Marine Debris Visual Identification Assessment

By

ZACHARY ANGELINI

Bachelor of Science, Environmental Engineering, University of New Hampshire, 2014

THESIS

Submitted to the University of New Hampshire

in Partial Fulfillment of

the Requirements for the Degree of

Master of Science

In

Civil Engineering

May, 2017

This thesis has been examined and approved by:

Dr. Nancy Kinner, Thesis Director

Professor of Civil Engineering and Environmental at the University of New Hampshire

Dr. Phil Ramsey, Committee Member

Lecturer of Mathematics and Statistics at the University of New Hampshire

Dr. Weiwei Mo, Committee Member

Assistant Professor of Civil and Environmental Engineering at the University of New

Hampshire

Original approval signatures are on file with the University of New Hampshire Graduate

School.

iii

ACKNOWLEDGEMENTS

This research was made possible by the support and funding of the University of

New Hampshire (UNH) Coastal Response Research Center (CRRC) and the

Commonwealth Scientific and Industrial Research Organization (CSIRO) Energy

Flagship. Dr. Nancy Kinner developed a relationship with CSIRO through her oil spill

response research. This relationship enabled me to intern with CSIRO for two, three month

periods. These internships, which included extensive field work for an environmental

baseline study of the Great Australian Bight, led to the research question investigated in

this thesis. The CRRC and CSIRO’s continued support during my thesis research was

crucial for the completion of the final project.

I would like to sincerely thank Dr. Nancy Kinner for her continued support and

expertise. I would also like to thank Dr. Kinner for providing me with the opportunity to

intern with CSIRO in 2014, as well Dr. Andrew Ross of CSIRO for supporting a second

internship in 2015. These were incredible opportunities to experience remote areas of the

world while contributing to research that can help protect and preserve Australia’s amazing

natural places.

I would like to thank all of my team members at CSIRO, including Dr. Andrew

Ross, Christine Trefry, Chris Dyt, Se Gong, Richard Kempton, Stephane Armand, and

Cameron White, as well as the University of Adelaide’s David Mckirdy and Alexander

Corrick. This team of research scientists all significantly assisted in gathering a robust

marine debris dataset during the environmental baseline study. This team, as well as Dr.

Denise Hardesty from CSIRO’s Ocean and Atmosphere Flagship and Dr. Nancy Wallace

iv

from the National Oceanic and Atmospheric Administration, contributed important thought

leadership that led to the focus of the thesis research.

I would like to thank Dr. Kenneth Fuld, former Dean of the College of Liberal Arts,

for providing his expertise on vision and visual perception. This insight was critical to

ensure a defendable design of experiment and explanation of results.

I would like to thank Justin Thibault, Masters Graduate of UNH’s Department of

Mathematics and Statistics, and Dr. Phil Ramsey, lecturer of Mathematics and Statistics at

UNH, for their immense support in the design of experiment and statistical data analysis

for this research.

I would like to thank my undergraduate research team: Andrew DeMeo, Pia

Marciano, Griffin Parodi, and Ben Sweeney. These students contributed significantly to

the execution of the thesis experiment.

I would also like to thank Dr. Weiwei Mo and Dr. Ramsey for serving on my thesis

committee and providing their feedback on this work.

Finally, I would like to sincerely thank my friends, family, and especially my

parents for their constant support in my academic career.

v

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ............................................................................................ iii

TABLE OF CONTENTS ................................................................................................. v

LIST OF TABLES ........................................................................................................... vi

LIST OF FIGURES ........................................................................................................ vii

LIST OF ACRONYMS ................................................................................................. viii

ABSTRACT ...................................................................................................................... ix

CHAPTER 1 ...................................................................................................................... 1

INTRODUCTION............................................................................................................. 1

CHAPTER 2 ...................................................................................................................... 5

METHODS ........................................................................................................................ 5

2.1 Experimental Set-up.................................................................................................. 5

2.2 Experimental Design ................................................................................................. 9

2.3 Experimental Runs with Human Subjects ................................................................ 9

2.4 Data Analysis .......................................................................................................... 13

CHAPTER 3 .................................................................................................................... 15

RESULTS AND DISCUSSION ..................................................................................... 15

3.1 Effects of Fragment Characteristics ........................................................................ 16

3.2 Effects of Beach Characteristics ............................................................................. 20

3.3 Effects of Observer Characteristics ........................................................................ 22

3.4 Models/Interactions................................................................................................. 23

CHAPTER 4 .................................................................................................................... 26

CONCLUSIONS AND RECOMMENDATIONS ........................................................ 26

4.1 Conclusions ............................................................................................................. 26

4.2 Recommendations for Future Experiments ............................................................ 28

REFERENCES ................................................................................................................ 29

APPENDICES ................................................................................................................. 33

Appendix A: Literature Review .................................................................................... 34

vi

LIST OF TABLES

Table 1: Munsell Color System Hue Symbol, Chroma, and Value Characterizations ...... 8 Table 2: Breakdown of Accurate Counts Based on Actual Number of Plastic Fragments

Present ............................................................................................................................... 19

Table 3: White Count Accuracy vs. Shell Density ........................................................... 21 Table 4: Clear Count Accuracy vs. Sand Color ................................................................ 22

vii

LIST OF FIGURES

Figure 1: Plan View of Beach Transects ............................................................................. 6

Figure 2: Observer Debris Logging Sheet ........................................................................ 12 Figure 3: Distribution of Under, Over, and Accurate Total Plastic Fragment Counts ..... 15 Figure 4: Distribution of Under, Over, and Accurate Blue Plastic Fragment Counts ...... 16 Figure 5: Distribution of Count Accuracy for Clear Plastic Fragments (a) and White Plastic

Fragments (b) .................................................................................................................... 18

Figure 6: Completion Time vs. Accuracy ......................................................................... 22 Figure 7: Observer Confidence ......................................................................................... 23 Figure 8: White Plastic Model Effect Summary Report ................................................... 24

Figure 9: Clear Plastic Model Effect Summary Report .................................................... 24 Figure 10: Blue Plastic Model Effect Summary Report ................................................... 25

viii

LIST OF ACRONYMS

CRRC: Coastal Response Research Center

CSIRO: Commonwealth Scientific and Industrial Research Organization

GAB: Great Australian Bight

MARPOL: International Convention for the Prevention of Pollution from Ships

NOAA: National Oceanic and Atmospheric Administration

RSP: Regional Seas Program

UNEP: United Nations Environment Program

UNH: University of New Hampshire

UV: UltraViolet

ix

ABSTRACT

MARINE DEBRIS VISUAL IDENTIFICATION ASSESSMENT

By

Zachary Angelini

University of New Hampshire, May 2017

Estimates of marine debris are often based on beach surveys conducted by

volunteers/citizen scientists. Few studies have documented the veracity of their

observations and the factors that may affect their accuracy. Our laboratory-scale

experiment identified potential sources of error associated with visual identification of

marine debris (1-2 cm long) during shoreline surveys of sand beaches. Characteristics of

the survey site (beach characteristics), observer (personal characteristics), and debris (color

and size) may be important factors to consider when analyzing data from shoreline surveys.

The results of this study show that the ability of individuals to accurately identify plastic

fragments depends on the plastic and sand color, and density of shell fragments, as well as

the time taken on the survey. Most suggestively, the high accuracy of blue plastic counts

(95%) and the under-counting of white (50%) and clear plastic fragments (55%) confirmed

the hypothesis that a significant amount of clear and white plastic fragments may be missed

during shoreline surveys.

1

CHAPTER 1

INTRODUCTION

Marine debris is a burgeoning global issue that threatens marine biota and

ecosystems, as well as human health, safety, and the economy (UNEP 2014). Globally,

plastics are the most common type of marine debris (Coe and Rogers, 1997; Derraik, 2002).

Ubiquitous in the marine environment, plastics have been documented in every part of the

ocean, including Arctic sea ice (UNEP, 2014). Jambeck et al. (2015) estimated that 275

million metric tons of plastic debris were produced by 192 coastal countries in 2010. Their

report suggested that 4.8 - 12.7 million metric tons of debris were input to the world’s

oceans that year. In 2014, Eriksen et al., (2014) estimated that 5.25 trillion plastic particles

were present in the oceans.

By 2025, the amount of plastic being input to the world’s oceans could increase by

an order of magnitude, unless significant steps are taken to improve waste management

infrastructure in coastal countries (Jambeck et al., 2015). In order to minimize marine

debris on a global scale, the problem needs to be recognized, understood, and addressed on

the international and local levels. In 2009, the United Nations Environment Program

(UNEP) developed the Regional Seas Programme (RSP) to respond to the marine plastic

debris issue and to collect and distribute information (Vegter et al., 2014). In total, there

are 18 regions recognized within the RSP. For the majority of these regions, minimal data

exist on the extent of plastic pollution. Furthermore, there are no data standards (i.e.,

standardized survey and reporting protocols, size classification, debris type classification),

and few countries acknowledge marine plastics as a problem (Vegter et al., 2014). Without

2

a scientifically-based global and local understanding of plastic pollution, it is likely that

there will be insufficient governmental and public motivation to make the changes urgently

needed.

One of the most common and well established methods for determining marine

plastic pollution is shoreline or “beach” surveying (Cheshire et al., 2009). Shoreline

surveys provide estimates of plastic loads in coastal waters (Thiel et al., 2013; Ribic et al.,

1992), and are far more economically viable than at-sea (ship-based) surveying (Morishige

et al., 2007). Shoreline surveys are often conducted by volunteers (citizen scientists) and

can be highly effective in increasing social awareness, a key factor in addressing the marine

plastics issue holistically (Vegter et al., 2014). With increased social awareness, the marine

plastics issue can be addressed in a variety of ways, including changes to consumer

behavior.

According to a recently published list of global research priorities to mitigate plastic

pollution impacts on wildlife, one of the greatest limitations to the quantification of marine

plastic debris loadings is the dependence on the human eye (Vegter et al., 2014). This issue

is not specific to citizen scientists. Hardesty et al. (2014) found that citizen scientist-

generated marine debris data are comparable to those collected by trained scientists.

In 2015, marine debris surveys were conducted as part of a collaboration between

the Commonwealth Scientific and Industrial Research Organization (CSIRO) (Perth,

Australia) and the University of New Hampshire (UNH) Coastal Response Research

Center (CRRC) (Durham, NH) along the coastline of the Great Australian Bight (GAB).

For the GAB survey, >64 kilometers of beach were surveyed and the most common debris

item found was small (1-2 cm) plastic fragments (Appendix B). Exposure to ambient

3

ultraviolet (UV) radiation, as well as the ocean’s physical, chemical, and biological

processes, results in plastic fragmentation and size reduction (UNEP, 2014). This leads to

the increased presence of meso-(0.5 - 2.5 cm) and micro-sized (0.1 - 0.5 cm) plastic

particles, which can be difficult to see.

Certain protocols exclude smaller size ranges from visual surveys and only log

debris greater than 2.5 cm, as this is the minimum disposal size allowed under MARPOL

(International Convention for the Prevention of Pollution from Ships) for ground shipment

of waste (Ribic et al., 1992). As awareness of meso- and micro-plastics has risen,

organizations, such as the U.S. National Oceanic and Atmospheric Administration

(NOAA), have developed a sieving method aimed at quantifying these plastic loads on

beaches (Lippiat et al., 2013). The NOAA method involves sediment core sampling,

sieving, and further laboratory analysis to determine meso- and micro-debris densities.

Alternatively, a protocol developed by CSIRO uses visual logging for all size

ranges starting at 0-1 cm, and continuing 1-2 cm, 2-4 cm, 8-16 cm and >16 cm (Hardesty

et al., 2014). This method was designed to allow citizen scientists to rapidly measure debris

in the meso- and micro-size range while simultaneously collecting macro debris data. If

the error associated with visual identification within the lower size ranges could be better

understood, then useful estimates of the uncertainty associated with these CSIRO data

could be generated. Through observations made during the GAB survey, a hypothesis was

developed that plastic debris characteristics (e.g., color); beach characteristics (e.g., sand

color, density of shell fragments); and observer characteristics (e.g., color-blindness,

height, age, sex, eyesight, survey speed) may have impacts on the accuracy of small plastic

fragment identification. Specifically, we hypothesized that under counting of white and

4

clear plastic fragments would occur on beaches with high shell densities, owing to visual

noise and reduced contrast between the plastic debris and its background surface.

Most current survey methodologies include quality assurance procedures

(Cheshire, 2009; Lippiatt et al., 2013; Hardesty et al., 2014), but they rely on sample re-

surveying by visual means, which does not address the fundamental errors in human visual

performance.

To study these issues, a laboratory experiment was conducted to directly quantify

the potential error associated with visual identification of plastic fragments (1-2 cm long)

during mock shoreline surveys. By quantifying the accuracy of counts associated with

different plastic fragment colors, beach and observer characteristics, and interactions

between these characteristics, visual estimates in the meso- and micro-size range plastics

can be better understood and the uncertainty associated with them further defined. This

understanding can be used to develop methods that allow for more reliable and rapid

measurements of meso- and micro-debris on sandy beaches. It can also be used to estimate

the uncertainty bounds associated with visual estimates made by citizen scientists, based

on known levels of accuracy associated with debris, beach, and/or surveyor characteristics.

5

CHAPTER 2

METHODS

Artificial beach transects were constructed containing known concentrations of

plastic fragments in a UNH environmental engineering laboratory. Volunteers were asked

to log plastic debris along the transects and the results were analyzed to determine the

impacts that debris characteristics, beach characteristics, and observer characteristics

have on the accuracy of visual identification.

2.1 Experimental Set-up

Four beach “transects” were constructed in the laboratory, each comprised of six 0.25

m2 black plastic trays of sand aligned in a row on the floor (Figure 1). Each transect was

located in its own section of the laboratory. Within each transect, plastic fragment and

beach characteristics varied.

6

Figure 1: Plan View of Beach Transects

The plastic fragments varied in chromaticity (appearing blue, white, clear

(translucent colorless)) and density (#/m2) in each tray. The colors were chosen based on

the preliminary GAB survey which found white and blue plastic fragments are the most

commonly occurring colors. Observations made during that survey led to our hypothesis

that blue fragments may be highly distinguishable and logged with greater accuracy, while

white and clear may be less distinguishable colors and logged with greater error. The 1-

2cm plastic fragments used in this study were provided by Poly Recovery (Portsmouth,

NH), a company that collects recycled plastics and processes them for post-consumer use.

They fell within the size range of 1-2 cm.

Each color fragment was characterized following the Munsell Color System

(Munsell, 1929) by hue symbol, chroma, and value (Table 1). Each hue is assigned a

number and letter. The letter stands for Red, Yellow, Green, Blue, Purple, or a combination

7

thereof. Numbers indicate the level of hue. Each hue and hue combination are divided into

four increments measured in differences of 2.5. Value represents a 0 to 10 scale of lightness

(0 = total blackness, 10 = total whiteness). Chroma, equivalent to saturation, is represented

from 0 to 12. These characteristics provide a standardized way to compare colors,

determine the perceptual distance between them, as well as measure color for repeatability

of the experiment.

The total plastic densities varied between 0 fragments/0.25 m2 to 18 fragments/0.25

m2. The latter resulted from a maximum of six fragments of each of the three colors in each

0.25 m2 tray. These densities represented the range encountered during the GAB survey.

The beach characteristics considered were sand color and shell density.

White/cream, gray, and brown sands were chosen as they are common hues. Hue symbol,

chroma, and value are shown in Table 1.

Three different shell densities were used to represent the common range found on

sand beaches: no shell, moderate shell (20 mL of shell fragments randomly spread over

0.25 m2), and high shell (80 mL randomly spread over 0.25 m2) densities. The naturally

weathered shells were collected from Wallis Sands Beach (Rye, NH). They were an

average of 1.1 cm (Range of 0.4 – 3.5 cm) by their longest length. Their average hue

symbol, chroma, and value are shown in Table 1.

8

Plastic Fragment Categories Hue Symbol Chroma Value

Blue 5B /6 6/

White 2.5Y /0 9/

Clear (on white/cream sand) 10Y /1 7/

Clear (on gray sand) 2.5Y /2 8/

Clear (on brown sand) 5Y /2 9/

Sand Categories Hue Symbol Chroma Value

White/Cream 10Y /1 9/

Gray 2.5Y /2 7/

Brown 5Y /4 9/

Shell Categories Hue Symbol Chroma Value

Average 5R /1 9/

Table 1: Munsell Color System Hue Symbol, Chroma, and Value Characterizations

Several parameters were controlled. All outside light was blocked from entering

the laboratory and the experimental area was continually illuminated with 6500K daylight-

balanced bulbs, in order to replicate the temperature of sunlight. The luminance reflected

off the trays of sand was an average of 206 cd/m2 (Range = 120 - 320 cd/m2). This value

is below the luminance measured in real daylight conditions (e.g., luminance on a cloudy

day is 1000 to 5000 cd/m2)(Beck, 1999), however, as long as sufficient time has been

provided for luminance adaptation, the human visual system calibrates itself to lower

luminance levels (Graham, 1965), eliminating the need to replicate the exact luminance of

daylight.

Upon visual inspection, the plastic fragments varied by specular reflectance as well

as color. In order to remove specular reflectance as a variable, the plastics were artificially

weathered for 30 minutes in a tumbler filled with sand. This resulted in fragments with

uniform visually reduced specular reflectance for all plastic fragments.

9

2.2 Experimental Design

The statistical software JMP Pro (Version 12; Cary, NC) was used to create a

custom design that allowed estimates of two-way interactions. This generated a design with

a default of 24 trays observed (6 trays in each of 4 transects) and no aliasing between any

of the main effects and the two-way interactions. No aliasing allows determination of

influence from all main effects and two-way interactions. The design automatically

randomized the values of each factor in each tray, as well the order of the trays. Each of

the 103 human subjects (observers) examined all of the trays in a random order.

2.3 Experimental Runs with Human Subjects

Observers were recruited from within the UNH community and included first

through fourth year undergraduate students, graduate students, faculty and staff. Upon

entering the laboratory, observers were asked to sign a letter of consent, following

protocols mandated by the UNH Institutional Review Board for the Protection of Human

Subjects in Research. Observers were given a written experimental protocol and were

asked to read it prior to receiving verbal instructions. They were asked to provide their age,

academic major, height, gender, and vision (near, far or normal sighted), use of corrective

lenses, and previous participation in marine debris surveys or beach clean-ups. Observers

conducted an online version of the Ishihara color blindness test (www.color-

blindness.com/ishihara-38-plates-cvd-test). The Ishihara test is usually conducted by an

eye care professional and the online version is problematic due to variations in computer

monitors (Flück, 2013). While the monitor we used was color calibrated and the test was

performed under the appropriate lighting, the color-blindness classification was still not

fully accurate. This test was conducted only as a preliminary test for color-blindness. After

10

completing the Ishihara test, observers recorded their score along a sliding scale with

categorical values of “None”, “Weak”, and “Moderate/Strong”. [N.B., Observers with

color blindness were not excluded from the study because observers used in real-world

marine debris projects are not screened for this trait.]

Each participant was verbally informed of the debris logging protocol. Observers

were shown NOAA’s debris photo manual (Lippiatt et al., 2013) for reference and were

told to specifically look for hard plastic fragments, greater than 1 cm in size, and with the

colors listed on their debris logging sheets (Figure 2) (Color categories replicated those

used by CSIRO). Observers were told to stand upright and walk beside the transects while

staying on a designated path located approximately 8 cm from the transects. Observers

were asked to log the color and number of plastics that they saw in each of the six trays

within each of the four transects. Observers were asked to be as accurate as possible,

logging only the plastics they saw. They were told to be careful not to log shell fragments

or any other natural debris items. After determining that observers understood the protocol,

their start time was recorded and they began logging plastic debris. (On average, observers

spent approximately 10 minutes completing the survey preparation procedures, allowing

sufficient time for visual calibration to the ~206 cd/m2 luminance before beginning the

survey) (Hecht, Haig, and Chase, 1937). After logging debris along each of the four

transects, total observer survey time was recorded and observers were asked to judge their

confidence in the accuracy of their results on a scale from 1-10 (10 = fully confident).

11

12

Figure 2: Observer Debris Logging Sheet

13

2.4 Data Analysis

Data distributions were generated for each of the variables (e.g., fragment color,

sand color). Each data point was representative of one tray, viewed by one observer. There

were 2472 counts (103 observers recording 24 trays each). From these distributions, initial

trends in the data were determined.

Factor and covariant significance on plastic counts were observed through

distributions, analysis of variance and simple modelling.

First, several covariates were removed due to a lack of information. “Color

blindness” was removed as a factor since only one of the observers, exhibited the condition

through the test provided. The covariant “age” was also removed, as only 17% of the

observers were between the ages of 30 and 62, whereas, 79% were between 18 and 24.

Similarly, “experience” and “amount of experience” were removed since there were so few

individuals who had participated in marine debris surveying or beach clean-ups. “Major”

(area of study) was also removed because the vast majority of observers were in the fields

of civil or environmental engineering.

Several factors were removed because they were not observed to have measurable

effects. These factors had a sufficient amount of data, however, did not indicate

significance (P > 0.3), nor did the interactions between them. The factors about the observer

removed included “height”, “class” (year or status at university), “corrected vision status”,

and “near, far, or normal sightedness”.

The factors for analysis were “plastic debris color”, “plastic debris density”, “sand

color”, and “shell density”, and their interactions, and “completion time”. These factors

14

were determined to have a potentially significant impact on the counts of plastic debris and

the observers’ accuracy for those counts and were selected for further analysis.

15

CHAPTER 3

RESULTS AND DISCUSSION

73% of the total plastic fragment counts were under-counts, 8% were over-counts,

and 19% were accurate (Figure 3). All over-counts were grouped into a single category due

to the occurrence of infinite numbers (i.e., recording any amount of plastic fragments when

none were present).

Figure 3: Distribution of Under, Over, and Accurate Total Plastic Fragment Counts

16

Data peaks at “Accurate”, “50%”, and “0%” are representative of observers

recording exactly the number of plastic fragments present, exactly half of the plastic

fragments present, and none of the plastic fragments present, respectively, in each tray. The

median of the observers’ counts was a 50% under-count.

3.1 Effects of Fragment Characteristics

Consistent with our hypothesis, blue plastic fragments were logged with high

accuracy (Figure 4), and therefore, were not a major contributor to the inaccuracy of the

total plastic fragment counts.

Figure 4: Distribution of Under, Over, and Accurate Blue Plastic Fragment Counts

Approximately 3.2% of blue plastic fragment counts were under-counts, 93.9%

were accurate counts, and 2.9% were over-counts.

Also consistent with our hypothesis from the GAB survey, the accuracy of clear

(Figure 5a) and white (Figure 5b) plastic counts were significantly different than that of

17

blue. Using t-tests, the differences between the accuracy of blue and white plastic counts

and blue and clear counts were highly significant (P-values < 0.0001). There was no

significant difference between the number of accurate white plastic counts and accurate

clear plastic counts (P-value =0.6018). There was also no significant difference between

the number of white plastic under-counts and clear plastic under-counts (P-value = 0.1617).

However, there was a significant difference between the number of white plastic over-

counts and clear plastic over-counts (P-value <0.001).

18

Figure 5: Distribution of Count Accuracy for Clear Plastic Fragments (a) and White

Plastic Fragments (b)

(a

)

(b

)

19

Approximately 55.2% of clear plastic fragment counts were under-counts, 42.4%

were accurate counts, and 2.4% were over-counts. A high percentage of under-counting

was expected due to the lack of hue contrast and resulting minimal perceptual distance

between the clear plastic fragments and sand (as determined by the Munsell Color System).

A low percentage of over counting was also expected due to the greater perceptual distance

between clear plastic fragments and shell fragments, minimizing the probability of

mistaken identity.

Approximately 50.5% of white plastic counts were under-counts, 42.2% were

accurate, and 7.3% were over-counts. Compared to clear plastic counts, a greater

percentage of white plastic counts were over-counts. It is probable that mistaken identity

between white plastic and white shell caused this. Consistent with our hypothesis from the

GAB survey, it is probable that a large amount of the under-counts for white plastic

occurred due to mistaken identity between plastic fragments and shell as well. While the

perceptual distance between white plastic and sand is relatively large (as determined by the

Munsell Color System), the perceptual distance between the white plastic fragments and

white shell is less.

The accurate counts for clear and white plastic fragments can be further explained

by the actual amounts of plastic fragments present (Table 2).

Plastic Color Number of Plastic Fragments Present

0 2 4 6

Clear 86% 5% 1% 8%

White 87% 1% 5% 7%

Table 2: Breakdown of Accurate Counts Based on Actual Number of Plastic Fragments

Present

20

The majority of accurate white and clear plastic counts (86% and 87%,

respectively) occurred when observers counted trays with zero plastic fragments present

(true zero counts); only 14% of all accurate clear plastic counts and 13% of all accurate

white plastic counts occurred when there were plastic fragments present.

3.2 Effects of Beach Characteristics

Strong trends were only observed between beach characteristics (i.e., shell density,

sand color) and the accuracies of white and clear plastic counts.

The majority of inaccurate white plastic counts were under-counts (Table 3). The

difference between the accuracy of white plastic counts in zero shell, moderate shell, and

high shell density conditions resulted in a p-value = 0.0025 (t-test). This indicated that the

accuracy associated with at least one of the shell density conditions was significantly

different. Further analysis (t-tests) showed that the accuracy of counts in high shell density

conditions was significantly different from zero shell and moderate shell density conditions

(P-value =0.0017 and P-value =0.0015, respectively). There did not appear to be a

significant difference between the accuracy of counts made in zero shell and moderate shell

density conditions (P-value =0.8790). Overall, significantly less accurate counts were

made for the high shell density condition than in the moderate and zero shell densities.

Observers’ counts of white plastic fragments were less accurate when more shell was

present, with a tendency towards under-counting. This result may be partially explained by

signal detection theory (Swets, Tanner, and Birdsall, 1961) in the following manner: As a

result of decreasing the signal (plastic fragments)-to-noise (shell fragments) ratio, through

increased shell densities, observer bias may have tended towards the expectation of shell

rather than plastic fragments, leading to greater under-counting.

21

Shell Density Categories Measurement Accuracy

Accurate counts Inaccurate counts (Under/over)

Zero shell 51% 49% (48/1)

Moderate shell 43% 57% (47/10)

High shell 33% 67% (55/12)

Table 3: White Count Accuracy vs. Shell Density

There also was a correlation between the accuracy of clear plastic counts and sand

color (Table 4). Using t-tests, the difference between the accuracy of clear plastic counts

for each of the sand colors resulted in P-value < 0.0001. Further analysis showed that

accuracy of counts was significantly different between all three colors (each of the P-values

were less than 0.0122). Brown sand resulted in the most accurate counts, while gray sand

resulted in the least accurate. One possible explanation for this result is that the clear plastic

fragments may have reflected ambient light in a way that, to the observers, their spectral

properties appeared closer to grey sand than the other sand colors.

22

Sand Color Measurement Accuracy

Accurate counts Inaccurate counts (Under/over)

Gray 39% 61% (58/3)

White/cream 40% 60% (56/4)

Brown 46% 54% (52/2)

Table 4: Clear Count Accuracy vs. Sand Color

3.3 Effects of Observer Characteristics

As completion time increased, so did median observer accuracy. The longer

observers took to log data, generally, the more accurate their counts (Figure 6). However,

regardless of how long observers spent observing the transects, their median results never

reach 100% accuracy (average observation time = 12.3 min).

Figure 6: Completion Time vs. Accuracy

23

The median observer confidence (Figure 7) in the accuracy of their results (6 out of

10) was surprisingly high when compared to the low percentage of total accurate counts

(19%). This may indicate that many observers would not have bent down to further inspect

the sand. This needs further investigation in an experiment where bending down is allowed.

Figure 7: Observer Confidence

3.4 Models/Interactions

In order to determine the most significant factors and interactions between factors

that affected overall counts, models were generated for white, clear, and blue plastic counts.

The distributions used in the generalized linear models were determined from observing

the distributions of counts of each color plastic. The generalized models, using Poisson

distributions with log link functions, characterized the variations in observations. To

determine which factors were most significant to the variability in counts of plastic, the

models considered the amount actually present. When first run, not all factors were

estimated indicating that reducing the models was necessary. A p-value cutoff of 0.30 was

used in addition to indications of no significance from other analyses (e.g., analysis of

24

variance). The models were only reduced to the point where all of the remaining factors

were estimated. Each step of the reduction removed a single factor or interaction.

Figure 8: White Plastic Model Effect Summary Report

For white plastic counts, the most significant factors were the interaction between

white plastic density and shell density, blue plastic density alone, and shell density alone

(Figure 8). The interaction between white plastic density and shell density was likely a

significant factor due to the increased potential for mistaken identity. Blue plastic density

and shell density were likely significant factors as they contributed to overall visual noise,

distracting the observer from identifying the white plastic fragments present.

Figure 9: Clear Plastic Model Effect Summary Report

For clear plastic counts, the most significant factors were completion time and the

interaction between clear plastic density and sand color (Figure 9). Median accuracy and

completion time were directly related. It is likely that observers who spent more time

searching the trays increased their opportunity to identify clear plastic fragments. There

25

was a significant difference between the accuracy of counts made in trays with different

sand colors. It makes sense that the interaction between clear plastic density and sand color

was a significant factor. Clear plastic was best detected when the observer took more time

and the sand was brown.

Figure 10: Blue Plastic Model Effect Summary Report

For blue plastic counts, the most significant factors were the interaction between

blue plastic density and shell density and shell density alone (Figure 11). Both factors

contribute to overall visual noise, potentially distracting the observer from identifying the

blue plastic fragments present. Still, the overall high accuracy of blue plastic counts (95%),

diminished the importance of any of the above factors.

26

CHAPTER 4

CONCLUSIONS AND RECOMMENDATIONS

4.1 Conclusions

To the best of our knowledge, this is the first study on the effects these

characteristics have on the accuracy of beach surveys. The high accuracy of blue plastic

counts and the under-counting of white and clear plastic counts confirmed the hypothesis

that a significant amount of clear and white plastic fragments may be missed during

shoreline surveys.

The accuracy of white plastic counts was most significantly affected by increasing

shell densities due to observer bias and the low contrast between white plastic fragments

and white shell fragments. The majority of the time, white plastic fragments were mistaken

as shell (under-counts), with fewer circumstances of shell being mistaken for white plastic

fragments (over-counts). The accuracy of clear plastic counts was most significantly

affected by the color of the sand; the clear fragments contrasting best with brown sand.

There was also a correlation between higher median accuracies and slower survey

speeds. However, despite speed, surveyors never reached full accuracy.

The results of this study demonstrate that visual identification of lower size ranges

during beach surveys should be coupled with advanced beach (sand and shell) and debris

(color) characterization. Specifically, characteristics including shell cover and sand color

should be noted, as well as debris color for each item identified. If certain debris colors can

27

be further associated with known degrees of under counting in certain beach conditions,

then visual estimates can be adjusted to account for this.

The results also suggest that pre-survey advanced identity training should be further

investigated as a means of minimizing under- and over-counting. This training could

involve survey leaders setting up visual calibration plots of sand, which include a variety

of plastic and natural debris items (i.e. shells). Before beginning the survey, observers

could log the plastic debris in the calibration plot. After logging, the results could be

discussed and distinguishing factors between plastic debris and natural debris identified.

The variables tested in this experiment represent only a portion of possible beach,

observer, or debris characteristics that may be encountered in shoreline surveys. There may

well be other variables (i.e. weather/daylight conditions) that will result in drastic

differences in accuracy. By further understanding and identifying these variables, marine

debris estimates can be put in perspective and potentially improved. New survey techniques

can work to address these issues in order to obtain more accurate meso- and micro-plastic

debris estimates from visual surveys.

By further understanding the accuracy of visual observations of marine debris and

addressing it, the huge resource potential that citizen scientists represent for collecting data

on small plastic fragments can be maximized, resulting in the most accurate data possible.

28

4.2 Recommendations for Future Experiments

Future experiments could increase variation in shell density, as well as shell size

and color. In addition, other natural debris (e.g., seaweed, wood, crustacean remains), could

allow for a more realistic representation of beach characteristics.

Additional colors of plastics should be studied. For example, there are ten color

categories used by CSIRO (Black, Blue/Purple, Brown, Clear/Translucent, White, Green,

Grey/Silver, Orange, Red/Pink, Yellow). It would also be prudent to include other sizes of

plastic fragments, since the issues described in this study may apply.

This study did not obtain a sufficient number of observers with the color-blindness

trait to analyze their accuracy compared to observers lacking this trait. Therefore,

additional studies should be conducted with larger sample sizes in order to determine if

this trait significantly impacts survey accuracy, either positively or negatively.

Finally, training procedures to minimize mistaken identity between small plastic

fragments and shells should be developed and tested to improve accurate fragment

detection.

29

REFERENCES

Beck, A., Werner, K., Gross, O., Fricke, J. 1999. Making Better Use of Natural Light With a

Light-redirecting Double Glazing System. Solar Energy, 66(3), 215-21.

Cheshire, A.C., Adler, E., Barbière, J., Cohen, Y., Evans, S., Jarayabhand, S., Jeftic, L., Jung,

R.T., Kinsey, S., Kusui, E.T., Lavine, I., Manyara, P., Oosterbaan, L., Pereira, M.A.,

Sheavly, S., Tkalin, A., Varadarajan, S., Wenneker, B., Westphalen, G., 2009.

UNEP/IOC Guidelines on Survey and Monitoring of Marine Litter. UNEP Regional

Seas Reports and Studies (120 pp.,

http://www.unep.org/regionalseas/marinelitter/publications/docs/Marine_Litter_Surve

y_and_Monitoring_Guidelines.pdf).

Coe, J. M., Rogers, D. (Eds.), 1997. Marine Debris Sources, Impacts, and Solution. Springer-

Verlag, New York.

Derraik, J. G. B., 2002. The pollution of the marine environment by plastic debris: a review.

Mar. Pollut. Bull. 44, 842–852.

Eriksen, M., Lebreton, L. C. M., Carson, H. S., Thiel, M., Moore, C.J., Borerro, J. C., 2014.

Plastic pollution in the world's oceans: More than 5 trillion plastic pieces weighing

over 250,000 tons afloat at sea. PLoS One 9, e111913.

Flück, D. Color Blind Essentials. Colblindor. 2013.

30

Flück, Daniel. 2013. "Ishihara 38 Plates CVD Test | Colblindor." Colblindor. Wordpress. n.d.

Web. 7-11 Mar. 2016. http://www.color-blindness.com/ishihara-38-plates-cvd-test/

Graham, C. H. Bartlett, N. R., Brown, J. L., Hsia, Y., Mueller, C. G., Riggs, L. A. 1965.

Vision and Visual Perception. New York (NY): John Wiley & Sons. Chapter 4, Some

Fundamental Data; p 76.

Hardesty, B. D., Wilcox, C., Lawson, T. J., Lansdell, M., Velde, T . 2014. Understanding the

effects of marine debris on wildlife. Final Report to Earthwatch Australia. CSIRO

Hecht, S., Haig, C., and Chase, A. M., The influence of light-adaptation on subsequent dark-

adaptation of the eye, J. Gen. Physiol. 20, 831 (1937).

Jambeck, J. R., Perryman, M., Geyer, R., Andrady, A., Wilcox, C., Siegler, T. R., Narayan,

R., Law, K. L., 2015. Plastic waste inputs from land into the ocean. Science. 347

(6223), 768-771.

Lippiatt, S., Opfer, S., Arthur, C., 2013. Marine debris monitoring and assessment:

recommendations for monitoring debris trends in the marine environment. NOAA

Technical Memorandum NOS-OR&R-46, pp. 1–88.

31

Morishige, C., Donohue, M. J., Flint, E., Swenson, C., Woolaway, C. 2007. Factors affecting

marine debris deposition at French Frigate Shoals, northwestern Hawaiian Islands

Marine National Monument, 1990−2006. Mar Pollut Bull 54: 1162−1169.

Munsell, A.H. Munsell book of color, defining, explaining, and illustrating the fundamental

characteristics of color / a revision and extension of "The atlas of the Munsell color

system". Munsell Color Company, Inc. Baltimore. 1929.

Swets, J.A., Tanner, W.P., and Birdsall, T.G. Decision processes in

perception. Psychological Review, 68, 301-340, 1961.

Ribic, C.A., Dixon, T.R. and Vining, I. (1992). Marine Debris Survey Manual. NOAA

Technical Report NMFS 108.

United Nations Environmental Programme, 2014. UNEP Year Book 2014. Emerging issues

in our global environment. In: United Nationals Environment Programme. Nairobi,

Kenya, 75pp.

Vegter, A. C., Barletta, M., Beck, C. Borrero, J., Burton, H., Campbell, M. L. M. F. Costa,

M. Eriksen, C. Eriksson, A. Estrades, K. V. K. Gilardi, B. D. Hardesty, J. A. Ivar do

Sul, J. L. Lavers, B. Lazar, L. Lebreton, W. J. Nichols, C. A. Ribic, P. G. Ryan, Q. A.

Schuyler, S. D. A. Smith, H. Takada, K. A. Townsend, C. C. C. Wabnitz, C. Wilcox,

32

L. C. Young, M. Hamann. 2014. Global research priorities to mitigate plastic

pollution impacts on marine wildlife. Endangered Species Research, 25(3), 225-247.

33

APPENDICES

34

Appendix A: Literature Review

Magnitude of Marine Debris Problem

Marine plastic pollution is an urgent and wide-spread issue. A recent report

estimated that 275 million metric tons of plastic debris were produced by 192 coastal

countries in 2010, with 4.8 - 12.7 million metric tons now in the world’s oceans (Jambeck

et. al, 2015). By 2025, this amount could increase by an order of magnitude, unless

significant steps are taken to improve waste management infrastructure in coastal countries

(Jambeck et. al, 2015).

Environmental, Social, and Economic Impacts of Marine Debris

The staggering amount of plastic debris currently in the world’s oceans has resulted

in a wide range of environmental, social, and economic damages. For example, at least 170

marine species have been documented ingesting plastics. (Carr 1987, Laist 1987, Bjorndal

et al. 1994, Derraik 2002, Ceccarelli 2009, Boerger et al. 2010, Jacobsen et al. 2010, Baulch

& Perry 2012, Fossi et al. 2012, Schuyler et al. 2012, Besseling et al. 2013). The species-

level impacts of this occurrence include gut perforation, gut impaction, dietary dilution,

interference with development, and toxin introduction. (Ryan 1988a, Bjorndal et al. 1994,

McCauley & Bjorndal 1999, Mader 2006, Teuten et al. 2009, van Franeker et al. 2011,

Gray et al. 2012, Tanaka et al. 2013).

Studies have shown that plastics not only can contain harmful plasticizers

introduced during manufacturing of the virgin material (Meeker et al. 2009), but once

present in the marine environment, they can adsorb heavy metals such as Copper and Zinc

(commonly leached into seawater from antifouling paint used in shipping) (Brennecke et

35

al. 2016). The measured heavy metal concentrations in marine plastics can reach toxic

levels, far above concentrations commonly found in the water column (Brennecke et al.

2016). Also, the hydrophobicity of persistent organic pollutants (POPs) enables their

adsorption on meso-/microplastic fragments and leads to concentrations on these plastics

that are multiple orders of magnitude higher than concentrations in sea water (Andrady,

2011).

Marine plastics also have a greater tendency to become bio-fouled by introduced

species (Whitehead et al. 2011), than by native species (Wyatt et al. 2005, Glasby et al.

2007, Tamburri et al. 2008). This characteristic, coupled with the durability and persistence

of plastic in the marine environment, results in the significant prospect of increased non-

native species transportation (Vegter et al. 2014).

Humans are not exempt as a species effected by marine debris. Social and economic

damages occur via disruption to shipping, tourism, and fisheries. In 2011, the total yearly

cost of debris-related damage to marine industries in the Asia-Pacific rim countries was

estimated to be US $1.26 billion per year (Hall 2000, McIlgorm et al. 2011). In South

Korea, a single period of intense rainfall during July 2011 resulted in significantly

increased coastal debris, leading to a 63% decrease in tourism (Jang et al. 2014). The

damages caused by fouling of water bodies go beyond the economics of social well-being

and include an increase in emotional stress, and loss of relaxation, insight, self-reflection

and creativity (White et. al, 2010).

Addressing the Marine Debris Problem

In order to improve waste management on a global scale, the problem needs to be

recognized, understood, and addressed on both international and local levels. The United

36

Nations Environment Program (UNEP) developed the Regional Seas Programme (RSP) to

respond to the marine debris issue (UNEP 2009) and to collect and distribute information.

(Vegter et al. 2014). In total, there are 18 regions recognized within the RSP, but only 12

have participated in UNEP supported marine debris activities (Vegter et al. 2014). The

majority of these regions have minimal data on the extent of the issue, do not have data

standards, and few acknowledge marine debris as a problem (Vegter et al. 2014).Without

a scientifically based global understanding of plastic pollution, there will be insufficient

governmental and public drive to make the changes urgently needed to address this

problem.

Marine Debris Surveying

One of the most common and established methods for marine debris data collection

is shoreline or “beach” surveying (Cheshire et al. 2009). Shoreline surveys are capable of

providing estimations of debris loads in coastal waters (Thiel et al. 2013), and are far more

economically viable than at-sea surveying (Morishige et al. 2007). Shoreline surveys are

often conducted by volunteers or “citizen scientists” and can be highly effective in

increasing social awareness of the issue (Vegter et al. 2014).

A recent study in Australia analyzed the difference between volunteer or “citizen

science” data from shoreline surveys and data collected by scientists from the

Commonwealth Scientific and Industrial Research Organization along the same shoreline

areas. The study found there to be no significant difference between the two datasets

(Hardesty et al. 2014). This highlighted the effectiveness of involving citizen scientists in

data collection. This effectiveness combined with the social and economic benefits of

involving volunteers, make beach surveys conducted by citizen scientists an important

37

method in addressing the marine debris issue. However, according to a recently published

list of global research priorities to mitigate plastic pollution impacts on wildlife, one of the

greatest limitations to the quantification of marine plastic debris loadings is the dependence

on the human eye (Vegter et al. 2014). The goal of this research is to understand the

potential error associated with visual identification of plastic fragments during shoreline

surveys.

38

References:

Andrady, A.L., 2011. Microplastics in the Marine Environment. Marine Pollution Bulletin.

62(8) 1596-1605.

Auman HJ, Woehler EJ, Riddle MJ, Burton H (2004) First evidence of ingestion of plastic

debris by seabirds at subAntarctic Heard Island. Mar Ornithol 32: 105−106.

Baulch S, Perry C (2012) A sea of plastic: evaluating the impacts of marine debris on

cetaceans. Paper SC/64/E10, 64th Meeting of the International Whaling

Commission Science Committee, Panama City.

Besseling E, Wegner A, Foekema EM, Van Den HeuvelGreve MJ, Koelmans AA (2013)

Effects of microplastic on fitness and PCB bioaccumulation by the lugworm

Arenicola marina (L.). Environ Sci Technol 47: 593−600.

Bjorndal KA, Bolten AB, Lagueux CJ (1994) Ingestion of marine debris by juvenile sea

turtles in coastal Florida habitats. Mar Pollut Bull 28: 154−158.

Boerger CM, Lattin GL, Moore SL, Moore CJ (2010) Plastic ingestion by planktivorous

fishes in the North Pacific Central Gyre. Mar Pollut Bull 60: 2275−227.

Brennecke D, Duarte B, Paiva F, Caçador I and Canning-Clode J, 2016. Microplastics as

vector for heavy metal contamination from the marine environment. Estuarine,

Coastal and Shelf Science, doi:10.1016/j.ecss.2015.1012.1003.

Carr A (1987) Impact of nondegradable marine debris on the ecology and survival outlook

of sea turtles. Mar Pollut Bull 18: 352−356.

Ceccarelli D (2009) Impacts of plastic debris on Australian marine wildlife. Report by

C&R Consulting to the Australian Government Department of the Environment,

Water, Heritage and the Arts, Canberra, p 1−7.

Cheshire A, Adler E, Barbière J, Cohen Y and others (2009) UNEP/IOC guidelines on

survey and monitoring of marine litter. UNEP Regional Seas Reports and Studies,

No. 186, IOC Technical Series No. 83.

Derraik JGB (2002) The pollution of the marine environment by plastic debris: a review.

Mar Pollut Bull 44: 842−852.

Fossi MC, Panti C, Guerranti C, Coppola D, Giannetti M, Marsili L, Minutoli R (2012)

Are baleen whales exposed to the threat of microplastics? A case study of the

Mediterranean fin whale (Balaenoptera physalus). Mar Pollut Bull 64: 2374−2379.

39

Glasby TM, Connell SD, Holloway MG, Hewitt CL (2007) Nonindigenous biota on

artificial structures: Could habitat creation facilitate biological invasions? Mar Biol

151: 887−895.

Gray H, Lattin GL, Moore CJ (2012) Incidence, mass and variety of plastics ingested by

laysan (Phoebastria im mutabilis) and black-footed albatrosses (P. nigripes)

recovered as by-catch in the North Pacific Ocean. Mar Pollut Bull 64: 2190−2192.

Hall, K. (2000) Impacts of marine debris and oil: economic and social costs to coastal

communities. KIMO, Shetland Islands Council, Lerwick.

Hardesty, B. D., Wilcox, C., Lawson, T. J., Lansdell, M., Velde, T . 2014. Understanding

the effects of marine debris on wildlife. Final Report to Earthwatch Australia.

CSIRO.

Jacobsen JK, Massey L, Gulland F (2010) Fatal ingestion of floating net debris by two

sperm whales (Physeter macrocephalus). Mar Pollut Bull 60: 765−767.

Jambeck, J. R., Perryman, M., Geyer, R., Andrady, A., Wilcox, C., Siegler, T. R., Narayan,

R., Law, K. L., 2015. Plastic waste inputs from land into the ocean. Science. 347

(6223), 768-771.

Jang YC, Hong S, Lee J, et al. 2014. Estimation of lost tourism rev-enue in Geoje Island

from the 2011 marine debris pollution event in South Korea. Mar Pollut Bull 81:

49–54.

Laist DW (1987) Overview of the biological effects of lost and discarded plastic debris in

the marine environment. Mar Pollut Bull 18: 319−326.

McCauley SJ, Bjorndal KA (1999) Conservation implications of dietary dilution from

debris ingestion: sublethal effects in post-hatchling loggerhead sea turtles. Conserv

Biol 13: 925−929.

McIlgorm A, Campbell HF, Rule MJ (2011) The economic cost and control of marine

debris damage in the AsiaPacific region. Ocean Coast Manage 54: 643−65.

Meeker JD, Sathyanarayana S, Swan SH (2009) Phthalates and other additives in plastics:

human exposure and associated health outcomes. Philos Trans R Soc Lond B Biol

Sci 364: 2097−211.

Morishige C, Donohue MJ, Flint E, Swenson C, Woolaway C (2007) Factors affecting

marine debris deposition at French Frigate Shoals, northwestern Hawaiian Islands

Marine National Monument, 1990−2006. Mar Pollut Bull 54: 1162−1169.

Ryan PG (1988a) Effects of ingested plastic on seabird feeding: evidence from chickens.

Mar Pollut Bull 19: 125−12.

40

Schuyler Q, Hardesty BD, Wilcox C, Townsend K (2012) To eat or not to eat? Debris

selectivity by marine turtles. PLoS ONE 7: e40884.

Tamburri M, Luckenbach M, Breitburg D, Bonniwell S (2008) Settlement of Crassostrea

ariakensis larvae: effects of substrate, biofilms, sediment and adult chemical cues.

J Shellfish Res 27: 601−608.

Tanaka K, Takada H, Yamashita R, Mizukawa K, Fukuwaka MA, Watanuki Y (2013)

Accumulation of plastic-derived chemicals in tissues of seabirds ingesting marine

plastics. Mar Pollut Bull 69: 219−22.

United Nations Environmental Programme, 2014. UNEP Year Book 2014. Emerging

issues in our global environment. In: United Nationals Environment Programme.

Nairobi, Kenya, 75pp.

van Franeker JA, Blaize C, Danielsen J, Fairclough K and others (2011) Monitoring plastic

ingestion by the northern fulmar Fulmarus glacialis in the North Sea. Environ Pollut

159: 2609−2615.

Vegter, A. C., Barletta, M., Beck, C. Borrero, J., Burton, H., Campbell, M. L. M. F. Costa,

M. Eriksen, C. Eriksson, A. Estrades, K. V. K. Gilardi, B. D. Hardesty, J. A. Ivar

do Sul, J. L. Lavers, B. Lazar, L. Lebreton, W. J. Nichols, C. A. Ribic, P. G. Ryan,

Q. A. Schuyler, S. D. A. Smith, H. Takada, K. A. Townsend, C. C. C. Wabnitz, C.

Wilcox, L. C. Young, M. Hamann. 2014. Global research priorities to mitigate

plastic pollution impacts on marine wildlife. Endangered Species Research, 25(3),

225-247.

White M, Smith A, Humphryes K, Pahl S, Snelling D, Depledge M (2010) Blue space: the

importance of water for preference, affect, and restorativeness ratings of natural

and built scenes. J Environ Psychol 30: 482−493.

Whitehead TO, Biccard A, Griffiths CL (2011) South African pelagic goose barnacles

(Cirripedia thoracica): substratum preferences and influence of plastic debris on

abundance and distribution. Crustaceana 84: 635−649.

Wyatt ASJ, Hewitt CL, Walker DI, Ward TJ (2005) Marine introductions in the Shark Bay

World Heritage Property, Western Australia: a preliminary assessment. Divers

Distrib 11:33-44.