flow rate (2%, 5%, 10%, 25%, 50%, 75%, 100% and 125% of the given design flow rate)

Influent particle gradation based on an NJCAT gradation (Total 2003).

Two gravimetric concentrations of NJCAT influent particles (100 mg/L and 300

mg/L) are examined in this analysis, in order to investigate the effect of influent solid

concentration on the separation performance of the baffled HS across the range of

hydraulic loading capacities considered. Table 2-3 presents the complete experimental

design.

Effluent sampling protocol

The sampling is conducted according the following procedure. During the test

running time, twenty representative effluent samples are taken manually as discrete

samples in 1L wide mouthed bottles. Samples are collected in duplicate at constant

intervals through the entire duration of the run at the effluent section of the unit. The

sampling interval times spanned from 2 minutes to 48 minutes for the range of operating

flow rates from 2% to 125% of design flow rate.

Supernatant sampling protocol

The sampling protocol examining the supernatant PSDs consists of taking one

duplicate sample at the geometric midpoint of the supernatant after overnight settling.

An investigation was performed to confirm the sampling methodology adopted is an

appropriate procedure to adequately quantify the PSD of PM remaining entrained within

the supernatant after a period greater than 8hrs. This analysis intended to verify the

PSD of PM remaining in suspension at various depths in the supernatant after overnight

32

settling after an event based test on the baffled HS. In particular, four PSD samples

were sampled from the draining supernatant of the baffled HS at 4 separate times and

were calculated so that the samples were taken at four evenly spaced intervals of height

of the draining supernatant. The result proved there is not a significant deviation among

the four samples in terms of supernatant incremental PSD after overnight settling.

During this same investigation, four 4L SSC samples taken at the same time as the

PSD sample were analyzed and the result determined that there was a significant

variation in the SSC throughout the vertical profile within the unit, thus it was concluded

that the best sampling approach for SSC would be a 4L composite SSC sample made

from four 1L samples taken at the same time as the PSD samples. The supernatant

mass represents approximately 7% of the event captured mass.

Mass recovery methodology and protocol

After the supernatant sample has been collected the wet slurry from the system is

recovered from the bottom of the unit by manually sweeping it through the washout into

buckets and taken to the laboratory where they are allowed to stand for quiescent

settling and dried in glass trays at 110 degrees Celsius in an oven. After the slurry

completely dries the dry silica is disaggregated and collected in pre-weighed glass

bottles and the gross weight is recorded to find the overall efficiency of the system

based on mass and for the mass balance. Laser diffraction analysis for the collected dry

sample is then performed to analyze the PSD of the captured PM.

Laboratory analyses

During a run a total of forty 1L samples were collected. The laboratory analyses

consisted of SSC, PSD of the aqueous phase and the dry phase of the captured mass

using the Malvern Mastersizer 2000, and the mass balance for the efficiency of the

33

system. Twenty samples were used for the SSC analysis and 20 for the PSD analysis.

The SSC analysis was performed as duplicate composite samples (A and B), while the

PSD analysis was performed on each 1L sample to enable quantification of the PSD

over the duration of the experimental run. These methodologies are elaborated below.

SSC methodology and protocol

SSC analysis was performed to quantify particle concentration for each effluent

composite sample as collected from each run and to calculate the effluent mass load for

the operating flow rates. In addition to the 10L composite samples, an additional 1L was

added to the event mean concentration (EMC) calculation to act as an initial sample to

accommodate the fact that the samples were taken as end interval samples instead of

mid-interval samples. Since the run was conducted under steady flow conditions and

the samples collected at equal intervals, the particle concentration in the composite

sample represents the mean effluent concentration. The protocol specifically followed

for this laboratory analysis is the ASTM D 3977 (2002).

PSD methodology and protocol

Effluent and supernatant PSD measurement in aqueous phase. A large

variety of instruments have been developed for particle size determination. A relatively

new, state of the art, high resolution method is a laser diffraction analyzer. The Malvern

Mastersizer 2000 is a commercial laser diffraction analyzer and was utilized in this

experimental analysis to characterize particle sizes. The instrument technology is based

on laser diffraction, occurring when a laser beam passing through a dispersion of

particles in air or in a liquid is diffracted at the particle surface. The angle of diffraction is

influenced by the size and the shape of the particle. As the particle size decreases, the

scattering angle increases (Jillavenkatesa et al. 2001). The Mastersizer 2000 detects

34

particle sizes in the range of 0.02 to 2000 μm assuming a spherical diameter. The 10

duplicate samples are analyzed independently in the Mastersizer 2000 as 20 one liter

samples. The sample collection, handling, and particle analysis procedure followed

Standard Method 2560 (Eaton et al. 1998).

Captured particulate PSD measurement. In order to representatively sub-

sample the dry mass the silica is uniformly mixed to obtain a sub-sample as

representative as is physically obtainable. Duplicate 20 gram samples are taken for the

dry phase of the laser diffraction analyzer. The dry dispersion cell is connected to the

laser diffraction analyzer and the dry sample is measured by forming a PM aerosol with

a high pressure, high velocity air stream. The PSDs measured are observed for stability

and averaged.

Head loss by manual measurement

The head loss through the system was measured by manual tape measurement.

This was accomplished by measuring the approximately static head over the inlet and

outlet sections of the fiberglass insert. These measurements were performed in the

same spatial location for each run and were trued to a level to account for the difference

in vertical height of the surface of the insert at the inlet and outlet sections.

Turbidity

Effluent and influent turbidity were measured by a portable turbidimeter probe (YSI

600OMS). Three point calibration curve was determined to calibrate the probe by using

the turbidimeter calibration standard solution. The turbidity of the effluent was

characterized throughout the run by an in-situ YSI 600OMS probe installed before each

run in the outlet riser. The average value of the turbidity once a steady state was

obtained was designated as the effluent turbidity. For this study, turbidic steady state is

35

defined as a turbidity measurement at 95% of the first maxima. Another in-situ turbidity

meter was installed in the inlet drop tee and to avoid the restriction of flow in the orifice

plate. This data recorded by this inlet turbidity meter was significantly besmirched by

turbulent flow in the inlet and possibly light scattering from the water surface above the

orifice plate. In addition, trends in the data from this turbidity meter indicated particle

settling on the surface of the small angle lens used in this device and further obscured

the data. Large scale turbidity tests were performed to measure the influent turbidity by

using a YSI turbidimeter. The test was carried out on 150L samples in which a specific

amount of NJCAT mass, depending on the concentration considered (100 mg/L and

300 mg/L), was added. Initially, a turbidity measurement of the background was taken.

Then, the NJCAT gradation mass was added to the sample. After vigorously mixing the

sample and ensuring the cleaning of the instrument lenses, two discrete turbidity

readings were recorded. The same procedure is followed for both sediment

concentration of 100 mg/L and 300 mg/L.

QA/QC

Verification of mass balance for each experimental run. The PM mass

balance was calculated from dried captured mass, effluent mass load, and supernatant

mass load. The mass balance error (MBE) criterion is ±10% MBE and determined by

Equation 2-15.

100

load massInfluent

load mass Infl.mass Captured mass Eff. (%)

MBE (2-15)

Efficiency calculation

The methods selected to estimate the particle removal efficiency of the unit are

described in the present section. The first approach used is efficiency ratio or percent

36

removal. This measure is based on the change in event mean concentration (ΔEMC),

and is defined in Equation 1-16.

100EMC INF

EMC EFF - EMC INF EMC

(2-16)

Where INF EMC is defined as the total influent solid mass loading divided by the total

treated influent volume and EFF EMC represents the average effluent SSC obtained

from composite samples collected in the field

The second methodology used to evaluate the removal efficiency is based on the

captured PM recovered from the unit (Equation 2-17).

100PMInfluent

PM captured of Mass Mass

(2-17)

CFD Model Dataset Creation

To investigate the PSD discretization effect, a computational dataset of PM

separation based on particle size is modeled for the HS units as a function of flow rate.

Datasets contain separation efficiencies calculated for PM from 1 to 1000 m in 1m

increments for a series of flow rates (0.36, 0.91, 1.8, 4.5, 9.1, 13.6, 18.1, 22.7, 28.3, and

34.0 L/s). Intermediate values were determined, when needed, by linear interpolation.

For all gradations particle sizes ranged from 1000 µm or 1µm and separation

efficiencies are based on this range of particle sizes.

PSD discretization is performed on a symmetric, gravimetric basis on an

arithmetic scale. For the first increment, DN = 1 and the entire PSD distribution was

represented by the PSD d50m. Subsequently for DN = 2, the PSD was subdivided into

two ranges and each was represented by the representative gravimetric median particle

diameter for each range, the d25m and d75m quartiles. Sequentially subdividing each

37

range further in a similar fashion resulted in PSD representation by 2(n-1) (n = 1,2,3…)

discrete particles giving DNs of 1, 2, 4, 8... The DN was increased until the model

converged sufficiently, as defined by the model discretization error. Advantages of

symmetric, gravimetric discretization is that each discrete particle size range is equally

weighted simplifying calculations of ΔEMC and that the discretization is relatively

independent of flow rate and HS. However, this discretization scheme could also

expend computational resources on particles that are larger than the particle completely

removed under given conditions. Other discretization schemes that are customized for

a particular unit operation over a more narrow flow rate range potentially offer greater

accuracy for similar DNs or lower computational effort by focusing computational

resources on a PM range of relevance for the unit operation and flow rate.

The computational model discretization error is error introduced into the modeled

ΔEMC by representing a continuous PM property, the PSD, with representative discrete

PM sizes. As the DN approaches infinity, the model will converge to a simulated

ΔEMC. The model PSD discretization error for a specific DN is characterized by an

absolute Relative Percent Difference (RPD) as compared to DN =∞, represented by a

surrogate DN_MAX (Equation 2-18).

|( ) ( ) |

( )

(2-18)

DN_MAX is the calculated surrogate of the theoretical model at DN = ∞ based on

a calculated overall mean RPD of DN_MAX and DN_MAX-1 of less than 0.1% for all

models and HS units.

38

Model Validation

To ensure that a CFD model produces physically representative results, modeled

PM separation is validated in the laboratory by pilot-scale physical models loaded by a

hetero-disperse regulatory gradation (23) for a series of flow rates (Baffled HS: 0.4, 0.9,

1.8, 4.5, 9.1, 13.6, 18.1, 22.7 L/s and Screened HS: 2.3, 5.7, 11.4, 17.2, 22.9, and 28.6

L/s). For the CFD model, the PSD was represented by the hetero-disperse NJDEP

gradation (Total 2003) with a d50m of 66.7µm (#8: k = 0.56, λ = 232.54, R2 = 0.99).

Model fit is examined based on mean and maximum absolute RPD (Equation 2-19).

|( ) ( )|

( )

(2-19)

An overall mean RPD of < 10% is considered to be an acceptable fit between the

physical and numerical models. An additional rubric identified as the bias error was

calculated to give an indication of the fit of the model to the physical model results. The

bias error is essentially the mean RPD excluding the absolute value in the calculation.

The bias error gives an overall indication of the tendency for the model to over- or

under-predict the physical results.

Results and Discussion

Baffled HS Performance

The treatment performance based on ΔEMC and ΔMass of the baffled HS as a

function of flow rate and for both aforementioned concentrations is reported in Table 2-4

and Figure 2-3. The figure also correlates a given experimental run to a MBE, which

confirms that each accepted run remained within ±10% MBE in order to fulfill the QC

protocol.

39

As expected the performance of the unit demonstrates exponentially increasing

mass removal as the flow rate tends towards zero flow. At 2% design flow rate, the unit

exhibits removal efficiency nearly equal to 90%. This efficiency rapidly decreases with

increased flow up to the flow rate of 142gpm at 50% design flow. At this point the rate

of decrease of efficiency with respect to the flow rate diminishes and eventually the

efficiency reaches a minimum of approximately 50% removal at about 355gpm or 125%

design flow.

This phenomenon corresponds to the exponentially decreasing settling velocity of

PM with respect to particle size. Since the influent PM gradation is heterodisperse and

contains a significant portion of finer particles it is expected that the unit, primarily using

the mechanism of gravitational settling, will not reach high efficiencies unless it is

operating at very low flow rates. For the higher flow rates the unit is primarily capturing

coarse PM while the fine material is passing through the unit.

In order to investigate the influence of the influent sediment concentration on the

overall removal efficiency of the system, the experimental results corresponding to the

concentrations of 100 mg/L and 300 mg/L are reported respectively for EMC and

Mass in Figure 2-4.

As depicted in the plots, the removal efficiency performed by the system for the

two NJCAT sediment concentrations considered (100 mg/L and 300 mg/L) across the

entire range of flow rates tested is nearly comparable. This outcome demonstrates that

for these dilute concentrations particle-particle interactions are negligible. Therefore, the

removal efficiency does not depend on the influent solid loading concentrations as long

as they remain within this range and can be considered dilute.

40

Model Validation Results

The criteria outlined for model discretization error is utilized and DN_MAX is

incremented for the screened and baffled HS for each of the nine PSDs until a mean

overall absolute modeled RPD of < 0.1% is obtained for each PSD tested. For these

HS units and PSDs, the criteria are satisfied with the DN = 128, a PSD discretization

greater than provided by most PM analysis.

Model validation was performed and results illustrated in Figure 2-1. The

screened HS model had an overall mean absolute RPD of 6.5%, a maximum absolute

RPD of 11.7% (at 2.3 L/s) and an over prediction bias error of 5.0%. The baffled HS

had an overall mean absolute RPD of 4.6% (100 mg/L) and 4.1% (300 mg/L), a

maximum absolute RPD of 9.2% (100 mg/L at 1.8L/s) and 7.3% (300 mg/L at 18.1 L/s),

and an over prediction bias error of 1.1% (100mg/L) and 0.7% (300mg/L). The low bias

likely indicates that a majority of the error is likely from physical data collection. The

CFD-modeled HS units satisfy the requirements for model validation.

PSD Discretization Results

Figures 2-5 and 2-6 illustrate the PSD discretization results for the screened HS

and the baffled HS, respectively. Results illustrate an exponential decline approaching

zero with increasing DN. The role of the gradation sorting on the error due to

discretization is apparent especially for low DNs. Of primary importance is the DN

required for acceptable convergence of the ΔEMC for a given DPM. For the six

gradations centered on 66.7 µm and 100 µm the maximum modeled RPD is less than

2.1% at a DN of 8. For a DN of 8 and the three gradations centered on 33.3µm, the

maximum modeled RPD is 7.5%. However, at a DN of 16 the maximum modeled RPD

is 1.1%.

41

The use of the 1µm PSD increment for each HS dataset facilitated subsequent

analysis of unit behavior. This high resolution CFD output generated large CFD particle

tracking files that are post-processed as batch script files and further processed through

user-defined coding. An advantage of utilizing this methodology is the time-intensive

CFD output did not require model re-creation to further explore unit behavior for

different PM size gradations that are within the limits of the original high resolution PSD

range; including sampled gradations from physical modeling. The drawback to this

methodology is the necessity of linear interpolation to evaluate PM sizes that are not

explicitly solved for by the CFD simulation. However, since the resolution of the output

is high compared to the required analysis resolution, this impact is minimal.

High resolution, per-particle size removal datasets offer a fundamental perspective

on the behavior of any unit operation, in this study, two similar HS units. These

datasets, illustrated in figures 2-7 and 2-8, provide a three-dimensional surface of

performance for different steady flow rates that essentially are a HS “fingerprint” in a

flow-PSD domain. Currently, regulatory requirements for approval of unit operations

such as these HS units rely on the physical modeling of a given unit, under specific flow

conditions, for a specific PM influent gradation. These gradations vary widely between

regulations as illustrated by the NJDEP (hetero-disperse) and Indianapolis (mono-

disperse) gradations in Figure 2-2. A disadvantage in such disparate regulations is the

required physical modeling needed for diverse regulatory frameworks. One solution is

for regulators to allow full-scale testing of a unit operation to physically model the per-

particle size behavior with a hetero-disperse PSD with high resolution PSD monitoring,

such as a laser diffraction analysis given a demonstration of mass and volume

42

balances. Such high resolution validation datasets can be utilized in CFD post-

processing to comply with diverse or regional regulatory frameworks while still providing

meta-performance indices such as ΔEMC.

Most urban drainage PSDs are hetero-disperse as represented by the NJDEP

regulatory PSD and therefore must be examined at DN levels that reasonably minimize

the RPD between measured and modeled results. However, a significant conclusion of

this study is that only mono-disperse gradations are reasonably measured and modeled

by a DN = 1; the d50m. Given a constant density across the PSD and a given flow field,

this result matches the expectation driven by particle fluid dynamic theory. Across the

narrow PM sizes in mono-disperse PSDs, particles will remain within the same flow

regime (turbulent/transitional/laminar) and the non-linear relationship of particle size to

particle drag force will have negligible impact on the final result due to the minimal

diameter range encountered with uniform PSDs. However, as the PSD becomes hetero-

disperse or dense, particles will no longer reside in the same flow regime and the non-

linear relationship between particle size and particle drag force greatly impacts the

accuracy of using the singular gravimetric median as the sole representative particle.

The uniformity of the gradation also influenced the rate of convergence with regard

to the DN, as hypothesized in this study. For uniform gradations, convergence is almost

immediate as when a PSD is well represented by the d50m. The convergence rate of the

medium gradations centered on 66.7 and 100 µm is located between the convergence

rate of the hetero-disperse and mono-disperse gradation in figures 2-5 and 2-6. This

result can be explained based on the hetero-dispersivity of the PSD. However, for the

finest PSDs, those centered on 33.3 µm, the convergence rate of the medium and

43

hetero-disperse gradations are more similar than for coarser PSDs illuminating the

additional influence of overall fineness of the gradation on the convergence rate. For

these fine PSDs a significant portion of the gradation is below the effective separation

potential of both HS units for a majority of the studied flow rates. This results in a

largely suspended PM with more opportunity for variability in modeled behavior because

of poorly represented gradations and low separation potential by both units.

The DN is the driving factor in the modeling discretization error as well as a

significant contributor to computational time. In general, doubling the DN doubled the

computational time while providing diminishing improvement in model accuracy.

Gradation uniformity and fineness impact the discretization error at DNs in the range of

1 to 8. As DNs increase beyond 16, the effect of gradation uniformity and overall

fineness are essentially negligible for all tested gradations as shown in figures 2-5 and

2-6. For the gradations centered on 66.7 µm and 100 µm model convergence is

achieved at a DN of 8 regardless of gradation uniformity. Based on the results, a

discretization at a DN of 8 to 16 is generally provides a discretization error that for many

applications can provide acceptable results for coarser gradations. For very fine

gradations, similar to the gradations centered on 33.3 µm, a DN of 16 to 32 is

recommended.

Power Law Model (PLM):

For this study mean and maximum RPD are modeled as a function of DN. Results

are examined with a power law model (PLM) (Cristina and Sansalone 2003) across all

granulometric variations (Equation 2-20).

( ) (2-20)

44

The power law constants and exponents, c and m, are summarized in Table 2-5

and the model fit is presented graphically in Figure 2-9. Physically, the power law

constant (10c) represents the RPD of the d50m. The m values (slopes) are all negative

and are an index for the rate of convergence to zero discretization error with increasing

DN. Higher negative slopes indicate a higher convergence rates and therefore lower

DN to provide a given RPD. The coefficients for the PLM are generated by the least

squares fit of the CFD RPD results as well as a linear transcription of the least squares

fit to the upper 95% confidence interval for a more conservative model. In general, the

power law model presented should extend well as an initial guidance for other particle

fluid systems that maintain the underlying assumptions of the provided computational

analysis.

Computational Time

Tables 2-6 and 2-7 present the processor computational time for executing the

steady state DPM particle tracking model for a few select flow rates as total processor

seconds for the baffled HS and screened HS, respectively. These results illustrate that

the computational time approximately doubles by doubling the DN. These results were

obtained using a Dell Precision 690 with dual quad core Intel Xeon E5345 CPUs at 2.33

GHz and 16 GB of RAM running Fluent 6.3.26 on Microsoft Windows XP Professional

x64.

45

Table 2-1. Table of cumulative gamma distribution modeled gradations

Gradation #

#1a #2a #3a

#4b #5b

#6b #7c #8c

#9c

k 162.9 177.2 163.0 2.3 2.3 2.3 0.6 0.6 0.6 λ 0.2 0.4 0.6 17.0 33.8 50.3 116.3 232.6 353.2 σI 0.1 0.1 0.1 1.0 1.0 1.0 2.6 2.6 2.6 ρ (g/cm3) 2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.7 d95m (µm) 37.8 75.3 113.4 88.4 176.8 262.8 241.9 483.2 730.6 d84m (µm) 35.9 71.8 107.9 62.5 125.2 186.1 128.9 257.3 388.4 d50m (µm) 33.3 66.7 99.9 33.3 67.0 99.4 33.2 66.1 99.3 d16m (µm) 30.8 61.8 92.3 15.1 30.6 45.4 3.8 7.5 11.2

d5m (µm) 29.2 58.8 87.6 8.1 16.3 24.2 0.5 0.9 1.4 a Uniformly distributed; b Gradations of medium distribution; c Hetero-disperse. The characteristics of selected gradations including gamma distribution parameters (k, λ), and the sorting coefficient (σI) that are utilized in this study. Gradation #8 is the gamma curve fit of the NJDEP gradation (R2 = 0.99). The other gradations were selected to systematically explore the effect of uniformity and d50m on discretization error and were chosen so that the uniform gradations are very well sorted (σI < 0.23), the medium gradations transect the boundary between moderately sorted and poorly sorted gradations (σI = 1.0), and the other hetero-disperse gradations (#7, #9) have similar sorting to the NJDEP distribution (#8). The distribution indices (percentiles) necessary for the calculation of the sorting coefficient are also included.

Table 2-2. Morsi and Alexander drag equation and coefficients for a sphere (1972) Reynolds Number K1 K2 K3

<0.1 24.0 0 0 0.1 < Re < 1.0 22.73 0.0903 3.69 1.0 < Re < 10.0 29.1667 -3.8889 1.222 10.0 < Re < 100.0 46.5 -116.67 0.6167 100.0 < Re < 1000.0 98.33 -2778 0.3644 1000.0 < Re < 5000.0 148.62 -4.75 X 104 0.357 5000.0 < Re < 10,000.0 -490.546 57.87 X 104 0.46 10,000.0 < Re < 50,000.0 -1662.5 5.4167 X 106 0.5191

46

Table 2-3. Hydrodynamic separator experimental run information and operational parameters

Run # Desig

n f

low

rate

(%)

Actu

al flo

w r

ate

(gp

m)

Targ

et

influ

ent

co

ncen

tration

(mg

/L)

Actu

al in

flu

en

t

co

ncen

tration

(mg

/L)

Influ

en

t m

ass

loa

d (

g)

Me

an

in

flu

ent

turb

idity (

NT

U)

1 50 149.1 300 266.1 9000 32.2 2 100 296.0 300 287.5 9986 30.8 3 75 224.3 300 288.4 10124 30.8 4 125 365.1 300 297.0 10124 30.8 5 100 294.0 100 97.8 3375 12.4 6 25 73.8 300 73.8 10124 32.2 7 125 364.6 100 98.5 3375 10.6 9 10 28.9 300 300.0 10124 30.8

11 10 29.3 100 97.8 3375 12.4 13 25 73.2 100 96.5 3375 12.4 14 75 220.1 100 96.9 3375 12.4 15 5 15.7 300 273.5 7817 30.8 16 2 5.4 100 106.0 1042 10.6 17 50 146.8 100 97.8 3375 10.6 18 5 16.6 100 86.3 2606 10.6 19 2 5.9 300 291.2 3127 32.2

Experimental matrix of treatment runs for the baffled HS unit loaded by NJCAT gradation under 100 and 300 mg/L and various operating flow rates

47

Table 2-4. Hydrodynamic separator performance

Run # Eff

luen

t

co

ncen

tration

(mg

/L)

Eff

luen

t m

ass

loa

d (

g)

Tota

l m

ass

ca

ptu

red

( g)b

E

MC

(%)

Ma

ss b

ala

nce

err

or

(%)

Me

an

tu

rbid

ity

red

uction

(NT

U)

1 104.8 3550 4803 60.6 7.2 2.7 2 121.3 4213 4760 57.8 10.1 0c 3 134.2 4709 5475 53.5 -0.6 0c 4 146.0 4977 4809 50.8 3.3 0c 5 48.0 1655 1802 51.0 -2.4 0.3 6 108.0 3749 7161 63.0 -7.8 2.8 7 46.3 1588 1747 53.0 1.2 1.5 9 84.6 2874 7947 71.8 -6.9 5.4

11 29.8 1028 2660 69.5 -9.3 4.1 13 36.0 1242 2225 62.7 -2.7 2.0 14 44.5 1551 1885 54.1 -1.8 0.5 15 62.1 1775 6119 77.3 -1.0 17.3 16 11.1 109 917 89.5 1.5 6.6 17 37.6 1295 2318 61.6 -7.1 3.3 18 14.8 448 2186 82.8 -1.1 7.4 19 30.5 328 2817 89.5 -0.6 20.6

a Effluent event mean concentration; b Total mass captured is the sum of suspended PM in supernatant and settled PM recovered as wet slurry from the unit; c The actual value is slightly less than 0, since mean turbidity reduction is within the range of instrument resolution it is considered essentially as zero.

Table 2-5. PM gradations with mean power law model parameters

Model parameter Un

ifo

rm

me

an

Me

diu

m

me

an

He

tero

-

dis

pe

rse

me

an

Un

ifo

rm

up

pe

r

95

%

Me

diu

m

up

pe

r

95

%

He

tero

-

dis

pe

rse

up

pe

r

95

%

Mean RPD: c -0.012 1.24 1.46 0.55 1.64 1.75

Mean RPD: m -1.07 -1.51 -1.66 -1.07 -1.51 -1.66

Maximum RPD: c 0.46 1.54 1.69 0.92 2.02 2.00

Maximum RPD: m -1.04 -1.43 -1.56 -1.04 -1.43 -1.56

The mean model parameters represent the power law parameters that are the result of the least square fit of the residuals. The upper 95% model parameters represent the upper bound of the 95% confidence interval for the parameter values.

48

Table 2-6. Discrete phase model computational time for the baffled hydrodynamic separator

Flow Rate (L/s) DN1 DN2 DN4 DN8 DN16 DN32 DN64 DN128

0.4 4 11 16 30 58 114 227 452

18.1 8 20 41 77 151 302 599 1200

Times are in seconds. The results show that doubling the DN generally doubles the computational time. This result combined with the diminishing convergence rate of the RPD for higher DNs demonstrates the exponentially increasing computational time to achieve the same improvement in the models RPD error due to discritization. Table 2-7. Discrete phase model computational time for the screened hydrodynamic

separator

Flow Rate (L/s) DN1 DN2 DN4 DN8 DN16 DN32 DN64 DN128

4.5 5 6 16 31 61 121 241 481

22.9 4 8 15 30 60 118 235 469

Times are in seconds. The results show that doubling the DN generally doubles the computational time. This result combined with the diminishing convergence rate of the RPD for higher DNs demonstrates the exponentially increasing computational time to achieve the same improvement in the models RPD error due to discritization.

49

Figure 2-1. Experimental validation of full scale units. Experimental validation of the change in event mean concentration for the screened HS (A) and baffled HS (B). CFD models from experiments conducted on full-scale 1.8m (6-ft) diameter units under laboratory conditions, loaded with the hetero-disperse NJDEP regulatory gradation (gradation #8). As required, the CFD model results were linearly interpolated to provide data points at concurrent flow rates for RPD calculation. Range bars show experimental mass balance recovery (%). Influent flow is demarcated by Q in each subfigure. Reported volumes are calculated from the water level under static conditions.

50

% f

ine

r by m

ass

0

20

40

60

80

100

Uniform - #1

( = 162.94, = 0.20)

Medium - #4

( = 2.28, = 17.01)

Hetero-disperse - #7

( = 0.57, = 116.32)

% f

ine

r by m

ass

0

20

40

60

80

100Uniform - #2

( = 177.19, = 0.38)

Medium - #5

( = 2.30, = 33.82)

Hetero-disperse - #8

( = 0.56, = 232.64)

Particle Diameter (mm)

1101001000

% f

ine

r by m

ass

0

20

40

60

80

100Uniform - #3

( = 163.07,= 0.61)

Medium - #6

( = 2.30, = 50.33)

Hetero-disperse - #9

( = 0.56, = 353.16)

% f

ine

r by m

ass

0

20

40

60

80

100Indianapolis PSD

New Jersey PSD

Figure 2-2. Cumulative PSDs utilized in the study. The cumulative PSDs selected to explore the effect of gradation dispersity (σI) and d50m on modeling error due to PSD discretization. The vertical axis represents the percentage by mass of particles that are finer than the diameter on the corresponding horizontal axis. There are three sets of three gradations. Each set of gradations is centered on a common focus (33.3µm, 66.7µm, or 100µm). Each gradation set contains gradations with σI = 0.11, 1.03, and 2.64. From an experimental perspective, the hetero-disperse gradation centered on 66.7µm (#8) is the gamma curve fit of the New Jersey (NJDEP) gradation (R2 = 0.99) and the uniform gradation centered on 100µm (#3: σI = 0.11, d50m = 100) is analogous to an OK-110 gradation (σI = 0.21, d50m = 110).

51

Actual Flow Rate (gpm)

0 71 142 213 284 355

E

MC

and

Mass (

%)

0

20

40

60

80

100

CInf

= 300 mg/L

MB

E (

%)

CInf

= 100 mg/L

Percentage of Design Flow Rate (0.64cfs)(%)

0 25 50 75 100 125

-10

-5

0

5

100 25 50 75 100 125

-10

-5

0

5

10

0 71 142 213 284 355

E

MC

and

Mass (

%)

0

20

40

60

80

100

Mass

EMC

CInf

= 100 mg/L

-10/100 -10/100

Mass

EMC

MB

E (

%)

Figure 2-3. Results from the full-scale experimental testing on the baffled HS. Baffled

HS treatment performance based on EMC and Mass for influent mass loading concentrations of 300 mg/L and 100 mg/L and matched to a corresponding mass balance error (MBE).

52

0 25 50 75 100 125

EM

C (

%)

0

20

40

60

80

100

Percentage of design flow rate (%)

0 71 142 213 284 355

CInf

= 300 mg/L

CInf

= 100 mg/L

Actual flow rate (gpm)

Percentage of Design Flow Rate (%)

0 71 142 213 284 355

0

20

40

60

80

1000 25 50 75 100 125

CInf

= 300 mg/L

CInf

= 100 mg/L

M

ass (

%)

Figure 2-4. Results comparing influent loading concentrations on the baffled HS. This figure demonstrates the baffled HS treatment performance by comparing

EMC efficiency and Mass efficiency across the two influent mass loading concentrations. The results show that the unit performs similarly for both influent mass loadings and that at the tested concentrations that particle-particle interaction is negligible validating the use of the discrete phase model without accounting for particle collisions.

53

Discretization Number (DN)

2 4 6 8 10 12 14 16

I = 0.11

(k = 163.07, = 0.61)

I = 1.03

(k = 2.30,= 50.33)

I = 2.65

(k = 0.56, = 353.16)

Discretization Number (DN)

2 4 6 8 10 12 14 16

I = 0.11

(k = 177.19, = 0.38)

I = 1.03

(k = 2.30, = 33.82)

I = 2.64

(k = 0.56, = 232.64)

Discretization Number (DN)

0 2 4 6 8 10 12 14 16

Ma

xim

um

RP

D

0

10

20

30

40

50

60

I = 0.11

(k = 162.94, = 0.20)

I = 1.03

(k = 2.28, = 17.01)

I = 2.63

(k = 0.57, = 116.32)

Me

an

RP

D

0

10

20

30

40

50

I = 0.11

(k = 162.94, = 0.20)

I = 1.03

(k = 2.28, = 17.01)

I = 2.63

(k = 0.57, = 116.32)

I = 0.11

(k = 163.07, = 0.61)

I = 1.03

(k = 2.30, = 50.33)

I = 2.65

(k = 0.56, = 353.16)

I = 0.11

(k = 177.19, = 0.38)

I = 1.03

(k = 2.30, = 33.82)

I = 2.64

(k = 0.56, = 232.64)

d50m

= 33.3 m

A

d50m

= 33.3 m

B

d50m

= 66.7 m

d50m

= 66.7 m

d50m

= 100 m

d50m

= 100 m

C

ED F

Figure 2-5. Computational results for the screened HS. Mean and maximum RPDs based on DN as calculated against DN = 128. In this manner, the RPD is a characterization of error due solely to PSD discretization. Results show rapidly decreasing error for higher DNs as well as the effect of σI at low DNs for the mono-disperse (σI = 0.11), moderately dispersed (σI = 1.03), and hetero-dispersed (σI = 2.64) gradations. Results for DNs higher than 16 are not displayed due to high convergence. The d50m is only a reasonable characterization and model input for mono-disperse gradations.

54

Discretization Number (DN)

2 4 6 8 10 12 14 16

I = 0.11

(k = 163.07, = 0.61)

I = 1.03

(k = 2.30, = 50.33)

I = 2.65

(k = 0.56, = 353.16)

Discretization Number (DN)

2 4 6 8 10 12 14 16

I = 0.11

(k = 117.19, = 17.01)

I = 1.03

(k = 2.30, = 33.82)

I = 2.64

(k = 0.56, = 232.64)

Discretization Number (DN)

0 2 4 6 8 10 12 14 16

Maxim

um

RP

D

0

10

20

30

40

50

60

I = 0.11

(k = 162.94, = 0.20)

I = 1.03

(k = 2.28, = 17.01)

I = 2.63

(k = 0.57, = 116.32)

Mean R

PD

0

10

20

30

40

50

= 0.11

(k = 162.94, = 0.20)

= 1.03

(k = 2.28, = 17.01)

I = 2.63

(k = 0.57, = 116.32)

I = 0.11

(k = 163.07, = 0.61)

I = 1.03

(k = 2.30, = 50.33)

I = 2.65

(k = 0.56, = 353.16)

I = 0.11

(k = 117.19, = 17.01)

I = 1.03

(k = 2.30, = 33.82)

I = 2.64

(k = 0.56, = 232.64)

d50m

= 33.3 m

A

d50m

= 33.3 m

B

d50m

= 66.7 m

d50m

= 66.7 m

d50m

= 100 m

d50m

= 100 m

C

ED F

Figure 2-6. Computational results for the baffled HS. Mean and maximum RPDs based on DN as calculated against DN = 128. In this manner, the RPD is a characterization of error due solely to PSD discretization. Results show rapidly decreasing error for higher DNs as well as the effect of σI at low DNs for the mono-disperse (σI = 0.11), moderately dispersed (σI = 1.03), and hetero-dispersed (σI = 2.64) gradations. Results for DNs higher than 16 are not displayed due to high convergence. The d50m is only a reasonable characterization and model input for mono-disperse gradations.

55

Figure 2-7. CFD per-particle size efficiency surfaces for both the screened HS (A) and the baffled HS (B). These surfaces represent a “fingerprint” of the performance of the solids separator at steady flow rates across the spectrum of designed flows. The top plane in each surface represents particles that are completely captured and the bottom plane in each surface represents particle sizes that are negligibly captured by the solids separators.

56

Figure 2-8. CFD per-particle size efficiency differential surface for the screened HS and the baffled HS. This represents the performance differential per particle size for the two units – positive indicating better performance from the baffled HS.

57

Discretization Number (DN)

0.1 1 10 100

Discretization Number (DN)

0.1 1 10

Discretization Number (DN)

0.1 1 10

Maxim

um

RP

D

0.001

0.01

0.1

1

10

Mean

RP

D

0.001

0.01

0.1

1

10

100

Baffled HS

Screened HS

Convergence Model (CM)

Upper 95% CM

Uniformity: Uniform

A

Uniformity: Uniform

B

Uniformity: Medium

Uniformity: Medium

Uniformity: Hetero-disperse

Uniformity: Hetero-disperse

C

ED F

R2

= 0.84 R2

= 0.95 R2

= 0.98

R2

= 0.88 R2

= 0.93 R2

= 0.97

Figure 2-9. Predictive results of the power law model for RPD with increasing DN. Cumulative baffled HS and screened HS convergence data modeled by the power law. Panels (a) through (c) are model mean RPDs in increasing order of hetero-dispersivity and panels (d) through (f) are model maximum RPDs in increasing order of hetero-dispersivity. The convergence model is based on a least squares linear regression of the log – log plot of the data. An upper bound model (Upper 95% CM) constructed from the standard deviation of the regression residuals is given for more conservative estimation of RPD.

58

CHAPTER 3 OVERALL RATE KINETICS MODEL OF SODIUM HYPOCHLORITE DEMAND BY THE

DISSOLVED AND PARTICULATE MATTER FRACTIONS IN URBAN RAINFALL-RUNOFF

Urban runoff is a significant source of hetero-disperse PM, chemical and

microbial loadings to receiving waters and combined sewer overflows. Effective

treatment and reuse options are needed given urban water demands and regulations

such as total maximum daily loads (Code of Federal Regulations 2001), and numeric

nutrient criteria in Florida (USEPA 2010). Urban runoff microbial loads can impair

receiving waters (Jin et al. 2004, USEPA 1984) and in reuse applications pose a

potential public health risk if untreated (DEC 2006). Chlorination is the most commonly

utilized disinfectant worldwide (Hrudey and Hrudey 2004) and is utilized to provide a

residual in reclaimed wastewater utilized for irrigation in the built environs. However,

the efficacy of chlorine disinfection is highly dependent on the available residual over

time (Chick 1908, Fair et al. 1948).

Many kinetic studies of free and total chlorine demand in source waters and

wastewater have been undertaken. Taras (1950) model chlorine demand of inorganic

and organic substances with a power law function. Hass and Karra (1984) evaluate

chlorine demand in wastewater utilizing first order, power law and parallel first-order

models. Clark describes chlorine demand in drinking waters utilizing a second order

model (1998) and, in conjunction with Sivaganesan, describes chlorine demand with a

parallel second-order model for raw and finished waters (2002). Huang and McBean

extend Clark‟s work on finished waters employing Bayesian statistics to determine

model parameters (2007). Warton et al. utilize piecewise first order functions and

dissolved organic carbon (DOC) to normalize models to provide dose independence

59

(2006). However, there are fewer chlorine demand kinetic models for urban runoff or

wet-weather combined sewer overflows (CSOs). While the potential for DBPs exists for

waters with natural organic matter (Rook 1977), an EPA study concludes that

chlorination/dechlorination is preferred in decentralized CSO treatment. The study cited

high costs of decentralized ozone generation for intermittent wet-weather events, PM

shielding in ultraviolet inactivation schemes, and the public health hazard associated

with Cl2 gas storage required for chlorine dioxide generation (USEPA 2003).

The constituents of urban runoff differ from potable water sources and wastewater.

Kim and Sansalone (2010) compared PM from untreated runoff to wastewater treatment

plant (WWTP) influent. The study demonstrated that influent PM was relatively fine

(median diameter: d50m = 26 μm), mono-disperse (80th to 20th percentile dispersivity

index: d80m/d20m = 3.1 μm), with a specific gravity of 1.5, and a volatile fraction of

suspended PM of 76% (a surrogate indicator of organic content) and is transport is

relatively steady with regular diurnal fluctuations in flow. In comparison, the urban runoff

gradation is relatively coarse (d50m = 136 μm), hetero-disperse (d80m/d20m = 50.5 μm),

with a specific gravity of 2.3, a volatile fraction of suspended PM of 27% and is

transported spatio-temporally by episodic events. It is hypothesized that these urban

runoff differences will provide distinct chlorine demand kinetics as compared to first

order decay models commonly utilized for wastewater (Haas and Karra 1984). It is

hypothesized that the temporally varying loading rate with multiple potential limiting

factors require a second order model, similar to (Clark and Savaganesan 2002), but

with the inclusion of PM demand, and correlation of the model parameters to runoff

chemistry parameters. Furthermore, as runoff can be the dominant volumetric factor in

60

combined sewerage flows, by a factor of 100 in a Philadelphia study (USEPA 1973),

chlorine demand kinetics in runoff has direct application to CSOs. Quantifying chlorine

demand of PM fractions allows decision-making with respect to PM separation and

advanced computational modeling (Dickenson and Sansalone 2009) of unit operations

and processes. Therefore the objective of this study is to model chlorine demand

kinetics in urban runoff and compare the kinetics for dissolved and PM fractions.

Results elucidate the primary phase or PM fractions in urban runoff for chlorine

demand.

Methodology

Catchment

The catchment for this study is an urban source area located in Gainesville, FL.

The land use is a 13,000 m2 carpark field site of which a 500 m2 catchment was

delineated, instrumented and monitored for this study (Figure A-2). The surface area of

the carpark is 75% asphalt pavement and 25% raised vegetated islands with mature

trees and delineated by vertical concrete curbs. The catchment has an average daily

traffic loading of 530 vehicles (observed). Physical (PM), chemical (nutrients, metals

and organic compounds) and microbial loading sources are anthropogenic (tire,

vehicular and pavement abrasion and urban litter) and biogenic (leaf litter, grass

clippings, insects, and small urban animals and bird feces). The catchment drains to a

catch basin modified to allow all flow to be diverted for full cross-sectional flow manual

sampling. Rainfall is measured with a tipping bucket rain gage and flow measured with

a calibrated Parshall flume, an ultrasonic transducer and data logger. Runoff is sampled

at volumetrically-spaced intervals.

61

PM Fractionation

During the monitoring phase of the study, four runoff events are sampled. The

hydrologic and PM indices are presented in Table 3-1. Approximately 100 L of runoff is

sampled for each event and immediately (30 minutes) transferred to the laboratory for

analysis. Prior to each of these events additional runoff from a previous event loading

the same catchment is collected and filtered (0.45 µm membrane). This filtered runoff

served as a dissolved matrix for re-suspending each PM fraction. The dissolved matrix

is autoclaved at 121°C for one hour to render the dissolved matrix biologically inert

(Consolidated MKII, manufacturer‟s guidelines). The runoff matrix is then stored at 4°C.

In this study “runoff matrix” refers to the filtered and autoclaved dissolved fraction of

runoff. PM is fractionated by wet sieving and microfiltration from the sampled 100 L

volume into gross solids (>4250 µm, Rushton 2007), sediment PM (4250 – 75 µm, Kim

and Sansalone 2008), settleable PM (75 – 25 µm, Kim and Sansalone 2008),

suspended PM (25 – 0.45 µm, Kim and Sansalone 2008), and dissolved fractions

(<0.45 µm, Kim and Sansalone 2008). The gross solids are dried and weighed to

characterize their contribution to runoff PM, but are not included in the batch reactors for

chlorine kinetic analysis. Sediment and settleable PM fractions were immediately re-

suspended in 1 L of runoff matrix as PM concentrate and 25 µm sieve filtrate is set

aside as the suspended PM fraction.

Batch Reactor Framework

The batch reactor framework for chlorine demand kinetics utilizing the first runoff

event consists of two replicate analysis of three sodium hypochlorite doses (15, 30, and

45 mg/L ± 2 mg/L) for each runoff fraction (sediment PM, settleable PM, suspended PM,

and the dissolved fraction) for a total of 24 batch reactors. Subsequently, three

62

additional storm events are sampled and batch reactors are run with a single

hypochlorite dose across the four runoff fractions with the hypochlorite dose varying for

subsequent events (nominally 15, 30, and 45 mg/L) for a total of 8 additional batch

reactors per storm. In addition to these reactors, control batch reactors are run to

determine the effects of autoclaving the runoff matrix and environmental losses to the

atmosphere by volatilization or UV breakdown of HOCl/OCl-. Finally, an additional 22

batch reactors are run at very dilute PM fractions to enhance the resolution of the

ultimate chlorine demand by PM. For each reactor, samples are taken at 1, 5, 10, 20,

40, 120, and 480 min. with an additional sample taken at 24 h for the dilute PM fraction

reactors.

Batch reactors, stirbars, and color spectrophotometer cuvettes are prepared to be

chlorine demand free by immersion in 60 mg/L HOCl for a minimum of one hour (Eaton

et al 1998). Subsequently, the glassware is triple rinsed with chlorine demand free

water (Barnstead NanoPure). Batch reactors containing the dissolved and suspended

fractions are filled with 1700 mL of their respective fractions and reactors containing

sediment and settleable PM fractions are reconstituted by adding 150 mL of PM

concentrate to a sufficient volume of runoff matrix to achieve a final volume of 1700 mL.

Prior to initiation of each reactor run, the batch reactors are brought to room

temperature of 25 °C ± 2 °C. The batch reactors are enclosed in an aluminum foil

jacket to eliminate free chlorine decomposition due to light and sealed with a lid to

reduce HOCl volatization to the atmosphere. The continuously mixed batch reactors

are dosed with sufficient stock hypochlorite solution to meet the requirements of the

experimental construct (mixing speed 1000 rpm for first 30 s, 350 rpm thereafter). The

63

stock hypochlorite solutions are formulated to be approximately 1000 mg/L and

standardized by sodium thiosulfate titration (Eaton et al. 1998). For these experiments

the stock hypochlorite solution ranged from 962 – 1017 mg/L.

The experimental construct for the first baseline storm event consisted of replicate

analysis of 3 sodium hypochlorite doses (nominally 15, 30, and 45 mg/L) across the 4

aforementioned stormwater fractions (sediment, settleable, suspended, and dissolved)

for a total of 24 batch reactors. Subsequently, three additional storm events were

sampled and batch reactors were run with a single sodium hypochlorite dose across the

4 stormwater fractions with the hypochlorite dose varying with each storm (nominally 15,

30, and 45 mg/L) for a total of 8 additional batch reactors per storm. In addition to

these reactors, control batch reactors were run to determine the effects of autoclaving

the stormwater matrix and environmental losses to the atmosphere by volatilization or

UV breakdown of HOCl/OCl-.

Batch Reactor Setup

Batch reactors, stirbars, and color spectrophotometer cuvettes were prepared to

be chlorine demand free by immersion in 60 mg/L HOCl for a minimum of one hour

(Eaton et al. 1998). Subsequently, the glassware was triple rinsed with chlorine

demand free water (Barndstead NanoPure). Batch reactors containing the dissolved

and suspended fractions were filled with 1700 mL of their respective fractions and

reactors containing sediment and settleable fractions were reconstituted by adding 150

mL of PM concentrate to sufficient stormwater matrix to achieve a final volume of 1700

mL. Prior to initiation of the experiment, the batch reactors were brought to room

temperature of 25 °C ± 2 °C in a water bath at 45 °C. The batch reactors were enclosed

in an aluminum foil jacket to eliminate free chlorine decomposition due to light and

64

sealed with a lid to reduce HOCl volatization to the atmosphere. Continuously mixed

batch reactors were dosed with sufficient stock hypochlorite solution to meet the

requirements of the experimental construct (mixing speed 1000 rpm for first 30 s, 350

rpm thereafter). The stock hypochlorite solution was formulated to be approximately

1000 mg/L and was standardized by sodium thiosulfate titration (Eaton et al. 1998). For

these experiments the stock hypochlorite solution ranged from 962 to 1017 mg/L.

Analytical Methods

Residual free chlorine is determined from reactor samples by measuring

absorbance at 530 nm with a color spectrophotometer (Hach DR2800) utilizing N,N-

diethyl-p-phenylenediamine (DPD, Hach Chemical) as the reagent (Eaton et al. 1998).

A custom absorbance curve is developed with prepared free chlorine standards to

ensure the accuracy of the supplied DPD batch as well as to extend the range of the

analysis. At each sampling interval, two aliquots of reactor constituents are removed for

residual free chlorine analysis to perform the method in replicate. The volume of the

aliquots were selected for each individual reactor and diluted to 10 mL with chlorine

demand free water to ensure that the sample is within the analytical range of the DPD

reagent and absorbance curve. In addition, at time zero, a third aliquot of sample is

removed to check for manganese oxide interference. Manganese oxide interference is

accounted for by adding 3 drops of 30 g/L potassium iodide to the analyte in the

cuvette, waiting one minute, and then adding three drops of 5 g/L sodium arsenite.

DPD is then added to the aliquot and the absorbance is measured at 530 nm. The

absorbance from the interference test is then subtracted from the previously measured

absorbance and the result utilized to calculate the residual free chlorine.

65

The PM granulometry for each reactor is determined by low angle laser light

scattering (LALLS, Malvern Mastersizer 2000) to ensure consistent loading conditions

within the reactors and validate the fractionation process. LALLS employs the principle

of Mie scattering to iteratively estimate the particle size distribution (PSD) (ISO 2009).

PSD determination is carried out in the aqueous phase and measurements are made in

triplicate to ensure that the measurements are reproducible and stable. Figure 3-1

presents the PM fractions visually as they are applied within the framework of this study.

PM mass loading within the batch reactors is determined as suspended sediment

concentration (SSC) method (ASTM 2002). The entire contents of the reactor are

filtered through a nominal 1 µm glass fiber filter, dried at 105 °C, cooled in a desiccator,

and weighed to the nearest 0.1 mg. The volatile fraction of the PM is calculated by

volatization at 550 °C, cooling in a desiccator, and weighed. Water chemistry

measurements of temperature, pH, specific conductivity, and oxidation and reduction

potential (ORP) are made utilizing calibrated electrodes and dissolved chemical oxygen

demand (CODd) is calculated utilizing color photospectrometry and the USEPA

approved Hach Method 8000 (Federal Register 1980).

Parallel Second Order Demand Model for Dissolved Phase

Urban stormwater runoff contains a complex matrix of dissolved inorganic and

organic species (Dean et al. 2005). Clark and Savaganesan‟s (2002) parallel second

order model demonstrated applicability to the complex chemistry of both raw and

finished waters. This model is chosen to represent the kinetics of free chlorine demand

in urban runoff. The model assumes that the chlorine demand kinetics is the result of

two parallel reactions.

66

(3-1)

In these reactions, , and pn are stoichiometric coefficients, Pn are products,

and are the concentration of free chlorine participating in each reaction and

and

are the concentration of reactants in the solution that react with the free chlorine. An

analytical solution of Equation 3-1 is Equation 3-2

( ) ( )

( )

( )( )

( )( )

(3-2)

(Note that in Equation 18 in Clark and Savaganesan (2002), and

( ) as presented in Equation 3-2). In this expression, Cl(t) = chlorine

concentration at time, t; Cl0 is the initial concentration of free chlorine ( +

); X =

( ); k1 and k2 are rate constants; and Rn is defined in Equation 3-3.

(3-3)

Substituting Equation 3-3 into Equation 3-2 gives Equation 3-4.

( )

(

)

(

)

( )(

)

( )(

)

(3-4)

To aid in the development and physical significance of the model a hypothetical ultimate

chlorine demand term is introduced, CB0 which is the sum of the initial concentrations of

and

and reacts on a one to one stoichiometric basis with free chlorine (

). Note that the 1:1 stoichiometric relationship gives . Utilizing these

relationships gives Equation 3-5.

67

( ) (

)

(

)

( )( ( ) ( )

)

( ) ( )

( )(

( ) ( )

) (3-5)

In addition, CB0 is assumed to be a fractional component represented by the dissolved

chemical oxygen demand (CODd). CODd is a reasonable model parameter given the

ease of measurement and that the basis of the measurement is an oxidation reaction.

Substituting CB0 = fCODd into the model of Clark and Savaganesan returns a solution of

the form in Equation 3-6.

( )

( )

( )( ) ( ) (3-6)

In this expression the physical parameters are the initial chlorine dose as

Cl0 [mg/L], CODd [mg/L], and time, t. The four model parameters are f, the fractional

component of the CODd that represents the ultimate chlorine demand (unitless), X

(unitless), the proportion of the chlorine demand that reacts quickly at rate k1, and k1

and k2 the quick and slow, respectively, reaction rate constants with the units of

[L1mg-1min-1]. Figure 3-2 illustrates the physical significance of the parameters of the

second order demand model for the dissolved phase.

To determine the model parameters, the non-linear curve fitting using the

Levenberg-Marquardt algorithm (Marquardt 1963) to estimate the parameters of the

non-linear functions as well as the parameter standard error. In addition, the model

parameters are combined for sixteen datasets to obtain globally best-fitting parameters

based on the catchment runoff data. The remaining datasets are utilized as non-

influencing verification datasets of the model.

68

Second Order Potential Driving Model for the PM Fractions

As a source water, urban rainfall-runoff contains a heterodisperse PM phase that

from a treatment perspective contains three significant PM fractions, suspended,

settleable, and sediment based on size and mechanistic delimiters. The chlorine

demand for each PM fraction is determined by subtracting the chlorine demand of the

dissolved runoff matrix from the overall chlorine demand of the batch reactor containing

a PM fraction. A model of the dissolved phase demand is determined from two replicate

reactors from the same dissolved runoff matrix source and dosed with the same initial

hypochlorite concentration.

Studies have demonstrated the applicability of a second order potential driving

model for kinetics of adsorption in runoff for dissolved metals (Liu et al. 2005) and

phosphorus (Wu and Sansalone 2011) for simulating overall mass transfer. The second

order potential driving model has the following form (Liu et al 2005).

( )( ) (3-3)

In this expression Ct is the concentration of free chlorine at time, t; Ce is the

concentration of free chlorine at equilibrium; St is the number of active reaction sites on

the PM at time, t; Se is the number of active reaction sites on the PM at equilibrium; and

kPM is the mass transfer rate constant. Equation 3-3 models the instantaneous free

chlorine demand from PM as a function of the available PM reaction sites and the

concentration of the available free chlorine. Substituting:

(3-4)

(3-5)

69

Substituting Equations 3-4 and 3-5 into Equation 3-3 and linearizing gives Equation 3-6

(Liu et al. 2005):

( )

(3-6)

In this expression is the sorbent(PM)/solute(HOCl) ratio. From Equation 3-6, the final

mass transfer form of the model (Equation 3-10) is obtained by substituting Equations 3-

7 through 3-9 in Equation 3-6 (Wu and Sansalone 2011).

( )

(3-7)

( )

(3-8)

(3-9)

Where W/V is the mass of PM [g] over the volume of the reactor [L], which is the SSC

[g/L]; qt [mg/g] is the ratio of the mass of free chlorine transferred to the mass of PM in

the reactor at time, t [min]; and qe [mg/g] is the ratio of the mass of free chlorine

transferred to the mass of PM in the reactor at equilibrium.

(3-10)

Equation 3-10 is the linearized second order potential driving model for the overall mass

transfer of free chlorine from the solute phase into surface reactions on constituent PM.

Initial values for kPM [g1mg-1min-1] and qe are determined experimentally by plotting t/qt

versus t. Resulting parameter values favor the experimental endpoint due to the

linearization, and are further refined by minimizing the normalized root mean square

error (NRMSE).

70

Model Evaluation

NRMSE is utilized to evaluate model performance.

√∑ ( )

(3-11)

In this expression Oi is the observed value at measurement i; Ei is the modeled value at

measurement i; n is the total number of measurements; and Clo is the initial chlorine

dose.

Results and Discussion

Control Reactors

The results from the control reactors demonstrate that no detectable

environmental losses of free chlorine due to UV light or volatization occur during the

duration of the batch reactor experiments (Figure A-3). In addition, autoclaving the

runoff matrix resulted in no detectable change to the model parameters (Figure A-4).

Kinetics Model for Dissolved Phase

Figure 3-3, panel A, and Table 3-2 summarize the kinetic model parameters for

the inter-event CSBR dataset of the dissolved phase (n = 16). The fractional

component of CODd reacting on a 1:1 basis with free-chlorine as chlorine demand, f, is

0.36 ± 0.025 (95% C.L). X, which represents the portion of fCODd that reacts rapidly

are 0.39 ± 0.035. k1 is 0.07 ± 0.033 (L1mg-1min-1) and k2 is (2.93 ±0.79) X 104 (L1mg-

1min-1). Figure 3-3 also demonstrates the predictive capability for the second order

demand model for the dissolved phase using four non-influencing datasets. The data,

second order model, and a 95% confidence bands for model parameters are shown.

From the figure, the multi-modal decay rates are visibly apparent from the datasets

confirming the selection of a parallel model for the analysis (c.f. Panel D). One of the

71

strengths of the parallel second order dissolved model is the characterization of the

initial phase of chlorine demand, thus, allowing the entire curve to be modeled as

opposed to modeling decay after an “instantaneous” demand. The 95% confidence

band illustrates the sensitivity of the model to the four model parameters with the

parameter sensitivity being f > X > k2 > k1.

The model parameters are developed on an inter-event basis. Thus, the given

model parameters and ranges represent loading characteristics which are consistent for

the catchment and event independent. The rate constants, k1 and k2, demonstrate that

there are parallel reactions with separate rate constants and that these rate constants

differ by two orders of magnitude. This difference provides additional validation for the

use of a parallel model in identification of two separate reaction rates of disparate

values. The parameter, X, indicates that approximately 39% of the chlorine demand is

exerted by dissolved components that react rapidly with free chlorine. This

proportionality is consistent (±3.5%) for the catchment across multiple events and

results from the dissolved inorganic and organic loads. Further research is needed to

elucidate the value of X for similar and dissimilar loadings on differing watersheds and,

in particular, to illuminate if X is related to the ratio of DOC to COD. The fraction, f, of

the COD that exerts a chlorine demand is consistent (±2.5%) across events for the

catchment. Similar to X, this fraction may be a result of the type and proportion of the

biogenic and anthropogenic loadings for a catchment.

The NRMSE of the dissolved model on the non-influencing datasets is < 6% in

each case. The use of non-influencing datasets is an important reliability indicator of

the second order chlorine demand model. As with any model, it is necessary to predict

72

the behavior of a system given initial conditions and system reaction rates. Predicting

the chlorine residual concentration over time in urban runoff can be determined by this

model. From a system design standpoint, this is a powerful tool in the development of

disinfection unit process in a runoff treatment train.

Figure 4-4 illustrates the intra-event mass transport of CODd which is a primary

parameter of the second order, parallel chlorine demand model summarized in Equation

2. The 21 August 2010 event is a low intensity, low volume, short duration event that is

flow limited with respect to CODd. In contrast the 27 September 2010 event is a high

intensity, high volume, long duration event of low previous dry hours (PDH). The long

duration of the low intensity falling limb of intermittent runoff results in dissolution of

particulate bound COD into the dissolved phase. This increases the mass loading of

the dissolved phase at the end of the event and accounts for the inverted cumulative

distribution of CODd during the tail end of the event. The 04 November 2010 event is of

moderate intensity, moderate duration, moderate volume, and 910 PDH. The extended

dry time increases buildup of CODd on the catchment and this event transports the

highest cumulative mass of CODd. The 16 November 2010 event is low volume, low

intensity and low duration. The cumulative distribution indicates that this event is

weakly mass limited. The inter-event variation of CODd load demonstrates the influence

of hydrologic parameters on transport and illustrates the intra- and inter-event temporal

variability of chlorine demand. As a result, a treatment water quality volume (WQV)

cannot be defined for CODd a priori and corroborates a previous study for CODd on a

disparate watershed (Sansalone et al. 2005) of differing land use.

73

PM Kinetic Model

The ultimate free chlorine demand of the reactor constituents is potentially limited

by the chlorine dose if the theoretical demand exceeds the dose value. For the second

order potential driving model for free chlorine demand this ultimate free chlorine

demand is represented by qe. Figure 3-5 presents the results of plotting the modeled qe

versus the maximum qe that that is available to the PM fraction, qe_max, given the initial

hypochlorite dose, the SSC of the reactor, and the dissolved kinetic chlorine demand.

As can be seen from the figure, for under-chlorinated reactors qe = qe_max as the

maximum chlorine demand per PM mass cannot exceed the available free chlorine in

the reactor. As the chlorine available in the reactor increases, the relationship between

qe and qe_max enters a transitional region, where increasing the qe_max continues to

increase the value of qe, but the values are not equal. This transitional region

asymptotically converges on a maximum value for qe which is the maximum mass

transfer of the chlorine out of the dissolved phase to reactions with PM. Suspended PM

achieves a maximum modeled value of approximately 350 Cl2/g PM, settleable material

achieves a maximum modeled value of approximately 320 mg/g, and sediment PM

reaches an asymptotic maximum of 720 mg/g. The proximity of the maximum qe value

of the suspended and settleable PM fractions is attributed to the similar siliceous, low

organic crystalline structure of these fractions. For the sampled events, PM in these

fractions have a low volatile fraction with a range of 24-58%. However, the sediment

PM fraction is much more organic in nature, by the surrogate measurement of the

volatile fraction of the SSC, and has volatile fractions from 59-75%.

The reaction rate constant, kPM, also varies with qe_max. This parameter governs

the rate of the reaction and lowers as qe_max increases. This is a result of the model

74

governing an overall mass transfer rate. For lower values of qe_max the chlorine reacts

with the easiest to access sites on the PM. For higher values of qe_max in addition to

reacting with the most accessible sites, the free chlorine also reacts through the macro-

pore structure of the PM involving a diffusion process that slows the overall reaction rate

as measured experimentally. Thus, the second order diffusion model does not account

for this variability. The range of kPM values for this experimental construct is 3.95 X 10-6

g1mg-1min-1 (super-chlorinated reactors) to 9.2 X 10-3 g1mg-1min-1. However, the

variability is the strongest for under-chlorinated reactors, thus appropriate design of the

chlorination process given the criteria of SSC allows for a predictable kPM.

Figure 3-6 demonstrates the model fit of the second order PM kinetic model. The

results indicate the model fit the experimental data with R2 values ranging from 0.89 to

0.97 for the different fractions. This indicates the applicability of the second order

potential driving model to the overall mass transfer of hypochlorite from the dissolved

phase to the particulate phase.

Table 3-3 presents a comparison of the modeled demand of the dissolved and

particulate fractions on an event mean basis for the monitored storms. The sediment

PM fraction accounts for greater than 55% of all potential hypochlorite demand for each

event and up to 93.6% of the potential hypochlorite demand for the 27 September 2010

event. In addition, the demand from all PM fractions for the monitored events represent

over 77% of the potential chlorine demand illustrating the potential benefits of treatment

by sedimentation or filtration prior to chlorination from a chlorine demand perspective.

The dissolved phase demand, represented by fCODd, exerts a range of hypochlorite

demand from 7.6 mg/L to 88.7 mg/L. Previous authors (Clark 1998) have linked

75

chlorine demand to disinfectant by product formation, thus, providing motivation for

demand removal prior to treatment with hypochlorite. Effective dissolved phase

demand reduction can be accomplished through the use of membrane treatment.

Overall the two models developed in this study provide a methodology for

estimating the kinetic chlorine demand of urban rainfall-runoff PM fractions and

dissolved phase. Results indicate a significant chlorine demand by each PM fraction,

although the predominance of the demand is from the sediment PM fraction of higher

organic content as compared to suspended or settleable PM fractions. This study

models the chlorine demand kinetics of PM fractions and the dissolved phase of runoff

by combining rainfall-runoff event monitoring utilizing CODd, the gravimetric-based PSD

of PM fractions, and a batch reactor framework for the generation of chlorine demand

parameters. Results enable the evaluation of hypochlorite disinfection as a unit process

for the conditioning of urban rainfall-runoff for reuse. Implementation of these results

allows the balancing of primary and secondary unit operation requirements for source

area runoff reuse with hypochlorite demand by each PM fraction and the dissolved

runoff phase.

76

Table 3-1. Summary of hydrologic and PM event mean concentration indices for captured events.

Event (2010)

PDH (h)

Rainfall duration

(min)

Total runoff

volume (L)

Qmed

(L/s) Qp

(L/s)

PM fractions (mg/L)

Suspended Settleable Sediment

7-Auga 24 48 2623 1.01 4.3 13.1 (3-50)

32.2 (8-99)

222.5 (6-21414)

21-Augb 83 31 299 0.03 1.5 2.2 (0.5-4)

36.8 (6-192)

301.1 (18-3295)

27-Sep 10 388 3842 0.01 10.9 44.5 (16-190)

50.0 (1-289)

874.1 (2-6035)

4-Nov 910 43 996 0.13 3.5 93.6 (15-319)

51.5 (4-225)

486.6 (5-18145)

16-Nov 286 34 307 0.01 1.8 123.2 (30-247)

137.8 (4-340)

332.2 (24-3208)

a The event on 7 August 2010 was utilized for stormwater matrix collection. b Baseline event. PDH is previous dry hours; Qmed is the median runoff flow rate; and Qp is the peak runoff flow rate. Table 3-2. Global parallel 2nd order demand model parameters for the dissolved phase.

Parameter Units Mean SEa

k1 L1mg-1min-1 0.070 0.015

k2 L1mg-1min-1 2.93 X 10-4 3.66 X 10-5

X - 0.39 0.016

f - 0.36 0.012 The global adjusted R2 = 0.988. The results indicate that the rate constant k1 is over two orders of magnitude greater than k2. In addition, X indicates that approximately 40% of the chlorine demand is exerted by dissolved components that react quickly with free chlorine. f indicates that the estimated ultimate chlorine demand is approximately 36%

of the CODd. aStandard Error = √ ⁄ , where s is the standard deviation of the sample

mean and n = 16.

77

Table 3-3. Hypochlorite event-based ultimate demand of urban stormwater fractions for the monitored storms.

COD

(mg/L) PM

(mg/L) PM HOCl Demand

(mg/L) HOCl Demand

(%)

Event (2010) CODd fCODd

Suspe

nde

d

Sett

lea

ble

Sed

iment

Suspe

nde

d

Sett

lea

ble

Sed

iment

Dis

solv

ed

Suspe

nde

d

Sett

lea

ble

Sed

iment

7-Aug 19.5 7.0 13.1 32.2 222.5 4.6 10.3 160.2 3.9 2.5 5.7 88.0

21-Aug 80.5 29.0 2.2 36.8 301.1 0.8 11.8 216.8 11.2 0.3 4.6 83.9

27-Sep 35.7 12.9 44.5 50.0 874.1 15.6 16.0 629.4 1.9 2.3 2.4 93.4

4-Nov 166.5 59.9 93.6 51.5 486.6 32.8 16.5 350.4 13.0 7.1 3.6 76.2

16-Nov 227.5 81.9 123.2 137.8 332.2 43.1 44.1 239.2 20.1 10.6 10.8 58.6

Calculations utilize f = 0.36, suspended qe = 350 mg/g, settleable qe = 320 mg/g, and sediment qe = 720 mg/g. Results indicate that the constituents of the sediment PM fraction exert the largest portion of HOCl demand on an event mean basis.

78

0.1 1 10 100 1000 10000

0

2

4

6

8

10

Settleable25 m < < 75 m

R2 Suspended PM

R4 Settleable PM

R6 Sediment PM

% f

ine

r b

y m

ass f

or

ea

ch

PM

fra

ctio

n

Particle diameter (m)

Suspended0.45 < < 25 m

Sediment75 m < < 4750 m

Figure 3-1. PSD of quintessential fractions from the batch reactors. The figure

demonstrates validation by laser diffraction analysis of the wet sieve fractionation procedure. Note that the upper bound of the laser diffraction analysis is 2mm.

79

0 100 200 300 400 5000

5

10

15

20

25

30

35

40

45

50

Batch Reactor (R20) data

Parallel 2nd Order Demand Model

HO

Cl/O

Cl- (

mg

/L)

Reaction Time (min)

Clo={HOCl/OCl

-}

o

CB0

=f(CODo)

X = C

1

B

CB0

k2

k1

C*

Figure 3-2. Physical representation of the parameters of the parallel 2nd order demand

model. Clo is the initial concentration of free chlorine; CB0 is the ultimate free chlorine demand of the reactor (note a/b = 1) which is estimated by a fractional component, f, of the reactor initial dissolved COD; X is the ratio of the ultimate chlorine demand that reacts quickly with the second order rate constant k1; (1-X) is the ratio of the ultimate chlorine demand that reacts slowly with the second order rate constant k2; and C* represents the remaining free chlorine concentration at equilibrium. Note that as a parallel model, the effects of the slower reaction, k2, are also exerted during the initial portion, but as the time-scale of this reaction is several orders of magnitude slower than the quick initial reaction, the initial interval is well illustrated by the parameters k1 and X.

80

Figure 3-3. Predictive fit of the dissolved fraction parallel 2nd order demand model. The

model parameters are derived from the global best fit of the experimental dataset and the demand model shown utilizes these parameters and the initial conditions of the reactor. Models are shown within a 95% C.L. The experimental data are four datasets excluded from the parameter estimation analysis to independently verify the derived model for the watershed. NRMSE is reported for each panel and show that for all models NRMSE is < 6%.

0

5

10

15

20

25

30

35

40

45

500 100 200 300 400

Time (min)

Storm Date: 21 August 2010

Co = 30.4 mg/L | COD = 91.18 mg/L

Parallel 2nd Order Demand Model

k1 = 0.0701 L

1mg

-1min

-1

k2 = 2.93 X 10

-4 L

1mg

-1min

-1

X = 0.39

f = 0.36

Model Parameter 95% C.L.

HO

Cl/O

Cl- (

mg

/L)

NRMSE: 4.7%

A

0 100 200 300 4000

5

10

15

20

25

30

35

40

45

NRMSE: 5.7%

7 August 2010

Co = 29.0 mg/L | COD = 33.5 mg/L

HO

Cl/O

Cl- (

mg

/L)

Time (min)

D

0 100 200 300 400

0

5

10

15

20

25

30

35

40

45

50

NRMSE: 3.7%

Storm Date: 4 November 2010

Co = 45.6 mg/L | COD = 148.2 mg/L

Time (min)

HO

Cl/O

Cl- (

mg

/L)

C

0 100 200 300 4000

5

10

15

20

25

30

35

40

45

NRMSE: 3.7%

Storm Date: 21 August 2010

Co = 43.4 mg/L | COD = 91.18 mg/L

HO

Cl/O

Cl- (

mg

/L)

Time (min)

C

81

Figure 3-4. Transient loading of CODd on the small urban catchment in north central

Florida. With respect to CODd the 21-Aug-2010 event illustrates an event that is flow limited and the extended event on 27-Sep-2010 illustrates potential dissolution of PM COD during the low flow period at the end of the event. The November storms illustrate mass limited events. This panoply of CODd loadings is indicatory of the nature of antecedent conditions and hydrologic parameters on CODd transport, thus, a treatment WQV cannot be defined for this parameter a priori.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0

Runoff

Normalized Cumulative Volume

16 November 2010 | Qp=1.8 L/s | COD

d max= 28.3 g

CODd cum

= 69.5 g | Vrunoff

= 307 L

No

rma

lize

d C

OD

d

0.0 0.2 0.4 0.6 0.8 1.00.0

0.2

0.4

0.6

0.8

1.0

1.24 November 2010 | Q

p=3.5 L/s | COD

d max= 52.9 g

CODd cum

= 166 g | Vrunoff

= 996 L

Normalized Cumulative Volume

Q/Q

p CODd

Cum. CODd

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0

Normalized Cumulative Volume

27 September 2010 | Qp= 10.9 L/s | COD

d max= 35.1 g

CODd cum

= 137 g | Vrunoff

= 3842 L

No

rma

lize

d C

OD

d

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0

Normalized Cumulative VolumeQ

/Qp

21 August 2010 | Qp= 1.5 L/s | COD

d max= 4.9 g

CODd cum

= 22.2 g | Vrunoff

= 299 L

82

0 1000 2000 3000 4000

0

200

400

600

800

Sediment PM

75 m < < 4750 m

Settleable PM

25 m < < 75 m

Suspended PM

0.45 m < < 25 m

qe (

mg

/g)

qe_max

(mg/g)

q e =

qe_m

ax

Figure 3-5. Maximum particle free chlorine demand. For under-chlorinated reactors, the

modeled chlorine demand at equilibrium, qe, is limited by the maximum chlorine available in the reactor, qe_max. For the over-chlorinated reactors, qe for the suspended PM reaches an asymptotic maximum of 350 mg/g, qe for the settleable PM reaches an asymptotic maximum of 320 mg/g, and qe for the sediment PM reaches an asymptotic maximum of 720 mg/g. In the reactors chlorinated near the plateau point there is a transitional region where qe increases up to the asymptotic maximum.

83

Figure 3-6. The modeling results of the second order PM chlorine demand model. The

results indicate an excellent model fit for all PM fractions (R2 > 0.89).

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Mo

de

led

qt/q

e

R2 = 0.94

Composite PM

Measured qt/q

e

D

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Sediment

75 m < < 4650 m

Mo

de

led

qt/q

e

Experimental qt/q

e

R2 = 0.97

C

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Mo

de

led

qt/q

e

Measured qt/q

e

R2 = 0.89

Settleable

25 m < < 75 m

B

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0

Measured qt/q

e

R2 = 0.93

Suspended

0.45 m < < 25 m

Mo

de

led

qt/q

e

A

84

CHAPTER 4 SODIUM HYPOCHLORITE DISINFECTION OF INDICATOR ORGANISMS

ASSOCIATED WITH URBAN STORMWATER PARTICLES

Water resources and reuse thereof are increasingly central themes of sustainable

development for the developed and developing world. Future water needs will be

addressed by integrating systems that reduce water consumption, reusing water

discharges and simultaneously moving towards hydrologic restoration. In urban areas

rainfall-runoff relationships have a significant impact on the hydrologic cycle and in such

areas have been subject to significant anthropogenic modification due to urban activities

such as traffic and the high imperviousness of such watersheds. Rainfall-runoff

transports particulate matter (PM), microbial, chemical and nutrient loadings and can

impair receiving waters (Heaney and Huber 1984, House et al. 1993). With respect to

hydrology urban modification of the rainfall-runoff relationship diverts a significant

volume of water annually from pre-developed pathways (Marselek et al. 1993) with the

commensurate increases in peak flow, volume and transported load. Such scenarios in

urban areas makes runoff a critical hydrologic component for an integrated

management approach involving low impact development and reuse. Hatt et al. (2006)

reports that the current research base is inadequate to support the present

implementation of urban runoff reuse. With respect to urban water reuse there is the

need for identification of PM-associated microbial distribution and the appropriate unit

operations and processes for PM separation and corresponding disinfection in order to

minimize public health risks.

The U.S. EPA has documented the use of indicator organisms as surrogate

indices of pathogenic organisms in receiving waters through epidemiologic correlations

of illness and issued ambient water chemistry criteria (USEPA 1984). Jin et al. (2006)

85

investigated indicator organism loadings by urban rainfall-runoff vectors into Lake

Pontchartrain and subsequent recreational water closings. Charaklis et al. (2005) has

documented that indicator organisms in urban runoff exist as both planktonic and

particle associated organisms, and Krometis et al. (2007) has documented that there is

no significant variation in intra-event partitioning of indicator organisms in runoff, but that

there is a significant variation in intra-event microbial loading rates. Finally, He et al.

(2008) examined runoff for reuse and found that a retention basin could produce water

of acceptable microbial levels during dry weather periods, but that runoff events

mobilized significant microbial loadings in excess of public reuse guidelines in Alberta,

Canada.

Chlorination has been in use for over a century and is the most common form of

disinfection in practice today (Hrudey and Hrudey 2004). However, urban rainfall-runoff

is a complex matrix of dissolved and heterodisperse PM fractions (Sansalone and Kim

2008, Kim and Sansalone 2010). With respect to the impact of PM on disinfection

LeChavallier et al. (1981) documented the hindering effect of PM, using turbidity as a

surrogate, on the disinfection of environmental surface water in Oregon. Berman et al.

(1988) documented similar findings for the organic PM in wastewaters and in addition

found that chlorine permeated smaller PM faster than PM of larger diameter. Dietrich

(2003) extended these findings and modeled the intra-particle transport of free chlorine

with a radial diffusion model. Winward et al. (2008) extrapolated this research to the

use of chlorine as a disinfectant for grey water reuse applications. In addition, studies

have modeled the kinetics of inactivation for bacteriological and protozoan organisms

utilizing the Chick-Watson (Chick 1908), the Hom (Hom 1972), a modified Hom (Finch

86

et al. 1993, Haas and Joffe 1994), and rational models (Gyurek and Finch 1998).

Contributions to this existing body of knowledge are the distribution of indicator

organisms as a function of PM fractions and the efficacy of chlorination as a function of

PM fractions in urban runoff.

Given the heterodispersivity, granulometry, organic content and distribution of

nutrients for urban runoff PM fractions (Dickenson and Sansalone 2009, Berretta and

Sansalone 2011) the present study examines the association of indicator organisms

with PM fractions and the efficacy of disinfection for these PM fractions. Specifically it is

hypothesized that organisms do not distribute equally across the PM gradation.

Furthermore it is hypothesized that disinfection efficacy is not equal for each PM

fraction.

Methodology

The catchment for this study is an urban source area located in Gainesville, FL.

The land use is a 13,000 m2 carpark field site of which a 500 m2 catchment was

delineated, instrumented and monitored for this study (Figure A-2). The surface area of

the carpark is 75% asphalt pavement and 25% raised vegetated islands with mature

trees and delineated by vertical concrete curbs. The catchment has an average daily

traffic loading of 530 vehicles (observed). Physical (PM), chemical (nutrients, metals

and organic compounds) and microbial loading sources are anthropogenic (tire,

vehicular and pavement abrasion and urban litter) and biogenic sources (leaf litter,

grass clippings, insects, and small urban animals and bird feces). The catchment

drains by sheet and gutter flow to a catch basin that was modified to allow all flow to be

diverted for full cross-sectional flow manual sampling during a monitored rainfall-runoff

event. Rainfall is measured with a tipping bucket rain gage and flows are measured

87

real-time with a calibrated Parshall flume, an ultrasonic transducer and data logger.

During a rainfall event runoff is sampled at volumetrically-spaced intervals in replicate

and each set of replicates composited to construct paired event-based replicates for an

event.

PM Fractionation

During the monitoring phase of the study a series of rainfall-runoff events were

sampled. The hydrologic and PM indices are presented in Table 4-1. Approximately

100 L of runoff was sampled for each event and immediately (within 30 minutes)

transferred to the laboratory for analysis. Prior to each of these events additional runoff

from a previous event loading the same catchment was collected and filtered (0.45 µm

membrane). This filtered runoff served as a dissolved matrix for re-suspending each

PM fraction. The dissolved matrix was autoclaved at 121°C for one hour to render the

dissolved matrix biologically inert. The runoff matrix was then stored at 4°C. In this

study “runoff matrix” refers to the filtered and autoclaved dissolved fraction of runoff.

PM was fractionated by wet sieving and microfiltration from the sampled 100 L volume

into, sediment PM (4250 – 75 µm, Kim and Sansalone 2008), settleable PM (75 – 25

µm, Kim and Sansalone 2008), and suspended PM (25 – 0.45 µm, Kim and Sansalone

2008). Sediment and settleable PM fractions were immediately re-suspended in 1 L of

runoff matrix as PM concentrate and filtrate from the 25 µm sieve is set aside as

suspended PM. For each of the events analysis included microbial enumeration for

total coliform, E. coli, fecal streptococcus, and enterococcus organisms partitioned to

each PM fraction in replicate composite samples for each runoff events. Batch reaction

testing elaborated the chlorine inactivation kinetics for total coliform partitioned to each

88

PM fraction (Kim and Sansalone 2008) of the runoff events. In all cases, microbial

samples are analyzed immediately or maintained at 4°C and analyzed within six hours.

Microbiological Enumeration

Microbiological enumeration of the organisms utilizes the multiple tube

fermentation, the most probable number (MPN) method. This method is selected due to

its applicability to turbid waters (Eaton et al. 1999). For the event microbial monitoring,

total coliforms, E. coli, fecal streptococcus and enterococcus organisms are enumerated

according to Standard Methods 9221B, 9223B and 9230B (Eaton et al 1998). For total

coliform organisms and E. coli, samples are inoculated aseptically into a five row by five

dilution tube bank of lauryl triptose broth amended with 4-methylumbelliferyl-β-D-

glucuronide (LTB-MUG) and incubated at 35°C. At 24 and 48 h the cultures were

checked for lactose fermentation (gas bubbles in an inverted vial) and fluorescence

under a 366 nm UV light. Lactose fermentation in LTB incubated at 35°C represents a

presumptive positive for total coliform organisms and fluorescence indicates β-

glucuronidase enzymatic activity (Feng and Hartman 1982) – confirming the presence

of E. coli. The most dilute row with all positive lactose fermenting tubes and all positive

tubes of higher dilution were transferred aseptically to brilliant green bile broth (BGB)

and incubated at 35°C. At 24 and 48 h the cultures were checked for lactose

fermentation and positive tubes represented confirmed positives for total coliform

organisms. Reference microbiological controls were simultaneously processed for

quality assurance including E. coli (ATCC: 25922, positive lactose fermentation, positive

UV fluorescence), E. aerogenes (ATCC: 13048, positive lactose fermentation, negative

UV fluorescence), E. faecalis (ATCC: 29212, negative lactose fermentation, negative

UV fluorescence), and a non-inoculated blank.

89

Fecal streptococcus and enterococcus organisms were enumerated using

samples inoculated aseptically into a five row by four dilution tube bank of azide

dextrose broth (ADB) and incubated at incubated at 35°C. At 24 and 48 hrs the cultures

were checked for turbidity with positive samples aseptically transferred to

Enterococcosel™ agar (BBL) for confirmation. Esculin hydrolysis on Enterococcosel™

agar results in a black halo around the colonies and is characteristic of fecal

streptococci (Isenberg et al 1970). Positive fecal streptococci plates are aseptically

transferred to 6.5% NaCl brain heart infusion broth (BHI) and incubated at 45°C, the

Sherman criteria for enterococcus organisms. At 24 and 48hrs the BHI broth was

checked for turbidity, with positive tubes indicating the presence of enterococcus

organisms. Reference microbiological controls were simultaneously processed for

quality assurance including E. coli (ATCC: 25922, negative ADB turbidity, negative

esculin hydrolysis, negative BHI turbidity), E. faecalis (ATCC: 29212, positive ADB

turbidity, positive esculin hydrolysis, positive BHI turbidity), and a non-inoculated blank.

For both bacteriological enumeration schemes, most probable numbers (MPNs) are

calculated according to standard method 9221C (Eaton et al 1998).

Batch Reactors

Table 4-1 outlines the batch reactor experimental matrix. For the runoff events, a

reference storm on 21 August 2010 was sampled and batch reactors are initialized in 2

replicate reactors (Figure A-1) for the three PM fractions at sodium hypochlorite doses

of 15, 30 and 45 mg/L for a total of eighteen batch reactors. Three additional storms

are sampled on 27-Sep-2010, 4-Nov-2010, and 16-Nov-2010 and are initialized in 2

replicate reactors across the three PM fractions at a single sodium hypochlorite dose of

30, 45, and 15 mg/L, respectively, for an additional eighteen batch reactors. For the

90

experimental analysis, additional rainfall-runoff is collected on 7-Aug-2010 and 5-Sept-

2010, micro-filtered through 0.45 µm nylon filters, and autoclaved at 121°C for 60

minutes. This filtered and sterilized rainfall-runoff is utilized to reconstitute the

separated PM into the batch reactors and is referred to as stormwater matrix.

PM is immediately fractionated from the rainfall-runoff sample by a sterile wet-

sieve procedure. Stormwater sample is poured through sterilized 4750, 75, and 25 µm

sieves. PM remaining on the 4650 µm sieve is considered gross solids (Rushton et al.

2007) and is dried and characterized, but not included in batch reactor experiments.

PM remaining on the 75 and 25 µm sieves are reconstituted in 1 L stormwater matrix as

sediment and settleable fractions, respectively, and stored at 4°C until reactor

initialization. PM passing through the 25 µm into a sterile container is considered

suspended material and is stored at 4°C until reactor initialization.

The batch reactors are 2 L nominal glass jars with a 4 cm stir-rod that were

prepared to be chlorine demand free (Eaton et al. 1998) and autoclaved at 121°C for 20

minutes. Sediment and settleable PM reactors are filled with 150 ml of respective PM

concentrate and 1650 ml of dissolved stormwater matrix and suspended PM reactors

are filled with 1800 ml of 25 µm sieve filtrate. At time zero, batch reactors are brought to

25°C ± 2°C, initial water quality measurements of pH, temperature, and conductivity are

recorded, a 100 ml aliquot of sample is removed for bacteriological analysis and the

reactor is dosed with a standardized sodium hypochlorite solution according to the

dosing schedule. At each time interval listed in Table 4-1, water quality measurements

are made, a 100 ml aliquot is removed for bacteriological analysis, and 2 replicate

aliquots are removed to analyze the free chorine residual. Sterile syringes and pipette

91

tips are used at each time interval and water quality electrodes are chlorine sterilized

before immersion in the reactor. In the interim between samples, an LDPE lined lid is

tightly secured on the reactors. Aliquots utilized for bacteriological analysis are

processed in a sterile blender at 22,000 rpm (Bar-maid, manufacturers reported

specifications) to dissociate particle attached organisms from PM (Borst and

Selvakumar 2003) prior to total coliform enumeration. A laboratory blank of sterile DI

water was also processed through the blender and microbiologically enumerated to

ensure no cross contamination from the blending procedure. In addition to total coliform

enumeration for the events on 27-Sep-2010, 4-Nov-2010, and 16-Nov-2010, the time

zero measurements for each batch reactor were additionally enumerated for E. coli,

Fecal Streptococcus, and Enterococcus to determine the partitioning of each organism

in the suspended, settleable, and sediment fractions. Following the experiment, batch

reactors are analyzed for total (SSC) and volatile (VSSC) suspended sediment

concentration (ASTM 2002) and particle size distribution (ISO 2009) to characterize the

PM loading and validate the wet sieve fractionation procedure. The granulometry of

each reactor is reported in Table 4-2.

Residual Chlorine

Residual free chlorine is analyzed utilizing a DR2800 (Hach Chemical) color

spectrophotometer measuring the absorbance of N,N-diethyl-p-phenylenediamine

(DPD, Hach Chemical) at 530 nm. Replicate 2 ml (5:1 dilution), 5 ml (2:1), or 10 ml

(1:1) aliquots, depending on anticipated residual chlorine, are removed from the batch

reactor and diluted to 10 ml with chlorine demand free water (Barndstead Nanopure). A

custom absorbance curve is obtained for the DPD reagent to increase the range of the

analysis and the measurement has an accuracy of ±0.1 mg/L for a 1:1 dilution ratio.

92

Results and Discussion

Microbiological controls and laboratory blanks ensured sterility of microbiological

growth media, sterility of the process, and appropriate media response to reference

organisms. The laboratory blender blank demonstrated no cross contamination

between samples for each analyzed event. In addition, the media‟s response to

reference organisms are utilized as comparators for UV fluorescence and turbidity

where appropriate.

Figure 4-1 presents the results from the monitoring study of event mean bacterial

densities of indicator organisms for twenty-five wet weather events. The results indicate

that, as expected, total coliform organisms are ubiquitous in the watershed and have a

median density in excess of 106 organisms per 100 ml. E. coli densities range from 30

MPN/100ml to greater than 104 MPN/100ml with a median density of 1300 MPN/100ml.

From these storms, E. coli represented <1% to 21% of total coliform organisms present.

The fecal streptococci densities ranged from 3000 MPN/100ml to 105 MPN/100ml with a

median of 5 X 104 MPN/100ml and the enterococcus densities ranged from 103 to 105

MPN/100ml with a median value of 104 MPN/100ml.

Epidemiological studies of disease due to contact exposure of the types

experienced in reuse applications with organisms transported by rainfall-runoff acting as

the etiological agents do not currently exist. The state of Florida has water quality

criteria regulations regarding the restricted and unrestricted reuse of reclaimed water.

However, the numeric bacteriological criteria in the regulations are specifically for the

reuse of water from domestic wastewater treatment facilities and do not apply to the

reuse of urban rainfall-runoff. In the event of the design and implementation of urban

rainfall-runoff reuse in an urban setting in Florida, the designer is required to

93

demonstrate that the reuse application will not impair water quality criteria at or near the

site of application and are required to demonstrate that the reuse application of urban

rainfall-runoff does not impair the public health or welfare (Personal communication,

Eric Livingston, Florida Dept. of Environmental Protection). This is accomplished

through the implementation of best management practices. However, it is useful in the

absence of specific, numeric regulatory guidance regarding the bacteriological water

quality of urban rainfall runoff as a source water to consult the regulations regarding

water quality criteria for recreational waters and wastewater reuse for restricted and

unrestricted urban reuse (He et al. 2008). The U.S. Environmental Protection Agency

(EPA) and many state agencies, including the Florida Department of Environmental

Protection, issue water quality criteria for recreational waters, waters utilized as feed

waters to drinking water treatment plants, and reuse/reclaimed waters. The EPA

bacteriological water quality criteria for freshwater recreational water use are, as

geometric means, 126 MPN/100ml for E. coli and 33 MPN/100ml for enterococci and for

brackish/saltwater recreational water use is 35 MPN/100ml for enterococci (USEPA

1984) – note that the EPA does not recommend the use of E. coli as an indicator

organism in brackish/saltwaters. As can be readily observed from Figure 4-1, urban

rainfall-runoff exceeds E. coli water quality criteria in 19/25 events and enterococcus

water quality criteria in 25/25 events when sampled at the head of the rainfall-runoff

conveyance system. Thus, for recreational waters, stormwater runoff functionally

impairs the bacteriological quality and the dilution ratio and organism die off governs

whether the recreational water as a whole will exceed the standard water quality criteria.

In Florida, the restricted and unrestricted reuse of wastewater requires high level

94

disinfection with an effluent bacteriological density of fecal coliforms below the detection

limit of 2 MPN/100 ml as a geometric mean and below 25 MPN/100 ml for any single

sample. As can be seen from Figure 4-1, urban rainfall-runoff exceeds the regulatory

requirement for water quality criteria for both E. coli (as a member of the fecal coliform

family) and enterococcus indicator organisms in regards to recreational waters and

urban reuse applications.

In addition, previous authors have suggested a link between the ratio of fecal

coliforms (FC) to fecal streptococcus (FS). High ratios of FC/FS, those greater than

one, indicated potential anthropogenic sources of fecal contamination, whereas low

ratios of FC/FS potentially indicate contamination from animal and bird sources

(Geldreich and Kenner 1969). This rubric, however, should only be applied to recent

bacterial loadings as the die-off rates for the indicator organisms vary (Maier et al.

2009). As fecal coliforms were not explicitly measured in this study a direct comparison

to the work of these authors cannot be made. However, as E. coli is the primary

bacteria from human origins that constitute fecal coliforms (Kott 1977), the ratio of E.

coli to FS is useful to comment on. For every observed event, the ratio of E. coli to FS

is < 1, corroborating the expected loadings of contamination from animal, bird or insect

sources. This result is substantiated by the potential microbial loading sources of the

watershed and corroborates Clausen et al. (1977) on the FC to FS ration for urban

stormwater. As a primary source, sampled from the entrance to municipal separate

storm sewer system, there are no identified anthropogenic fecal influences. There are

no nearby septic systems and no potential cross connections due to the location of

sampling. Thus, as observed, any fecal contamination is from small animals, birds and

95

insects. However, as such, this fecal contamination is pertinent to human health as

there is the potential for cross species communication of disease such as giardiasis

(Majewska 1994) and cryptosporidiosis (Beach 2008) and as wet-weather events have

been shown to mobilize waterborne pathogens present in the environmental watershed

(Hrudey and Hrudey 2004).

Batch Reactor Results

The results from the batch reactor experiments illuminate the level of shielding

actuated by PM on particle associated organisms. Figure 4-2 presents the results from

the batch reactor inactivation experiments which demonstrate that the finer particles of

the suspended and settleable fractions are readily permeated by the hypochlorite and a

high level of inactivation is demonstrated in each case. Corroborating this finding,

Figure 4-3 plots the log removal of the 04-Nov-2010 event in the suspended, settleable

and sediment PM fractions and shows the rate of the inactivation (Panels A and B). In

all cases, the initial rate of inactivation rapidly reaches a plateau of maximum

inactivation. For the suspended and settlable fractions, this plateau occurs near the

maximum removal rate observable by the experiment. This denotes that the

hypochlorite readily penetrates the PM in those fractions at 45 mg/L and that shielding

of associated organisms is not occurring. In addition, a percentage of maximum log

removal is displayed and is defined as the observed log removal over the maximum

potential log removal for the reactor. For the sediment fraction (Panel C), the reactors

rapidly achieve a maximum log inactivation of 20-60% reactor inactivation potential

which corresponds to final bacterial densities of approximately 103 MPN/100 ml for

these reactors (Figure 4-2). Thus, there is shielding of particle associated organisms in

the sediment PM fraction.

96

Figure 4-4 extends these results to demonstrating the log removal of the

sediment fractions for various initial concentrations of hypochlorite. As can be readily

observed from the figure, increasing the hypochlorite dose in the reactor increases the

final level of inactivation achieved. This result corroborates the work of other authors

who find similar results in highly organic grey-waters (Winward et al. 2008), but is

tempered by the fact that shielding is observed at all hypochlorite doses.

The shielding in the sediment PM fraction may be attributed to several governing

factors. The primary driver of shielding is the large particle size which represents a

physical barrier to disinfectant penetration. In addition, the volatile fraction of PM is a

surrogate indicator of the organic content. During the monitored events the volatile

fraction (VF) of the SSC for the sediment PM was in the range of 59-79%, whereas the

VF of the suspended and settleable PM was in the range of 24-58%. Thus, the

carbonaceous material may represent a localized chlorine sink conveying additional

protection to particle associated organisms. Finally, the constituents of the suspended

and settleable PM fractions contain more siliceous, crystalline material than the

sediment PM fraction. The internal pore-structure of the crystalline material may be less

hospitable to particle associated organisms, resulting in a surface orientation for

associated organisms. This potentially explains the similarity of the rate of disinfection

for the suspended and settleable fractions with respect to particle diameter.

PM shielding of associated organisms evokes a deleterious effect on inactivation

by chlorination of sediment PM. However, chlorination at all levels was able to

significantly inactivate PM associated organisms in the suspended and settleable

fractions. PM fractions are separated into suspended, settleable, and sediment

97

categories from a functional, treatment perspective. Sediment material readily

separates by gravitational quiescent settling. Settleable material is the material that will

settle in 60 min in an Imhoff cone and suspended material is the PM remaining in the

supernatant and 60 minutes. Thus, these categorical definitions indicate the

applicability to treatment of urban rainfall-runoff by gravitational settling. The correlation

of these categorical labels to particle size ranges is imperative to the quantification and

descriptive analysis that is required in many studies. This correlation also enables

extrapolations of these PM ranges to treatment process of differing nature, such a

filtration. Many implementations of urban rainfall-reuse will require the detention of

stormwater for storage before reapplication as reuse water. This is advantageous in

that the detention of the rainfall-runoff simultaneously provides clarification of PM –

especially of the sediment PM exuding the strongest shielding capability. In addition,

the use of a diatomaceous earth or sand filter pre-chlorination can reduce the effects of

particle associated shielding as well as the required chlorine dose.

Indicator Organism Partitioning

Figure 4-5 presents the gravimetric density of the indicator organism partitioning to

rainfall runoff. The bacterial density for each enumerated organism is the highest in the

suspended fraction. In general, this is followed by the settleable and the sediment

fractions in relative orders of magnitude of bacterial density. Of particular note is the

low E. coli density of less than 25 MPN/mg PM for the sediment fractions as compared

to the densities of the suspended and settleable fractions. However, this result is

contrasted with the fact that the sediment fraction of urban stormwater runoff has the

largest PM mass of the three categories on an event mean basis. For the monitored

storms, the suspended fraction had a PM geometric mean of 41.8 mg/L, the settleable

98

had an SSC geometric mean of 47.2 mg/L and the sediment fraction had a SSC

geometric of 572.5 mg/L. Table 4-3 presents the percentage of mobilized organisms

associated with each PM fraction as weighted by PM. E. coli is most mobilized in the

suspended and settleable fractions for the monitored events, whereas enterococcus

organisms are most mobilized in the suspended and sediment fractions.

In conclusion, wet weather events result in the bacteriological mobilization of

indicator organisms on a small, impervious urban catchment in north central Florida.

Batch reactor experiments on particle associated coliforms demonstrated that

organisms in the suspended and settleable fractions are readily inactivated at the

applied hypochlorite doses, but that particle associated coliforms in the sediment

fraction are shielded by host PM. In sediment PM, microbial inactivation increased with

increasing hypochlorite dose. However, disinfectant shielding is observed at all

hypochlorite doses and pretreatment and removal of sediment PM is recommended for

any practical design implementation of hypochlorite inactivation for urban reuse

purposes.

99

Table 4-1. Batch reactor experimental matrix of PM fractions, HOCl dose, and event date.

Two replicate reactors are utilized for each experiment to ensure reproducibility. x represents batch reactor initial concentration for the event.

HOCl Dose Sediment Settleable Suspended

(mg/L): 15 30 45 15 30 45 15 30 45

Time

(min)

Contact time (mg-min/L)

0 0 0 0 0 0 0 0 0 0

1 15 30 45 15 30 45 15 30 45

5 75 150 225 75 150 225 75 150 225

10 150 300 450 150 300 450 150 300 450

20 300 600 900 300 600 900 300 600 900

40 600 1200 1800 600 1200 1800 600 1200 1800

120 1800 3600 5400 1800 3600 5400 1800 3600 5400

480 7200 14400 21600 7200 14400 21600 7200 14400 21600

Event Date

21-Aug-2010

x x x x x x x x x

27-Sep-2010

x

x

x 4-Nov-2010

x

x

x

16-Nov-2010 x x x

100

Table 4-2. Batch reactor particle granulometry.

Event

HOCl PM d10a d50

b d90c d[4,3]

d

(2010) Fraction (mg/L) (mg/L) (µm) (µm) (µm) (µm) 2

1 A

ug

ust

20

10

Suspended 15 50.6 2.1 10.1 29.8 13.7

Suspended 15 24.5 1.6 7.7 29.0 16.0

Suspended 30 25.9 1.7 6.9 39.0 15.4

Suspended 30 105.5 4.1 15.1 42.4 23.5

Suspended 45 47.9 2.2 9.3 24.4 19.8

Suspended 45 22.1 1.9 8.2 27.2 14.2

Settleable 15 188.6 14.9 44.3 90.7 49.3

Settleable 15 176.3 14.5 39.5 91.0 49.1

Settleable 30 139.2 12.9 42.8 69.9 48.2

Settleable 30 206.8 14.5 44.2 93.0 49.9

Settleable 45 147.9 13.4 43.1 89.7 48.2

Settleable 45 197.8 14.0 43.6 90.7 51.7

Sediment 15 284.8 53.8 244.9 958.3 390.4

Sediment 15 787.6 62.4 323.5 427.0 480.2

Sediment 30 - - - - -

Sediment 30 419.8 72.6 268.0 1059.9 438.1

Sediment 45 211.4 39.3 193.9 948.5 366.1

Sediment 45 391.6 66.3 310.8 1120.4 467.1

27

Se

pt 2

01

0 Suspended 30 36.8 2.2 9.5 25.1 12.0

Suspended 30 30.2 2.3 9.9 28.0 15.7

Settleable 30 187.0 13.0 39.2 83.2 44.4

Settleable 30 149.4 11.8 37.7 82.1 45.8

Sediment 30 330.9 41.8 389.1 1165.7 501.9

Sediment 30 285.1 50.8 393.7 1288.6 561.8

4 N

ov 2

01

0 Suspended 45 32.9 2.1 9.4 29.6 14.0

Suspended 45 39.6 2.2 10.0 28.5 14.9

Settleable 45 83.1 8.4 35.0 61.9 41.6

Settleable 45 75.7 8.1 35.8 100.5 53.2

Sediment 45 330.4 54.0 203.5 887.7 347.0

Sediment 45 353.8 38.4 189.3 842.5 334.9

16

No

v 2

010 Suspended 15 89.3 1.8 8.7 27.3 15.8

Suspended 15 98.0 1.9 9.8 30.0 14.5

Settleable 15 67.3 5.9 27.8 50.3 33.2

Settleable 15 67.5 7.7 40.6 245.0 88.7

Sediment 15 272.0 73.3 207.9 670.7 302.6

Sediment 15 232.2 75.9 218.4 769.9 344.0 a Characteristic particle size of 10% finer by volume. b Characteristic particle size of 50% finer by volume. c Characteristic particle size of 90% finer by volume. d De

Brouckere volumetric mean: ∑ , which is analogous to the number mean volume size.

101

Table 4-3. Event mobilization of indicator organisms and percentage of transported organisms associated with each PM fraction.

27 Sept 2010

4 Nov 2010

16 Nov 2010

27 Sept 2010

4 Nov 2010

16 Nov 2010

[%] [%] [%]

Su

spe

nd

ed PM g 170.8 92.3 36.2 4.6 15.0 20.8

T. Coliform MPN•104 549 1303 179 55.4 69.4 77.4

E. Coli MPN•104 32.3 1.7 11.9 95.1 32.6 94.6

F. Strep MPN•104 89.1 417 48.3 51.1 62.1 83.4

Enterococcus MPN•104 20.4 107 10.9 71.1 47.5 75.4

Se

ttle

ab

le PM g 192.0 38.4 40.4 5.2 6.3 23.3

T. Coliform MPN•104 19.1 179 28.6 1.9 9.5 12.4

E. Coli MPN•104 0.5 2.6 0.7 1.4 50.5 5.2

F. Strep MPN•104 4.8 70.0 4.2 2.8 10.4 7.2

Enterococcus MPN•104 1.9 18.8 2.5 6.5 8.4 17.1

Se

dim

ent PM g 3357 483.5 97.5 90.2 78.7 56.0

T. Coliform MPN•104 424 395 23.9 42.7 21.1 10.3

E. Coli MPN•104 1.2 0.9 <0.1 3.5 16.9 0.2

F. Strep MPN•104 81.0 185 5.4 46.2 27.5 9.3

Enterococcus MPN•104 6.4 99.0 1.1 22.4 44.1 7.5

Percentages are weighted by PM loading. Results indicate that organisms are highly mobilized in the suspended fraction relative to mobilization in the settleable and sediment fractions, with the exception of E. coli during the 04 November 2010 event.

102

T. C

oli

form

E. C

oli

F. S

trep

toco

ccus

Ente

roco

ccus10

0

101

102

103

104

105

106

L3

L2

L1

MP

N/1

00

mL

A

B C

Mean

10%1%

25%

50%

90%

Legend

75%

99%

Figure 4-1. Event mean most probable number per 100 mL box-plot for twenty-five wet

weather events on a small urban watershed in north central Florida. Comparative EPA regulatory guidance for freshwater recreational ambient bacteriological density is shown as (A), 126 MPN/100 ml (geometric mean) for E. coli. Comparative regulatory guidance in Florida for unrestricted urban reuse for wastewater effluent is shown as (B), 25 MPN/100 ml (single sample) for fecal coliforms. Comparative regulatory guidance for brackish/saltwater recreational use is shown as (C), 35 MPN/100 ml for Enterococcus organisms. Comparative Australian regulatory guidelines for urban runoff reuse are: (L1), <1 MPN/100 ml for non-potable residential reuse; (L2), <10 MPN/100 ml for reuse in area with un-restricted access; and (L3), <1000 MPN/100 ml for reuse in areas with restricted access.

103

Figure 4-2. Hypochlorite inactivation kinetics of particle associated coliform organisms

on suspended, settleable, and sediment PM. Results indicate shielding of bacteria on sediment PM throughout the duration of the experiment and rapid inactivation of particle associated coliform organisms on the settleable and suspended fractions. Sediment fractions maintained bacterial densities on the order of 103 MPN/100 ml at the end of the 8 h HOCl reactor experiment.

4 November 2010C

o = 45 mg/L

0 100 200 300 400 500

Reactor Time (min)

Event:16-Nov-2010C

o = 15 mg/L

0 100 200 300 400 500

Reactor Time (min)

27 September 2010C

o = 30 mg/L

0 100 200 300 40010

0

101

102

103

104

105

106

MP

N/1

00

mL

Reactor Time (min)

21 August 2010C

o = 45 mg/L

100

101

102

103

104

105

106

MP

N/1

00

mL

21 August 2010C

o = 30 mg/L

100

101

102

103

104

105

106

0 100 200 300 400

Reactor Time (min)

MP

N/1

00

mL

Suspended

Settleable

Sediment

Detection

Limit

21 August 2010C

o = 15 mg/L

104

Figure 4-3. Log removal of particle associated coliforms for the 04-Nov-2010 (Panel A, B) event with an initial hypochlorite dose of 45 mg/L. Results indicate that particle associated coliforms on suspended and settleable PM and rapidly achieve maximum log removal. Particle associated coliforms on sediment material (Panel C) only achieve 20-60% of the maximum potential log removal for the reactor, demonstrating particle shielding of associated coliforms.

0 100 200 300 400 500

Time (min)

15 mg/L (16-Nov-10)

30 mg/L (27-Sep-10)

45 mg/L (4-Nov-10)

15 mg/L (21-Aug-10)

30 mg/L (21-Aug-10)

45 mg/L (21-Aug-10)

Sediment

0 100 200 300 400

0

20

40

60

80

100

% M

ax L

og

Rem

ov

al

Time (min)

Suspended

Settleable

Sediment CB

A-6

-5

-4

-3

-2

-1

0

10 5000 10000 15000 20000

Ct (mgmin/L)

{HOCl/OCl-}

o = 45 mg/L

Suspended

Setteable

Sediment

Rem

ov

al (

log

-un

its)

A

105

0 5000 10000 15000 20000

-3

-2

-1

0 15 mg/L

30 mg/L

45 mg/L

Re

mo

va

l (l

og

-un

its)

Ct (mgmin/L)

{HOCl/OCl-}

o

Figure 4-4. Log removal of particle associated coliforms on sediment PM across the

inoculation doses of 15, 30, and 45 mg/L. Results indicate that increasing chlorine doses penetrate sediment PM with increasing efficaciousness. Oscillations in removal are attributed to the effects of sampling a heterodisperse particle size distribution within the sediment fraction.

106

Susp

ended

Set

tlea

ble

Sed

imen

t

Susp

ended

Set

tlea

ble

Sed

imen

t

Susp

ended

Set

tlea

ble

Sed

imen

t

Susp

ended

Set

tlea

ble

Sed

imen

t100

101

102

103

104

105

106 EnterococcusF. Strep.E. coli

MP

N/m

g o

f P

M

T. Coliforms

Figure 4-5. Partitioning of particle associated organisms to suspended, settleable, and

sediment PM fractions. For each organism, the suspended PM fraction contains the highest bacterial density followed by the settleable and sediment fractions. In particular, the sediment PM fraction, which exhibits the greatest organism shielding potential contains the lowest density of the indicator E. coli ( < 25 MPN/mg PM) and enterococcus ( < 300 MPN/mg PM) organisms.

107

CHAPTER 5 ADVANCED COMPUTATIONAL MODELING OF FREE CHLORINE DEMAND AND DISINFECTION IN UNIT OPERATIONS AND PRECESSES LOADED BY URBAN

STORMWATER

Urban rainfall-runoff is a water which has come under increasing scrutiny for an

integrated management approach (Heaney and Sample 2000). Development in the

United States and elsewhere has resulted in increased volumetric transport of water

with constituent microbiological, particulate, and nutrient loadings to receiving waters

(House et al 1993). In order to reverse this trend, technologies, research, and

integrated management systems need to continue to be developed to reduce, treat, and

reuse urban rainfall-runoff. Urban rainfall-runoff exhibits temporally varying water

volume and quality and transports constituent particulate, microbial, heavy metal, and

nutrient environmental loadings in both dissolved and particulate fractions (Sansalone

and Kim 2008, Sansalone and Buchberger 1997, Christina and Sansalone 2003, Kim

and Sansalone 2010). The implementation of a source for reuse effectuates the need

for the consideration of public health and safety.

Chlorination as a form of microbial inactivation is the oldest chemical oxidizing

reaction utilized for the public health of drinking waters and waste waters and is

currently the most widely used inactivation process world-wide (Hrudey and Hrudey

2004). From the earliest use of this process, researchers documented the importance

of the concentration of disinfectant over time, the contact time (CT), to the level of

microbial inactivation (Chick 1908, Fair et al. 1948). Continued studies developed

generalized batch inactivation models as the Hom model (Hom 1972) and the

incomplete gamma Hom model (Haas and Joffe 1994) as organisms were shown to

exhibit disparate inactivation kinetics as compared to the Chick-Watson model and for

108

utilization on water with disinfectant demand. Extension of batch reactor data to full

scale flow through reactors has been demonstrated (Haas et al. 1998) as well as the

utilization of reactor residence time distributions (RTD) to determine reactor CT values

(Bellamy et al. 1998). The most advanced modeling technique of inactivation kinetics in

potable water applied to date is computational fluid dynamics (CFD) (Greene et al.

2004, Baawain et al. 2006, Goula et al. 2008, Greene et al. 2004) whereby it is utilized

to improve reactor design and model first order chlorine demand. CFD is the numerical

solution of the fundamental equations of fluid motion involving the simulation of transient

flow fields, chemical reactions, and particle fate and transport in spatially complex

geometry.

In urban stormwater, CFD has enhanced the modeling of PM separation for

transient flows (Sansalone and Pathapati 2009) and heterodisperse particle size

distributions (Dickenson and Sansalone 2009) as well as re-entrainment of PM by

scouring mechanisms (Pathapati and Sansalone 2011). In addition, in Chapter 3, urban

stormwater batch reactor experimentation indicated parallel second order chlorine

demand kinetics for the dissolved fraction and a second order potential driving force

model for particulate fractions. As a result, there is the requisite need to formulate

discretized finite-rate chlorine kinetic dissolved and PM equations for modeling transient

and complex flows encountered in urban stormwater runoff.

Objectives

The objective of this study is to discretize the analytical parallel second order

dissolved and potential driving PM finite-rate free chlorine demand models for utilization

in CFD. The computational dissolved, PM, and composite CPD model of sodium

hypochlorite kinetic demand in urban rainfall-runoff is validated by batch reactor data.

109

Methodology

CFD is the numerical solution of the fundamental partial differential equations

that govern fluid flow and particle and chemical transport. CFD can be implemented in

both Lagrangian, fluid particle tracking schemes, or Eulerian, fluid flux through control

volume schemes. CFD is capable of modeling both laminar and turbulent flows, where

the turbulent flow characteristics are numerically simulated through direct numerical

simulation (DNS) or modeled by solving the bulk equations of motion coupled with a

turbulent flow closure model, an example of which is the Reynolds averaged Navier-

Stokes (RANS) equations with a variant of the k-ε two equation turbulent model.

Contradistinguishing between DNS and RANS are computational time and turbulent

scale resolution. DNS is computationally expensive, particularly at high Reynolds

numbers and for many industrial flows (White 2006). RANS simulations are

computationally less expensive than DNS, but the simulations generalize detailed

turbulent structure information. A third available option is a large eddy simulation (LES).

LES employs a scaling filter that delineates turbulent eddy length scales larger than the

scaling filter for direct numerical resolution and models turbulent structures smaller than

the filter length scale. This technique has been shown to be superior to RANS k-ε

models in disinfection flow through reactors (Wols et al. 2010). The LES filtered

continuity (Equation 5-1) and momentum (Equation 5-2) equations are:

(5-1)

*( )

(

)+ (5-2)

110

where ρ is fluid density; xi is the ith direction vector; is the filtered velocity in the

ith direction; is the filtered pressure; is the kinematic fluid viscosity; and is the

turbulent viscosity. In the present implementation, the Smagorinsky model is utilized to

model the turbulent viscosity of the filtered turbulent eddies:

( ) | | (5-3)

where Cs = 0.1 is the Smagorinsky constant; Ls is the sub-grid characteristic filter

length scale of the finite volume mesh; and | | is the local strain rate tensor. The reader

is encouraged to refer to Lesieur and Metais (1996) for a more comprehensive

development of the LES-Smagorinsky turbulent model.

To model the transport of PM, a mixed mode Eulerian-Lagrangian reference frame

is utilized where the fluid velocity and pressure flow fields are modeled in an Eulerian

reference frame and the PM is modeled as discrete particles in a Lagrangian reference

frame. PM transport modeled in the discrete phase is integrated across the fluid

velocity and pressure flow fields modeled in the Eulerian reference frame. This scheme

does not account for particle influence on the velocity and pressure flow fields and, thus,

is restricted to dilute fluid flows of < 10% volume fraction (VF) (Brennen 1996). Even

with this restriction, many flow situations, including the batch reactors in this present

study, are successfully modeled as dilute flows (the concentrations of interest for the

present study are less than 1% as VF).

In the discrete phase, PM transport and fate is simulated by integrating Newton‟s

second law (Equation 5-4) for representative particles across the numerical fluid

domain.

111

( ) ( )

(5-4)

(5-5)

| |

(5-6)

(5-7)

The formulation of equation (5-4) is particle acceleration

equal to the

summation of the forces per unit particle mass. The quantity ( ) is drag force

per unit particle mass; and the quantity ( )

is buoyancy/gravitational force per unit

particle mass. Equation (5-6) is the definition of the relative Reynolds number for flow

around a sphere. In equation (5-5) and (5-6) ρ is the fluid‟s density; ρp is the particle‟s

density; dp is particle diameter; vp_i is particle velocity in the ith direction; vi is the

localized fluid velocity in the ith direction; and µ is the dynamic viscosity. Equation (5-7)

is the drag coefficient for spherical particles with the constants K1, K2, and K3 defined in

Morsi and Alexander (1972). In a transient Eulerian-Lagrangian solution framework the

discrete phase is numerically solved by iteration in interwoven intervals with the

Eulerian solution space.

The computational modeling of the free chlorine concentration, transport, and

decay is a coupled set of Eulearian-Lagrangian equations extending the analytical work

of dissolved and particulate chlorine demand in urban stormwater in Chapter 3 whereby

chlorine reactions are broken into three component reactions: a second order, fast

112

demand reaction with the dissolved fraction (rate constant: k1), a second order, slow

demand reaction with the dissolved fraction (rate constant: k2), and a potential driving

PM model (rate constant: kpm). The kinetic governing equations implemented in

Chapter 3 are the second order parallel dissolved model (Equation 5-8) and the PM

potential driving model (Equation 5-9):

(5-8)

( ) (5-9)

where Ct is the total free chlorine concentration; Cf is the fast acting free chlorine

concentration; Cs is the slow acting free chlorine concentration; DMDf is the

concentration of the chlorine demand reacting with the fast acting free chlorine; DMDs is

the concentration of the chlorine demand reacting with the slow acting free chlorine; qe

is the mass of free chlorine consumed in PM surface reactions per mass PM at

equilibrium; and qt is the mass of free chlorine consumed in PM surface reactions per

mass PM at time t.

The differential equation set is formulated on the following assumptions:

The dissolved reactions (fast/slow) are parallel reactions without

interaction

The dissolved chlorine reactions involve a 1:1 reaction with a theoretical

chlorine demand (fast/slow)

The initial dissolved chlorine demand is a fraction, f, of the dissolved

chemical oxygen demand, CODd

PM chlorine demand is driven by a chlorine reaction potential and is

limited by the local available free chlorine

113

Potential particulate chlorine demand dissolution into the dissolved

fraction is accounted for in the PM demand

At a uniform temperature, the fast dissolved free chlorine demand reaction is

modeled by a combined Eulerian-Lagrangian mass transport equation:

( )⏟

( )

⏟

( )

⏟

⏟

⏟

(5-10)

where YCl-F is the mass fraction of the free chlorine in the fast reaction; YDMD-F is

the mass fraction of the chlorine demand in the fast reaction; ρ is the species density; ui

is the local velocity vector; D is the diffusion coefficient of the species; µt is the turbulent

viscosity; Sct is the turbulent Schmidt number; k is the dissolved second order rate

constant with units [L3M-1T-1]; cpm is the local concentration of PM as represented by

discrete particles; kpm is the PM potential driving model rate constant with units

[M1M-1T1]; and

( ( )

) (5-11)

where

is limited by the lower value of the PM potential driving model and the

local free chlorine concentration. qe is the mass of free chlorine consumed in PM

surface reactions per mass PM at equilibrium with units [M1M-1]; qt is the mass of free

chlorine consumed in PM surface reactions per mass PM at time t with units [M1M-1];

and is defined as Equation 5-12.

(5-12)

where YCl-T is the sum of the mass fractions of the free chlorine in the fast and

slow (YCl-S) reactions and ensures the proportionate removal of free chlorine from the

114

dissolved fractions. The convection, turbulent diffusion, and dissolved demand terms

are solved in Eulerian space, and the PM demand is solved in Lagrangian space with

the time rate of change solved in both reference frames. Similarly, the mass transport

equation for the slow reaction is defined in Equation 5-13:

( )

( ) (

)

( )

(5-13)

and two additional transport equations for chlorine demand for j = F,S are defined

as Equation 5-14.

( )

( ) (

)

(5-14)

Numerical requirements for the mixed transport and reaction Eulerian-Lagrangian

dissolved and PM species model require that the PM chlorine demand does not exceed

the local species mass within the numerical cell within the Lagrangian timestep

advancement when the reactor is not in an overall limiting condition (when the

volumetric mean ρYCl-T >>

). Qualitatively, this is governed by mesh spacing, overall

chlorine concentration, PM mass loading, the number of representative particles in the

discrete phase, and the timestep of the simulation. In the absence of experimental data

validating data, the solution must demonstrate mesh, particle number, and timestep

independence.

The numerical Eulerian-Lagrangian kinetic chlorine model is computationally

simulated in ANSYS Fluent 13.0 where the demand reaction terms in the above

115

equations are programed in C computer code and compiled as external dynamic linked

libraries.

Batch Reactor Setup and Initialization

The batch reactors are modeled as three-dimensional continuously stirred batch

reactors (CSBRs) of 1700 ml volume with a diameter of 140 mm, height of 130 mm, and

a flat bottom with a 20 mm fillet with the sidewalls. The mesh has 84 thousand

tetrahedral cells resulting in a mean cell volume of 0.02 ml. The CSBRs consist of three

distinct fluid zones – a 50 ml free chlorine injection zone, a rotating zone containing the

stir rod, and a bulk fluid zone. The mixing within the batch reactor is motivated by a 6

mm by 40 mm stir rod set within a hemispherical rotating mesh at the bottom center of

the reactor. The hemispherical moving mesh has a sliding interface with the bulk fluid

zone and rotates at 350 rpm. The free chlorine injection zone is contained within the

1700 ml of the batch reactor, which is illustrated in Figure 5-1. The reactor is initialized

with all species fractions set to zero and the velocity and pressure field are brought to

periodic steady state by solving 60 s of flow time at a 1 s timestep. The initial values of

the chlorine species are then patched to the free chlorine injection zone to mimic the

physical hypochlorite injection of the physical CSBRs with the concentration of the

patched zone scaled over the volume of the reactor to provide a mean concentration of

the initial experimental value which was nominally 15, 30, or 45 mg/L depending on the

modeled batch reactor. Chlorine demand species that react on a 1:1 basis are patched

to the CSBR domain where YDMD-F = X•fCODd/ρ and YDMD-S = (1-X)•fCODd/ρ where

CODd is the dissolved chemical oxygen demand and X and f are model parameters is

the second order dissolved model defined for a small urban catchment in Gainesville,

FL in Table 5-1. For CSBRs with PM interaction, 1400 representative discrete particles

116

are injected at t = 61s along the viewing plane that bisects the CSBR in Figure 5-1.

Timesteps for the CSBR are 1 s for the first minute and 5 s thereafter.

CSBR Validation

CSBR validation is performed on a dataset of 9 experimental batch reactors from

Chapter 3 with runoff from a small paved urban catchment for events that occurred

during the fall of 2010. A control batch reactor with hypochlorite addition to chlorine

demand free water is used as a control for the mixing of the simulation. Four of these

batch reactors elaborate performance on dissolved urban rainfall-runoff and validate the

CFD finite-rate parallel dissolved kinetic model and the remaining batch reactors

validate the PM potential driving model and the composite PM and dissolved model.

The criteria of the model performance for the CFD species model is the normalized root

mean square error (NRMSE):

√∑ ( )

(5-15)

where Oi is the observed value at measurement i; Ei is the modeled value at

measurement i; n is the total number of measurements; and Co is the initial chlorine

dose. For the experimental dataset, n = 8. The NRMSE defined as in equation (5-15)

illuminates the model performance relative to the initial chlorine dose of the reactor. For

both the dissolved and PM laden reactors, the volumetric mean of YCl-T of the reactor is

utilized as the estimate of the chlorine concentration at the sample time.

Results and Discussion

Figure 5-3 presents a histogram mixing analysis of the CSBR of the initial mixing

phase of a simulated hypochlorite injection, of reactor average value Co, with no

117

chlorine demand. Initially 96.5% of the cells within the CSBR contain no chlorine

species with the remainder containing the high chlorine dose of the injection. At 5 s, the

89% of the CSBR volume elements contain chlorine concentration with ±5 mg/L of Co.

At 15s, 100% of the CSBR volume elements contain a chlorine concentration with ±1

mg/L of Co. The experimental batch reactor found the target Co concentration at the

earliest sample time of 1 min and the CFD simulation corroborates this finding.

Figure 5-4 presents the validation of the parallel dissolved model under disparate

initial conditions. The correlation of the dissolved chlorine demand in urban stormwater

to the CODd is apparent comparing panels C and D, which illustrates a high CODd, high

demand sample, and a low CODd, low demand sample, respectively. Overall the data

is illustrative of the second order nature of the dissolved demand wherein the kinetic

rate is dependent on the concentration of both the free chlorine and dissolved demand

and may be limited by either constituent depending on the initial chlorine dose and the

water quality of the sample. In each case in the dissolved CSBRs there is a clear initial

demand followed by a chlorine demand of slower timescale and the CFD kinetic model

correctly predicts the exhaustion of the fast rate demand portion with respect to the

experimental data. The NRMSE of the reactors are 3.3%, 4.0%, 3.8%, and 5.4%, for

panels A through D, respectively. The modeled reactors utilized in the validation of the

CFD implementation of the second order parallel analytical rate expression for the

dissolved phase were non-influential in the derivation of the model constants. Thus,

these reactors are examples of the predictive capability of the CFD model on typical

loadings given the CODd of the runoff and Co for the small urban catchment in

Gainesville, FL with the characteristic model parameters in Table 5-1. These

118

characteristic parameters are a result of the typical event particulate and dissolved

loadings of the catchment and were found to be consistent on an inter-event basis.

Figure 5-5 presents the validation of the potential driving PM kinetic model with

CSBRs laden with PM in a dissolved stormwater matrix. The dissolved demand is

subtracted from the total demand to produce the PM demand in the batch reactor. This

PM demand is modeled by the potential driving model which is governed by the

available reaction sites on the PM mass within the CSBR and the available free chlorine

dose. These characteristics are modeled through qe and kpm. NRMSEs for the CFD PM

kinetic demand model are 7.1%, 7.9% 2.6% and 10.3%, for panels A through D with a

slight modification where Co = qe in the determination of the NRMSE. As can be seen

from the figure, the timescale of the PM potential driving model is on the order of the

slow second order dissolved demand reaction. It is also important to note that the

capacity of PM chlorine demand is high with respect to PM mass and represents a

pronounced potential chlorine sink in urban rainfall-runoff at even low PM loadings.

The results from the composite particulate and dissolved CFD kinetic model are

presented in Figure 5-6. The NRMSEs for the composite model are 2.4%, 4.3%, 2.6%,

and 3.9% for panels A through D, thus the composite model is capable of reproducing

the experimental CSBR kinetic reaction utilizing all three simultaneous reactions.

The extension of the CFD composite kinetic model to a flow through chlorine

contactor for urban rainfall runoff incorporates a few important considerations. PM

transport is essential in modeling urban stormwater disinfection processes as sediment

PM shielding of associated organisms has been established in stormwater runoff in

Chapter 4. Thus, a modeled unit operation and process must ensure that all particles

119

greater than 75 µm are captured by the unit. The dissolved species transport model

should not require significant modifications for flow through reactors as implemented in

this study. To directly model bacteriological, viral, or protozoan transport and

inactivation, a generalized scalar equation can be solved utilizing a kinetic inactivation

sink term such as the Hom model (Greene et al. 2004) overlaying the composite CFD

kinetic model presented in this study. The analytical potential driving model, from which

the CFD model is derived for the PM surface reaction, is derived for initial hypochlorite

doses to find the chlorine demand at equilibrium. Additional research investigating the

maximum qe values for PM fractions with continual exposure to low doses of free

chlorine is warranted to investigate the performance of the PM model under continually

limiting chlorine application. However, even with this limitation, the implementation of

the PM demand model in this study would either equal the reaction under a continual

low-dose chlorination experiment or exceed the value and remain a conservative

estimator of the chlorine concentration in this case.

120

Table 5-1. Model parameters for the dissolved parallel second order and PM potential driving force equations.

Dissolved

PM

k1 0.07 L1mg-1min-1

Reactor V4 R13 V5 V1 k2 2.9 L1mg-1min-1

kpm 0.83 2.19 1.00 1.80 g1mg-1min-1

f 0.39 -

qe 102 105 85 154 mg/g X 0.36 -

121

Figure 5-1. Physical batch reactor showing stirplate, aluminum foil jacket, and water

quality electrodes. Utilized reactor volume is 1700 ml and the stirplate is set to 350 rpm. The batch reactor is sealed tightly with a lid when water quality measurements are not being made.

122

Figure 5-2. Illustration of the fluid zones within the batch reactor. The hemispherical

rotating zone is shown containing the stir rod at a single frame. The hypochlorite injection region is a cylindrical zone bisected by the viewing plane and is represented by the red region. The bulk fluid zone is shown in blue. Reactor volume is 1700 ml and is comprised of approximately 41 thousand tetrahedral cells.

HOCl Injection Region

Bulk Fluid Zone

Stir Rod in Rotating Fluid

Zone

123

Figure 5-3. Histogram analysis of the computational mesh CFD free chlorine

concentration during the initial mixing phase in a batch reactor with an overall initial Co = 45 mg/L. After 5s of mixing, 90% of the computational mesh exhibits a free chlorine concentration within ±5 mg/L. After 15 s of mixing, 100% of the computational mesh exhibits a free chlorine concentration within ±1 mg/L. The uniformity of the chlorine concentration in the reactor enables increasing the timestep of the simulation.

<-5 -5 -4 -3 -2 -1 Co 1 2 3 4 5 >50

20

40

60

80

p

df

(%)

Cell Deviation from Co (mg/L)

Elapsed Time: 15s

<-5 -5 -4 -3 -2 -1 Co 1 2 3 4 5 >50

20

40

60

80

pd

f (%

)

Cell Deviation from Co (mg/L)

Elapsed Time: 10s

<-5 -5 -4 -3 -2 -1 Co 1 2 3 4 5 >5

0

20

40

60

80

100

pd

f (%

)

Cell Deviation from Co (mg/L)

Elapsed Time: 5s

0

20

40

60

80

100<-5 -5 -4 -3 -2 -1 Co 1 2 3 4 5 >5

Cell Deviation from Co (mg/L)

pd

f (%

)

Elapsed Time: 0s

124

Figure 5-4. Comparison of the second order CFD dissolved demand model with

experimental results. NRMSE values are reported and indicate that the CFD model accurately (NRMSE < 6%) accounts for the complex reaction dynamics of the urban stormwater demand reactor containing dissolved matrix of disparate initial water quality conditions and demand.

0

5

10

15

20

25

30

0 100 200 300 400

Reactor Time (min)

Storm Date: 21-Aug-2010

Co = 30.4 mg/L | COD

d = 91.18 mg/L

CFD Model

k1 = 0.0701 L

1mg

-1min

-1

k2 = 2.93 X 10

-4 L

1mg

-1min

-1

X = 0.39

f = 0.36

HO

Cl/O

Cl- (

mg

/L)

NRMSE: 3.3%

A

0 100 200 300 4000

5

10

15

20

25

30

35

40

45

NRMSE: 5.4%

Storm Date: 7-Aug-2010

Co = 29.0 mg/L | COD

d = 33.5 mg/L

CFD Model

HO

Cl/O

Cl- (

mg

/L)

Reactor Time (min)

D

0 100 200 300 400

0

5

10

15

20

25

30

35

40

45

50

NRMSE: 3.8%

Storm Date: 4-Nov-2010

Co = 45.6 mg/L | COD

d = 148.2 mg/L

CFD Model

Reactor Time (min)

HO

Cl/O

Cl- (

mg

/L)

C

0 100 200 300 4000

5

10

15

20

25

30

35

40

45

NRMSE: 4.0%

Storm Date: 21-Aug-2010

Co = 43.4 mg/L | COD

d = 91.18 mg/L

CFD Model

HO

Cl/O

Cl- (

mg

/L)

Reactor Time (min)

B

125

Figure 5-5. Comparison of the second order potential driving PM CFD model with

experimental results.

0

20

40

60

80

1000 100 200 300 400

Reactor Time (min)

Storm Date: 27-Sep-2010

Co = 31.8 mg/L | V4

Settleable | PM = 149.4 mg/L

CFD Model

qt (

mg

/g)

NRMSE: 7.1%

A

0 100 200 300 4000

20

40

60

80

100

120

140

160

NRMSE: 10.3%

Storm Date: 27-Sep-2010

Co = 29.0 mg/L

Suspended | PM = 36.18 mg/L

CFD Model

qt (

mg

/g)

Reactor Time (min)

D

0 100 200 300 400

0

20

40

60

80

100

NRMSE: 2.6%

Storm Date: 4-Nov-2010

Co = 31.5 mg/L | V5

Sediment | PM = 148.2 mg/L

CFD Model

Reactor Time (min)

qt (

mg

/g)

C

0 100 200 300 4000

20

40

60

80

100

120

140

160

NRMSE: 7.9%

Storm Date: 21-Aug-2010

Co = 46.6 mg/L | R13

Settleable | PM = 197.8 mg/L

CFD Model

qt (

mg

/g)

Reactor Time (min)

B

126

Figure 5-6. Comparison of the composite dissolved and PM CFD model with batch

reactor data. NRMSEs and RPDs are reported and are less than 5% in each case. Results validate the finite rate CFD kinetic model developed for free chlorine demand in urban stormwater.

0

5

10

15

20

25

30

35

40

45

500 100 200 300 400

Reactor Time (min)

Storm Date: 27-Sep-2010

Co = 30.4 mg/L | COD = 18.2 mg/L

Settleable | PM = 91.18 mg/L

CFD Model

HO

Cl/O

Cl- (

mg

/L)

NRMSE: 2.4%

A

0 100 200 300 4000

5

10

15

20

25

30

35

40

45

NRMSE: 3.9%

Storm Date: 27-Sept-2010

Co = 32.0 mg/L | COD = 46.7 mg/L

Suspended | PM = 36.1 mg/L

CFD Model

HO

Cl/O

Cl- (

mg

/L)

Reactor Time (min)

D

0 100 200 300 400

0

5

10

15

20

25

30

35

40

45

50

NRMSE: 2.6%

Storm Date: 27-Sep-2010

Co = 31.6 mg/L | COD = 18.2 mg/L

Sediment | PM = 330.9 mg/L

CFD Model

Reactor Time (min)

HO

Cl/O

Cl- (

mg

/L)

C

0 100 200 300 4000

5

10

15

20

25

30

35

40

45

NRMSE: 4.3%

Storm Date: 21-Aug-2010

Co = 46.6 mg/L | COD = 43.3 mg/L

Settleable | PM = 197.8 mg/L

CFD Model

HO

Cl/O

Cl- (

mg

/L)

Reactor Time (min)

B

127

CHAPTER 6 CONCLUSION

Urban stormwater particle transport and disinfection reactions are complex

phenomena with coupled transport and reaction kinetics across both solid and liquid

phases. In addition, stormwater volumetric and particle transport are the result of

stochastic rainfall events that render the volumetric and particulate matter (PM) loading

difficult to determine or estimate a priori.

Free Chlorine Kinetics

Dissolved Phase Reaction Kinetics

The reaction kinetics of the dissolved phase of the urban stormwater runoff exhibit

parallel second order characteristics. The ultimate modeled chlorine demand of the

water in a batch reaction is determined to correlate well with the dissolved chemical

oxygen demand (CODd). CODd is a simple and expedient analytical procedure and it is

possible to process many samples simultaneously. However, certain formulations of

COD reagents generate hazardous waste, thus the suitability and regulatory usability of

non-hazardous waste formulations remains to be determined.

The bimodal proportionality of the parallel reaction of the second order kinetics is

stable on the inter-event reference frame for the small paved urban watershed utilized in

this study for runoff collection. This result is expected to be function of the ratio of the

inorganic to organic loadings on the watershed. This loading ratio can be expressed in

terms of the dissolved organic carbon ratio (DOC) to CODd. This ratio has been

determined to be consistent and a characteristic parameter of the watershed. However,

as the experimental matrix only included a single catchment, the full elaboration of the

relationship between the bimodal proportionality of the reaction is beyond the resolution

128

of the experimental construct. Thus, future research regarding the relationship of this

proportionality factor, X, to DOC/CODd for similar and dissimilarly loaded catchments is

warranted.

Particulate Kinetics

The reaction kinetics of free chlorine with the various particulate matter (PM)

fractions mobilized by urban stormwater runoff are well modeled by a second order

potential driving model. This model relates the kinetic parameters to the available

surface reaction sites and a driving potential of available free chlorine. As free chlorine

is an added component to the kinetic reference frame, there are three general resulting

cases for reaction whereby the batch reactor is under-chlorinated, super-chlorinated, or

transitionally chlorinated. There remains a case where the reactor is severely

under-chlorinated, where the dissolved demand immediately quenches all available

chlorine and this is most suitably modeled by the dissolved model without particulate

consideration. In the under-chlorinated model the parameter of maximal mass transfer,

qe, is limited by the maximum value of chlorine mass available for reaction with the

particulate phase, qe max. qe max is the initial mass of chlorine inoculated into the reactor

less the chlorine that reacts with the dissolved phase normalized of the PM mass

contained within the reactor. In the super-chlorinated reactors, the chlorine mass

transfer is limited by the available reaction sites on the PM. This in effect exhausts the

chlorine demand of the PM fraction of the reactor. This maximum sorbance per PM unit

mass is characteristic of the PM fraction with the suspended and settleable fraction

exerting similar maximal demand due to their similar low volatile fractions and siliceous

composition. The maximum sorbance per PM unit mass for the sediment fractions is

129

double the sorbance for the suspended and settleable fraction and is attributed to the

high volatile fraction and, thereby, the increased organic density of that fraction.

Computational Modeling

PM Fate and Transport

The dispersivity of the granulometry of the PM phase influences the computational

accuracy of a Lagrangian-Eulerian computational fluid dynamics (CFD) model. The

fundamental fluid mechanics driving this phenomenon are non-linear drag forces, with

respect to particle diameter, exerted on mobilized PM entrained in the flow. In the

framework of a Lagrangian-Eularian PM fate and transport model, PM is represented by

discrete particle sizes and the non-linear nature of the mechanics reduces the

computational accuracy of the when the simulated particle sizes inadequately represent

a disperse influent gradation.

The computational structure for the investigation involves a convergence study

that incrementally increases the number of representative particle sizes, the

discritization number (DN), by a factor of 2 until solution difference is negligible –

achieved at a DN of 128. Average relative percent difference (RPD) calculations

ascertain the influence of gradation dispersivity on computational accuracy.

Mono-disperse, or uniformly disperse gradations, are well represented by a median

particle size, the d50m, which correlates to a DN of 1. However, gradations of medium

dispersivity and heterodisperse gradation require a DN of 8 to 16, depending on overall

gradation fineness, for computational modeling. Overall gradation fineness influences

the DN requirement mechanistically as the non-linearity of particle drag forces increases

for very small particle diameters. Thus, a DN of 8 suffices for particle granulometries

130

with d50m = 66.7 µm and d50m = 100 µm, however, a DN of 16 is required for finer

gradations near a d50m of 33.3 µm.

The Lagrangian-Eulerian CFD model also enabled the development of a novel

CFD rubric for the performance evaluation of hydrodynamic and baffled separators. A

composite dataset of the per-particle removal efficiency across a spectrum of flow rates

generates a performance surface for the unit. The performance surface is akin to a unit

“fingerprint” and is unique to the fate and transport behavior of the unit given a particle

density. Mathematical comparison of the performance surfaces of two distinct

separators yields a multi-dimensional, differential performance evaluation that extends

the depth found in typical performance studies utilizing a singular gradation. In addition,

the data from the performance surface can be utilized to model the efficiency of a unit

operation given a specific gradation without the need for additional computational time.

CFD Free Chlorine Reaction Kinetics

Advanced modeling of the fate, transport, and kinetics of free chlorine in a

transient simulation of urban stormwater runoff is possible in CFD. The coupled

dynamics of the chlorine kinetics with the dissolved and PM phases under transient

loading conditions generate a dynamic physical-chemical network with multiple reaction

pathways. CFD models the fundamental equations of motion for PM, chemical species,

and fluid flow. The analytical models developed in Chapter 3 are successfully ported to

discrete equivalents in differential space for computational modeling in CFD and

validated on the dataset generated in the CSBRs for rainfall-runoff events captured in

the fall of 2010 in Gainesville, FL and generated NRMSEs of less than 6% in all cases.

Advanced CFD modeling utilizing validated CFD models allows for the integrated

and iterative design of treatment systems without the need for pilot scale testing of each

131

design variant. This reduces research and development expenses for such testing and

permits exploration of non-traditional design implementations which were previously

cost prohibitive.

132

APPENDIX A ADDITIONAL FIGURES

Figure A-1. Continuously stirred batch reactor (CSBR) schematic.

133

Figure A-2. Schematic of monitored urban sub-catchment in Gainesville, FL showing contributing impervious surface.

134

0 500 1000 1500

0

10

20

30

40

50

15 mg/L

45 mg/L

HO

Cl/O

Cl- (

mg

/L)

Reaction Time (min)

Figure A-3. Control CSBRs showing hypochlorite kinetics in Nanopure DI for 8 h at 15 mg/L and 24 h at 45 mg/L. Results indicate no detectable environmental loss of hypochlorite due to UV or volatilization during the experimental timeframe (p < 0.05).

135

0 100 200 300 400 500

0

10

20

30

40

50 30 mg/L

30 mg/L

30 mg/L Autoclaved

30 mg/L Autoclaved

HO

Cl/O

Cl- (

mg

/L)

Reaction Time (min)

Figure A-4. Control CSBRs comparing autoclave sterilized and non-autoclave sterilized stormwater Matrix. Results indicate that there is not a significant difference between autoclaved and non-autoclaved matrix (p < 0.05).

136

LIST OF REFERENCES

American Society for Testing and Materials (ASTM) (2002). “Standard Test Methods for Determining Sediment Concentration in Water Samples,” (ASTM Designation: ASTM Standard D 3977-97(2002)), ASTM, West Conshohocken, PA.

Baawain, M. S., El-Din, M. G., and Smith, D. W. (2006). “Computational fluid dynamics application in modeling and improving the performance of a storage reservoir used as a contact chamber for microorganism inactivation.” Journal of Environmental Engineering Sciences, 5, 151-162.

Beach, M. J. (2008). “Waterborne: Recreational Water.” Cryptosporidium and Cryptosporidiosis, 2nd Ed., R. Fayer and L., eds., CRC Press: Florida 2008; 335–370.

Bellamy, W. D., Finch, G. R., and Haas, C. N. (1998). Integrated disinfection design framework. American Water Works Research Foundation, Denver, CO.

Berman, D., Rice, E. W., and Hoff, J. C. (1988). “Inactivation of particle-associated coliforms by chlorine and monochloramine.” Applied and Environmental Microbiology, 54(2), 507-512.

Berretta, C. and Sansalone, J. (2011). “Hydrologic transport and partitioning of phosphorus fractions.” Journal of Hydrology, In Press.

Boccelli, D. L., Small, M. J., and Diwekar, U. M. (2004). “Treatment plant design for particulate removal: Effects of flow rate and particle characteristics.” Journal of the American Water Works Association, 96(11), 77-90.

Borst, M. and Selvakumar, A. (2003). “Particle-associated microorganisms in stormwater runoff.” Water Research, 37, 215-23.

Brennen, C. (2005). Fundamentals of multiphase flow, Cambridge Univ. Press, New York.

Characklis, G. W., Dilts, M. J., Simmons III, O. D., Likirdopulos, C. A., Krometis, L. A., and Sobsey, M. D. (2005). “Microbial partitioning to settleable particles in stormwater.” Water Research, 39(2005), 1773-1782.

Chellum, S. and Wiesner, M. (1997). “Particle back-transport and permeate flux behavior in crossflow membrane filters.” Environmental Science and Technology, 31(3), 819-824.

Chick, H. (1908). “An investigation of the laws of disinfection.” The Journal of Hygiene, 8(1), 92-158.

137

Chizek, A. R., Characklis, G. W., Krometis, L., Hayes, J. A., Simmons III, O. D., Lonardo, S. D., Alderisio, K. A., and Sobsey, M. D. (2008). “Comparing the partitioning behavior of Giardia and Cyrptosporidium with that of indicator organisms in stormwater runoff.” Water Research, 42(2008), 4421-4438.

Clark, R. M. (1998). “Chlorine demand and TTHM formation kinetics: A second-order model.” Journal of Environmental Engineering, 124(1), 16-24.

Clark, R. M. and Savaganesan, M. (2002). “Predicting chlorine residuals in drinking water: Second order model.” Journal of Water Resources Planning and Management, 128(2), 152-161.

Clausen, E. M., Green, B. L., Litsky, W. (1977). “Fecal Streptococci: Indicators of Pollution.” In Bacterial Indicators / Heath Hazards Associated with Water, ASTM STP635, Hoadley, A.W., Dutka, B. J., Eds, American Society for Testing and Materials, 247-264.

Code of Federal Regulations (2001). “Total maximum daily loads (TMDL) and individual water quality-based effluent limitations.” Part 130.7, Title 40.

Cristina C. and Sansalone J. (2003). “First flush, power law and particle separation diagrams for urban storm-water suspended particulates.” Journal of Environmental Engineering, 129 (4), 298-307.

Dean, C. M., Sansalone, J. J., Cartledge, F. K., Pardue, J. H. (2005). “Influence of hydrology on rainfall-runoff metal element speciation.” Journal of Environmental Engineering, 131(4), 632-642.

Delfino, J. (1977). “Effect of river discharge and suspended sediment on water-quality in the Mississippi River.” Journal of Environmental Science and Health Part A, 12(3), 79-94.

Department of Conservation (DEC) NSW (2006). Managing Urban Stormwater: Harvesting and Reuse. Sydney, Australia.

Dickenson, J. and Sansalone J. (2009). “Discrete Phase Model Representation of Particulate Matter (PM) for Simulating PM Separation by Hydrodynamic Unit Operations.” Environ. Sci. Technol., 43, 8220-8226.

Dietrich, J. P., Basagaoglu, H., Loge, F. J., and Ginn, T. R. (2003). “Preliminary assessment of transport processes influencing the penetration of chlorine into wastewater particles and the subsequent inactivation of particle-associated organisms.” Water Research, 37(2003), 139-149.

Eaton A. D., Clesceri L. S. and Greenberg A. E. (1998). “Standard Methods for the Examination of Water and Wastewater.” 20th Edition. American Public Health Association, American Water Works Association and Water Environment Federation.

138

Fair, Geyer, and Okun (1968). Water and Wastewater Engineering, Volume 2, John Wiley and Sons, Inc., New York.

Fair, G. M.; Morris, J. C., Chang, S. L., Weil, I., and Burden, R. P. (1948). “The Behavior of Chlorine as a Water Disinfectant.” Journal of the American Water Works Association, 40(10), 1051-1061.

Feng, P. C. S. and Hartman P. (1982). “Fluorogenic Assays for Immediate Confirmation of Eschericha coli.” Applied and Environmental Microbiology, 43(6), 1320-1329.

Ferziger, J. and Peric, M. (2002). Computational methods for fluid dynamics, Springer, Berlin, Germany.

Finch, G.R., Black, E. K., Gyürék, L., and Belosevic, M. (1993). “Ozone Inactivation of Cryptosporidium parvum in Demand Free Phosphate Buffer Determined by In Vitro Excystation and Animal Infectivity.” Applied and Environmental Microbiology, 59(12), 4203-4210.

Fluent Inc., (2006). Fluent 6.3 User's Guide, Lebanon, New Hampshire, USA.

Folk, R. L., Ward, W. C. (1957). “Brazos River Bar: A study in the significance of grain size parameters.” Journal of Sedimentary Petrololgy, 27, 3-27.

Geldreich, E. E. and Kenner, B. A. (1969). “Concepts of Fecal Streptococci in stream pollution.” Water Pollution Control Federation, 41(8), R336-R352.

Goula, A. M., Kostoglou, M., Karapantsios, T., and Zouboulis, A. I. (2008). “A CFD methodology for the design of sedimentation tanks in potable water treatment. Case study: the influence of a feed flow control baffle.” Chemical Engineering Journal, 140(2008), 110-121.

Greene, D. J., Haas, C. N., and Farouk, B. (2006). “Computational Fluid Dynamics Analysis of the Effect of Reactor Configuration on Disinfection Efficiency.” Water Environ. Res., 78, 909-919.

Greene, D. J., Farouk, B., and Haas, C. N. (2004). “CFD design approach for chlorine disinfection processes.” Journal of the American Water Works Association, 38, 3449-3457.

Gyürék, L. L. and Finch G. R. (1998). “Modeling water treatment chemical disinfection kinetics.” Journal of Environmental Engineering, 124(9), 783-793.

Haas, C. N. and Joffe, J. (1994). “Disinfection under dynamic conditions: modification of Hom‟s model for decay.” Environmental Science and Technology, 28, 1367-1369.

Haas C. N., Joffe, J., Heath, M., Jacangelo, J., Anmangandla, U. (1998). “Predicting disinfection performance in continuous flow systems from batch disinfection kinetics.” Water Science and Technology, 38(6), 171-179.

139

Haas, C. N. and Karra, S. B. (1984). “Kinetics of wastewater chlorine demand exertion.” Journal of the Water Pollution and Control Federation, 56(2), 170–173.

Hathaway, J. M. Hunt, W. F., and Simmons, O. D. (2010). “Statistical evaluation of factors affecting indicator bacteria in urban storm-water runoff.” Journal of Environmental Engineering, 136(12), 1360-1368.

Hatt, B. E., Deletic, A., Fletcher, T. D. (2006). “Integrated treatment and recycling of stormwater: a review of Australian practice.” Journal of Environmental Management, 79(2006), 102-113.

Hazen, A. (1904). “On sedimentation.” Transactions of the American Society of Civil Engineers, 52, 45-71.

He, J., Valeo, C., Chu, A., and Neumann, N. F. (2008). “Water quality assessment in the application of stormwater reuse for irrigating public lands.” Water Quality Research Journal of Canada, 43(2/3), 145-159.

Heaney, J. P. and Huber, W. C. (1984). “Nationwide assessment of urban runoff impact on receiving water quality.” Water Resources Bulletin, 20(1), 35-42.

Heaney, J.P. and Sample, D. (2000). “Stormwater Storage-Treatment-Reuse Systems.” In Innovative Urban Wet-Weather Flow Management Systems. Field, R., Heaney, J.P., Pitt, R. and R. Field, Eds. Technomics Publishing Co., Lancaster, PA.

Hom, L.W. (1972). “Kinetics of chlorine disinfection in an ecosystem.” Journal of the Sanitary Engineering Division, 98(1), 183-194.

House, M. A., Ellis, J. B., Herricks, E. E., Hvitved-Jacobsen, T., Seager, J., Lijklema, L., Aalderink, H., Clifforde, I. T. (1993). “Urban drainage – impacts on receiving water quality.” Water Science Technology, 27(12), 117-158.

Hrudey, S. E., and Hrudey, E. J. (2004). Safe drinking water: Lessons from recent outbreaks in affluent nations, IWA Publishing, London, UK.

Huang, J. J. and McBean, E. A. (2007). “Using Bayesian statistics to estimate the coefficients of a two- components second order chlorine bulk decay model for a water distribution system.” Water Research, 41(2007), 287-294.

Isenberg, H.D., Goldberg, D., Sampson, J. (1970). “Laboratory Studies with a Selective Enterococcus Medium.” Applied Microbiology, 20(3), 433-436.

International Organization for Standardization (ISO). ISO 13320. “Particle size analysis – Laser diffraction methods.” Geneva, 2009.

Jakobsen, H.A. (2008). Chemical reactor modeling: Multiphase reactive flows, Springer Berlin, Heidelberg, Germany.

140

Jillavenkatesa A., Dapkunas S. J. and Lum L. S. H. (2001). “Particle size characterization”, Special Publication 960-1, National Institute of Standards and Technology, Washington D.C.

Jin, G., Englande, A. J., Bradford, H., Jeng, H. (2004). “Comparison of E. coli, Enterococci, and Fecal Coliform as indicators for brackish water quality assessment.” Water Environment Research, 76, 245–255.

Kim, J. and Sansalone, J. J. (2008). “Event-based size distributions of particulate matter transported during urban rainfall-runoff events.” Water Research, 42(2008), 2756-2768.

Kim, J. and Sansalone, J. J. (2010). “Representation of particulate matter COD in rainfall-runoff from paved urban watershed.” Water, Air and Soil Pollution, 205, 113–132.

Kott, Y. (1977). “Current concepts of indicator bacteria.” Bacterial Indicators / Heath Hazards Associated with Water, ASTM STP635, A. W. Hoadley and B. J. Dutka, eds., American Society for Testing and Materials, Philadelphia, PA, 3-13.

Krometis, L. A., Characklis, G. W., Simmons III, O. D., Dilts, M. J., Likirdopulos, C. A., and Sobsey, M. D. (2007). “Intra-storm variability in microbial partitioning and microbial loading rates.” Water Research, 41(2007), 506-516.

Launder, B. E. and Spalding, D. B. (1974). “The numerical computation of turbulent flows.” Computer Methods in Applied Mechanical Engineering, 3, 269-289.

LeChevallier, M. W., Evans, T. M., and Seidler, R. (1981). “Effect of turbidity on chlorination efficiency and bacterial persistence in drinking water.” Applied and Environmental Microbiology, 42(1), 159-167.

Lesieur, M. and Métais, O. (1996). “New trends in large-eddy simulations of turbulence.” Annu. Rev. Fluid Mech., 28, 45-82.

Lin, H., Ying, G., and Sansalone, J. (2009). “Granulometry of non-colloidal particulate matter transported by urban runoff.” Water, Air and Soil Pollution, 198, 269-284.

Liu, D., Sansalone, J. J., Cartledge, F. K. (2005). “Adsporption kinetics for urban rainfall-runoff metals by composite oxide-coated polymeric media.” Journal of Environmental Engineering, 131(8), 1168-1177.

Lünsdorf, H., Erb, R. W., Abraham, W. R., and Timmis, K. N. (2000). “„Clay Hutches‟: a novel interaction between bacteria and clay minerals,” Environmental Microbiology, 2(2), 161-168.

Maier, R. M., Pepper, I. L., and Gerba, C. P. (2009). Environmental Microbiology. Academic Press: Burlington, MA.

141

Majewska, A. C. (1994). “Successful experimental infections of a human volunteer and Mongolian gerbils with Giardia of animal origin.” Transactions of the Royal Society of Tropical Medicine and Hygiene, 88(3), 360-362.

Marquardt (1963). “An algorithm for least-squares estimation of nonlinear parameters.” Journal of the Society for Industrial and Applied Mathematics, 11(2), 431-441.

Marselek, J., Barnwell, T. O., Geiger, W., Grottker, M., Huber, W. C., Saul, A. J., Schilling, W., and Torno, H. C. (1993). “Urban drainage systems: Design and operation.” Water Science Technology, 27(12), 31-70.

McManus, D. A. (1966). “A criticism of certain usage of the Phi notation.” Journal of Sedimentary Petrology, 33, 670-674.

Morsi, S. A. and Alexander, A. J. (1972). “An investigation of particle trajectories in two-phase flow systems.” Journal of Fluid Mechanics, 55(2), 193-208.

NJDEP (2003). “Total Suspended Solids Laboratory Testing Procedure.” New Jersey Department of Environmental Protection, Office of Innovative Technology and Market Development, Trenton, NJ <http://www.state.nj.us/dep/dsr/bscit/TestProcedure_Dec%2703_.pdf>

Pathapati, S., Sansalone, J. (2009). “CFD modeling of a stormwater hydrodynamic separator.” Journal of Environmental Engineering, 135(4), 191-202.

Patruno, L. E, Dorao, C. A., Svendsen, H. F., and Jakobsen, H. A. (2009). On the modeling of droplet-film interaction considering entrainment, deposition and breakage processes. Chemical Engineering Science, 64, 1362-1371.

Pougatch, K., Salcudean, M., Chan, E., and Knapper, B. (2009). “A two-fluid model of gas-assisted atomization including flow through the nozzle, phase inversion, and spray dispersion.” International Journal of Multiphase Flow, 35, 661-675.

Powell, J. C., Hallam, N. B., West, J. R., Forster, C. F., and Simms, J. (2000). “Factors which control bulk chlorine decay rates.” Water Research, 34(2000), 117-126.

Rook J. J. (1977). “Chlorination Reaction of Fulvic Acids in Natural Waters.” Environmental Science and Technology, 11(5), 478-482.

Rushton, B., England, G., Smith, D. (2007). “Proposed Guidelines for Monitoring Stormwater Gross Solids.” ASCE Conference Proceedings, (243)58, DOI:10.1061/40927(243)58.

Sansalone, J. J. and Buchberger, S. G. (1997). “Characterization of solid and metal element distributions in urban highway stormwater.” Water Science and Technology, 36, 155-160.

142

Sansalone, J. J., Hird, J. P., Cartledge, F. K., Tittlebaum, M. E. (2005). “Event-Based Stormwater Quality and Quantity Loadings from Elevation Urban Infrastructure Affected by Transportation.” Water Environment Research, 77(4), 348-365.

Sansalone, J. J. and Kim, J. (2008). “Transport of particulate matter fractions in urban source area pavement surface runoff.” Journal of Environmental Quality, 37,1883-1893.

Sansalone, J. J. and Pathapati, S.-S. “Particle dynamics in a hydrodynamic separator subject to transient rainfall-runoff.” Water Resour. Res., 45, doi:10.1029/2008WR007661.

Sansalone, J. J. and Ying, G. (2008). “Partitioning and granulometric distribution of metal leachate from urban traffic dry deposition particulate matter subject to acidic rainfall and runoff retention.” Water Research, 42(15), 4146-4162.

Shih, T.-H., Liou, W. W., Shabbir, A., Yang, Z., and Zhu, J. (1995). “A new κ-ε eddy-viscosity model for high Reynolds Number turbulent flows - model development and validation.” Computers Fluids, 24(3), 227-238.

Small, M. and DiToro D. (1979). “Stormwater treatment systems”, Journal of Environmental Engineering, 105(3), 557-569.

Taras, M. J. (1950). “Preliminary studies on the chlorine demand of specific chemical compounds.” Journal of the American Water Works Association, 42, 462-472.

Tchobanoglous, G., Burton, F. L., and Stensel, H. D. (2003). Wastewater Engineering: Treatment and Reuse, 4th Ed., McGraw-Hill, Boston.

USEPA (1973). “Microstraining and Disinfection of Combined Sewer Overflows – Phase II.” EPA-R2-73-124, Washington, DC.

USEPA (1980). “Approval of Alternate Water Pollutant Testing Procedure; Chemical Oxygen Demand (COD).” Federal Register, 454(78), 26811-26812.

USEPA (1984). “Ambient Water Quality for Bacteria.” EPA440/5-84-002, Washington, DC.

USEPA (2003). “CSO Disinfection Pilot Study: Spring Creek CSO Storage Facility Upgrade.” EPA/600/R-01/07, Cincinnati, Ohio.

USEPA (2004). “Guidelines for Water Reuse: Chapter 4.” EPA/625/R-04/108, Washington, DC.

USEPA (2010). “Water Quality Standards for the State of Florida‟s Lakes and Flowing Waters.” Federal Register, 75 (233), 75762-75807.

143

Warton, B., Heitz, A., Joll, C., Kagi, R. (2006). “A new method for calculation of the chlorine demand of natural and treated waters.” Water Research, 40(2006), 2877-2884.

White, F. Viscous Fluid Flow. 3rd Ed. McGraw-Hill, New-York, NY, 2006, 441.

Winward, G. P., Avery, L. M., Stephenson, T., and Jefferson, B. (2008). “Chlorine disinfection of grey water for reuse: Effect of organics and particles,” Water Research, 42(2008), 483-491.

Wols, B. A., Uijttewaal, W., Hofman, J., Rietveld, L., and van Dijk, J. (2010). “The weakness of a k-ε model compared to a large-eddy simulation for the prediction of UV dose distributions and disinfection.” Chemical Engineering Journal, 162, 528-536.

Wu, T., Sansalone, J. J. (2011). “Mass transfer to filter media from aqueous matrices II: Kinetics.” Journal of Environmental Engineering, in Review.

144

BIOGRAPHICAL SKETCH

Joshua Dickenson was born and raised in Jacksonville, FL and was

homeschooled from 3rd through 12th grades. In 2005, Joshua graduated Cum Laude

from the University of Florida with a Bachelor of Science in Mechanical Engineering.

After his bachelor‟s, Joshua spent 7 months in Bundibugyo, Uganda working with a

Christian non-governmental organization implementing water development projects.

While there he developed a passion for providing clean water to the poorest of the

world. In 2007, Joshua matriculated at the University of Florida in the Environmental

Engineering Sciences Department in a combination master‟s and doctoral program. In

May 2010, Joshua received a Master of Engineering degree from the University of

Florida. In May 2011, Joshua received Doctor of Philosophy Degree in Environmental

Engineering and Science from the University of Florida. Joshua pursues life to the

fullest, loves his family deeply, enjoys deep, intimate relationships, and ultimately owes

all to God and His Son, Jesus Christ.