e ologi ally relevant quantitative ethods for ...the gbr. as a result of the governments evolving...

1

ECOLOGICALLY RELEVANT, QUANTITATIVE

METHODS FOR MEASURING PESTICIDE

REDUCTION FOR THE GREAT BARRIER REEF

A THESIS SUBMITTED IN FULFILMENT OF THE REQUIREMENT FOR THE DEGREE OF

MASTERS OF APPLIED SCIENCE (RESEARCH)

FACULTY OF SCIENCE AND ENGINEERING,

SCHOOL OF MATHEMATICAL SCIENCES,

QUEENSLAND UNIVERSITY OF TECHNOLOGY

Rachael Anne Smith

Bachelor of Environmental Biology (Honours) Doctor of Philosophy

University of Technology, Sydney

Principal supervisor: Kerrie Mengersen Associate supervisor: Kate Helmstedt

External co-supervisor: Michael Warne

2018

2

i Keywords

Pollutant Loads, Mixtures, Toxic Equivalency Factor, Relative Potency, Multiple species, Great Barrier

Reef, Photosystem II Herbicides, multisubstance Potentiallly Affected Fraction, Reef Water Quality

Protection Plan, Reef 2050 Water Quality Improvement Plan, agriculture, coral, seagrass, pesticide

targets, global curve fit.

3

ii Abstract

Poor water quality transported to the Great Barrier Reef (GBR) from agricultural lands in Queensland

catchments is impacting the health and resilience of the Reef’s ecosystems. In an effort to halt

further decline of GBR ecosystems, the Australian and Queensland governments implemented

management plans with water quality improvement targets for reducing nutrients, sediments and

pesticides (primarily sourced from agriculture) transported to the Reef; the Reef Water Quality

Protection Plans (2003, 2009, 2013) and more recently the Reef 2050 Water Quality Improvement

Plan (2017-2022). Nested within these plans are monitoring and evaluation programs that measure

the progress made towards achieving these targets, ultimately assessing the improvement in water

quality gained from improved land management practices. This thesis aims to develop quantitative

methods for measuring the progress towards the targets set for reducing pesticides transported to

the GBR.

As a result of the governments’ evolving and adaptive management strategy, the pesticide targets

have advanced since their introduction in 2009 to be more ecologically relevant. The original

pesticide target was to reduce the loads (tonnes per year discharged to the Great Barrier Reef) of

five priority photosystem II herbicides by 50%. However, this target, and the methods used to

measure progress towards achieving these targets, did not account for two fundamental aspects of

ecological risk: (i) ecological risk is dependent on the concentration and temporal and spatial

exposure of a pollutant which is not measured by the annual loads of pesticides; and (ii) individual

pesticides have varying levels of toxicity and should not be quantified with equal weighting. In an

effort to develop more ecologically relevant pesticide targets, a toxicity-based load (toxic load)

approach was developed that could measure progress against the first revision of the pesticide

targets (Reef Water Quality Protection Plan, 2013), and is presented in this thesis. The method is

adapted from well-known mixture toxicity methods, the toxic equivalency factor (TEF) and relative

potency (ReP) methods, that allows the user to simply weight each of the herbicide loads based on

their toxicity (the TEF), thus transforming them to a relative toxicity scale and allowing them to be

aggregated to a total toxic load. It was important that the TEF was derived from the RePs of multiple

species to ensure the distribution of responses in the ecological community was represented. Two

novel tests were therefore introduced that determined a TEF representative of the distribution of

RePs and that produced the most robust and environmentally relevant toxic loads. In doing so, a

reduction in the total toxic load would mean a commensurate reduction in ecosystem impact. The

method was validated with pesticide monitoring data collected from the major catchments

discharging to the GBR.

4

In 2017, the Draft Reef 2050 Water Quality Improvement Plan was released with updated pesticide

targets. The targets changed from a load-based measure, to a species protection value (i.e. at least

99% species protection for GBR waterbodies), similar to what is used in the National, State and GBR

water quality guidelines. The change was a further advancement towards ensuring the targets were

environmentally relevant. However, it meant new methods were needed to quantify progress

towards the target. The new method needed to represent the response of an ecological community

(i.e. multiple species), the cumulative effects from multiple pesticides, and expand to also include

other pesticides, not just the PSII herbicides. For these reasons, the multisubstance Potentially

Affected Fraction (msPAF) method was investigated. The results found that, through some

modifications of the method, the percent of species affected by mixtures of pesticides could be

estimated that are both robust and environmentally relevant. These results are presented in this

thesis.

5

TABLE OF CONTENTS

i Keywords ........................................................................................................................................... 2

ii Abstract ............................................................................................................................................ 3

iii List of Figures................................................................................................................................... 8

iv List of Tables .................................................................................................................................. 10

v Statement of Original Authorship .................................................................................................. 11

vi Publications and presentations arising from the research presented in this thesis ..................... 12

vii Acknowledgements ...................................................................................................................... 13

Chapter 1: Introduction ........................................................................................................................ 14

1.1 Rationale for thesis ..................................................................................................................... 14

1.2 Research Objective and Aims ...................................................................................................... 16

1.3 Thesis Outline .............................................................................................................................. 17

1.4 References .................................................................................................................................. 18

Chapter 2: Literature Review ................................................................................................................ 20

2.1 Water quality management for the Great Barrier Reef .............................................................. 20

2.1.1 Background ................................................................................................................... 20

2.1.2 Pesticides in the Great Barrier Reef .............................................................................. 20

2.1.3 Photosystem II herbicides ............................................................................................. 21

2.2 History of water quality management targets for the GBR ........................................................ 22

2.1.4 Pesticide targets ............................................................................................................ 23

2.3 Measuring progress towards the targets .................................................................................... 25

2.1.5 Pesticide Loads .............................................................................................................. 25

2.1.6 Species Protection ......................................................................................................... 26

2.4 Mixtures ...................................................................................................................................... 27

2.1.7 Concentration addition ................................................................................................. 28

2.1.8 Toxic Equivalency Factor (TEF) and Relative Potency (ReP) methods .......................... 30

2.1.9 Multisubstance Potentially Affected Fraction .............................................................. 30

2.5 References .................................................................................................................................. 31

Chapter 3: An Improved Method for Calculating Toxicity-Based Pollutant Loads: Part 1. Method

Development......................................................................................................................................... 39

3.1 Abstract ....................................................................................................................................... 39

3.2 Introduction ................................................................................................................................ 40

3.3 Definitions of key terms .............................................................................................................. 42

6

3.4 General method for calculating toxic loads ................................................................................ 43

3.4.1 Nominating the reference chemical ............................................................................. 44

3.4.2 Collating and screening toxicity data ............................................................................ 45

3.4.3 Determining the quality of toxicity data ....................................................................... 46

3.4.4 Calculating relative potencies (RePs) ............................................................................ 46

3.4.5 Fitting a cumulative distribution function to ReP values .............................................. 47

3.4.6 Selecting the percentile of the ReP cumulative distribution function to calculate the

toxic equivalency factors .............................................................................................................. 47

3.4.7 Testing toxic equivalency factors and toxic loads for environmental relevance and

robustness ..................................................................................................................................... 49

3.4.8 Adopting the TEFs that generate the most relevant and robust toxic loads ................ 52

3.5 Conclusions ................................................................................................................................. 52

3.6 Acknowledgements ..................................................................................................................... 53

3.7 References .................................................................................................................................. 53

Chapter 4: An Improved Method for Calculating Toxicity-Based Pollutant Loads: Part 2. Application

to Contaminants Discharged to the Great Barrier Reef, Queensland, Australia. ................................. 57

4.1 Abstract ....................................................................................................................................... 57

4.2 Methods ...................................................................................................................................... 60

4.2.1 Pesticide loads data ...................................................................................................... 60

4.2.2 Ecotoxicity data ............................................................................................................. 61

4.2.3 Reference chemicals and ReP distributions .................................................................. 62

4.3 Results and Discussion ................................................................................................................ 64

4.3.1 Ecotoxicity data and the reference chemicals .............................................................. 64

4.3.2 Relative Potency Cumulative distribution functions..................................................... 66

4.3.3 Toxic Equivalency Factors ............................................................................................. 67

4.3.4 Environmental relevance and robustness .................................................................... 70

4.3.5 Differences between loads and toxic loads .................................................................. 72

4.4 Conclusions ................................................................................................................................. 74

4.5 Acknowledgements ..................................................................................................................... 75

4.6 References .................................................................................................................................. 76

Chapter 5: Advancements to the msPAF method for assessing mixtures of photosystem II herbicides

.............................................................................................................................................................. 80

5.1 Abstract ....................................................................................................................................... 80

5.2 Introduction ................................................................................................................................ 80

5.3 Summary of Terms ...................................................................................................................... 84

5.4 Methods ...................................................................................................................................... 84

5.4.1 Collation of ecotoxicity data ......................................................................................... 84

7

5.4.2 Testing the msPAF method ........................................................................................... 85

5.4.3 Tests for parallelism ...................................................................................................... 88

5.4.4 Validation ...................................................................................................................... 88

5.5 Results and Discussion ................................................................................................................ 90

5.5.1 Local curve estimates .................................................................................................... 90

5.5.2 Case Study – testing 𝛽 to estimate msPAF. .................................................................. 92

5.5.3 Tests for parallelism ...................................................................................................... 92

5.5.4 Global curve estimates .................................................................................................. 95

5.5.5 Comparison with Guideline Values ............................................................................... 97

5.6 Conclusions ............................................................................................................................... 102

5.7 References ................................................................................................................................ 104

Chapter 6: Discussion and Conclusions .............................................................................................. 108

6.1 Summary of results and research impact ................................................................................. 108

6.2 Recommendations for Future Research ................................................................................... 110

6.3 Reference .................................................................................................................................. 111

Appendix A: Supplementary Material from Chapter 3 ...................................................................... 114

Appendix B: Supplementary Material for Chapter 5 .......................................................................... 118

8

iii List of Figures

Figure 2.1: Development of pesticide targets for water quality management set by the Reef Water

Quality Protection Plans (Australian Government and Queensland Government, 2003, 2009 and

2013) and Draft Reef 2050 Water quality Improvement Plan (2017–2022) (Australian Government

and Queensland Government, 2017). ................................................................................................... 23

Figure 2.2: Illustration of parallel response curves of three different chemicals with the same mode of

action. The relative difference in median values (𝑬𝑪) of each chemical indicates their relative

potencies from which a toxic equivalency factor can be derived. ........................................................ 29

Figure 3.1 Flow diagram of steps for calculating toxic loads using a modified TEF approach. ............ 44

Figure 3.2 Probability density function of log ReP values. The distribution of ReP values for chemical A

sit principally to the left of the reference chemical (i.e. log ReP = 0) while the distribution of ReP

values for chemical B sit principally to the right of the reference chemical. The shaded area

represents the ReP values for 95% of species for chemical A and B and are calculated using the 5th

and 95th percentiles, respectively. ......................................................................................................... 48

Figure 3.3 Key steps within steps six and seven of the Toxic Loads general method. .......................... 49

Figure 4.1 Cumulative distribution functions (CDF) of the relative potency values for (A)

ametryn:diuron, (B) hexazinone:diuron, (C) diuron:atrazine, and (D) tebuthiuron:atrazine. Symbols for

ReP values represent freshwater (○), marine (●), and estuarine (Δ) species. Blue dotted lines indicate

equal toxicity, i.e. ReP = 1. All plots generated by BurrliOZ V2 (Barry and Henderson, 2014). ............ 67

Figure 4.2 Percent contributions of five PSII herbicides to the toxic load of the mixture (TLi:TLmix)

calculated using different toxic equivalence factors (in turn calculated using different percentiles of

the relative potency cumulative distribution functions). Contributions are for (A) the Fitzroy River, (B)

Barratta Creek (Haughton Catchment) and (C) the Pioneer River, estimated using 2010–11 data. .... 69

Figure 4.3 Difference (expressed as a percentage) in the total toxic loads for each catchment (2010–

11) calculated using different percentiles of the relative potency (ReP) cumulative frequency

distributions (diuron equivalents) compared to the toxic loads calculated using the 50th percentile

(TLmix,50). ............................................................................................................................................... 70

Figure 4.4 Scores of environmental relevance (●) and robustness (○) as a function of the percentile of

the relative potency (ReP) cumulative distribution function (CDF). The percentiles tested were the

50th, 70th (or 30th), 75th (or 25th), 80th (or 20th) and 95th (or 5th). ................................................ 71

Figure 4.5 Percent contributions of the constituents of a mixture of five PSII herbicides calculated in

terms of (A) the load (Li:Lmix) and (B) the toxic load (TLi:TLmix) for nine Great Barrier Reef catchments in

2011–12. Toxic loads were calculated using the diuron equivalent TEFs presented Table 3. .............. 73

Figure 4.6 The 2011–2012 annual Total Loads (A) and Total Toxic Load (B) for nine Great Barrier Reef

catchments. Total toxic loads were calculated using the bolded TEFs (diuron equivalent) in Table 3. 74

Figure 5.1 Illustration of the concentration addition approach for calculating the multisubstance

Potentially Affected Fraction. The left hand graph shows species sensitivity distributions (SSDs) of

three chemicals with the same mode of action but different relative toxicities. The median value (𝑬𝑪)

of each SSD is used to rescale the SSDs to Hazard Units (HUs) so that the SSDs are brought together

and centre at x=1 (right hand figure). A single cumulative distribution function is then fitted to the

HUs of the three chemicals to generate a SSD for the mixture of the three chemicals (SSDmix). .......... 82

Figure 5.2 Logistic cumulative frequency distributions of the 13 photosystem II herbicides showing the

variation in slopes and potency. ........................................................................................................... 91

Figure 5.3: Relationship between the number of ecotoxicity values per species sensitivity distribution

(SSD) and the number of parallel SSD pairs. Dotted line represents a linear trendline. ....................... 94

Figure 5.4 Global species sensitivity distribution based on ecotoxicity data of 13 PSII herbicides

(atrazine and terbuthylazine, fluometuron, metribuzin, propazine, tebuthiuron, simazine, hexazinone,

9

ametryn, prometryn, diuron, and bromacil). The black line represents the logistic cumulative

distribution function fitted using the global curve fit method, blue lines represent the 95% confidence

bands and the red lines represent the 95% prediction bands. .............................................................. 96

Figure 5.5 Global species sensitivity distributions for two groups of PSII herbicides: (A) group A - two

PSII herbicides (atrazine and terbuthylazine) and (B) Group B - eleven PSII herbicides (fluometuron,

metribuzin, propazine, tebuthiuron, simazine, hexazinone, ametryn, prometryn, diuron, and

bromacil). The black line represents the logistic cumulative distribution function fitted using the

global curve fit method, blue lines represent the 95% confidence bands and the red lines represent

the 95% prediction bands. Arrows in (A) represent points of inflection. .............................................. 97

Figure 5.6: Comparison of the protective concentration for 99% of species (PC99) estimated using the

global curve fits to herbicides in Groups A and B (black X), local curve fits (grey X), with the proposed

default guideline values (King et al., 2017) for freshwater species (blue dot with blue 95% confidence

intervals) and marine species (green dot with green 95% confidence intervals). .............................. 100

Figure A.1 Cumulative distribution functions of the relative potency values for ametryn and atrazine,

hexazinone and atrazine, hexazinone and ametryn, tebuthiuron and diuron, tebuthiuron and

ametryn, and tebuthiuron and hexazinone. ....................................................................................... 116

Figure 8.1: Comparison of the protective concentration for 95% of species (PC95) estimated using the

global logistic curve fits – GC11 and GC2 (black X), local logistic curve fits (grey X), and those estimated

by King et al 2017 for freshwater species (blue dot with blue 95% confidence intervals) and marine

species (green dot with green 95% confidence intervals). .................................................................. 119

10

iv List of Tables

Table 3.1 Example of corresponding percentiles of the relative potency cumulative distribution

function (ReP CDF) for chemicals less toxic and more toxic than the reference chemical. ................... 48

Table 5.1 Model parameters derived from the local curve fit of a logistic cumulative distribution

function to the ecotoxicity data of 13 individual PSII herbicides and summary statistical information

(n, R2 and normality). The PSII herbicides are presented in order from highest to lowest n. ................ 91

Table 5.2: Results of the atrazine case study – the maximum (A) and minimum (B) msPAF (%) values

calculated from set atrazine concentrations (0.1 to 34.72(α) g/L) estimated using β (Equation 5-12)

of 12 paired atrazine mixture combinations. Reference msPAF(%) values (A) were calculated from the

local atrazine SSDs. The variance was assessed by calculating the squared difference of the msPAF(%)

values (A to B) relative to the reference (C) according to Equation 5-13. ............................................. 92

Table 5.3: Test for parallelism between paired species sensitivity distributions fitted with logistic

cumulative distribution functions. Grey shading indicates pairs of SSDs that are parallel................... 94

Table 5.4 Logistic cumulative frequency curve parameters (α and β) and their fit (R2 value) to

ecotoxicity data of different groupings of the PSII herbicides and using the global fit and Traas et al.

(2002) methods. .................................................................................................................................... 97

Table 5.5: Observed differences between the local and global protective concentration (PC) values

and the marine and freshwater (FW) default guidelines values, the possible causes of the observed

differences and, based on these, the concluding reason for the observed differences. ..................... 101

Table A.1 Sources of herbicide ecotoxicity data used in the calculation of relative potencies. .......... 114

Table A.2 P-values from Kruskal Wallis tests applied to the relative potency values of species

belonging to freshwater, marine and estuarine ecosystems for all combinations of test chemicals

with atrazine and diuron as reference chemicals. .............................................................................. 117

Table B.1 Proposed freshwater and marine pesticide default guideline values for the protection of at

least 99% of species, their reliability and the distribution model used to generate the values. ......... 118

Table B.2 Proposed freshwater and marine pesticide default guideline values for the protection of at

least 95% of species, their reliability and the distribution model used to generate the values. ......... 119

The work contained in this thesis has not been previously submitted to meet requirements for an

award at this or any other higher education institution. To the best of my knowledge and belief, the

thesis contains no material previously published or written by another person except where due

reference is made.

Signed

Date

Statement of Original Authorship

QUT Verified Signature

12

vi Publications and presentations arising from the research presented in this thesis

Publications

Chapters 3 and 4 from this thesis have been published in peer-reviewed science journals:

Smith RA, Warne MStJ, Mengersen K and Turner RDR. 2017. An improved method for calculating

toxicity‐based pollutant loads: Part 1. Method development. Integ Environ Manag Assess 13(4):

746-753.

Smith RA, Warne MStJ, Mengersen K, Turner RDR. 2017. An improved method for calculating toxicity‐

based pollutant loads: Part 2. Application to contaminants discharged to the Great Barrier Reef,

Queensland, Australia. Integ Environ Manag Assess 13(4): 754–764.

The methods developed from Chapter 5 has been used in the following publication:

Waterhouse J, Brodie J, Tracey D, Smith R, Vandergragt M, Collier C, Petus C, Baird M, Kroon F, Mann

R, Sutcliffe T, Waters D, Adame F. 2017. Scientific Consensus Statement 2017: A synthesis of the

science of land-based water quality impacts on the Great Barrier Reef, Chapter 3: The risk from

anthropogenic pollutants to Great Barrier Reef coastal and marine ecosystems. Brisbane (QLD),

Australia: State of Queensland, 2017.

Conference Presentations

The methods developed in Chapters 3 and 5 have been presented at international science

conferences:

Smith RA, Warne MStJ, King OK, Turner RDR, Shaw M, Mann R. 2017. Ecologically Relevant Pesticide

Targets for the Reef Water Quality Protection Plan 2017. SETAC Australasia Conference,

September 2017, Gold Coast, Australia.

Smith RA, Warne MStJ, Mann R, Turner RDR, Mengersen K. 2016. Quantifying and communicating the

ecological risk of pesticides for the Reef Water Quality Protection Plan. Oral presentation, SETAC

Australasia Conference, 4–7 October, Hobart, Australia.

Smith R, Delaney K, Turner R, Mengersen, Warne MStJ. 2014. Pesticide Toxic Loads. Presentation to

the Reef Water Quality Protection Plan Independent Science Panel, 1st August, Brisbane,

Australia.

13

vii Acknowledgements

The work produced in this thesis was undertaken in collaboration with the Queensland Government

under the supervision of Kerrie Mengersen, Michael Warne, Fiona Harden and Kate Helmstedt.

I took this Masters on after already completing a PhD to improve my skills in data analysis and my

understanding of statistical methods, but in the end I gained many more skills and understandings. I

will walk away from this Masters with a greater appreciation of logical thinking, a lot of which I

learnt from my principal supervisor Kerrie Mengersen. I thank Kerrie for her supervision, her sound

and logical advice stands out in particular, which helped me navigate the complex, multilayered

problems I was trying to solve. I always walked out of her office clear and confident about my

research and motivated to keep going.

Through this thesis, I have also gained new and better understandings of ecotoxicology, particularly

in dealing with the complex nature of chemical mixtures. I can attribute a lot of this new knowledge

and understanding to my external supervisor Michael Warne, particularly the white board sessions

scribbling diagrams and equations and a few airport/plane trips brainstorming on the back of

envelopes.

I would also like to thank the Water Quality Investigations team (Department of Environment and

Science). The pesticide monitoring data used in this research was provided by the Water Quality and

Investigations team. But more importantly, the support of the team was fundamental in delivering

this work.

I would also like to thank the Office of the Great Barrier Reef (OGBR) for supporting the research and

helping to deliver the methods developed here to the Paddock to Reef program. Through the OGBR,

I was provided the opportunity to present these methods to the Reef Independent Science Panel

who endorsed their use in the Paddock to Reef program.

I would also like to thank Olivia King for her hard work and dedication for leading the development

and delivery of the pesticide species sensitivity distributions and proposed default water quality

guidelines. Without these, the methods in Chapter 5 could not have been developed.

I would also like to thank my family and friends, and Kerrie and Michael for their positive support

that helped me through the long hours and maintain the drive to finish, particularly while working

full-time. In particular I would like to thank Kim and Dick Barnes, Billy Thomson and Cameron Harris.

And lastly a special note to thank Nicolas Schmidt for his help with equation writing.

14

Chapter 1: Introduction

1.1 Rationale for thesis

This thesis was carried out over a four year period from 2014 – 2018. Overlapping this period were

two major government policy updates for managing poor water quality affecting the health and

resilience of the Great Barrier Reef (GBR); the 2013 Reef Water Quality Protection Plan (Reef Plan)

was an update of the 2009 Reef Plan and was released at the end of 2013 (Australian Government and

Queensland Government, 2013), and four years later, the Draft Reef 2050 Water Quality Improvement

Plan 2017–2022 was released (Australian Government and Queensland Government, 2017). The

updates of these policy documents included changes to the way pesticide pollution was targeted for

reduction. How this pesticide reduction is quantified is the rationale for this thesis.

The Reef Plan 2009 set a Reef-wide pesticide reduction target1 (along with nutrient and sediment

targets) to improve water quality discharged from Queensland catchments to the GBR. These reef-

wide water quality targets quantified “the amount of improvement to be achieved in water quality

parameters including nutrient, pesticide and sediment loads”, with “loads” referring to the tonnage

of each of the three pollutant types discharged to the GBR annually (Australian Government and

Queensland Government, 2009). The load-based targets were set in order to achieve the long-term

goal of the Reef Plan, which was based on ecological protection, i.e. "to ensure that by 2020 the

quality of water entering the Reef from adjacent catchments has no detrimental impact on the

health and resilience of the Reef" (Australian Government and Queensland Government, 2009).

However, there were two fundamental problems with the quantification of the pesticide target that

made it incompatible for measuring progress towards achieving the long-term goal: (i) pesticide

loads are not directly related to ecological impact (or risk) of pesticides – ecological risk is quantified

by the concentration of the chemical in the environment, the duration and spatial extent of

exposure and the sensitivity of the exposed biota (US EPA, 1998), and (ii) pesticides are not all equal

in their ecological risk and cannot be grouped as the same – individual pesticides exhibit varying

levels of toxicity and therefore, biota will be more sensitive to some pesticides than others.

In 2013, it was widely agreed that the pesticide targets should be more ecologically relevant. At the

time, the Australian and Queensland governments’ plan for improving Reef water quality, the Reef

Water Quality Protection Plan 2013 (Reef Plan, 2013), had a long-term goal ‘to ensure that by 2020

1 The pesticide reduction target set for the Reef Water Quality Protection Plan 2009 was ‘minimum 50 per cent reduction in pesticides at the end-of-catchments’ (Australian Government and Queensland Government, 2009).

15

the quality of water entering the reef from broadscale land use has no detrimental impact on the

health and resilience of the Great Barrier Reef’ (Australian Government and Queensland Government,

2013). In terms of pesticides, this meant that the water entering the GBR must not cause any

measurable detrimental effects on the health and resilience of the GBR. The 2013 pesticide target2

was an improvement from the 2009 target; the relative toxicities of different pesticides had to be

included in the measurements when determining what progress had been made towards the target.

To do so, required the development of new methods for calculating loads of pesticides based on their

toxicity to the GBR ecosystems.

However, there was still an obvious disconnect between the pollution reduction targets and the long-

term goal of Reef Plan 2013 – the load-based (annual total mass) targets did not provide a direct link

between water quality improvement and ecosystem health. The toxicity of any chemical and the risk

that it poses to aquatic ecosystems is determined by the concentration of the chemical and the

duration that organisms are exposed to the chemical, not the annual mass of the contaminants. While

load-based targets were successful in driving and measuring reductions in the total amount of

pesticides discharged to the GBR, progress to meeting this target did not necessarily equate to an

equivalent reduction in the toxicity or harm on ecosystems and, was therefore potentially misleading

in terms of achieving the Reef Plans’ long-term goal. To reduce the potential impacts pesticides were

having on the GBR, water quality improvements needed to focus on the most toxic waters that

contained the highest concentrations of the most toxic pesticides, not necessarily the catchments with

the largest loads – as these could be dominated by low toxicity pesticides.

In 2017, the Australian and Queensland governments commenced a fourth update of their policies

and plans relating to water quality improvement for the Reef (outline in the Draft Reef 2050 Water

Quality Improvement Plan; Australian Government and Queensland Government, 2017), in which the

pesticide target was revised and updated3. In this new Plan, the pesticide target moved away from a

load-based target and was brought into line with other, well-established policy documents for the

management of toxicants in aquatic ecosystems e.g. Environmental Protection (Water) Policy 2009

and the ANZECC and ARMCANZ (1994) National Water Quality Management Strategy. But ultimately,

the target was changed to align with the long-term goal of improving water quality for the GBR. In

2 The pesticide reduction target set for the Reef Water Quality Protection Plan 2013 was ‘at least a 60 per cent reduction in the end-of-catchment pesticide loads in priority areas’ (Australian Government and Queensland Government, 2013). 3 The pesticide reduction target set for the Reef 2050 Water Quality Improvement Plan 2017 was ‘to protect at least 99% of species at end of catchments’ (Australian Government and Queensland Government, 2017).

16

doing so, another major shift in the methods used to measure progress towards the targets was

required.

Measuring the progress made towards the Reef Water Quality Protection Plan and Reef 2050 Water

Quality Improvement Plan (the Plans) targets, resulting from improvements in agricultural

management practice, is vital for determining the success of the Plans. Progress towards the goal and

targets is reported through an annual Reef Report Card. This provides information that enables

partners and stakeholders to evaluate, prioritise and continuously improve the efficiency and

effectiveness of the implementation of the Plans at GBR-wide and regional scales. This thesis is the

culmination of developing the methods used to measure the progress towards the targets, firstly for

Reef Plan 2013, and secondly for the Reef 2050 Water Quality Improvement Plan.

1.2 Research objective and aims

The primary objective of this thesis was to deliver scientifically and statistically robust and ecologically

relevant quantitative methods to measure the progress towards achieving the pesticide targets for

Reef Water Quality Protection Plan 2013 and the Reef 2050 Water Quality Improvement Plan that can

be easily adopted and used by managers of GBR water quality.

The aims of this thesis were to:

1. Develop a quantitative method to calculate a total photosystem II (PSII) herbicide load from

the loads of individual PSII herbicides that accounts for their differences in toxicity to measure

progress towards the 2013 pesticide target. This aim is addressed in Chapter 3.

2. Validate the method developed in addressing the first aim, using monitoring data from the

GBR, to prove the most robust and environmentally relevant method has been developed.

This aim is addressed in Chapter 4.

3. Develop a quantitative method to determine the percent of species affected from mixtures of

PSII herbicides in GBR ecosystems, to measure progress towards the 2017 pesticide target.

This aim is addressed and validated in Chapter 5.

The research for this thesis ran in parallel with policy developments of the Australian and Queensland

governments. Changes to policy for improving water quality for the GBR was a major driver for the

research undertaken, as was the need to ensure the highest quality of research, transparency and

defensibility of the methods. There was also a strong desire for the methods to be developed in a

manner that could be easily interpreted and adopted by scientists from other fields without a strong

mathematical or statistical background.

17

The research reported in this thesis was undertaken in collaboration with the Queensland

Government, Department of Environment and Science (DES; previously the Department of Science,

Information Technology and Innovation and the Department of Environment and Heritage

Protection).

1.3 Thesis outline

As this is a thesis by publication, Chapters 3 – 5 have been written in the standard format for

publication within a peer-reviewed science journal, and therefore stand on their own as independent

works. With that being said, Chapters 3 and 4 are intrinsically linked to be published as ‘sister’ papers

within the same journal at the same time; Chapter 3 as the method development and Chapter 4 as

the validating case study. Thus, much of the relevant literature is reviewed within each chapter, and

the methods, results, discussion, conclusions and references specific to the topic are also contained

within each chapter. For this reason there is some overlap between these chapters.

Chapter 2 presents a broad overview of the relevant literature to supplement the literature reviewed

in individual chapters. The literature review provides background context to the issues that this thesis

aims to resolve. This includes the environmental impacts of poor water quality on the health and

resilience of the GBR, pesticide contamination transported from agriculture in Queensland

catchments to the GBR, the water quality improvement targets set by the Australian and Queensland

governments to protect the GBR, and how progress towards meeting these targets is quantitatively

measured.

Chapter 3 presents the ‘Toxic Loads’ method; a quantitative method developed to measure the

progress towards reducing the loads of multiple pesticides. In this chapter the concept of mixture

toxicology is examined and methods published in the literature are investigated for their application

to the problem. This chapter has been intentionally written in a generic form so that the methods can

easily be applied to various scenarios where pollutant loads are calculated, i.e. the method is not

restricted to calculating pesticide loads for the conditions of the GBR.

Chapter 4 then demonstrates how the toxic load method can be modified and optimised for a

particular problem or scenario, in this example, calculating the total loads of multiple PSII herbicides

discharged annually to the GBR. A key outcome of this chapter was the development of two novel

tests that allowed the methods from Chapter 3 to be assessed for robustness and environmental

relevance, hence fulfilling the second research aim of the thesis.

18

Chapter 5 further examines the concepts and methods used in mixture toxicology and assesses the

use of a second approach, the multisubstance Potentially Affected Fraction (msPAF), for measuring

progress towards the pesticide targets. This chapter explores the msPAF method in detail for the

suitability of its intended use with PSII herbicides. The chapter is presented in an iterative structure

for testing, reporting and adapting each step of the msPAF method to suit the data limitations of the

PSII herbicides. The modifications to the method are validated with two small case studies, also

presented in the chapter.

Chapter 6 provides a final discussion that summarises the key findings of the research presented in

this thesis and how the aims were met. The significance of the research in context of its incorporation

into key GBR State and regional management documents is briefly summarised. The chapter also

considers what additional research is required that was beyond the scope of this thesis.

Supplementary material for Chapter 3 and Chapter 5 are presented in Appendix A and Appendix B,

respectively. Lastly, a bibliography is presented at the end of the thesis that compiles the list of all

references cited in this thesis. As the thesis has been structured for publication of individual chapters,

a reference list has also been provided at the end of each chapter.

1.4 References

ANZECC and ARMCANZ [Australia and New Zealand and Environment and Conservation Council and

Agriculture and Resource Management Council of Australia and New Zealand]. 1994. National

water quality management strategy. Canberra (ACT), Australia: Australian and New Zealand

Environment and Conservation Council and Agriculture and Resource Management Council of

Australia and New Zealand.

Australian Government and Queensland Government. 2009. Reef Water Quality Protection Plan 2009

For the Great Barrier Reef World Heritage Area and adjacent catchments. Brisbane (QLD), Australia:

Reef Water Quality Protection Plan Secretariat, the State of Queensland.

Australian Government and Queensland Government. 2013. Reef Water Quality Protection Plan 2013.

Securing the health and resilience of Great Barrier Reef World Heritage Area and adjacent

catchments. Brisbane (QLD), Australia: Reef Water Quality Protection Plan Secretariat, the State of

Queensland.



19


Queensland.

Australian Government and Queensland Government. 2017. Draft Reef 2050 Water Quality

Improvement Plan 2017–2022. Reef Water Quality Protection Plan Secretariat, Brisbane.

King OC, Smith RA, Mann R and Warne MStJ. 2017. Proposed aquatic ecosystem protection guideline

values for pesticides commonly used in the Great Barrier Reef catchment area: Part 1—2,4-D,

Ametryn, Diuron, Glyphosate, Hexazinone, Imazapic, Imidacloprid, Isoxaflutole, Metolachlor,

Metribuzin, Metsulfuron-methyl, Simazine, Tebuthiuron. Brisbane (QLD), Australia: Department of

Science, Information Technology and Innovation, 294pp.

King OC, Smith RA, Warne MStJ, Frangos JS and Mann R. 2017. Proposed aquatic ecosystem protection

guideline values for pesticides commonly used in the Great Barrier Reef catchment area: Part 2—

Bromacil, Chlorothalonil, Fipronil, Fluometuron, Fluroxypyr, Haloxyfop, MCPA, Pendimethalin,

Prometryn, Propazine, Propiconazole, Terbutryn, Triclopyr and Terbuthylazine. Brisbane (QLD),

Australia: Department of Science, Information Technology and Innovation, 209pp.

US EPA [United States Environmental Protection Agency]. 1998. Guidelines for ecological risk

assessment (Volume 2). Washington DC, USA: US EPA. EPA/630/R-95.

20

Chapter 2: Literature Review

2.1 Water quality management for the Great Barrier Reef

2.1.1 Background

The need to protect the Great Barrier Reef (GBR) is obvious. It is the largest living structure in the

world, a tropical coral reef system comprised of hundreds of thousands of marine and coral species,

making it one of the most unique ecosystems on Earth (Deloitte, 2013; Schaffelke et al., 2017). It is

also facing global-scale pressures from climate change impacts, local and frequent disturbances (e.g.

cyclones and flooding) of increased intensity, and poor water quality from land-runoff (Schaffelke et

al., 2017).

Poor water quality is principally a result of agricultural development of the adjacent Queensland

catchments since European settlement, introducing man-made chemicals (i.e. pesticides) and

generating loads of sediment and nutrients above natural levels (Brodie et al., 2008; Australian

Government and Queensland Government, 2009, 2013). Other sources include urban centres, mines,

waste water treatment plants, harbours, dredging. Poor water quality can impact the GBR in two ways:

(1) direct impacts to reef biota from exposure to pollutants transported out to the marine

environment and (2) indirect impacts generated from the exposure of biota within connecting coastal

ecosystems (e.g. rivers and estuaries) affecting the services that these ecosystems provide to the GBR

(Brodie et al., 2013; Cappo and Kelley, 2001).

Pesticides are transported in runoff from a range of agricultural industries including grazing, sugar

cane, horticulture, plantation forestry, pasture, cropping and cotton (Devlin et al., 2015). They have

been identified as one of the main pollutants that pose a direct threat to GBR biota (Australian

Government and Queensland Government, 2009, 2013, 2017).

2.1.2 Pesticides in the Great Barrier Reef

Pesticide exposure in the GBR region is known to extend from creeks and rivers (Smith et al., 2012;

Davis et al., 2012; O’Brien et al., 2013), to freshwater wetlands (Devlin et al., 2015), estuaries (Davis

et al., 2012; O’Brien et al., 2013), and the marine environment (Kapernick et al., 2006, 2007; Bartkow

et al., 2008; Bentley et al., 2012; Gallen et al., 2013, 2014). Since 2009, a range of pesticides with

various modes of action (MoA) have been detected in GBR ecosystems marine and coastal ecosystems

(Devlin et al., 2015). Water quality monitoring in these ecosystems has detected the residues of 55

different pesticides: 29 herbicides; 16 insecticides; 4 fungicides and 6 herbicide metabolites (Devlin et

al., 2015). Such an array of pesticides with different MoA means that many different types of

21

organisms within the ecosystem are potentially impacted, depending on which pesticide they are

exposed to. In addition, the types of pesticides and their exposure characteristics vary between

ecosystems and regional areas (Devlin et al., 2015), indicating that ecological risk from pesticides will

vary spatially based on these characteristics.

Mixtures of pesticides are most commonly observed in catchment monitoring, where up to 26

pesticides have been detected together in a single grab sample (Smith et al., 2012; Devlin et al., 2015)

with an average of 6 pesticides detected per sample (Warne et al., in prep). The detection of pesticides

in mixtures indicates potential cumulative impacts on GBR ecosystem. The 2008 Scientific Consensus

Statement identified that this was a key uncertainty in understanding the causal relationship between

pesticide contamination and ecosystem health (Brodie et al., 2008).

In 2009, five photosystem II (PSII) herbicides were identified as the priority pesticides for targeted

reduction in land management practices, these were ametryn, atrazine, diuron, hexazinone and

tebuthiuron, which are predominately sourced from sugarcane and cattle grazing land uses (Australian

Government and Queensland Government, 2009; Devlin et al., 2015). Despite efforts to increase

adoption of better pesticide management practices, PSII herbicides still contribute the highest loads

and highest ecological risk, out of all herbicide detected, to the GBR (Turner et al., 2012, 2013; Wallace

et al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017; Devlin et al., 2015;

Waterhouse et al., 2017).

2.1.3 Photosystem II herbicides

Photosystem II herbicides act by binding to the receptor sites on the QA/QB plastoquinone molecules

located on photosystem II of the photosynthetic pathway of phototrophic species (University of

Hertfordshire, 2013; Wilson et al., 2000). By binding to these receptor sites, the PSII herbicide

molecules prevent electron transport that drives photosynthesis. This specific MoA means that

terrestrial and aquatic plants including algae are highly sensitive to these chemicals, whereas other

species are relatively insensitive (King et al., 2017a, b). Thus, PSII herbicides present a risk to some key

GBR species, including zooxanthellae (the symbiotic algae of coral) and seagrass (Jones and Kerswell,

2003; Negri et al., 2005; Haynes et al., 2000; Gao et al., 2011), as well as benthic microalgae and

crustose coralline algae (Magnusson et al., 2010; Harrington et al., 2005, respectively).

22

2.2 History of water quality management targets for the GBR

The development of water quality management targets for the protection of the GBR commenced in

2003 and has since been through three subsequent updates, the latest to be released shortly in 2018.

The targets are outlined in policy plans, prepared jointly by the Australian and Queensland

governments, to manage poor water quality from terrestrial pollutants for the GBR. The historical

development of these plans and their pesticide reduction targets are summarised in Figure 2.1.

The plans, the Reef Water Quality Protection Plans 2003, 2009 and 2013 (Australian Government and

Queensland Government, 2003, 2009 and 2013) and the more recent Reef 2050 Water Quality

Improvement Plan (Australian Government and Queensland Government, 2017), are designed for

adaptive management. That is, the plans and the programs that monitor and evaluate their targets,

continuously improve and adapt such that management and policy keeps up to date with the latest

and most relevant science and evidence for protecting the Reef. As such, the water quality

management targets have evolved to become more ecologically relevant and focused.

The long-term goal of the plans has remained the same: “the quality of water entering the lagoon

from broadscale land use has no detrimental impact on the health and resilience on the GBR”. With

this goal in mind, the water quality targets are set to achieve this goal. Fundamental to achieving this

long-term goal in the first and second plans was the assumption that a reduction in a pollutant load

will lead to a commensurate improvement in ecosystem health. However, when the pollutant load

consists of a mixture of chemicals with different toxicities, meeting the load reduction target may not

achieve the long-term goal of Reef Plan (2013).

23

Figure 2.1: Development of pesticide targets for water quality management set by the Reef Water

Quality Protection Plans (Australian Government and Queensland Government, 2003, 2009 and

2013) and Draft Reef 2050 Water quality Improvement Plan (2017–2022) (Australian Government

and Queensland Government, 2017).

2.1.4 Pesticide targets

Pesticide reduction targets for GBR water quality management weren’t introduced to the Reef Water

Quality Protection plans until 2009 and focused on just five ‘priority’ PSII herbicides. Since that time,

progress has been made to continually improve the targets and the way they are measured to ensure

they are more ecologically relevant and encompass more pesticides. The changes to the pesticide

reduction targets are outlined in Figure 2.1 and discussed below. The changes to the methods used to

measure progress towards achieving the targets are the focus of this thesis and will be discussed in

detail in Chapters 3 – 5.

The first pesticide reduction target (Australian Government and Queensland Government, 2009) was

a load-based target, similar to the nutrient and sediment targets, to reduce 50% of end-of-catchment

pesticide loads. This was to be achieved through the implementation of best management practices

in agriculture, with a primary focus on sugar cane and grazing lands (Australian Government and

Queensland Government, 2009). At the time, measuring progress towards the pesticide reduction

target involved a simple aggregation of the annual loads of the five priority PSII herbicides. However,

24

pesticides can vary in their toxicity to organisms over several orders of magnitude (Smith et al., 2012;

Devlin et al., 2015; King et al., 2017a, b). Thus, accounting for these differences in toxicity in the load

calculations would bring greater alignment between measuring progress towards the pesticide target

and the long-term goal of the Reef Plan.

In 2013 the pesticide targets were changed to account for all pesticides and set to achieve a 60%

reduction in the total pesticide load by 2018 (Australian Government and Queensland Government,

2013). However, to measure the progress towards this target, it was specified that the total pesticide

load would be calculated to account for the differences in the toxicity between the pesticides.

Developing the methods to do this was the focus of Chapter 3, and their validation is provided in a

case study of the GBR presented in Chapter 4.

While the 2013 targets were a step in the right direction towards more ecologically relevant targets,

there was still a major issue with these targets; the biological effect caused by a pesticide is correlated

with the concentration of the pesticide in an ecosystem, not its load. Typically, the toxicity of

pesticides are expressed in terms of the concentration that causes a measured effect on organisms;

for example, the LC50 is the concentration of a chemical that is lethal to 50% of test organisms. In

terms of Reef Plan’s (2013) reduction targets, achieving the target of a sixty per cent reduction in the

total pesticide load did not necessarily equate to a similar reduction in the effect or impact of the

pesticides on the GBR. For example, it would be logical in trying to achieve the pesticide reduction

target, to focus efforts on reducing pesticides with the highest loads (e.g. atrazine and tebuthiuron;

Turner et al., 2012, 2013; Wallace et al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al.,

2017). However, atrazine and tebuthiuron have relatively low toxicities compared to some other

pesticides. Thus, if the reduction target was met by focussing principally on reducing the loads of these

two pesticides, the total impact from all pesticides would not be reduced by much. In terms of

reducing the biological effects caused by pesticides it would be more effective to focus on those

catchments that have the most toxic waters, not the largest loads.

For this reason, the pesticide targets were revised in 2017 for the Reef 2050 Water Quality

Improvement Plan 2017–2022 (Brodie et al., 2017; Australian Government and Queensland

Government, 2017). They were developed to ensure consistency with the long-term goal of Reef

Water Quality Protection Plan 2013 (Australian Government and Queensland Government, 2013), the

Environmental Protection (Water) Policy 2009 and the ANZECC and ARMCANZ (1994) National Water

Quality Management Strategy. These targets were intentionally set in alignment with the National

Water Quality Management Strategy (ANZECC and ARMCANZ, 1994) framework of assigning

environmental values to a waterbody. The environmental values set for the waterbodies of the Great

25

Barrier Reef are in accordance with the National, Queensland State and Great Barrier Reef water

quality guidelines (ANZECC and ARMCANZ, 2000; DEHP, 2009; GBRMPA, 2010), which state that

waterbodies of high ecological value (such as the Great Barrier Reef Marine Park and World Heritage

Area) should be protective of at least 99% of species.

Thus, the 2017 pesticide target was set to “achieve concentrations of all pesticides in water bodies of

the GBR that will be protective of at least 99% of species, where the toxic impacts of all pesticides in

the water body are considered collectively” (Brodie et al., 2017). This means that water bodies in the

Great Barrier Reef should not be exposed to any pesticide or combination of pesticides of high enough

magnitude and duration that will cause adverse effects on more than 1% of species. Hence, to

measure the progress towards meeting these targets, the percent of species affected by all pesticides

would need to be calculated. Developing the methods to do this was the focus of Chapter 5.

2.3 Measuring progress towards the targets

2.1.5 Pesticide Loads

Pesticide loads quantified for the Paddock to Reef program were a measure for estimating the mass

of pesticides transported annually from agriculture to the GBR (Turner et al., 2012, 2013; Wallace et

al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017). Improved management

practices should equate to a reduced load of pesticides transported to the GBR. Thus, the measure of

pesticide loads is a valuable management tool. However, the interpretation of load reduction and

ecological improvement is not necessarily a relative conversion.

The Paddock to Reef program has two methods for assessing pesticide loads; those calculated from

monitoring data and those calculated from modelling data. Only loads calculated using monitoring

data will be discussed here as it is this data that has been used in case studies (Chapter 4) of this thesis.

Pesticide monitoring data are gathered from point-in-time sampling from end-of-system sites within

Reef catchments. The monitoring data, measured in concentration units, are coupled with the

discharge volume (flow m3s-1) of the water body at the same point in time of sampling (± 1 h) (Turner

et al., 2012, 2013; Wallace et al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017).

Samples are limited (on average 30 per year per site) relative to the number of flow measurements

(every hour = 8760 per year per site). Thus, to calculate the annual load, statistical models such as the

linear interpolation model (eWater, 2015), are used to infill missing concentration data with the flow

data. In doing so, a concentration value can be attributed to each flow measurement and aggregated

to generate a single load value for the year. As such, the estimated loads hold a high degree of

26

uncertainty. In addition, they are also highly influenced by the prevailing climatic conditions that

occurred within the year of monitoring (Devlin et al., 2015). For example, high rainfall years typically

have higher annual loads than drought years (Turner et al., 2012, 2013; Wallace et al., 2014, 2015,

2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017).

Monitored pesticide loads have been estimated annually since 2009–2010 for up to 15 catchments as

part of the GBR Catchment Loads Monitoring Program (Turner et al., 2012, 2013; Wallace et al., 2014,

2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017). Initially, only the five priority PSII

herbicides were estimated. Loads of other pesticides also detected in catchments have been

estimated since 2011–12 (Wallace et al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al.,

2017).

As a result of the dominant land use types within a catchment, the resulting pesticide loads

transported to the Reef vary between catchments and natural resource management (NRM) regions.

The loads vary in both the absolute magnitude of the loads, as well as the relative contribution of each

PSII herbicide to the load (Turner et al., 2012, 2013; Wallace et al., 2014, 2015, 2016; Garzon‐Garcia

et al., 2015; Huggins et al., 2017; Devlin et al., 2015). For example, diuron and atrazine generally

dominate the load in catchments with high sugarcane land use (e.g. Pioneer Catchment), however the

magnitude of this load can be small compared to loads from grazing catchments (Turner et al., 2012,

2013; Wallace et al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017). The pesticide

load from catchments with extensive grazing areas and other non‐sugarcane land‐uses (e.g. Burdekin,

Fitzroy and Burnett catchments) is predominantly comprised of tebuthiuron and atrazine and these

catchments have some of the largest loads (by kg) estimated (Turner et al., 2012, 2013; Wallace et al.,

2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017).

2.1.6 Species Protection

To measure impact in situ that has been directly caused by pesticides is difficult because of the

complexity in differentiating pesticide impact from the impact or responses from other variables and

stressors (Schäfer et al., 2007). An alternative and commonly used approach is to use ecotoxicity data

generated from laboratory and field experiments as an indicator of species sensitivity (Posthuma et

al., 2002). Ecotoxicity data in the form of effect concentration (ECx) data, describes the concentration

of a chemical required to cause a defined effect on a sample population of a species, e.g. the

concentration that effects 10% of the population. When EC data are available for a sample of species

representing an ecological community, the data can be used to describe how the ecological

27

community may respond when exposed to a given concentration of the chemical; i.e. by using the

statistical distribution of the EC data to predict the distribution of species’ sensitivity across the

community (Aldenburg et al., 2002). Through fitting a cumulative distribution function (CDF) to the EC

data and thus generating a species sensitivity distribution (SSD), we can predict the fraction of species

potentially affected (the potentially affected fraction; PAF) by a defined concentration of a chemical.

Alternatively, the concentration of a chemical that is protective of a given fraction of species can also

be determined, i.e. environmental quality criteria (e.g. water quality guideline values) (Posthuma et

al., 2002). This approach assumes amongst others that the sensitivity of the sample of species is

representative of the ecosystem under assessment (Posthuma et al., 2002).

The SSD technique is used to determine the default guideline values (formerly referred to as trigger

values) of the Australian and New Zealand Guidelines for Fresh and Marine Water Quality (ANZECC

and ARMCANZ, 2000; Warne et al., 2018). Recently, SSDs for 28 pesticides, including 13 PSII herbicides,

relevant to the GBR were developed to calculate default guideline values for the protection of the GBR

and coastal ecosystems (King et al., 2017a, b).

2.4 Mixtures

Using default guideline values to protect the GBR marine and coastal ecosystems falls short due to the

prevalence of pesticide mixtures in many of these ecosystems (Smith et al., 2012; Lewis et al., 2012;

Devlin et al., 2015; Warne, in prep). Pesticide mixtures produce a cumulative effect in organisms that

can be additive, synergistic or antagonistic (Warne and Hawker, 1995). Thus, to assess the ecological

risk of pesticides individually using default guideline values, we are likely to underestimate the real-

world effects in GBR ecosystems (Smith et al., 2012; Davis et al., 2012; Davis et al., 2013; Davis et al.,

2014).

The effect of multiple pesticides on an organism is dependent on the MoA of the pesticides. For those

with the same MoA, for example PSII herbicides, the effect on an organism has been shown to conform

to the concentration addition (CA) model of joint action (discussed below) (Berenbaum, 1985). That

is, one PSII herbicide can act as another PSII herbicide and they only differ based on their relative

toxicities. Mixtures of chemicals with different MoA follow the model of independent action and can

be estimated using the response addition approach (Traas et al., 2002).

There are a few, commonly used methods that have been developed based on the concentration

addition and response addition models, including the toxic equivalency factor (TEF) and

multisubstance-Potentially Affected Fraction (msPAF) methods (De Zwart and Posthuma, 2005). The

28

TEF method is a relatively simple method that estimates mixture effects for chemicals with the same

MoA (De Zwart and Posthuma, 2005). The msPAF approach is more complex and can be used to

calculate mixtures with both the same and different MoA (Traas et al., 2002). Given that PSII herbicides

have the same MoA, further discussion will focus only on the concentration addition approach.

2.1.7 Concentration addition

Photosystem II herbicides have the same MoA on photosynthetic organisms (University of

Hertfordshire, 2013) and therefore we could assume that the concentration addition model would be

the most appropriate model to determine the effect of mixtures of PSII herbicides on phototrophic

species. This has been verified in various laboratory studies, for example the toxicity of mixtures of

diuron, atrazine, simazine and tebuthiuron to the Queensland estuarine microalgae Navicula sp. and

Nephroselmis pyriformis conformed to the concentration addition model (Magnusson et al., 2010).

Faust et al. (2001) also demonstrated the concentration addition model as a valuable tool for reliably

predicting the hazard of PSII herbicide mixtures. They found that even the low, non-significant

concentrations of single PSII herbicides contribute to the overall toxicity, and recommended that all

risk assessments where PSII herbicide mixtures were present, should estimate the risk based on the

mixture rather than the individual herbicides (Faust et al., 2001).

The concentration addition model has an underlying assumption of parallel response curves (e.g.

dose- or concentration-response curves as well as SSDs) when estimating the effect of a mixture,

although this has been disputed by some authors (e.g. Faust et al., 2001). Both the TEF and msPAF

approaches assume that there is parallelism between response curves (Safe, 1998, Traas et al., 2002).

Parallel concentration-response curves of a pair of substances has been used as proof that the two

substances were similar, that is, they act as dilutions of the same substance (Fleetwood et al., 2015).

Hence, parallelism is commonly used in pharmacology with bioequivalency studies to identify whether

a test drug has a similar MoA as a reference drug (Finney, 1979; Walsh et al., 2005), as well it is used

to estimate the relative potency of the test compared to the reference drug (Hester and Harrison

1999). Two curves are parallel if their slopes (β) are equivalent and the only difference between them

is their placement along the x-axis (Faust et al., 2001; De Zwart, 2002; Chèvre et al., 2006). For

example, Figure 2.2 illustrates the response curves (SSDs) of three chemicals, with the same MoA,

where the slopes (β) of the curves are the same but the median of the effect concentrations (𝐸�̃�)

differs. It is this concept that the TEF and relative potency (ReP) methods are based upon.

29

Figure 2.2: Illustration of parallel response curves (species sensitivity distributions) of three different

chemicals with the same mode of action. The relative difference in the median of the effect

concentration values (𝐄�̃�) of each chemical indicates their relative potencies from which a toxic

equivalency factor can be derived.

Parallelism of dose- or concentration-response curves are most well-documented, however

parallelism in SSDs has also been examined. De Zwart (2002) theorised and demonstrated that the

SSDs of chemicals with similar MoA have equivalent β, however this was strongly dependent on the

number and variety of species tested (De Zwart, 2002). The msPAF method for concentration addition

developed by Traas et al. (2002) relies on calculating a single common β to fit the SSDs of multiple

chemicals with similar MoA, and therefore assumes parallelism of the SSDs.

The assumption of parallel response curves does not always carry through however. Faust et al. (2001)

found that concentration-response curves of some PSII herbicides were not parallel. Chèvre et al.

(2006) similarly found that parallelism was not consistent between all PSII herbicide SSDs tested and

the βs of the SSDs were significantly different. Chèvre et al. (2006) addressed this issue by applying a

common best-fit slope for all the PSII herbicides and concluded that there was undoubtedly a need

for further research on SSD parameters regarding their use in regulation. Faust et al. (2001) said that

the findings of non-parallel curves was not too surprising given that there was a temporal and spatial

gap between the measured inputs and outputs with many opportunities for other factors to

contribute such as differences in the uptake, the intracellular partitioning and accumulation, the

binding to unspecific sites or the biotransformation of different PSII herbicides. However, the authors

(Faust et al., 2001) concluded that these differences did not impair the predictive value of

concentration addition.

30

2.1.8 Toxic Equivalency Factor (TEF) and Relative Potency (ReP) methods

The toxic equivalency factor (TEF) and relative potency (ReP) methods are used to derive a factor to

express the potency of a test chemical relative to a reference chemical. The terms TEF and ReP seem

to be used interchangeably in the literature. In this thesis, these terms reflect the approach, most well-

known in ecotoxicology, used to assess mixtures of dioxin-like chemicals (van den Berg et al., 2006;

Haws et al., 2006; US EPA, 2008). Thus, ReP describes the relative difference between two chemicals

calculated from the same study design of a single species (Putzrath, 1997; Compton and Sigal, 1999),

and TEF calculates the difference from the distribution of RePs of multiple species (Finley et al., 2003;

Haws et al., 2006). These terms are defined in more detail in Chapter 3.

By converting concentrations of chemicals present in a mixture by their TEF or ReP, the concentrations

are weighted relative to the reference chemical and this permits the effects of the mixture

constituents to be compared or combined on an equitoxic basis. In the same manner, the loads of

pollutants can also be weighted easily using these methods allowing them to be aggregated to a single

pollutant load (Pedersen et al., 2006). Thus, in Chapters 3 and 4, the TEF and ReP methods are explored

in more detail for their application in weighting and aggregating the loads of five PSII herbicides to a

single pesticide load for the purpose of measuring the progress towards the 2013 loads-based

pesticide targets (Figure 2.1).

The toxic equivalency method has been used in the past for assessing the hazard of PSII herbicide

mixtures (Kennedy et al., 2012; Smith et al., 2012). Kennedy et al. (2012) developed diuron-equivalent

factors for several PSII herbicides from multiple species. At the time of reporting Kennedy et al. (2012)

noted that the data for which the TEFs had been derived were somewhat limited and recommended

more work to be done in the area, particularly in terms of increasing the number of species. Smith et

al. (2012) developed atrazine-equivalent values from single species that demonstrated the variation

in mixture impacts to different species. The variation in RePs between species has been well

documented (Putzrath, 1997; Compton and Sigal, 1999) and the cause for some criticism of the TEF

method (De Zwart and Posthuma, 2005; Compton and Sigal, 1999). The methods developed in Chapter

3 aim to resolve this issue.

2.1.9 Multisubstance Potentially Affected Fraction

The msPAF method, as originally published by Traas et al. (2002), uses the species sensitivity

distributions of individual contaminants and applies either a concentration addition or response

addition model to determine the percent of species affected by a mixture of contaminants. The

capacity to use both the concentration and response addition models with this method is an

31

advantage over the TEF method as the risk of pesticides with both similar and dissimilar MoA can be

estimated in the same metric (De Zwart and Posthuma, 2005). While pesticides with dissimilar MoA

are not the focus of this thesis, there are pesticides present in GBR ecosystems with dissimilar MoAs

(Devlin et al., 2015; Waterhouse et al., 2017) that will in future need to be incorporated in risk metrics

and the methods used to measure progress towards the Reef 2050 Water Quality Improvement Plan

pesticide targets (Figure 2.1).

Species sensitivity distributions form the basis of the msPAF method (Traas et al., 2002). Thus, the

resulting output of the method is reported as percent of species affected (or the inverse – percent of

species protected), analogous to water quality guideline values (ANZECC and ARMCANZ, 2000; Warne

et al., 2018). Therefore, the msPAF method provides an appropriate platform for measuring progress

towards the Reef 2050 Water Quality Improvement Plan pesticide reduction targets, i.e. to protect

more than 99% of species (Figure 2.1).

The disadvantage of the msPAF method is that generating SSDs is a significant amount of work and

considerably more time consuming than generating TEFs. In parallel to this thesis, the SSDs for 28

pesticides relevant to the GBR (King et al., 2017a, b) were generated that provided the mechanisms

to develop the msPAF method. The methods and outcomes of those SSDs are not presented in this

thesis, however a detailed description of the methodology can be found in Warne et al. (2018) and

the results have been recently published in King et al. (2017a, b). The msPAF method published by

Traas et al. (2002) is explored in Chapter 5 of this thesis where a detailed description can be found.

2.5 References

Aldenberg T, Jaworska JS, Traas TP. 2002. Normal species sensitivity distributions and probabilistic

ecological risk assessment. In: Posthuma L, Suter II GW, Traas TP, editors. Species Sensitivity

Distributions in Ecotoxicology. Boca Raton (FL), USA: Lewis Publishers. pp 49–102.






ANZECC and ARMCANZ. 2000. Australian and New Zealand Guidelines for Fresh and Marine Water

Quality, vol. 1. The Guidelines. Canberra (ACT), Australia: Australian and New Zealand Environment

32

and Conservation Council and Agriculture and Resource Management Council of Australia and New

Zealand.

Australian Government and Queensland Government. 2003. Reef Water Quality Protection Plan: For

catchments adjacent to Great Barrier Reef World Heritage Area, October 2003. Brisbane (QLD),

Australia: Reef Water Quality Protection Plan Secretariat, the State of Queensland.







Queensland.



Bartkow M, Dunn A, Komarova, T, Paxman, C, Mueller J. 2008. Monitoring of organic chemicals in the

Great Barrier Reef Marine Park and selected tributaries using time integrated monitoring tools:

October 2008. Report to the Great Barrier Reef Marine Park Authority (Marine Monitoring

Programme), National Research Centre for Environmental Toxicology (EnTox).

Bentley C, Devlin M, Paxman C, Chue KL, Mueller J. 2012. Pesticide monitoring in inshore waters of

the Great Barrier Reef using both time‐integrated and event monitoring techniques (2011 ‐ 2012).

The University of Queensland, The National Research Centre for Environmental Toxicology (Entox).

Berenbaum MC. 1985. The expected effect of a combination of agents: the general solution. Journal

of Theoretical Biology 114, 413‐431.

Brodie J, Binney J, Fabricius K, Gordon I, Hoegh-Guldberg O, Hunter H, O’Reagain P, Pearson R, Quirk

M, Thorburn P, Waterhouse J. 2008. Synthesis of evidence to support the scientific consensus

statement on water quality in the Great Barrier Reef. Brisbane (QLD), Australia: Department of

Premier and Cabinet, the State of Queensland.

Brodie J, Waterhouse J, Schaffelke B, Kroon F, Thorburn P, Rolfe P, Johnson J, Fabricius K, Lewis S,

Devlin M, Warne MStJ, McKenzie L. 2013. 2013 Scientific Consensus Statement: Land use impacts

on Great Barrier Reef water quality and ecosystem condition. Brisbane (QLD), Australia: Reef Water

Quality Protection Plan Secretariat, the State of Queensland 2013.

33

Brodie J, Baird M, Waterhouse J, Mongin M, Skerratt J, Robillot C, Smith R, Mann R, Warne MStJ. 2017.

Development of basin-specific ecologically relevant water quality targets for the Great Barrier Reef.

TropWATER Report No. 17/38, James Cook University. Brisbane (QLD), Australia: the State of

Queensland, 68 pp.

Cappo M, Kelley R. 2000. Connectivity in the Great Barrier Reef World Heritage Area: an overview of

pathways and processes. In: Wolanski (editor). Oceanographic Processes of Coral Reefs: Physical

and Biological Links in the Great Barrier Reef. Boca Raton (FL), USA: CRC Press. pp.161–187.

Chèvre N, Loepfe C, Singer H, Stamm C, Fenner K, Escher B. 2006. Including mixtures in the

determination of water quality criteria for herbicides in surface water. Environ Sci Tech, 40(2): 426–

435.

Compton R, Sigal EA. 1999. The use of toxic equivalency factors (TEFs) in ecological risk assessment:

Strengths and limitations. Hum Ecol Risk Assess, 5: 33–42.

Davis AM, Lewis SE, Bainbridge ZT, Glendenning L, Turner R, Brodie JE. 2012. Dynamics of herbicide

transport and partitioning under event flow conditions in the lower Burdekin region, Australia. Mar

Poll Bull, 65: 182‐193.

Davis AM, Thorburn PJ, Lewis SE, Bainbridge ZT, Attard SJ, Milla R, Brodie JE. 2013. Environmental

impacts of irrigated sugarcane production: Herbicide run‐off dynamics from farms and associated

drainage systems. Agriculture, Ecosystems and the Environment, 180: 123‐135.

Davis AM, Lewis SE, O’Brien DS, Bainbridge ZT, Bentley C, Mueller JF, Brodie JE. 2014. Water resource

development and high value coastal wetlands on the lower Burdekin floodplain, Australia. In:

Wolanski, E, editor. Estuaries of Australia in 2050 and beyond. Dordrecht, Germany: Springer.

pp. 223–245

Deloitte Access Economics. 2013. Economic contribution to the Great Barrier Reef. Townsville (QLD),

Australia: Great Barrier Reef Marine Park Authority.

DEHP [Department of Environment and Heritage Protection]. 2009. Queensland Water Quality

Guidelines, Version 3. Brisbane (QLD), Australia: Department of Environment and Heritage

Protection, 184pp.

Devlin M, Lewis S, Davis A, Smith R, Negri A, Thompson M and Poggio M. 2015. Advancing our

understanding of the source, management, transport and impacts of pesticides on the Great

Barrier Reef 2011–2015. A report for the Queensland Department of Environment and Heritage

34

Protection. Cairns (QLD) Australia: Tropical Water and Aquatic Ecosystem Research (TropWATER),

James Cook University, 134pp.

De Zwart D, Posthuma L. 2005. Complex mixture toxicity for single and multiple species: Proposed

methodologies. Environ Toxicol Chem 24: 2665–2676.

eWater Cooperative Research Centre. 2015. Water Quality Analyser v2.1.2.4. Canberra (ACT),

Australia: eWater Cooperative Research Centre.

Faust M, Altenburger R, Backhaus T, Blanck H, Boedeker W, Gramatica P, Hamer V, Scholze M, Vighi

M, Grimme LH. 2001. Predicting the joint algal toxicity of multi-component s-triazine mixtures at

low-effect concentrations of individual toxicants. Aq Toxicol 56(1):13–32.

Finley B, Connor K, Scott P. 2003. The use of toxic equivalency factor distributions in probabilistic risk

assessments for dioxins, furans, and PCBs. J Toxicol Environ Health Part A 66: 533–550.

Finney D. 1979. Bioassay and the Practice of Statistical Inference. International Statistical Review /

Revue Internationale de Statistique 47(1): 1–12.

Fleetwood K, Bursa F, Yellowlees A. 2015. Parallelism in Practice: Approaches to Parallelism in

Bioassays. PDA J Phar Sci Tech 69(2): 248–63.

Gallen C, Devlin M, Paxman C, Banks A, Mueller J. 2013. Pesticide monitoring in inshore waters of the

Great Barrier Reef using both time-integrated and event monitoring techniques (2012–2013).

Brisbane (QLD), Australia: National Research Centre for Environmental Toxicology University of

Queensland (Entox), University of Queensland.

Gallen C, Devlin M, Thompson K, Paxman C, Mueller J. 2014. Pesticide monitoring in inshore waters of

the Great Barrier Reef using both time-integrated and event monitoring techniques (2013–2014).



Gao Y, Fang J, Zhang J, Ren L, Mao Y, Li B, Zhang M, Liu D, Du M. 2011. The impact of the herbicide

atrazine on growth and photosynthesis of seagrass, Zostera marina (L.), seedlings. Mar Poll Bull 62:

1628–1631.

GBRMPA [Great Barrier Reef Marine Park Authority]. 2010. Water Quality Guidelines for the Great

Barrier Reef Marine Park. Townsville (QLD), Australia: Great Barrier Reef Marine Park Authority,

109pp.

35

Harrington L, Fabricius K, Eaglesham G, Negri A. 2005. Synergistic effects of diuron and sedimentation

on photosynthesis and survival of crustose coralline algae. Mar Poll Bull 51: 415‐427.

Haws LC, Su SH, Harris M, DeVito MJ, Walker NJ, Farland WH, Finley B, Birnbaum LS. 2006.

Development of a refined database of mammalian relative potency estimates for dioxin-like

compounds. Toxicol Sci 89: 4–30.

Haynes D, Michalek-Wagner K. 2000. Water quality in the Great Barrier Reef World Heritage Area:

past perspectives, current issues and new research directions. Mar Poll Bull 41: 428–434.

Harrison R. 1999. Endocrine disrupting chemicals. Cambridge: RSC.

Jones RJ, Kerswell AP. 2003. Phytotoxicity of photosystem II (PSII) herbicides to coral. Mar Ecol Prog

Ser 261: 149–159.

Kapernick A, Shaw M, Dunn A, Komarova T, Mueller J, Carter S, Eaglesham G, Schaffelke B, Bass D,

Haynes D. 2006. River pesticide loads and GBR lagoon pesticide data. Attachment A, Task 2.6.

Report to the Great Barrier Reef Marine Park Authority (Marine Monitoring Programme). Brisbane

(QLD), Australia: National Research Centre for Environmental Toxicology (EnTox).

Kapernick A, Shaw C, Dunn A, Komarova T, Mueller J, Carter S, Eaglesham G, Alberts V, Masters B,

Rohde K, Packett R, Prange J, Haynes D. 2007. River pesticide loads and GBR lagoon pesticide data

(2006–2007). Report to the Great Barrier Reef Marine Park Authority (Marine Monitoring

Programme). Brisbane (QLD), Australia: National Research Centre for Environmental Toxicology

(EnTox).

Kennedy K, Devlin M, Bentley C, Lee‐Chue K, Paxman C, Carter S, Lewis SE, Brodie J, Guy E, Vardy S.

2012. The influence of a season of extreme wet weather events on exposure of the World

Heritage Area Great Barrier Reef to pesticides. Mar Poll Bull 64: 1495–1507.

King OC, Smith RA, Mann R and Warne MStJ. 2017a. Proposed aquatic ecosystem protection guideline





King OC, Smith RA, Warne MStJ, Frangos JS and Mann R. 2017b. Proposed aquatic ecosystem

protection guideline values for pesticides commonly used in the Great Barrier Reef catchment area:

Part 2—Bromacil, Chlorothalonil, Fipronil, Fluometuron, Fluroxypyr, Haloxyfop, MCPA,

36

Pendimethalin, Prometryn, Propazine, Propiconazole, Terbutryn, Triclopyr and Terbuthylazine.

Brisbane (QLD), Australia: Department of Science, Information Technology and Innovation, 209pp.

Lewis SE, Schaffelke B, Shaw M, Bainbridge ZT, Rohde KW, Kennedy K, Davis AM, Masters BL, Devlin

MJ, Mueller JF, Brodie JE. 2012. Assessing the additive risks of PSII herbicide exposure to the Great

Barrier Reef. Mar Poll Bull 65: 280–291.

Magnusson M, Heimann K, Quayle P, Negri AP. 2010. Additive toxicity of herbicide mixtures and

comparative sensitivity of tropical benthic microalgae. Mar Poll Bull 60(11): 1978–1987.

Negri AP, Vollhardt C, Humphrey C, Heyward A, Jones R, Eaglesham G, Fabricius K. 2005. Effects of the

herbicide diuron on the early life history stages of coral. Mar Poll Bull 51: 370–383.

O’Brien DS, Lewis SE, Davis A, Gallen C, Paxman C, Smith R, Warne MStJ, Mueller J, Brodie JE. 2013.

Pesticide exposure within the Barratta Creek Catchment. Cairns (QLD), Australia: Report to the Reef

Rescue Water Quality Research and Development Program. Reef and Rainforest Research Centre

Limited, 58pp.

Pedersen JA, Yeager MA, Suffet IH. 2006. Organophosphorus insecticides in agricultural and residential

runoff: Field observations and implications for total maximum daily load development. Environ Sci

Technol 40: 2120–2127.

Posthuma L, Traas TP, Suter II GW. 2002. General introduction to species sensitivity distributions. In:

Posthuma L, Suter II GW, Traas TP, editors. Species Sensitivity Distributions in Ecotoxicology. Boca

Raton (FL), USA: Lewis Publishers. pp 3–11.

Putzrath RM. 1997. Estimating relative potency for receptor-mediated toxicity: re-evaluating the

toxicity equivalence factor (TEF) model. Regul Toxicol Pharmacol 25: 68–78.

Safe S. 1998. Development validation and problems with the toxic equivalency factor approach for risk

assessment of dioxins and related compounds. J Animal Sci 76(1): 134–141.

Schäfer RB, Caquet T, Siimes K, Mueller R, Lagadic L, Liess M. 2007. Effects of pesticides on community

structure and ecosystem functions in agricultural streams of three biogeographical regions in

Europe. Sci Tot Environ 382(2–3): 272–85.

Schaffelke B, Collier C, Kroon F, Lough J, McKenzie L, Ronan M, Uthicke S, Brodie J. 2017. Scientific

Consensus Statement 2017. Scientific Consensus Statement 2017: A synthesis of the science of

land-based water quality impacts on the Great Barrier Reef, Chapter 1: The condition of coastal

and marine ecosystems of the Great Barrier Reef and their responses to water quality and

disturbances. Brisbane (QLD), Australia: State of Queensland.

37

Smith R, Middlebrook R, Turner R, Huggins R, Vardy S, Warne MStJ. 2012. Large-scale pesticide

monitoring across Great Barrier Reef catchments–paddock to reef integrated monitoring,

modelling and reporting program. Mar Poll Bull 65: 117–127.

Traas TP, Van de Meent D, Posthuma L, Hamers T, Kater BJ, De Zwart D, Aldenberg, T. 2002. The

potentially affected fraction as a measure of ecological risk. In: Posthuma L, Suter II GW, Traas TP,

editors. Species Sensitivity Distributions in Ecotoxicology. Boca Raton (FL), USA: Lewis Publishers.

p 315–344.

Turner RDR, Huggins R, Wallace R, Smith RA, Vardy S, Warne MStJ. 2013. Total suspended solids,

nutrient and pesticide loads for rivers that discharge to the Great Barrier Reef: Great Barrier Reef

Loads Monitoring 2010 – 2011. Water Sciences Technical Report 13, 1. Brisbane (QLD), Australia:

Department of Science, Information Technology, Innovation and the Arts. Water Sciences Technical

Report 13, 1.

University of Hertfordshire. 2013. The Pesticide Properties Data Base (PPDB). Developed by the

Agriculture and Environment Research Unit (AERU), University of Hertfordshire, 2006–2013.

Available from: http://sitem.herts.ac.uk/aeru/ppdb/en/Reports/27.htm Accessed 13 May 2016

US EPA [United States Environmental Protection Agency]. 2008. Framework for Application of the

Toxicity Equivalence Methodology for Polychlorinated Dioxins, Furans, and Biphenyls in Ecological

Risk Assessment. Washington DC, USA: Office of the Science Advisor, US EPA. US EPA 20460.

van den Berg M, Birnbaum LS, Denison M, De Vito M, Farland W, Feeley M, Fiedler H, Hakansson H,

Hanberg A, Haws L, Rose M, Safe S, Schrenk D, Tohyama C, Tritscher A, Tuomisto J, Tysklind M,

Walker N, Peterson RE. 2006. The 2005 World Health Organization reevaluation of human and

mammalian toxic equivalency factors for dioxins and dioxin-like compounds. Toxicol Sci 93, 223–

241.

Wallace R, Huggins R, Smith RA, Turner RDR, Vardy S, Warne MStJ. 2014. Total suspended solids,

nutrient and pesticide loads (2011 – 2012) for rivers that discharge to the Great Barrier Reef –

Great Barrier Reef Catchment Loads Monitoring Program 2011–2012. Brisbane (QLD), Australia:

Department of Science, Information Technology, Innovation and the Arts.


nutrients and pesticide loads (2012–2013) for rivers that discharge to the Great Barrier Reef – Great

Barrier Reef Catchment Loads Monitoring Program 2012–2013. Brisbane (QLD), Australia:


http://sitem.herts.ac.uk/aeru/ppdb/en/Reports/27.htm%20Accessed%2013%20May%202016

38

Wallace R, Huggins R, King O, Gardiner R, Thomson B, Orr DN, Ferguson B, Taylor C, Smith RA, Warne

MStJ, Turner RDR, Mann RM. 2016. Total suspended solids, nutrient and pesticide loads (2014–

2015) for rivers that discharge to the Great Barrier Reef – Great Barrier Reef Catchment Loads

Monitoring Program. Brisbane (QLD), Australia: Department of Science, Information Technology,

Innovation and the Arts.

Walsh, C. 2004. Pharmacology Drug Actions and Reactions. (7. ed.). New York (NY), USA: Informa

Healthcare.

Warne MStJ, Hawker DW. 1995. The number of components in a mixture determines whether

synergistic and antagonistic or additive toxicity predominate: the funnel hypothesis. Ecotoxicol

Environ Saf 31: 23–28.

Warne MStJ, Batley GE, van Dam RA, Chapman JC, Fox DR, Hickey CW, Stauber JL. 2018. Revised

Method for Deriving Australian and New Zealand Water Quality Guideline Values for Toxicants.

Brisbane (QLD), Australia: Department of Science, Information Technology and Innovation.

Warne MStJ, Smith RA, Turner RDR. In prep. Analysis of Mixtures of Pesticides Discharged to the Great

Barrier Reef, Australia. To be submitted to Environmental Pollution.






Wilson PC, Whitwell T and Klaine SJ. 2000. Metalaxyl and simazine toxicity to and uptake by Typha

latifolia. Arch Environ Contam Toxicol 39: 282–288.

39

Chapter 3: An Improved Method for Calculating Toxicity-Based

Pollutant Loads: Part 1. Method Development.

3.1 Abstract

Pollutant loads are a means for assessing regulatory compliance and setting targets to reduce

pollution entering receiving waterbodies. However, a pollutant load is often comprised of multiple

chemicals, which may exert joint toxicity on biota. When the ultimate goal for assessing pollutant

loads is to protect ecosystems from adverse effects of toxicants, then the total pollutant load needs

to be calculated based on the principles of mixture toxicology. In this paper, an improved method is

proposed to convert a pollutant load to a toxicity-based load (toxic load) using a modified toxic

equivalency factor (TEF) derivation method. The method uses the relative potencies (RePs) of multiple

species to represent the response of the ecological community. The TEF is calculated from a percentile

of a cumulative distribution function (CDF) fitted to the RePs. The improvements permit the

determination of which percentile of the CDF generates the most environmentally relevant and robust

toxic loads. That is, environmental relevance ensures that a reduction in the toxic load is likely to result

in a corresponding improvement in ecosystem health and robustness ensures that the calculation of

the toxic loads is not biased by the reference chemical used. The improved methodology will therefore

ensure that correct management decisions will be made and ultimately, a reduction in the toxic load

will lead to a commensurate improvement in water quality.

KEY WORDS Pollutant Loads, Mixtures, Toxic Equivalency Factor, Relative Potency, Multiple species.

40

3.2 Introduction

The total mass of pollutants (loads) is often used by regulators and natural resource managers for

compliance, licensing, and water pollution reduction and control programs (e.g., US EPA, 2000; Raha,

2007; Hardy and Koontz, 2008; Australian Government and Queensland Government, 2013). For

example, the United States’ Clean Water Act requires the development of total maximum daily load

(TMDL) allocations for waterbodies to ensure water quality standards are maintained (US EPA, 2000).

Additionally, the Reef Water Quality Protection Plan (Reef Plan) (Australian Government and

Queensland Government, 2013) sets load-based reduction targets for suspended sediment, nutrients

and pesticides entering the Great Barrier Reef (GBR), located off the east coast of Queensland,

Australia. By meeting the load reduction targets, the 2013 Reef Plan aims to achieve its long-term goal

to ‘ensure that by 2020 the quality of water entering the Reef from broadscale land use has no

detrimental impact on the health and resilience of the GBR’ (Australian Government and Queensland

Government, 2013). In both these examples, there is an inherent assumption that reduction in a

pollutant load will lead to improved ecosystem health. Ensuring this is true becomes complicated

when the pollutant load consists of multiple chemicals which exert joint toxicity on biota.

Load based targets are an appropriate means of reducing and controlling the amount of pollutants

transported to receiving water bodies, however the ecological effects of chemicals are generally

controlled by their inherent toxicity, concentration in the ecosystem and the duration of exposure. A

load reduction of a single chemical is likely to result in some proportional improvement in ecosystem

health and therefore has a degree of environmental relevance. However, when multiple chemicals

with different toxicities are present, meeting load reduction targets, that place equal weighting on

each chemical, may not have the same level of environmental relevance. For example, a measured

percent reduction in the load of a low toxicity chemical will not lead to the same improvement in

ecosystem health as an equivalent reduction in the load of a more toxic chemical. As such load-based

pollutant reduction targets could lead to poorly targeted allocation of resources (effort and dollars)

and/or perverse environmental outcomes.

Weighting the mixture constituents of a load to be more environmentally relevant can be achieved by

incorporating techniques derived from mixture toxicology into the load calculations. The relative

potency (ReP) and toxic equivalency factor (TEF) approaches are well-known for their use with

mixtures of dioxin-like chemicals (van den Berg, et al. 2006; Haws et al., 2006; US EPA, 2008) and are

well-suited for incorporating into pollutant load calculations (Pedersen et al., 2006). The TEF and ReP

methods generate a factor for individual chemicals that can be easily applied to the load calculation,

and will weight the constituents according to their relative toxicity. This permits the effects of the

mixture constituents to be compared or combined on an equitoxic basis. This approach was first

41

demonstrated by Pedersen et al. (2006) who used the RePs of organophosphorus insecticides based

on a single species to weight the total daily maximum loads of mixtures of organophosphate

insecticides in order to ‘assess the potential ecotoxicological significance of their combined presence’.

However, the ReP of one chemical to another varies between species (Putzrath, 1997; Compton and

Sigal, 1999) and, in addition, there is an implicit uncertainty in extrapolating a TEF from one species to

the response of a whole community (De Zwart and Posthuma, 2005). Therefore, in order for pollutant

loads to have a better alignment with potential ecological effects, the RePs of multiple species should

be used to calculate the ‘toxic load’.

A probabilistic approach using a statistical distribution of RePs from a representative group of species

has been suggested (Finley et al., 2003; Haws et al., 2006) to calculate RePs and TEFs. While this

approach will account for multiple species, in many cases a single value from a distribution is preferred

or required (van den Berg et al., 2006). In the case of toxic load calculations, a single value is

preferable, particularly for ease of calculation and communication for compliance and licensing.

Deriving a single value from a representative distribution of species’ ReP values is comparable to

deriving environmental quality guidelines from species sensitivity distributions (SSDs). A SSD is a

cumulative distribution function (CDF) which describes the variation in the sensitivities of a sample of

species that occur in an ecosystem to a toxicant or mixture of toxicants. One of the assumptions of

SSD methods is that the sensitivity of the sample of species is representative of the assessed ecological

community (Posthuma et al., 2002). A SSD determines the concentration of a chemical that should

theoretically protect p % of species (termed either the hazardous concentration to the selected

percent of species e.g. HC5 or the protective concentration e.g. PC95). Defining the context of a

selected percentile of a ReP CDF is more complex and does not necessarily relate to environmental

protection. Specifically, a selected percentile (p) of the ReP CDF represents p% of species for which

the relative potency of the test chemical is up to x times more toxic than the reference chemical

(assuming the reference chemical is less toxic than the test chemical). Thus, there is still a question

remaining as to which percentile of the ReP CDF should be used to derive a TEF that is environmentally

relevant.

In the case of dioxin-like chemicals the World Health Organisation used a combination of unweighted

ReP distributions, expert judgment, and point estimates to derive the TEFs (van den Berg et al., 2006).

As a result, the TEFs derived from a range of percentiles that principally fell within the 50th and 75th

percentiles of the ReP distribution, with the majority being closer to the 75th percentile in order to be

“health protective” (van den Berg et al., 2006). Environment Canada and Health Canada (2001)

calculated TEFs for nonylphenol and its ethoxylates by taking the mean of the ReP values. Similarly,

Kennedy et al. (2010) calculated diuron TEFs for other herbicides by averaging the RePs of coral and

42

microalgae. In all of these cases however, there was no indication as to whether the percentile used

to calculate the TEFs provided an appropriate degree of protection (i.e. environmental relevance).

Furthermore, there was no test to demonstrate that the results weren’t biased by the chosen

reference chemical and the same environmental outcome would be achieved if the reference chemical

was changed, i.e. environmental robustness.

This paper proposes a method that converts annual pollutant loads to toxic loads using a modified TEF

approach that includes tests to maximise the environmental relevance and robustness of the TEF

values. Another paper (Smith et al., 2017) tests the applicability of the new method to a case study –

pollutants discharged from agricultural land to the Great Barrier Reef.

3.3 Definitions of key terms

The definitions of key terms used in this paper are provided below.

Cumulative Distribution Function (CDF) - describes the probability that a variable will be equal to or

less than a specified value. In the case of SSDs, the CDF curve describes the distribution of ecotoxicity

data to a chemical in which species are ranked from the most to the least sensitive (Posthuma et al.,

2002). From a CDF, we can therefore determine the percentage of species that would theoretically

experience adverse effects by any specified concentration of a chemical.

Load (L) - the estimated mass (e.g. t, kg) of a chemical that passes a specified point in a waterway.

Loads are usually calculated on a daily, event or annual basis. The load of an individual chemical (i) is

referred to in this paper as Li. The load of a mixture of chemicals (Lmix) is equal to the sum of the loads

of each chemical in the mixture.

Matched toxicity data - toxicity data from studies conducted within the same laboratory where

multiple chemicals are tested under the same test conditions to a consistent set of organisms.

Relative Potency (ReP) - is the estimate of the potency of the chemical being considered (henceforth

referred to as the ‘test’ chemical) relative to a reference chemical, that both cause a specified toxic

effect in a population, organism, cell or biochemical reaction (US EPA, 2008). The test and reference

chemicals must have the same mode of action (MoA), have parallel concentration-response curves,

and conform with the concentration addition (CA) model of joint action (Safe, 1998). The toxicity data

of the test and reference chemicals used to calculate a ReP must be derived from the same species

and study design, preferably using ‘matched toxicity data’. For one test chemical, there may be a range

of ReP values calculated from multiple species, and multiple ReP values for one species based on

different ecotoxicity endpoints.

43

Toxic equivalence factor (TEF) – the estimate (based on one or more studies) of the potency of a test

chemical relative to a reference chemical that causes a toxic effect. In this paper TEFs are determined

using a cumulative distribution function of the ReP values of multiple species.

Toxic Load (TL) - the product of the TEF for each chemical in a mixture multiplied by the load of each

individual chemical. The toxic load of chemicals with the same MoA can then be summed to generate

a toxic load for the mixture (TLmix). The TL is expressed as an equivalent mass of the reference chemical,

e.g. diuron equivalents (Eq).

3.4 General method for calculating toxic loads

The steps of the proposed general method for calculating TLs are presented in Figure 3.1. The name

‘general method’ was used because it outlines the ‘general’ process, however modifications may be

required to optimise the method for particular jurisdictions or for specific sites (see Smith et al., 2017,

for an example of this). The general method is largely adapted from published procedures used for

calculating RePs and TEFs (Safe, 1998; US EPA, 2008), and generating SSDs for water quality guideline

(WQG) derivation (e.g. Warne, 2001; Warne et al., 2015).

The principle variation to these earlier methods for determining the TEF from a cumulative distribution

function (CDF) is the inclusion of two novel procedures to ensure the most environmentally relevant

and robust TLs are generated. The two new procedures are a test for environmental relevance and a

test for robustness (Step 6, Figure 3.1).

44

Figure 3.1 Flow diagram of steps for calculating toxic loads using a modified TEF approach.

3.4.1 Nominating the reference chemical

Calculating RePs requires a reference chemical to which the potency of the test chemicals being

considered can be compared. Published methods for calculating RePs of particular classes of chemicals

have a suggested standard reference chemical. For example, 2,3,7,8-tetrachlorodibenzo-p-dioxin

45

(2,3,7,8 TCDD) or the polychlorinated biphenyl congener 126 (PCB 126) are the standard reference

chemicals for dioxin-like chemicals (van den Berg, 2006; US EPA, 2008), and estradiol is the traditional

reference for endocrine disruptor chemicals with estrogen mediated receptor responses (e.g.

Gutendorf and Westendorf, 2001). The Ontario Ministry of the Environment (OME) was the first to

propose the TEF methodology (OME, 1984; Haws et al., 2006) and suggested the most toxic and well-

studied member (2,3,7,8 TCDD) of the dioxin-like class of chemicals should be used as the reference

chemical. However, the ‘most toxic’ chemical may not be easy to define for many chemical classes,

due to inter-species variations in toxicity. Even in the case of the dioxin-like chemicals, more recent

research has demonstrated that 2,3,7,8-TCDD is not the most toxic chemical to all species, with other

dioxin-like chemicals having potencies up to four times that of 2,3,7,8-TCDD (Finley et al., 2003).

Furthermore, a fixed reference chemical may not be suitable in all circumstances. For example, it may

be important in communicating results to use a reference chemical that is one of the mixture

constituents present at a site. Ultimately, we consider that the most important requirement in

nominating a reference chemical is that there are sufficient matched data sets between the reference

chemical and each of the mixture constituents to generate reliable ReP CDFs. A reliable CDF is one

that is based on toxicity data that meets the minimum data requirements to generate a CDF (for details

see Step 5). The best reference chemical is therefore the chemical that can make reliable ReP CDFs

with the most chemicals in the suite of chemicals being considered.

3.4.2 Collating and screening toxicity data

Toxicity data required for calculating the RePs of multiple species can be sourced from the scientific

literature. The suggested methods for collating and screening suitable toxicity data for calculating

RePs are as follows:

Toxicity data should be collected from published peer-reviewed studies, published laboratory

reports and/or ecotoxicity databases – un-published data should only be used as a last resort

and if a copy of a document stating the method used is made publicly available (Warne et al.,

2015);

Preference should be given to using matched toxicity data sets (see earlier definition). This

overcomes differences in the data due to variable test conditions and intra-laboratory variation.

Where there are insufficient matched data available, data from separate studies can be used

providing key test conditions were identical or similar (e.g. identical species, test conditions,

measures of toxicity, endpoints).

If the MoA of a chemical is specific to a group of organisms, only species belonging to the target

group should be used (Warne et al., 2015). For example, photosystem II (PSII) herbicides are far

46

more toxic to photosynthetic organisms and therefore only phototrophic species should be

used to calculate RePs.

Preference should be given to using median lethal (LC50) or effect (EC50) concentration data as

these are the points in the concentration-response curve with the least error. This is more

important than using a more sensitive measure of toxicity such as the EC10 or NOEC type data

that are mainly preferred in deriving WQGs (ANZECC and ARMCANZ, 2000; EC, 2011; Warne et

al., 2015).

Preference should be given to toxicity data with exposure durations that are relevant to the

MoA of the chemical and test organism.

Preference should be given to using toxicity data that measure ecologically relevant endpoints

– those that affect the ecological competitiveness of a species (e.g. lethality, immobilisation,

growth, development, population growth, and reproduction – ANZECC and ARMCANZ, 2000;

Warne et al., 2015) or are relevant to the MoA of the chemical.

3.4.3 Determining the quality of toxicity data

All toxicity data that pass the screening process should be assessed by a data quality checking scheme

similar to what is used for generating water quality guidelines (e.g. Klimisch et al., 1997; Durda and

Preziosi, 2000; Hobbs et al., 2005; Schneider et al., 2009; Agerstrand et al., 2014 ; Isigonis et al., 2015;

Warne et al., 2015). Only data of sufficient quality should be used to derive RePs and TEFs.

3.4.4 Calculating relative potencies (RePs)

The ReP is calculated according to Equation 3-1.

Equation 3-1

𝑅𝐸𝑃𝑖 =𝐸𝐶𝑥𝑟

𝐸𝐶𝑥𝑖

where, ECxi is the concentration of chemical ‘i’ that effects x% of a population of the test organism,

and ECxr is the concentration of the reference chemical that effects x% of a population of the test

organism for the same endpoint. We recommend using the midpoint of the concentration-response

curve (i.e. the EC/LC50) as the least amount of error generally occurs at this point in the curve.

A ReP of 1 indicates that the test and reference chemicals are equally toxic, a ReP value < 1 indicates

that the test chemical is less toxic than the reference chemical, and a ReP value > 1 indicates that the

test chemical is more toxic than the reference chemical.

47

3.4.5 Fitting a cumulative distribution function to ReP values

This step adopts the approach of Finley et al. (2003) and Haws et al. (2006) by fitting the ReP data to

a CDF, analogous to the SSD concept. If multiple toxicity tests have been conducted or multiple

experimental conditions were employed (e.g. multiple endpoints measured, different test durations)

then species may have multiple ReP values. A single ReP value is required to represent each species in

the CDF. Data reduction methods such as those used to derive one toxicity value per species in SSDs

(e.g. van de Plassche et al., 1993; Warne et al., 2015) should be used in calculating ReP values. It is

recommended that at least the minimum data requirements for generating a SSD are used when

fitting ReP values to a CDF. Note however, that these minimum data requirements vary with the

jurisdiction. For example, the minimum data requirement for deriving water quality guidelines in

Australia and New Zealand is toxicity data for at least five species from at least four phyla (ANZECC

and ARMCANZ, 2000), in the USA acute toxicity data from species belonging to at least eight different

taxonomic groups and chronic toxicity data for species belonging to at least three different taxonomic

groups are required (Stephan et al., 1985) and in Europe data for at least ten species that belong to at

least eight taxa are required (EC, 2011).

The single ReP values for each species should be collated and analysed using a CDF method (e.g.

BurrliOZ V2 (Barry and Henderson, 2014) and SSD Generator (US EPA, 2012)). This approach, will

permit the calculation of TEF values from the CDF for different percentiles of the ReP CDF.

3.4.6 Selecting the percentile of the ReP cumulative distribution function to calculate the toxic

equivalency factors

It is important to realise that ReP CDFs can be located to the right of REP equals one (i.e. the test

chemical is more toxic than the reference chemical), to the left of one (i.e. the test chemical is less

toxic than the reference chemical), or either side of one (i.e. for some species the test chemical is

more toxic than the reference chemical and for other species it is equal to or less toxic than the

reference chemical). Thus, to determine the TEF that is ‘representative’ of a defined percentile of

species, we need to adjust our calculations to account for the position of the species’ ReP values

relative to the reference chemical (i.e. REP = 1). To illustrate the implications of the above, the CDFs

of the ReP values of a chemical that is less toxic (Chemical A) and more toxic (Chemical B) than a

reference chemical are presented in Figure 3.2. The distribution of ReP values for chemical A sit

principally to the left of the reference chemical (the logarithm of the ReP value for the reference

toxicant is 0 i.e. log10 1 = 0) while the distribution of ReP values for chemical B sit principally to the

right of the reference chemical. Thus, the TEF that represents the ReP values of chemical B (relative

to the reference chemical) for ≤ 95% of species, can be calculated using the 95th percentile of the

distribution (shaded area B). However, the TEF that represents the ReP values of chemical A (relative

48

to the reference chemical) for ≤ 95% of species would be calculated using the 5th percentile of the

distribution (shaded area A) (as this chemical is generally less toxic than the reference chemical). Table

1 provides examples of percentiles that could be used to calculate TEF values depending on the toxicity

of the test chemical relative to the reference chemical.

Figure 3.2 Probability density function of log ReP values. The distribution of ReP values for chemical A sit

principally to the left of the reference chemical (i.e. log ReP = 0) while the distribution of ReP values for

chemical B sit principally to the right of the reference chemical. The shaded area represents the ReP values

for 95% of species for chemical A and B and are calculated using the 5th and 95th percentiles, respectively.

Table 3.1 Example of corresponding percentiles of the relative potency cumulative distribution function (ReP

CDF) for chemicals less toxic and more toxic than the reference chemical.

Percentiles of the ReP CDF

Less toxic than the reference

chemical

More toxic than the reference

chemical

5 95

10 90

20 80

50 50

80 20

90 10

95 5

The next step examines which percentile of the ReP CDF should be selected to calculate the TEFs and

subsequently the TLs. The percentile that is selected determines the value of the TEF which in turn

determines the contribution of each constituent to the TLmix, and therefore, will determine which

49

chemicals become the focus of management action. For example, in Smith et al. (2017) it was

demonstrated that, depending on which percentile was selected, the relative contributions of the

mixture constituents to the TLmix varied by up to 40%. Such a large variation in results could lead to

different management actions and outcomes depending on which percentile was used. Therefore, a

percentile needs to be selected that will generate environmentally relevant and robust TLs.

We recommend an iterative approach (Figure 3.1) in which a percentile is selected to calculate a TEF,

the TLs are then calculated from the TEFs, and lastly the TLs are tested for environmental relevance

and robustness. This process is repeated for different percentiles until the percentile that generates

the optimal (i.e. the most environmentally relevant and robust) set of TEFs and TLs is determined.

Figure 3.3 Key steps within steps six and seven of the Toxic Loads general method.

3.4.7 Testing toxic equivalency factors and toxic loads for environmental relevance and

robustness

3.4.7.1 Calculating toxic loads

The TLmix is calculated by first generating the TLs of each of the mixture constituents using the

following equation:

50

Equation 3-2

𝑻𝑳𝒊,𝒑 = 𝑻𝑬𝑭𝒊,𝒑 × 𝑳𝒊

where TLi,p is the toxic load of chemical i for percentile p of the ReP CDF of chemical i, TEFi,p is the TEF

corresponding to the pth percentile of the ReP CDF of chemical i, and Li = the load of chemical i (kg or

tonnes).

The TLs for each constituent are then summed using the CA model of joint action to calculate TLmix,

i.e.:

Equation 3-3

𝑇𝐿𝑚𝑖𝑥,𝑝 = ∑ 𝑇𝐿𝑖,𝑝𝑖

where, TLmix,p is the toxic load of the mixture for the pth percentile of the ReP CDF, and TLi,p is the toxic

load of chemical i for the pth percentile of the ReP CDF.

The ratio of the TL of each mixture constituent to the TLmix can then be calculated, i.e.:

Equation 3-4

TLi,p:TLmix,p

3.4.7.2 Test for environmental relevance

The test for environmental relevance compares TLi,p:TLmix,p values generated from a selected

percentile, against a similar ratio calculated from an independent method for estimating mixture

toxicity. The multisubstance-potentially affected fraction (ms-PAF) method (see Traas et al., 2002 for

a detailed description) is the recommended independent method.

The TEF method outlined in this paper and the ms-PAF method are similar in that both are probabilistic

techniques that use a sample of the population to generate a CDF (SSD in the case of the ms-PAF). The

ms-PAF method differs from the TEF method by estimating the percent of species that would be

affected by the mixture in question. An advantage of using ms-PAF as an independent method is that

the toxicity data required for producing SSDs do not have to be matched, as with RePs. This means

that larger toxicity datasets with a better representation of species and phyla are often available to

use with the ms-PAF method, and therefore, a more reliable estimate of an ecosystem response is

produced. Unlike the TEF method in which there is a preference for EC/LC50 toxicity values, the No

Observed Effect Concentration (NOEC), No Effect Concentration (NEC), and/or EC/LC10 values are

often preferred for generating SSDs particularly for environmental conservation (e.g. Warne et al.,

51

2015). For this reason, we recommend the use of NOEC/NEC/EC/LC10 toxicity values for ms-PAF

calculations.

The ms-PAF method (for chemicals with the same MoA) is normally a two-stage process, however in

this study we only need to conduct the first stage; calculating the hazard units (HUs). The HUs of each

mixture constituent are calculated from their individual SSDs, according to Traas et al. (2002)

(Equation 3-5);

Equation 3-5

𝐻𝑈𝑖 =𝐶𝑖

�̃�𝑖𝑗

where HUi is the hazard unit for chemical i, Ci is the concentration of chemical i in a sample, and �̃�𝑖𝑗 is

the median EC/LCx (e.g. EC10) of species j to m exposed to chemical i.

The HU values of the mixture constituents are then summed resulting in a hazard unit for the mixture

(HUmix) (Equation 3-6).

Equation 3-6

𝐻𝑈𝑚𝑖𝑥 = ∑ 𝐻𝑈𝑖

𝑖

Using the above equations the contribution of each constituent to the HUmix can be determined

(HUi:HUmix) and compared to the corresponding contribution calculated using the TL method (i.e.

TLi,p:TLmix,p). This should be done for each of the selected percentiles. The percentile which generates

TLi:TLmix ratios most similar to the HUi:HUmix ratios for all constituents would be considered the most

environmentally relevant.

3.4.7.3 Test for robustness

It is important to select the percentile of the ReP CDF which, for any mixture constituent, generates

equal contributions to the TLmix irrespective of the reference chemical used. By doing this, it means

that changing the reference chemical will not change the contribution of each constituent and the

overall assessment of the risk the mixture poses. This would be advantageous if, for example, more

toxicity data became available for one of the other mixture constituents, the constituents of the

mixture changes, or the usual reference chemical is in the process of being phased out or its use

restricted. In any case, the contributions of the mixture constituents to the TLmix should remain the

same and not depend on the reference chemical chosen. Therefore, the percentile which generates a

TLi:TLmix ratio most similar amongst multiple reference chemicals would be considered the most

robust.

52

3.4.8 Adopting the TEFs that generate the most relevant and robust toxic loads

Depending on the case being examined the percentile could be selected to optimise either the

environmental relevance or the robustness or both. We recommend the latter. In the case study

presented in Smith et al. (2017), environmental relevance decreased with increasing percentile of the

ReP CDF while robustness increased with increasing percentile of the ReP CDF. This meant that no

percentile existed that scored the highest in both tests. Therefore, the percentile was selected which

had the highest possible scores from both tests by fitting regression models to the scores of each test

and identifying where the two regression lines met. The two models intersected close to the 75th

percentile, therefore this value was chosen as the percentile to generate the TEFs and hence calculate

the TLs. Coincidently, the 75th percentile was the preferred percentile by WHO when generating TEFs

for dioxin (van den Berg et al., 2006). Until, more case studies similar to Smith et al. (2017) have been

conducted it will not be possible to determine if the 75th percentile is the best to use universally.

3.5 Conclusions

Calculating toxic loads makes ecotoxicological sense when dealing with mixtures of toxicants and will

inform water quality managers on which chemicals and regions they should focus their actions, leading

to better allocation of resources. In this study we developed an environmentally relevant and robust

method for calculating toxic loads using a modified TEF approach. This paper proposes that converting

loads of a chemical mixture to toxic loads was more environmentally relevant than placing equal

weighting on each of the mixture constituents. However, care must be taken in selecting which

percentile of the ReP CDF is used to calculate the toxic loads, as the percentile chosen can have a

marked effect on the environmental relevance and robustness of the toxic load, the relative

magnitude of the toxic load, and the relative contribution of each constituent to the toxic load. Hence,

using the wrong percentile could lead to misguided management decisions, e.g. reducing a toxic load

by more than what is needed, which could have unnecessary economic costs, or not reducing a toxic

load enough which could lead to environmental degradation. A systematic approach that determines

the best percentile, such as that developed in this study, is necessary.

The toxic load method is not without limitations. The requirement for matched ecotoxicity data sets

means that this method may not be suitable for some chemicals, particularly newer ones where

ecotoxicity data are limited. In addition, the requirement for chemicals to have the same MoA and

parallel concentration-response curves means that this method would not be applicable for some

pollutant mixtures.

53

3.6 Acknowledgements

The authors are grateful to Andrew Negri (Australian Institute of Marine Science, Townsville) and

Stephen Lewis (James Cook University, Townsville) for reviewing the manuscript.

3.7 References

Ågerstrand M, Edvardsson L, Rudén C. 2014. Bad reporting or bad science? Systematic data evaluation

as a means to improve the use of peer-reviewed studies in risk assessments of chemicals. Human

and Ecological Risk Assessment: An International Journal, 20(6), 1427-1445.




Queensland.




Zealand.

Barry S and Henderson B. 2014. Burrlioz 2.0. Canberra, Australia: Commonwealth Science and

Industrial Research Organisation. Available from: https://research.csiro.au/software/burrlioz/ .

Accessed December 24, 2014.


Strengths and limitations. Hum Ecol Risk Assess, 5: 33-42.


methodologies. Environ Toxicol Chem 24: 2665-2676.

Durda JL, Preziosi DV. 2000. Data quality evaluation of toxicological studies used to derive

ecotoxicological benchmarks. Hum Ecol Risk Assess 6: 747-765.

EC [European Commission]. 2011. Common implementation strategy for the Water Framework

Directive (2006/60/EC). Guidance document no. 27. Technical guidance for deriving environmental

quality standards. Brussels, Belgium: European Commission.

Environment Canada and Health Canada. 2001. Priority substances list assessment report:

Nonylphenol and its ethoxylates. Canada: Environment Canada and Health Canada, Minister of

Public Works and Government Services.

https://research.csiro.au/software/burrlioz/

http://www.tandf.co.uk/journals/titles/10807039.asp

54


assessments for dioxins, furans, and PCBs. J Toxicol Environ Health Part A 66, 533-550.

Gutendorf B, Westendorf J. 2001. Comparison of an array of in vitro assays for the assessment of the

estrogenic potential of natural and synthetic estrogens, phytoestrogens and xenoestrogens.

Toxicol 166, 79-89.

Hardy SD, Koontz TM. 2008. Reducing nonpoint source pollution through collaboration: policies and

programs across the US states. Environ Manage 4: 301-310.



compounds. Toxicol Sci 89: 4-30.

Hobbs DA, Warne MStJ, Markich SJ. 2005. Evaluation of criteria used to assess the quality of aquatic

toxicity data. Integr Environ Assess Manage 1: 174-180.

Klimisch HJ, Andreae M, Tillmann U. (1997). A systematic approach for evaluating the quality of

experimental toxicological and ecotoxicological data. Regul Toxicol Pharmacol 25: 1-5.

Kennedy K, Paxman C, Dunn A, O'Brien J, Mueller JF. 2010. Monitoring of organic chemicals in the

Great Barrier Reef Marine Park and selected tributaries using time integrated monitoring tools

(2008-2009). Brisbane (QLD), Australia: National Research Centre for Environmental Toxicology

University of Queensland (Entox), University of Queensland.

OME [Ontario Ministry of the Environment] 1984. Scientific criteria document for standard

development. No. 4–84. Polychlorinated Dibenzop- dioxins (PCDDs) and polychlorinated

dibenzofurans (PCDFs). Ontario, Canada: Intergovernmental Relations and Hazardous

Contaminants Coordination Branch, Ontario Ministry of the Environment.

Pedersen JA, Yeager MA, Suffet IH (Mel). 2006. Organophosphorus insecticides in agricultural and

residential runoff: Field observations and implications for total maximum daily load development.

Environ Sci and Technol 40:2120-2127.

Posthuma L, Traas TP, Suter GW II. 2002. General introduction to species sensitivity distributions. In:

Posthuma L, Suter GW II, Traas TP, editors. Species Sensitivity Distributions in Ecotoxicology. Boca

Raton (FL), USA: Lewis Publishers. p 3-11.



55

Raha D. 2007. Paradigm Shift in Water Environment Protection in New South Wales. Aust J Water

Resour 11: 67-78.

Safe SH. 1998. Hazard and risk assessment of chemical mixtures using the toxic equivalency factor

approach. Environ Health Perspect, 106(Suppl 4): 1051-1058.

Schneider K, Schwarz M, Burkholder I, Kopp-Schneider A, Edler L, Kinsner-Ovaskainen A, Hartung T,

Hoffmann S. 2009. “ToxRTool”, a new tool to assess the reliability of toxicological data. Toxicol Lett

189: 138-144.



Queensland, Australia. Integ Environ Manag Assess 13(4): 754–764.

Stephan CE, Mount DI, Hansen DJ, Gentile JH, Chapman GA, Brungs WA. 1985. Guidelines for deriving

numerical national water quality criteria for the protection of aquatic organisms and their uses.

Washington DC, USA:US EPA. US EPA Report No. PB-85-227049.

Traas TP, Van de Meent D, Posthuma L, Hamers T, Kater BJ, De Zwart D, Aldenberg T. 2002. The



pp 315-344.

US EPA [United States Environmental Protection Agency]. 2000. Revisions to the Water Quality

Planning and Management Regulation and Revisions to the National Pollutant Discharge

Elimination System Program in Support of Revisions to the Water Quality Planning and

Management Regulation. Washington DC, USA: US EPA. 65 Federal Register, 43: 586.




US EPA [United States Environmental Protection Agency]. 2012. Species Sensitivity Distribution

Generator, Ver 1. Cincinnati (OH), USA: US EPA, Office of Research and Development, National

Center for Environmental Assessment. Available from:

http://www.epa.gov/caddis/da_software_ssdmacro.html

van de Plassche EJ, Polder MD, and Canton JH. 1993. Derivation of maximum permissible

concentrations for several volatile compounds for water and soil. Bilthoven, The Netherlands:

National Institute of Public Health and Environment Protection. Report No. 679101 008.


56



Walker N, Peterson RE. 2006. The 2005 World Health Organization reevaluation of human and

mammalian toxic equivalency factors for dioxins and dioxin-like compounds. Toxicol Sci 93: 223-

241.

Warne MStJ. 2001. Derivation of the ANZECC and ARMCANZ Water Quality Guidelines for Toxicants.

Australas J Ecotoxicol 7: 123 – 136.

Warne MStJ, Batley GE, van Dam RA, Chapman JC, Fox DR, Hickey CW, Stauber JL. 2015. Deriving

Australian and New Zealand Water Quality Guideline Values for Toxicants. Brisbane (QLD),

Australia: Department of Science, Information Technology and Innovation.

57

Chapter 4: An Improved Method for Calculating Toxicity-Based

Pollutant Loads: Part 2. Application to Contaminants Discharged to

the Great Barrier Reef, Queensland, Australia.

4.1 Abstract

Pollutant loads are widely used to set pollution reduction targets and assess regulatory compliance

for the protection of receiving waterbodies. However, when a pollutant load is comprised of a mixture

of chemicals, reducing the overall load (mass) will not necessarily reduce the toxicity by a similar

amount. This can be overcome by setting targets based on toxicity-based loads (toxic loads), where

the load is modified according to the relative toxicity (expressed as toxic equivalency factors— TEFs)

of each toxicant. Here we present the second paper of a two-part series in which a case study is used

to demonstrate the application of the toxic load method proposed in Part 1. The toxic load method

converts a pollutant load, comprised of multiple chemicals, to a toxicity-based load (toxic load), using

a modified TEF approach. The modified approach is based upon the cumulative distribution of relative

potency (ReP) estimates of multiple species, and is further improved on previously published TEF

methods with the inclusion of two tests to select the percentile of the cumulative ReP distribution

which generate TLs: that align with an independent mixture method (test for environmental

relevance); and are independent of the reference chemical used (test for robustness). Here, the TL

method is applied to mixtures of pesticides that are discharged from agricultural land to the Great

Barrier Reef (GBR) in order to test its utility. In this case study, the most environmentally relevant and

robust TLs were generated using the 75th percentile of the ReP cumulative distribution. The results

demonstrate that it is essential to develop pollution reduction targets based on toxic loads, and

making progress to meeting them will lead to a commensurate reduction in toxic effects caused by

toxicants in waters of the GBR.

KEY WORDS Pollutant Loads, Great Barrier Reef, Photosystem II Herbicides, Toxic Equivalency Factor,

relative potency.

58

4.2 Introduction

Continued concerns about the long-term health and sustainability of the Great Barrier Reef (GBR)

(Haynes and Michalek-Wagner, 2000; Brodie et al., 2008; Brodie et al., 2013) led to the Australian and

Queensland governments jointly establishing the Reef Water Quality Protection Plan (Reef Plan) in

2003 and revising it in 2009 and 2013 (Australian Government and Queensland Government, 2003,

2009, 2013). A key element of Reef Plan was setting aspirational pollutant load reduction targets and

agricultural land management practice targets to improve the quality of water entering the GBR from

adjacent agricultural land. The pollution load reduction targets included ‘at least a 60 per cent

reduction in end-of-catchment pesticide loads in priority areas’ by 2018 (Australian Government and

Queensland Government, 2013). The long-term goal of these load reduction targets was to ‘ensure

that by 2020 the quality of water entering the Reef from broadscale land use has no detrimental

impact on the health and resilience of the GBR’ (Australian Government and Queensland Government,

2013). Fundamental to achieving this long-term goal is the assumption that a reduction in a pollutant

load will lead to a commensurate improvement in ecosystem health. However, when the pollutant

load consists of a mixture of chemicals with different toxicities, meeting the load reduction target may

not achieve the long-term goal of Reef Plan (2013).

Load based targets are often used by regulators and natural resource managers as a means for

quantifying and controlling pollutant inputs under compliance, licensing, and water pollution

reduction and control programs (US EPA, 2000; Raha, 2007; Hardy and Koontz, 2008; Australian

Government and Queensland Government, 2013). However, it is well known that the ecological

impact of a toxicant is determined by the concentration of the chemical in the environment, the

duration and spatial extent of exposure and the sensitivity of the exposed biota (US EPA, 1998). A

pollutant load on the other hand, is the total mass of a pollutant over a given time period (e.g.

tonnes/year).

Two chemicals present together in a catchment with equal concentrations would have equal annual

loads; in terms of meeting a load-based pollutant reduction target, they would therefore be

interpreted as being equally important. However, if the two chemicals have different toxicities then

their ecological impact would be different. Reducing the load of a chemical with a low toxicity will not

lead to the same improvement in ecosystem health as an equivalent load reduction of a chemical with

a high toxicity (Smith et al., 2017). In general, we can assume that ecosystem health would improve

with a reduction in the load of chemicals, however, the improvement would not necessarily be of a

similar magnitude as the decrease in the load (Smith et al., 2017).

59

In the case of the GBR, a total pesticide load is currently quantified by simply summing the loads of

five priority photosystem II (PSII) herbicides which co-occur in Queensland catchments that discharge

to the GBR (Waters et al., 2014). Thus, to meet the Reef Plan pollutant reduction target for pesticides,

a 60% reduction in total pesticide load is required (Australian Government and Queensland

Government, 2013), irrespective of the relative amounts of the five herbicides that are reduced (i.e.

Reef Plan focusses on reducing the loads of ametryn, atrazine, diuron, hexazinone and tebuthiuron).

To achieve this target, it would be logical to focus management efforts on the pesticides with the

largest loads. However, if these pesticides had relatively low toxicities compared to the other

pesticides present, the toxicity to the GBR would not be reduced by the same amount as the reduction

in the total pesticide load. Thus, for managers to ensure that reductions in the pollutant load of a

mixture will result in a proportional reduction in toxicity or adverse environmental effects, the relative

toxicities of mixture constituents need to be considered.

The Toxic Load (TL) method, presented in Smith et al. (submitted to IEAM), uses a modified toxic

equivalency factor (TEF) approach similar to that which has been used for dioxin-like chemicals (van

den Berg et al., 2006; Haws et al., 2006; US EPA, 2008). The TEF is a factor which expresses the potency

of a test chemical relative to a reference chemical, and the modified approach determines this factor

based upon a cumulative distribution of relative potency (ReP) estimates calculated from multiple

species. Similar to the TEF, the ReP estimates also express the potency of the test chemical relative to

a reference chemical, however the ReP estimate is based on a specific toxic effect of a species (at the

population, organism, cellular or biochemical levels) (US EPA, 2008). The ReP estimate should only be

calculated for instances where the toxic effect on the species from the test and reference chemicals

is through the same mode of action (MoA), therefore the concentration-response curves are parallel,

and the concentration addition (CA) model of joint action can be applied (Safe, 1998). The TEF is then

calculated from a selected percentile of the cumulative distribution function (CDF) of the RePs (Smith

et al., 2017). In order to determine which percentile of the CDF should be used, two novel tests were

also introduced to determine the environmental relevance and robustness of the TLs generated from

the selected percentile (Smith et al., 2017). In summation, the TL method aims to transform the loads

of each of the load constituents to a relative toxicity scale such that their weightings are proportional

to their potential ecological impact, and thus they can be summed to calculate a total toxicity-based

load.

In this paper we aim to show that by converting the pollutant loads to a TL using the method proposed

in Smith et al. (2017), load reduction targets can have a greater alignment to improvements in

ecosystem health. In addition, we also aim to demonstrate the capacity to optimise the standard TL

60

calculation method to individual scenarios, and determine the TEFs for each of the GBR priority PSII

herbicides to generate environmentally relevant and robust TLs.

4.3 Methods

The Smith et al (submitted to IEAM) method for TL calculation was followed and adapted for the Great

Barrier Reef case study. The definitions of key terms and formulas referred to in the methods below

can be found in Smith et al. (submitted to IEAM). In brief, ReP estimates for a test chemical relative to

a reference chemical are first determined for multiple species from ecotoxicity data published in the

literature. It is ensured, where possible, that the data used to calculate the RePs are derived from

preferably the same study and/or laboratory, and at a minimum, the same endpoint, exposure period

and test conditions are used. A CDF is fitted to the RePs of multiple species for the test chemical

(relative to the reference chemical). The CDF is considered reliable based on its fit to the ReP values

and if it meets the minimum data requirements, i.e. ReP values from five species and four phyla

(Warne et al., 2015). TEFs are then calculated from the CDF based on a selected percentile, the

pollutant loads are multiplied by the TEF to generate the TLs, and lastly the TLs are tested for their

environmental relevance and robustness (see below for a more detailed explanation). This last step is

repeated, testing a range of TEFs (calculated from different percentiles of the CDF) for generating the

most environmentally relevant and robust TLs. Greater detail of these steps, specific to this case study,

are presented below.

4.3.1 Pesticide loads data

Annual loads (Ls) of five priority PSII herbicides (i.e. ametryn, atrazine, diuron, hexazinone and

tebuthiuron) from nine catchments (i.e., Johnstone, Tully, Herbert, Haughton, Burdekin, Pioneer,

Plane, Fitzroy and Burnett) were obtained from Turner et al. (2013) and Wallace et al. (2014). These

five PSII herbicides all have the same MoA; inhibition of electron transport at the plastiquinones QA

and QB in the D1 protein of the electron transport chain which drives photosynthesis in phototrophic

species (Jones, 2005). These type of herbicides have previously been shown to follow the CA model

(Faust et al., 2001; Backhaus et al., 2004; Magnusson et al., 2010).

For all catchments apart from the Haughton and Plane, the load of the main river (of the same name)

was used. In the Haughton and Plane catchments, the loads for Barratta Creek and Sandy Creek

respectively, were used.

Loads data for two years (2010 – 11 and 2011 – 12) were chosen to represent inter-annual variation

(that may influence calculations) from a longer-term monitoring program (2009 – present). The

methods for pesticide monitoring, sample analysis and load calculations are detailed in Turner et al.

(2013) and Wallace et al. (2014).

61

4.3.2 Ecotoxicity data

Ecotoxicity data were sourced, collated and quality checked for calculation of RePs for individual

species according to Smith et al. (2017). The US EPA Ecotoxicity Database (US EPA, 2015) and the

scientific literature were searched for appropriate toxicity data. Only data published since 1980 were

accepted due to limitations in the experimental design and methods in earlier work (Warne, 2001). In

addition, only phototrophic species were considered as PSII herbicides specifically inhibit

photosynthesis. Toxicity to non-phototrophic organisms would be exerted through a different mode

of action (MoA) and therefore would not be suitable to calculate ReP values (for an explanation see

Warne et al., 2015 and Smith et al., 2017). Studies using freshwater, estuarine and marine

phototrophic species were considered: (i) to ensure there were sufficient data to generate cumulative

distribution functions (CDFs) for ReP values; and (ii) so the resulting TEF values would be relevant to

all three aquatic ecosystems through which the pesticides are transported. The non-parametric

Kruskal Wallis test was used to assess whether there were significant differences in the ReP values of

species from freshwater, marine and estuarine ecosystems. A non-parametric test was used due to

the small sample sizes (i.e. sample sizes ranged from 5 – 18) and unequal group sizes.

Species of the GBR from important functional groups, such as coral (and their symbiotic zooxanthellae)

and seagrass, were, where available, also included in the data set of each herbicide. In the case where

no ecotoxicity data for ecologically relevant endpoints were available for these groups of species, data

measuring photosynthetic inhibition were permitted (photosynthesis is not considered as an

ecologically relevant endpoint for the derivation of the Australian and New Zealand Guidelines for

Fresh and Marine Water Quality, i.e. those that affect the ecological competitiveness of a species e.g.

lethality, immobilisation, growth, development, population growth, and reproduction or are relevant

to the MoA of the chemical (ANZECC and ARMCANZ, 2000; Warne et al., 2015)).

All ecotoxicity data were quality checked using the approach of Hobbs et al. (2005). Matched data sets

(toxicity data from studies conducted within the same laboratory where multiple chemicals are tested

under the same test conditions to a consistent set of organisms (Smith et al., 2017) were preferentially

selected, however, in cases where the minimum data requirements for generating the CDF (five

species from four phyla; Warne et al., 2015) could not be met, data from different studies were

considered. In these cases, experimental conditions of the studies were compared and if they were

sufficiently similar, the data were used. There was no preference for the duration of exposure in the

toxicity tests, except to be ≥ 24 h. If data from multiple exposure periods for a species were available,

the geometric mean of the RePs was calculated.

62

4.3.3 Reference chemicals and ReP distributions

All five priority PSII herbicides were assessed for their suitability to be the reference chemical (for an

explanation see Smith et al., 2017). Chemicals that met the minimum data requirements for

generating the CDF were considered to be more suitable as the reference chemical than those that

did not.

The CDFs of ReP values for each combination of two PSII herbicides (e.g. ametryn and atrazine,

ametryn and diuron, ametryn and hexazinone, ametryn and tebuthiuron) were generated using

Burrlioz 2.0 (Barry and Henderson, 2014). This software fits either a log-logistic (Equation 4-1) or a

Burr Type III distribution (Equation 4-2) to the ReP data, depending on the number of data points. The

log-logistic distribution is the default when there are 7 or fewer data (in this case ReP values) and the

Burr Type III distributions are used when there are 8 or more data.

Equation 4-1

𝑦 =1

1 + (x/α)−β

Equation 4-2

y = (1 + x−c)−k

4.3.3.1 Testing toxic equivalency factors and toxic loads for environmental relevance and robustness

4.3.3.1.1 Calculating toxic loads

A TEF is derived from the ReP CFD and is used to calculate the TLs. The TEF is the maximum ReP of a

selected percentile of species. Hence, the value of the TEF is dependent on the selected percentile of

the CFD, and therefore, so is the value of the TL, as shown from Equations 1 and 2 from Smith et al.

(submitted to IEAM):

Equation 4-3

𝑇𝐿𝑖,𝑝 = 𝑇𝐸𝐹𝑖,𝑝 × 𝐿𝑖

where TLi,p is the toxic load of chemical i for percentile p of the ReP CDF of chemical i, TEFi,p is the TEF

corresponding to the pth percentile of the ReP CDF of chemical i, and Li = the load of chemical i (kg or

tonnes); and,

Equation 4-4

𝑇𝐿𝑚𝑖𝑥,𝑝 = ∑ 𝑇𝐿𝑖,𝑝

𝑖

63

where, TLmix,p is the toxic load of the mixture for the pth percentile of the ReP CDF, and TLi,p is the toxic

load of chemical i for the pth percentile of the ReP CDF.

Therefore, we examined the degree of variation in the calculated TLs based on a range of selected

percentiles: the 50th, 70th, 75th, 80th and 95th percentiles if the test chemical was more toxic than the

reference chemical and the 50th, 30th, 25th, 20th and 5th percentiles respectively, when the test chemical

was less toxic than the reference chemical (the explanation for this is provided in Smith et al.

submitted to IEAM). The above percentiles were used to calculate an annual TLi for each PSII herbicide

and annual TLmix for each catchment.

4.3.3.1.2 Test for environmental relevance and robustness

The tests for environmental relevance and robustness (Smith et al., 2017) were developed to jointly

determine which percentile of the ReP CDF should be used to calculate the TEFs. Firstly, for the TLs to

be environmentally relevant, the ratio of TLi to TLmix must be commensurate with the relative toxicity

of chemical i compared to the toxicity of the mixture. This ensures that a reduction in TLi will generate

a relative improvement in ecosystem health. To do this, the TLi,p:TLmix,p (where p is the percentile being

tested) is compared to a similar ratio calculated using an independent method (the multi-substance

potentially affected fraction, ms-PAF, method (Traas et al., (2002)). Secondly, the test for robustness

determines the percentile which generates the most consistent ratio of TLi:TLmix irrespective of the

reference chemical used (Smith et al., 2017).

To test which TEFp generated the most environmentally relevant TLs, we followed the first stage of

the ms-PAF procedure for response addition (outlined in Traas et al., 2002) to calculate hazard units

(HUs). The methods for calculating the HUs of individual chemicals (HUi) and mixtures (HUmix) are

described in detail in Smith et al. (2017). In summary, species sensitivity distributions (SSDs) for the

five PSII herbicides were generated using the methods of Warne et al. (2015), from which the HU for

each PSII herbicide was calculated, according to Traas et al. (2002) (Equation 4-5):

Equation 4-5

𝐻𝑈𝑖 =𝐶𝑖

�̃�𝑖𝑗

where HUi is the hazard unit for chemical i, Ci is the concentration of chemical i in a sample, and �̃�𝑖𝑗 is

the median EC/LCx (e.g. EC10) of species j to m exposed to chemical i.

The HU values of the PSII herbicides in a mixture are then summed resulting in a hazard unit for the

mixture (HUmix) (Equation 4-6):

64

Equation 4-6

𝐻𝑈𝑚𝑖𝑥 = ∑ 𝐻𝑈𝑖

𝑖

The contribution of each constituent to the HUmix (HUi:HUmix) were then calculated and compared to

the TLi,p:TLmix,p.

From the monitored PSII herbicide concentrations (Turner et al., 2013; Wallace et al., 2014), we

calculated Ls, TLs and HUs based on the percentiles previously described, and compared the results.

A simple scoring method was applied which gave a score of 1 when the TLi:TLmix value (for a selected

percentile) was within 1 S.D. of the average HUi:HUmix and a score of 0 when it was outside this range.

For each percentile, the percent of cases (catchments × years) with a score of 1 was determined. The

percentile with the highest score was regarded as the one which produced the most environmentally

relevant TLs. The Li:Lmix ratios were also compared against the average HUi:HUmix (± 1 S.D.) ratios and

similarly scored in order to assess the environmental relevance of the summed Ls not converted to a

TL.

A similar scoring method was applied for the robustness test. For each percentile, the TLi:TLmix ratio

was calculated using multiple reference chemicals. The difference in the TLi:TLmix between reference

chemicals was calculated and a difference ≤ 5% was given a score of 1. For each percentile, the percent

of cases (catchments × years) with a score of 1 was then determined. The percentile with the highest

score was regarded as the one which calculated the most robust TLs.

4.4 Results and Discussion

4.4.1 Ecotoxicity data and the reference chemicals

Matched ecotoxicity data were sourced from twenty-three studies found in the literature (see Table

A.1 in the supplementary material for the references). There were a limited number of studies in which

ecotoxicity data for species from important functional groups of the GBR and other local species could

be used (based on the data selection criteria). The GBR species included for which there were data

were; the seagrass species, Halodule uninervis and Zostera muelleri (Flores et al., 2013), and

zooxanthellae of the coral species, Acropora millepora (Negri et al., 2011). Other local species that had

suitable data were Nephroselmis pyriformis and Navicula sp. (Magnusson et al., 2008), both estuarine

benthic microalgae. Ecotoxicity data for zooxanthellae from the coral species Seriatopora hystrix

(Jones and Kerswell, 2003) were also used when limited data were available. Data for this species were

65

not generally used as the test conditions did not meet all the preferred test requirements; i.e. the

endpoint measured was photosynthetic inhibition and the exposure period was < 24 h.

The number of species and phyla for which matched toxicity data were available for each combination

of test and reference chemical are presented in Table 1. For ease of discussion, combinations of test

and reference chemicals will be referred to using the following format — test:reference, e.g.

atrazine:diuron, where atrazine is the test chemical and diuron is the reference chemical. In all

test:reference combinations involving tebuthiuron, there were insufficient matched data to meet the

minimum data requirements, although the combinations with diuron, atrazine and hexazinone were

close, i.e. five species from three phyla. Combinations between ametryn and hexazinone also had

insufficient data to produce reliable ReP CDFs. For all other combinations (excluding tebuthiuron)

involving atrazine and diuron as the reference chemicals, reliable ReP CDFs were generated. Thus,

atrazine and diuron were considered the best chemicals for use as a reference chemical.

Statistically significant differences in ReP values of species from different ecosystem types

(freshwater, marine and estuarine) were tested on all combinations of test chemicals with atrazine

and diuron as the reference chemicals (Table A.2). In all cases there were no significant differences,

which is further demonstrated in the CFDs presented in Figure 4.1 and Figure A.1). Previous studies

have compared the sensitivity of marine and freshwater species to different toxicants and the

evidence is inconclusive and contradictory (e.g., Hutchinson et al., 1998; Leung et al., 2001; Wheeler

et al., 2002), and all have at least one of the following limitations: limited data for each comparison;

limited chemicals being compared; the fresh and marine toxicity data having markedly different

organism composition, inappropriate species toxicity data being used (e.g. comparing the sensitivity

distribution for all species to chemicals with specific MoA). We note the guidance provided on

combining toxicity data for freshwater and marine in the Technical Guidance Document for Deriving

Environmental Quality Standards (EC, 2011). This states that ‘in principle, ecotoxicity data for

freshwater and saltwater organisms should be pooled for organic compounds’ but this ‘must be

tested, except where a lack of data makes a statistical analysis unworkable’ (EC, 2011).

To determine whether atrazine or diuron was the better reference chemical, the total number of

species and phyla that constituted the CDFs for all combinations of test and reference chemicals (for

atrazine and diuron) were compared. While atrazine had slightly more data compared to diuron as the

reference chemical (i.e., ametryn:atrazine had data for one more species than ametryn:diuron and

hexazinone:atrazine had data for one more phyla than hexazinone:diuron, Table 1), these were not

sufficiently large differences to prefer atrazine over diuron as the reference chemical. Therefore, all

subsequent calculations were done using both chemicals as the reference chemical.

66

4.4.2 Relative Potency Cumulative distribution functions

The ReP CDFs (Figure 4.1 and the rest can be found in the Supplementary Material (Figure A.1))

indicated that the relative toxicities of the five priority PSII herbicides to the majority of phototrophic

species were, in decreasing order, diuron, ametryn, hexazinone, atrazine and tebuthiuron. In most

cases, the majority of ReP values (one per species) in each plot, lay either to the left or right of ReP =

1 (Figure 4.1B–D). However, in one case, ametryn:diuron (Figure 4.1A), species’ ReP values were

positioned on both sides of ReP = 1, indicating that diuron was more toxic than ametryn for slightly

more than half the species (~65%).

The approximately equal potencies between ametryn and diuron caused some problems when

choosing which percentile of the ReP CDF to use for TL calculations (Smith et al., 2017). The tests for

environmental relevance and robustness of different percentiles of the ametryn:diuron ReP CDF (data

not shown) found that the 50th percentile should be used. This phenomenon is likely to hold true for

all test chemicals with approximately equal toxicities to the reference chemical.

67

Figure 4.1 Cumulative distribution functions (CDF) of the relative potency values for (A) ametryn:diuron, (B)

hexazinone:diuron, (C) diuron:atrazine, and (D) tebuthiuron:atrazine. Symbols for ReP values represent

freshwater (○), marine (●), and estuarine (Δ) species. Blue dotted lines indicate equal toxicity, i.e. ReP = 1. All

plots generated by BurrliOZ V2 (Barry and Henderson, 2014).

4.4.3 Toxic Equivalency Factors

The TEFp values were extracted from all ReP CDFs with atrazine, diuron and hexazinone as the

reference chemicals (Table 2). The test for robustness requires all chemicals to be tested as reference

chemicals, however, the reliability of the ReP CDFs of hexazinone, ametryn and tebuthiuron (as

reference chemicals) was low. It was decided to also include hexazinone as a reference chemical in

the robustness test to validate the results of the diuron and atrazine tests; hexazinone was slightly

68

more reliable than ametryn as tebuthiuron:hexazinone had data for five species whereas

ametryn:tebuthiuron only had data for four (Table 1)

The values of the TEFs varied over the range of percentiles tested. In some cases the differences were

large, for example, ametryn:atrazine TEFs differed by a factor of approximately eight. In other cases,

the differences were much smaller; tebuthiuron:atrazine TEFs differed by a factor of 2.4. These

variances in the TEFs resulted in differences in the contributions each PSII herbicide had to the TLmix

(i.e. the TLi:TLmix ratio), the degree of which was dependant on the catchment (Figure 4.2). For

example, in the Fitzroy River (Figure 4.2A), there was up to a 40% difference in the diuron contribution

to TLmix depending on which percentile was selected. Similarly, the results from Barratta Creek

(Haughton Catchment) showed that the contribution of diuron varied by up to 24% between

percentiles (Figure 4.2B). In contrast, the Pioneer River results showed only small differences between

percentiles in the TLi:TLmix values (Figure 4.2C). The reason for these differences between catchments

is likely a result of the relative composition of the herbicides on a mass basis, which is discussed later.

These observed differences in the TLi:TLmix in some catchments could lead to different management

outcomes depending on the percentile used, which is why it is necessary to select a percentile that is

environmentally relevant and robust.

There were also notable differences in the calculated TLmix values depending on which percentile was

used. The TLmix values (diuron equivalent, i.e. using diuron as the reference chemical) calculated from

the 50th percentile (i.e. TLmix,50) were the largest and decreased as the percentile used to calculate the

TLs increased (Figure 4.3). The greatest difference in a TLmix calculation from one site was 83% at

Burdekin River (2010–11) (Figure 4.3). Substantial differences in TLmix estimates such as these could

influence management actions. For example, a 60% reduction of TLmix,50 (diuron equivalent) would

require a greater reduction in the mass of pesticides than a 60% reduction of TLmix,95. Conversely, the

TLmix values calculated using atrazine and hexazinone as the reference chemicals showed the opposite

trend; the TLmix values increased from TLmix,50 to TLmix,95. These results highlight the need for testing

which percentile should be used to calculate the TEFs.

69

Figure 4.2 Percent contributions of five PSII herbicides to the toxic load of the mixture (TL i:TLmix) calculated

using different toxic equivalence factors (in turn calculated using different percentiles of the relative potency

cumulative distribution functions). Contributions are for (A) the Fitzroy River, (B) Barratta Creek (Haughton

Catchment) and (C) the Pioneer River, estimated using 2010–11 data.

0% 20% 40% 60% 80% 100%

50th

70th

75th

80th

95th

TLi:TLmix (Diuron Eq.)

Pe

rce

nti

le o

f R

eP

CD

FA.

0% 20% 40% 60% 80% 100%

50th

70th

75th

80th

95th

TLi:TLmix (Atrazine Eq.)

Pe

rce

nti

le o

f R

eP

CD

F

B.

0% 20% 40% 60% 80% 100%

50th

70th

75th

80th

95th

TLi:TLmix (Diuron Eq.)

Pe

rce

nti

le o

f R

eP

CD

F

Ametryn

Atrazine

Diuron

Hexazinone

Tebuthiuron

C.

70

Figure 4.3 Difference (expressed as a percentage) in the total toxic loads for each catchment (2010–11)

calculated using different percentiles of the relative potency (ReP) cumulative frequency distributions (diuron

equivalents) compared to the toxic loads calculated using the 50th percentile (TLmix,50).

4.4.4 Environmental relevance and robustness

For the environmental relevance test, both atrazine and diuron were tested as the reference chemical

and the scores were combined to give one score for each percentile tested. Scores from the

environmental relevance and the robustness tests showed an inverse relationship (Figure 4.4). The

environmental relevance score was highest at the 50th percentile in which 98% of the tested cases

agreed with the HUs calculated by the independent method (ms-PAF) and decreased as the percentile

increased. Conversely, robustness was highest at the 95th percentile (94%) and lowest at the 50th

percentile (81%) (Figure 4.4). Accordingly, it was decided that the optimal percentile of the ReP CDF

to use in calculating TEFs would have the highest possible scores from both tests. To determine this,

the scores were fitted with regression models and the percentile where the two regressions

intersected was defined as the optimal for calculating TLs. A 5-parameter sigmoid model and a linear

model were fitted to the environmental relevance (R2 = 0.99) and robustness scores (R2 = 0.99),

respectively (SigmaPlot v12.5, Systat Software, San Jose, CA). The models intersected close to the 75th

percentile, which scored 92% for environmental relevance and 89% for robustness (Figure 4.4).

Although these scores were not the highest recorded individually for each test, they were the highest

values for both scores at the same percentile. Fortuitously, this recommended percentile to calculate

TEFs for the GBR case study is the same as that used by the World Health Organisation (van den Berg

71

et al., 2006). But for other case studies this may not be the case. It is too early to say whether the

similarities in these results are an indication that the 75th percentile can be used universally for any

case study. Once more case studies are tested using the methods outlined in Smith et al. (2017) a

more solid conclusion can be made.

Figure 4.4 Scores of environmental relevance (●) and robustness (○) as a function of the percentile of the

relative potency (ReP) cumulative distribution function (CDF). The percentiles tested were the 50th, 70th (or

30th), 75th (or 25th), 80th (or 20th) and 95th (or 5th).

The Li:Lmix ratios had a lower environmental relevance score compared to the corresponding ratios for

TLs. Only 53% of cases agreed with the HU results (data not shown) which clearly demonstrates the

need to weight pollutant loads based on their toxicity to ensure they are environmentally relevant.

These results demonstrate that using TLs rather than Ls to quantify load reduction targets would more

likely lead to a relative improvement in ecosystem health.

The high scores from the robustness test demonstrated that the method would provide the same

result irrespective of which herbicide was used as the reference chemical. Therefore, the selection of

reference chemical for calculating the TLs for the GBR, i.e. atrazine or diuron, had to be based on other

factors. For example, atrazine is more commonly detected across catchments (Smith et al., 2012) and

recent restrictions of diuron application may mean that diuron becomes less relevant to the region’s

water quality issues in the future. On the other hand, historically, diuron has been used as a reference

chemical in expressing the concentration of PSII herbicides in catchments (Lewis et al., 2012) and

marine waters of the GBR (Gallen et al., 2013, 2014). Given that stakeholders in the GBR region are

72

familiar with diuron equivalent concentrations, it makes sense to continue with this approach. Thus,

to calculate the TL and TLmix values for this case study, diuron was used as the reference chemical and

the 75th percentiles of the ReP CDFs were used to derive the TEFs, except in the case of ametryn:diuron

in which the 50th percentile was used.

4.4.5 Differences between loads and toxic loads

The recommended TEFs (Table 2) were used to derive TL and TLmix values for the nine GBR catchments

and were compared to their corresponding Lmix values. Substantial differences were found in the

percent contributions of each PSII herbicide to the mixture depending on whether this was

determined in terms of load (Li:Lmix) or toxic load (TLi:TLmix) (Figure 4.5). Overall, the contribution of

diuron to the TLmix was larger than its contribution to the Lmix. This is logical given diuron is more toxic

than atrazine, hexazinone and tebuthiuron. Conversely, due to the low toxicity of atrazine, its

contribution to the TLmix was considerably smaller than to the Lmix where it was the dominant herbicide.

For example, at Barratta Creek (Haughton Catchment), where atrazine had the greatest contribution

to Lmix (77%), its contribution to the TLmix was reduced to just 11%. For the Fitzory and Burdekin

catchments, the contribution of the tebuthiuron load was also considerably reduced when it was

converted to a TL. Tebuthiuron is used in the grazing industry and therefore contributes a large

proportion of the load in catchments dominated by grazing, i.e. Burdekin and Fitzroy. However, due

to the low toxicity of tebuthiuron compared to the other PSII herbicides, its contribution to the TLmix

is minor. Converting Ls to TLs therefore should change the focus of pesticide management. For

example, in the Burdekin and Fitzroy catchments management action should shift from reducing

tebuthiuron loads to reducing diuron loads, even though the mass is considerably less, in order to

achieve the maximum reduction in toxicity and maximum increase in ecosystem health.

Converting Ls to TLs also changes the relative differences in the magnitude of the loads between

catchments. The Fitzroy and Burdekin catchments have high Lmix values compared to the other

catchments, but low TLmix values (Figure 4.6A and Figure 4.6B, respectively). This is because they are

dominated by the least toxic of the five PSII herbicides; atrazine and tebuthiuron. Catchments with a

higher dominance of diuron, e.g. Tully, Herbert and Pioneer rivers, have a higher TLmix relative to the

other catchments. The differences observed here between the Ls and TLs have important implications

for prioritising catchment areas for management and using the wrong method could lead to less

effective allocation of resources and smaller progress to achieving the overall goal of Reef Plan.

73

Figure 4.5 Percent contributions of the constituents of a mixture of five PSII herbicides calculated in terms of

(A) the load (Li:Lmix) and (B) the toxic load (TLi:TLmix) for nine Great Barrier Reef catchments in 2011–12. Toxic

loads were calculated using the diuron equivalent TEFs presented Table 3.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

L i:L

mix

A.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

TLi:T

L mix Tebuthiuron

Hexazinone

Diuron

Atrazine

Ametryn

B.

74

Figure 4.6 The 2011–2012 annual Total Loads (A) and Total Toxic Load (B) for nine Great Barrier Reef

catchments. Total toxic loads were calculated using the bolded TEFs (diuron equivalent) in Table 3.

4.5 Conclusions

The results from this paper, (i) confirmed the proposition of Smith et al. (2017) that pollutant load

mixtures should be converted to a toxic load to ensure ecosystem protection, (ii) determined the

optimal percentile of the ReP CFD for calculating PSII herbicide TLs for the GBR, and (iii) demonstrated

the case-specific optimisation of the method.

75

Weighting the mixture load constituents based on their toxicity changed the contribution of each of

the PSII herbicides to the mixture load, and the magnitude of the loads for each catchment. In doing

so, the toxic loads were shown to be more environmentally relevant than loads following verification

by an independent mixture method (ms-PAF). Converting loads to toxic loads will better inform water

quality managers on which chemicals and regions they should focus their actions, leading to a more

environmentally beneficial allocation of resources.

Determining which percentile to use to calculate the TEF from the ReP CDF was shown to be a critical

step in the TL method. The tests for environmental relevance and robustness provided a systematic

approach for selecting the optimal percentile to calculate TLs. The 75th percentile was, for Great

Barrier Reef catchments, determined to be the optimal percentile for improvements in ecosystem

health and ensuring that load reductions were not biased by the toxicity selection of the reference

chemical. However, it is not certain that the 75th percentile would be optimal in all case studies, and

each should be assessed individually.

As was demonstrated in this paper, the TL method can be optimised for each case study, i.e.: a specific

chemical mixture; the choice of reference chemical; inclusion of species from important functional

groups of the GBR and other local species; and the TEF percentile - selected for environmental

relevance, robustness, or both, as was demonstrated here. However, it may not be applicable to all

pollutant load scenarios. For one, there is a requirement for matched ecotoxicity data sets to calculate

the RePs. For some chemicals, particularly newer ones, ecotoxicity data may be limited and calculating

enough ReP values to meet the minimum data requirements set for fitting the CDF may not be

feasible. In addition, the ReP and TEF methods are limited to chemicals with the same MoA, i.e. those

with parallel concentration-response curves. Thus for mixtures of chemicals with different MoA,

calculating ReP values between two chemicals with non-parallel concentration-response curves would

incur a degree of error depending on the variation in the shapes of the concentration-response curves.

Thus it is not recommended to use this method for these types of mixtures, unless the error can be

quantified.

4.6 Acknowledgements

The authors are grateful to Andrew Negri (Australian Institute of Marine Science, Townsville) and

Stephen Lewis (James Cook University, Townsville) for reviewing the manuscript. In addition, we

would like to acknowledge the financial contributions of the Queensland Department of Natural

Resources and Mines for the collection of pesticide concentration data used in this study.

76

4.7 References










Queensland.

Backhaus T, Faust M, Scholze M, Gramatica P, Vighi M, Grimme LH. 2004. Joint algal toxicity of

phenylurea herbicides is equally predictable by concentration addition and independent action.

Environ Toxicol Chem 23(2): 258-264.














low-effect concentrations of individual toxicants. Aq Toxicol 56(1):13-32.

Flores F, Collier CJ, Mercurio P, Negri AP. 2013. Phytotoxicity of four photosystem II herbicides to

tropical seagrasses. PloS One 8: e75798.


77


Great Barrier Reef using both time-integrated and event monitoring techniques (2012-2013).








programs across the US states. Environ Manage 4: 301–310.







toxicity data. Integr Environ Assess Manage 1: 174–180.

Hutchinson TH, Scholz N, Guhl W. 1998. Analysis of the ECETOC aquatic toxicity (EAT) database IV –

Comparative toxicity of chemical substances to freshwater versus saltwater organisms.

Chemosphere 36(1): 143–153.


Ser 261: 149–159.

Jones R. 2005. The ecotoxicological effects of Photosystem II herbicides on corals. Mar Poll Bull 51(5):

495–506.

Leung KMY, Morritt D, Wheeler JR, Whitehouse P, Sorokin N, Toy R, Holt M, Crane M. 2001. Can

saltwater toxicity be predicted from freshwater data? Mar Poll Bull 41(11): 1007–1013.




78

Magnusson M, Heimann K, Negri AP. 2008. Comparative Effects of Herbicides on Photosynthesis and

Growth of Tropical Estuarine Microalgae. Mar Poll Bull 56: 1545–1552.



Negri AP, Flores F, Röthig T, Uthicke S. 2011. Herbicides increase the vulnerability of corals to rising

sea surface temperature. Limnol Oceanogr 56: 471–485.


Resour 11: 67–78.




Smith RA, Warne MStJ, Mengersen K and Turner RDR. 2017. An improved method for calculating


746–753.

Traas TP, Van de Meent D, Posthuma L, Hamers T, Kater BJ, De Zwart D, Aldenberg T. 2002. The

potentially affected fraction as a measure of ecological risk. In: Posthuma L, Suter GW II, Traas TP,


p 315–344.



Loads Monitoring 2010–2011. Brisbane (QLD), Australia: Department of Science, Information

Technology, Innovation and the Arts. Water Sciences Technical Report 13, 1.






US EPA [United States Environmental Protection Agency]. 2015. ECOTOX User Guide: ECOTOXicology

Database System. Version 4.0. [cited 2012/09/09]. Available: http://www.epa.gov/ecotox/



http://www.epa.gov/ecotox/

79

Walker N, Peterson RE. 2006. The 2005 World Health Organization re-evaluation of human and


241.









Australia: Department of Science, Information Technology, Innovation and the Arts.

Waters DK, Carroll C, Ellis R, Hateley L, McCloskey GL, Packett R, Dougall C, Fentie B. 2014. Modelling

reductions of pollutant loads due to improved management practices in the Great Barrier Reef

catchments – Whole of GBR. Toowoomba (QLD), Australia: Queensland Department of Natural

Resources and Mines. Technical Report 1.

Wheeler JR, Leung KMY, Morritt D, Sorokin N, Rogers H, Toy R, Holt M, Whitehouse P, Crane M. 2002.

Freshwater to saltwater toxicity extrapolation using species sensitivity distributions. Env Tox Chem

21(11):2459–2467.

80

Chapter 5: Advancements to the msPAF method for assessing

mixtures of photosystem II herbicides

5.1 Abstract

It is recognised that pesticide contamination must be reduced to improve water quality transported

from Queensland catchments to the Great Barrier Reef. Phototsystem II (PSII) herbicides are the

most prevalent pesticides detected in GBR ecosystems and recognised as the highest risk herbicides

to key GBR species such as coral and seagrass. The presence of these herbicides in mixtures has

called for the development of a robust and environmentally relevant method that can calculate their

additive risk and demonstrate improvements in water quality with the adoption of improved

agricultural management practices. Here we investigate the use of the multisubstance Potentially

Affect Fraction method (msPAF) published by Traas et al. (2002) to calculate the risk (percent of

species affected) of mixtures of 13 PSII herbicides commonly found in GBR ecosystems. As these

herbicides have the same mode of action (MoA), the concentration addition approach was applied

to the species sensitivity distributions (SSDs) published in King et al. (2017a, b). It was found that the

SSDs violated assumptions of a normal distribution and parallelism, therefore the msPAF method

was modified to account for these. The main modification was to generate global PSII herbicide SSDs

from two groups of SSDs separated based on parallelism that can calculate the msPAF of any mixture

combination of the 13 PSII herbicides. The result allows for robust estimates of the percent of

species affected by mixtures of PSII herbicides.

5.2 Introduction

Management efforts to protect the Great Barrier Reef (GBR) have highlighted the necessity to

improve water quality transported from agricultural land to the Reef. Photosystem II (PSII) inhibiting

herbicides have been identified as one of the priority pollutant groups for improving GBR water

quality (Australian Government and Queensland Government, 2009; Waterhouse et al., 2017).

Photosystem II herbicides are used widely in sugar cane and other intensive cropping land uses, have

high mobility and are found regularly in agricultural runoff discharged via catchments to the GBR

lagoon (Smith et al., 2012; Devlin et al; 2015; Turner et al., 2012, 2013; Wallace et al., 2014, 2015,

2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017). Photosystem II herbicides have a direct mode

of action (MoA) on photosynthetic organisms; binding to plastiquinones QA/QB in photosystem II,

thereby blocking electron transport that drives photosynthesis (University of Hertfordshire, 2013;

Wilson et al., 2000). Thus, they present a risk to some key GBR species, including zooxanthellae (the

81

symbiotic algae of coral) and seagrass (Jones and Kerswell, 2003; Negri et al., 2005; Haynes et al.,

2000; Gao et al., 2011). In addition, PSII herbicides are often detected in GBR ecosystems together as

mixtures and with other types of pesticides (Smith et al., 2012; Smith et al., 2017; Warne et al., in

prep).

Water quality guideline values are used internationally, including in Australia (where they are

referred to as default guideline values; DGVs), to assess the hazard of single pesticides to aquatic

ecosystems by providing a concentration that is protective of P% of species (ANZECC and ARMCANZ,

2000; Warne et al., 2018). The desired level of protection defines P and is attributed to the values of

the ecosystem; for example, at least 99% species protection is sought for high ecological value

ecosystems and 95% species protection for ecosystems that are slightly to moderately disturbed

(ANZECC and ARMCANZ, 2000; Warne et al., 2018). However, DGVs only provide protection levels

for single toxicants, whereas, up to 26 pesticides have been detected together in GBR catchments

with an average of 6 pesticides detected per sample (Warne et al., in prep).

Recently, species sensitivity distributions (SSDs) were developed for ‘proposed’ pesticide DGVs for

the GBR (King et al., 2017a, b). Species sensitivity distributions are cumulative distributions that

describe the sensitivity of a set of species to a contaminant (Posthuma et al., 2002). The sensitivity of

each species in an SSD is determined from ecotoxicological data for the species, i.e. an effective

concentration (ECp) that elicits a response from p% of the population. The ecotoxicity data that are

available for a range of species (generally sourced from the literature) are collated and used to

represent the distribution of species sensitivity in an ecosystem, from which the distribution

parameters can be estimated (Posthuma et al., 2002). Australian DGVs are preferentially generated

from SSDs compiled from chronic EC10 data – the concentration that elicits a response from 10% of

the population of a species (ANZECC and ARMCANZ, 2000; Warne et al., 2018). The development of

SSDs for 13 PSII herbicides (King et al., 2017a, b), commonly detected in GBR ecosystems also

provided an opportunity to estimate the risk posed by mixtures of these herbicides using the

multisubstance potentially affected fraction (msPAF) method (Traas et al., 2002).

The msPAF method has several advantages over other methods such as the toxic equivalency or

hazard unit methods. The advantages are as follows:

(i) It provides an assessment of the impact of multiple species (as opposed to single

species);

(ii) It covers mixtures of chemicals with similar and dissimilar MOA;

(iii) It provides an estimate of the risk expressed as a % of species affected that is directly

relatable to water quality guideline protection levels; and,

82

(iv) It uses SSDs, which are also used to generate toxicant water quality DGVs (Warne et al.,

2018).

The msPAF method incorporates the concentration addition (CA) and response addition (RA)

approaches to estimating the toxicity of mixtures (Traas et al., 2002), depending on whether the

mixture constituents have similar or dissimilar MoA, respectively. The CA method combines the SSDs

of each constituent within a mixture and thereby can estimate the msPAF of the mixture by

generating a single SSD representative of the constituents with the same MoA (Figure 5.1). The SSDs

are combined by rescaling the ECp values of each local SSD to a relative scale by dividing them by

their median ECp value (𝐸�̃�) (Figure 5.1, left graph) such that the median equals 1 and the new units

on the x-axis are dimensionless and referred to as hazard units (HUs). The distribution of each SSD is

shifted along the x-axis to centre at x=1 (Figure 5.1, right graph).

Figure 5.1 Illustration of the concentration addition approach for calculating the multisubstance Potentially

Affected Fraction. The left hand graph shows species sensitivity distributions (SSDs) of three chemicals with

the same mode of action but different relative toxicities. The median value (𝐄�̃�) of each SSD is used to rescale

the SSDs to Hazard Units (HUs) so that the SSDs are brought together and centre at x=1 (right hand figure). A

single cumulative distribution function is then fitted to the HUs of the three chemicals to generate a SSD for

the mixture of the three chemicals (SSDmix).

A single SSD for the mixture constituents (SSDmix; Figure 5.1, right graph) is then generated by

estimating the parameters of a logistic cumulative distribution function that represents the pooled

HUs of the mixture constituents. Based on the methods of Traas et al. (2002), the SSDmix can be

mathematically defined by two parameters, �̅� – the slope calculated from the mean of the individual

slopes (β) of the individual constituent’s SSDs, and α – the mean of the log ECp data, which by

definition should equal 0 after rescaling the SSDs to HUs (Traas et al., 2002).

83

The msPAF of an environmental sample can be estimated by converting the concentration of each

mixture constituent in the sample to HUs by dividing the concentration in the sample by the

corresponding 𝐸�̃� of the constituent. The HUs of each constituent in the sample are then summed

(∑ 𝐻𝑈) and fed into the logistic CDF describing the SSDmix. For the msPAF CA method published by

Traas et al. (2002) to provide confident estimates of risk, the ECp values within each SSD should be

normally distributed, the βs of the mixture constituents should also be normally distributed, and the

SSDs should be parallel.

However, the SSDs of pesticides can often be of low reliability due to limited availability of

appropriate ecotoxicity data and a poor fit of the SSD to the toxicity data (Warne et al., 2018; King et

al., 2017a, b). The 13 PSII herbicide SSDs reported in King et al. (2017a, b), were preferentially

generated from chronic EC10 data, which was limited for a number of herbicides. In cases where

there is insufficient EC10 data, no observed effect concentration (NOEC) values are used or,

equivalent chronic EC10 values are estimated by converting acute data to chronic data using an

acute to chronic ratio and/or EC50, EC25 and lowest observed effect concentration (LOEC) values are

converted to an estimated EC10 value (Warne et al., 2018; King et al., 2017a, b). Including NOEC or

converted data in the SSD could violate the assumptions of a normal distribution and parallelism

previously mentioned, but even if they do not they will introduce additional uncertainty into the

calculation of the SSDs.

To deal with the potential violations described above, a proposed alternative approach is to

calculate the HUs from α, rather than 𝐸�̃�, determined from fitting a logistic CDF to the normalised

(log transformed) distribution of the ECp data. This will ensure homogeneity in the distribution of

the ECp values either side of log(x)=0 and therefore, greater cohesion between the local SSDs when

they are pooled to generate the SSDmix. An alternative to estimating �̅�, is to use a global curve fitting

approach that uses least-squares optimisation to determine the CDF parameters from the pooled

HUs. As with ordinary least-squares regression, the global curve fit method determines the

parameter values that minimize a sum of squares.

Thus the aim of this study was to determine the parameters of the logistic CDF that generate the

most robust and environmentally relevant estimates of the msPAF for PSII herbicides for the

management of water quality impacting on GBR ecosystems. Specific objectives were to:

i. To test whether the method published by Traas et al. (2002) could be used to achieve the

aim of the study for 13 PSII herbicides commonly detected in GBR ecosystems;

ii. To examine alternative methods to the Traas et al., (2002) method that would generate the

most robust and environmentally relevant msPAF estimates;

84

iii. Validate the developed methods with case studies to ensure the msPAF estimates are both

robust and environmentally relevant.

5.3 Summary of Terms

SSDmix – Species sensitivity distribution (SSD) representing a mixture of chemicals generated from

the toxicity data of only the mixture constituents.

Global SSD - Species sensitivity distribution (SSD) generated from fitting a global curve fit to the

toxicity data of a sample of chemicals with the same MoA (mode of action), which represents all

chemicals from that MoA. For example, the global PSSII herbicide SSD is the SSD

Local curve fit - the fitting of a logistic CDF (Equation 5-2) to an individual data set, i.e. separate

logistic CDF parameters (α and β) to represent the SSD of each PSII herbicide; and

Global curve fit - the fitting of a logistic CDF (Equation 5-2) to multiple data sets, i.e. one set of α and

β to represent the SSDs of multiple PSII herbicides.

5.4 Methods

To achieve the study objectives, we used an iterative approach of testing and adapting the msPAF

method to best suit the PSII herbicide ecotoxicity data. This approach involved:

1. Testing the appropriateness of using α of the log transformed EC values rather than the 𝐸�̃�,

to calculate the HUs;

2. Test for robustness - a validation case study to test the validity of �̅� for generating the

SSDmix;

3. Testing the individual PSII herbicide SSDs for parallelism;

4. Using the global curve fit method to generate one global PSII herbicide SSD to be used for all

possible PSII herbicide mixtures;

5. Test for environmental relevance - validate the best method for calculating msPAF against

the default GVs for individual PSII herbicides.

The methods involved in each step are detailed below.

5.4.1 Collation of ecotoxicity data

Ecotoxicity data (the effective concentration of a chemical that causes a defined toxic response from

a species) for marine and freshwater phototrophic species were collated from King et al. (2017a, b)

85

to generate SSDs for 13 PSII herbicides. There were toxicity data for a total of 113 species; 86 were

freshwater species and 27 were marine species. The SSDs for each herbicide were constructed from

ecotoxicity data for both freshwater and marine species, where available, as the GBR ecosystems

cover fresh, estuarine and marine habitats. An advantage of using both fresh and marine water

ecotoxicity data is that it increased the number of data used in the SSDs and therefore increased the

statistical robustness of the SSDs. The number of species in each SSD varied depending on the

availability of data and ranged from six species (propazine) to 56 species (atrazine). According to

Warne et al. (2017) the minimum data requirements for generating a SSD to calculate DGVs is five

species belonging to four taxonomic groups, although data for more species are preferred. All 13

SSDs generated in this study met these requirements.

Preferentially, SSDs used to generate Australian and New Zealand water quality guidelines are

generated using chronic EC10 data, i.e. the concentration of a toxicant that effects 10% of the

population of a species (Warne et al., 2018). However, due to the paucity of EC10 data for some PSII

herbicides, other toxicity values were used; NOEC, and converted LOEC, EC50, EC25 and acute data

as permitted by Warne et al. (2017). The type of ecotoxicity data (chronic vs acute), the number of

species for which there were data and the fit of the selected statistical distribution to the data

determine the reliability of the SSDs and the DGVs generated from them (as defined by Warne et al.,

2018). Proposed DGVs for the 13 PSII herbicides were also collated from King et al. (2017a, b), which

were generated from the same ecotoxicity data. The proposed DGVs and their reliability are

presented in Table B.1 and Table B.2 (Supplementary Material).

5.4.2 Testing the msPAF method

5.4.2.1 Calculation of HUs

Two methods were assessed for calculation of the HUs. The first, used the median EC value (EC̃) as

the denominator as per Traas et al. (2002), that is:

Equation 5-1

𝐻𝑈𝑖𝑗

=𝐸𝐶𝑖

𝑗

𝐸𝐶𝑖̃

where, i = 1 to m herbicides, j = 1 to v species, HU is the hazard unit, EC is the effective

concentration and EC̃ is the median EC.

86

The alternative method was developed for SSDs where the ECp data do not follow a normal

distribution. In this case, the mean of the distribution (α) was estimated from a logistic CDF fitted to

the log(ECp) data using SigmaPlot (Systat Software, Inc., San Jose, CA). The logistic CDF is described

by:

Equation 5-2

𝑦 =1

1 + 𝑒−(

𝑥−𝛼𝛽

)

where, x = log(ECp), α is the mean of the distribution and β is the slope.

As α is determined from the log transform of ECp, to calculate the HUs from ECp we need to back

transform from the log scale, such that:

Equation 5-3

𝛼′ = 𝑒𝛼

The HUs were then calculated according to the following equation:

Equation 5-4

𝐻𝑈𝑖𝑗

=𝐸𝐶𝑖

𝑗

𝛼′𝑖

Where, i = 1 to m herbicides, j = 1 to v species, and EC is the effective concentration.

5.4.2.2 Estimating msPAF

Multiple approaches were examined for estimating msPAF from mixtures of PSII herbicides. The first

approach for calculating the msPAF follows the methods outlined by Traas et al. (2002). A local curve

fit to the log transformed HUs was conducted for each of the 13 PSII herbicides using SigmaPlot

v14.0, from which the local α and β’s were derived.

According to Traas et al. (2002) the parameters for the SSDmix can be estimated from the parameters

derived from the local curve fits of each of the mixture constituents: i.e., α=0 as it is the mean of the

log(HU)’s, and; the slope can be calculated from the mean of the local βs, i.e. �̅�. Thus,

87

Equation 5-5

𝑦 = 1

1 + 𝑒−(

𝑥

�̅�)

where x = log(ECp), �̅� is the slope.

To then calculate the msPAF for an environmental sample containing a mixture of PSII herbicides,

the concentration of each herbicide in the mixture is converted to the HU scale using EC̃ or 𝛼′ as the

denominator as in Equation 5-4, as follows:

Equation 5-6

𝑆𝑎𝑚𝑝𝑙𝑒 𝐻𝑈𝑖 =𝐶𝑖

𝐸𝐶𝑖̃

or,

Equation 5-7

𝑆𝑎𝑚𝑝𝑙𝑒 𝐻𝑈𝑖 =𝐶𝑖

𝛼′𝑖

where, i = 1 to m herbicides, C = concentration of herbicide i in the environmental sample, EC̃ is the

median EC, and 𝛼′ is calculated according to Equation 5-3. The HU for the whole mixture in the

environmental sample (Sample HUmix) is then calculated from summing the HUs of each herbicide, as

follows:

Equation 5-8

𝑆𝑎𝑚𝑝𝑙𝑒 𝐻𝑈𝑚𝑖𝑥 = ∑ 𝑆𝑎𝑚𝑝𝑙𝑒 𝐻𝑈

𝑖

The msPAF for the environmental sample (msPAFsample) can then be calculated as follows:

Equation 5-9

𝑚𝑠𝑃𝐴𝐹𝑠𝑎𝑚𝑝𝑙𝑒 = 1

1 + 𝑒−(

𝑆𝑎𝑚𝑝𝑙𝑒 𝐻𝑈𝑚𝑖𝑥

�̅�)

88

The second approach, global curve fit, is similar but fits a single logistic CDF (Equation 5-5) to the

pooled toxicity data of multiple herbicides and estimates the global CDF parameters, α and β, using

the least squares method. The global curve fit was carried out using SigmaPlot v14.0. The derived

global α and β values were then used to estimate the msPAF.

Equation 5-10

𝑚𝑠𝑃𝐴𝐹𝑠𝑎𝑚𝑝𝑙𝑒 = 1

1 + 𝑒−(

𝑆𝑎𝑚𝑝𝑙𝑒 𝐻𝑈𝑚𝑖𝑥−𝛼𝛽

)

5.4.3 Tests for parallelism

The msPAF CA approach assumes SSDs are parallel for chemicals with the same MoA. This means

that β is common across all PSII herbicide SSDs. Therefore, the βs of the CDF of each PSII herbicide in

a mixture should not be significantly different from the β of the global curve fitted to the pooled

toxicity data of the mixture constituents. The assumption that β was the same for each pair of

herbicides was tested using the sum of squares F test (p ≤ 0.05) (Chèvre et al., 2006; Fleetwood et

al., 2015). The F test is a classical statistical test that compares two models by assessing the

difference in the sum of squares (Fleetwood et al., 2015).

The global β was tested for equivalence to the local β of each combination of herbicide pairs, based

on the following equation:

Equation 5-11

𝐹 =𝑆𝑆𝐺𝑙𝑜𝑏𝑎𝑙 − 𝑆𝑆𝐿𝑜𝑐𝑎𝑙

𝐷𝐹𝐺𝑙𝑜𝑏𝑎𝑙 − 𝐷𝐹𝐿𝑜𝑐𝑎𝑙

𝑆𝑆𝐿𝑜𝑐𝑎𝑙

𝐷𝐹𝐿𝑜𝑐𝑎𝑙⁄

where, SS = sum of squares, DF = degrees of freedom, global = global curve fit, Local = local curve fit.

5.4.4 Validation

Case study: testing �̅� to estimate msPAF

To test the Traas et al. (2002) approach of using �̅� to calculate msPAF (Equation 5-9), the SSDmix for

12 paired mixtures of atrazine with a second PSII herbicide were generated and assessed for the

89

variations in the msPAF outputs. For each mixture, �̅� was calculated as a pairwise average of the

slopes of two herbicides, one being atrazine, i.e.

Equation 5-12

�̅� =𝛽𝑎𝑡𝑟𝑧 + 𝛽ℎ𝑒𝑟𝑏

2

where 𝛽𝑎𝑡𝑟𝑧 is the slope of atrazine and 𝛽ℎ𝑒𝑟𝑏 is the slope of one of the other 12 PSII herbicides.

The msPAF(%) was then calculated for each mixture combination according to Equation 5-9 with x =

[atrazine]1, [atrazine]2, … [atrazine]8, where [atrazine]8 = 𝛼 of the atrazine local SSD. As a reference,

the msPAF(%) was calculated from the local atrazine SSD for each [atrazine]. The variance was

assessed by calculating the squared difference of msPAF(%) of the reference relative to the atrazine

paired mixtures, i.e.;

Equation 5-13

var(x) = ∑(xi − xj)

2

N

𝑁

𝑖=1, 𝑖≠𝑗

where xi is the ith msPAF(%) value, xj is the msPAF(%) value which corresponds to atrazine (i.e. the

reference value), and the total number of msPAF values is N=13.

Comparison with proposed guidelines values for single chemicals

Based on the results of this study that determined the best approach for estimating the msPAF of

PSII herbicides (i.e. the global curve fit), the protective concentration (PC) values were calculated for

99% and 95% of species, i.e. the PC99 and PC95, respectively. These were compared against the

PC99 and PC95 estimated from each herbicide’s local SSD and the King et al. (2017) proposed DGVs.

To determine the PC99 and PC95, the logistic function was rearranged for x, with y = 0.99 and 0.95,

respectively, as follows:

Equation 5-14

𝑥 = −𝛽(𝑙𝑛(1 − 𝑦) − 𝑙𝑛(𝑦))−∝

90

5.5 Results and Discussion

5.5.1 Local curve estimates

Species sensitivity distributions for each of the 13 PSII herbicides are shown in Figure 5.2 and the

derived local parameters along with R2 values are presented in Table 5.1. The R2 values indicated the

logistic CDF was a good fit for each of the individual PSII SSDs, the poorest fit of the model was for

hexazinone SSD (hexazinone R2 = 0.89).

Tests for normality (Shapiro-Wilk test; p ≥ 0.05) indicated the SSDs for individual PSII herbicides were

generally not normally distributed (Table 5.1). The EC̃ value was calculated and compared against

the local α’ for each herbicide (Table 5.1). In some cases, the difference between the EC̃ and the

local α was substantial (up to 42% of α). Thus, the local α values were used to calculate the HUs.

There were notable differences in the slopes of the SSDs suggesting that the SSDs were not parallel

(Figure 5.2). The βs did vary considerably between PSII herbicides ranging from 0.162 to 0.820 (Table

5.1), with an average of 0.387. For chemicals with similar MoAs, the βs should be approximately

equal for the SSDs to be parallel. In addition, to use the mean (i.e. �̅�) to represent the slope of the

SSDmix, the local βs should be normally distributed, however, this was not the case (Shapiro-Wilk

test; p = 0.024). The variations in β may be due to differences in the robustness of the SSDs, however

there did not seem to be a relationship between the deviation from the mean (�̅�) and n (Table 5.1).

91

Figure 5.2 Logistic cumulative frequency distributions of the 13 photosystem II herbicides showing the

variation in slopes and potency.

Table 5.1 Model parameters derived from the local curve fit of a logistic cumulative distribution function to

the ecotoxicity data of 13 individual PSII herbicides and summary statistical information (n, R2 and normality).

The PSII herbicides are presented in order from highest to lowest n.

PSII Herbicide n Median ecotoxicity

value

(g/L)

α’

(g/L)

β R2 Test for Normality (p value)

Atrazine 56 20.78 34.72 0.6787 0.95 < 0.001

Diuron 46 2.99 3.28 0.3732 0.98 < 0.001

Simazine 23 83.00 82.11 0.2969 0.96 < 0.001

Metribuzin 19 6.20 6.83 0.2503 0.95 < 0.001

Terbutryn 19 5.60 9.71 0.8201 0.97 < 0.001

Terbuthylazine 18 12.93 11.14 0.4695 0.98 < 0.001

Ametryn 17 5.20 5.24 0.3629 0.99 < 0.001

Prometryn 9 8.00 8.00 0.3631 0.96 < 0.001

Hexazinone 8 6.46 7.84 0.3265 0.89 < 0.001

Tebuthiuron 7 56.00 52.22 0.2854 0.98 0.004

Bromacil 7 17.00 15.70 0.3844 0.96 < 0.001

Fluometuron 7 124.09 140.77 0.1620 0.91 0.136

Propazine 6 14.50 14.26 0.2629 0.96 0.01

Average 0.387

92

5.5.2 Case Study – testing �̅� to estimate msPAF.

The results from this case study demonstrated the estimated percent of species affected attributed

to atrazine alone, varied depending on which other herbicide was present in the mixture (Table 5.2).

According to the CA model, this should not be the case for chemicals with similar MoA, as the toxic

effects are additive, not synergistic or antagonistic (Faust et al., 2001). That is, the percent of species

affected by a given concentration of atrazine should be constant irrespective of the other PSII

herbicides present (assuming concentrations aren’t at the top end of the SSD). If the SSDs for the 13

PSII herbicides were parallel, the percent of species affected by x g/L of atrazine would not vary

across the tested mixtures. However, there is a notable difference in the results relative to the % of

species affected determined from the atrazine SSD (Table 5.2). The variance in these differences also

changed as the concentration of atrazine increased to α (Table 5.2).

The local βs of the mixture constituents determines �̅�, and due to the variance in the local βs of the

PSII herbicides, �̅� can vary considerably depending on which local βs are used. For this reason, a

possible solution would be to use just one set of global parameters to calculate msPAF that

represented all PSII herbicides, i.e. parameters derived from a global PSII herbicide SSD, to be used

irrespective of which PSII herbicides were present in an environmental sample. This would be

possible if the local SSDs were parallel.

Table 5.2: Results of the atrazine case study – the maximum (A) and minimum (B) msPAF (%) values calculated

from set atrazine concentrations (0.1 to 34.72(α) g/L) estimated using �̅� (Equation 5-12) of 12 paired atrazine

mixture combinations. Reference msPAF(%) values (A) were calculated from the local atrazine SSDs. The

variance was assessed by calculating the squared difference of the msPAF(%) values (A to B) relative to the

reference (C) according to Equation 5-13.

Atrazine Concentrations

(g/L) 0.1 5 1 5 10 20 30

34.72 (α)

A. Maximum msPAF% – paired atrazine mixtures 3.3 8.0 11.5 24.9 33.1 42.5 48.3 50.7

B. Minimum msPAF% – paired atrazine mixtures 0.2 1.3 2.6 12.2 22.2 36.8 47.0 50.0

C. msPAF% – atrazine reference 2.3 6.2 9.4 22.4 31.1 41.3 47.7 50.0

Variance 2.6 13.0 22.8 42.0 29.1 6.5 0.1 0.4

5.5.3 Tests for parallelism

Based on the results from the assessments reported above, in particular the observation that the

SSDs may not be parallel, even though the herbicides have the same MoA, we statistically compared

the local SSDs of each herbicide pair for parallelism. These tests revealed that only 58% of the tested

93

pairs were parallel (Table 5.3). In fact, no single herbicide was parallel to all other tested herbicides.

The 13 PSII herbicides could be separated into two main groups. The atrazine and terbutryn SSDs

were found to be parallel to one another but not parallel to any other herbicide and were therefore

grouped on their own and hereafter termed Group A. The 11 remaining herbicides were, on average,

parallel to seven out of the other 10 PSII herbicides within the group and hereafter termed Group B.

The groupings are also confirmed by the differences in local β values, where the βs for atrazine and

terbutryn (0.6787 and 0.8201), are considerably higher than the βs of the remaining 11 herbicides

(Table 5.3). Fluometuron had a distinctly lower β than the rest of the herbicides, nonetheless, it was

still parallel to seven other herbicides in the second group (Table 5.3).

Theoretically the SSDs for the 13 PSII herbicides should all be parallel as they all have the same MOA

(De Zwart, 2002; Traas et al., 2002; Chèvre et al., 2006). Deviations from parallelism in response

curves of PSII herbicides were also reported by Faust el. (2001) and Chèvre et al. (2006). Possible

explanations for the deviations: other secondary MoAs for some of the herbicides, the ecotoxicity

data deviating from the true values due to the use of conversion factors, inconsistencies in p

between NOEC and EC10, variations in the end-points used, small sample sizes, the wrong CDF

model fitted to the data, or a combination of these. The possibility of a secondary MoA is discussed

in the next section.

De Zwart (2002) suggested that the number of species in an SSD determine the reliability of the

shape of the curve, with a recommendation of 25 to 50 observations for each SSD. Therefore, the

relationship between the number of parallel matches and number of observations (n) in each SSD

was examined (Figure 5.3). A small sample size generally resulted in a poorer fit of the SSD to the

ecotoxicity data and larger confidence intervals (King et al., 2017a, b). However, a lack of data did

not seem to be the determining factor for non-parallel matches here. There was instead a negative

relationship between n and the number of parallel matches between herbicide SSDs – although the

low R2 value (0.43) indicates a high level of uncertainty in this relationship.

The deviations from parallelism could also be due to the use of various measures of toxicity i.e.

NOEC and converted acute, LOEC, EC50 and EC25 data, which would carry with it a level of error

compared to the true EC10 value for a species. In addition, using EC10 and NOEC values that sit at

the bottom end of a concentration-response curve, also have an inherent level of uncertainty due to

the nature of variability within the concentration-response models – i.e. variability in measurements

is highest at the top and bottom of the curve with the least variability in the middle of the curve.

Generating SSDs based solely on chronic EC50 values is likely to generate SSDs closer to the shape of

94

their “true” CDF and result in a potential improvement in parallelism of SSDs for chemicals with

similar MOAs.

Table 5.3: Test for parallelism between paired species sensitivity distributions fitted with logistic cumulative

distribution functions. Grey shading indicates pairs of SSDs that are parallel.

PSII Herbicide β Te

rbu

tryn

Atr

azin

e

Terb

uth

ylaz

ine

Bro

mac

il

Diu

ron

Pro

met

ryn

Am

etry

n

Hex

azin

on

e

Sim

azin

e

Teb

uth

iuro

n

Pro

paz

ine

Met

rib

uzi

n

Flu

om

etu

ron

Terbutryn 0.820 A

Atrazine 0.679 A

Terbuthylazine 0.470 B

Bromacil 0.384 B

Diuron 0.373 B

Prometryn 0.363 B

Ametryn 0.363 B

Hexazinone 0.327 B

Simazine 0.297 B

Tebuthiuron 0.285 B

Propazine 0.263 B

Metribuzin 0.250 B

Fluometuron 0.162 B

A = PSII herbicides belonging to Group A based on their parallelism; B = PSII herbicides belonging to Group B

based on their parallelism.

Figure 5.3: Relationship between the number of ecotoxicity values per species sensitivity distribution (SSD)

and the number of parallel SSD pairs. Dotted line represents a linear trendline.

R² = 0.4306

0

2

4

6

8

10

12

0 10 20 30 40 50 60

Nu

mb

er o

f p

aral

lel p

airs

Number of ecotoxicity values

95

5.5.4 Global curve estimates

Based on the results from the case study reported above that demonstrated large variation in results

based on the constituents of the SSDmix, we wanted to test the concept of generating one global PSII

herbicide SSD that could represent mixtures of all combinations of PSII herbicides. A similar idea was

suggested by De Zwart who said that a common β represented chemicals from the same MoA and

this common β could be used in cases where toxicity data for a chemical was too limited to derive an

SSD (2002). To do this, the PSII ecotoxicity data were fitted using the global fit method. However,

given the results from the test for parallelism, two approaches were tested: (i) the global curve fit to

the HUs of all 13 PSII herbicides (Figure 5.4); and (ii) the global curve fits to two groups of herbicides

based on their parallelism – Group A consisting of two herbicides, and Group B consisting of 11

herbicides (Figure 5.5A and B).

The slopes (β) and R2 values of the logistic models are presented in Table 5.4 along with �̅� and R2

calculated using the Traas et al. (2002) method for comparison. A visual inspection of the global

curve fits showed narrower 95% prediction bands for approach (ii) i.e. global curves fitted to groups

A and B compared with approach (i) a global curve fitted to the full data set of 13 herbicides (Figure

5.5 and Figure 5.4), respectively). This, along with the high R2 values (Table 5.4), indicated that

separating the data based on parallel groups generates better fits of the global fit models. The HUs

of the 13 PSII herbicides were also fitted with the logistic CDF model generated using the Traas et al.

(2002) approach (�̅�) which demonstrated the poorest fit (lowest R2 value) of the three tested

methods. However, it should be noted that the fit using the Traas et al. (2002) method was only

marginally worse than the fit using the global fit method on data for all 13 PSII herbicides.

The slope of the global curve fitted to the herbicides in Group B (Figure 5.5B) was steeper (β =

0.3387; Table 5.4) than that for group A (β = 0.7267; Table 5.4, Figure 5.5B). The data distribution in

Figure 5.5A indicates a possible bimodal response – characterised by two inflection points in the

curve (identified by the arrows). The slope at the first point of inflection is steeper that the slope of

the remainder of the data set and may in fact be similar or parallel to the slope of Group B (Figure

5.5B). A bimodal distribution would indicate a secondary MoA (King et al., 2017a, b). However, the

species in this dataset are all phototrophic species which would suggest one MoA – blocking of the

electron transport chain. Secondary MoAs have been reported for some PSII herbicides, but these

relate to endocrine disruption in animal species, e.g. atrazine. However, for phototrophic species,

endocrine disruption is not a possible MoA. Discerning the cause of the bimodal distribution would

help to determine whether the more sensitive species should be used on their own in the atrazine

and terbutryn SSDs, which could ultimately result in the alignment of these SSDs with the other 11

96

PSII herbicides belonging to Group B. In doing so, one global PSII herbicide SSD could be generated

to reliably represent all PSII herbicides.

In summary, these results indicated the second approach, i.e. first separating the ecotoxicity data

into two groups before applying a global curve fit, was more robust. Hence, this second approach

was used in the following section; comparing the risk estimates from the global SSDs against the

proposed DGVs.

Figure 5.4 Global species sensitivity distribution based on ecotoxicity data of 13 PSII herbicides (atrazine and

terbuthylazine, fluometuron, metribuzin, propazine, tebuthiuron, simazine, hexazinone, ametryn, prometryn,

diuron, and bromacil). The black line represents the logistic cumulative distribution function fitted using the

global curve fit method, blue lines represent the 95% confidence bands and the red lines represent the 95%

prediction bands.

97

Figure 5.5 Global species sensitivity distributions for two groups of PSII herbicides: (A) group A - two PSII

herbicides (atrazine and terbuthylazine) and (B) Group B - eleven PSII herbicides (fluometuron, metribuzin,

propazine, tebuthiuron, simazine, hexazinone, ametryn, prometryn, diuron, and bromacil). The black line

represents the logistic cumulative distribution function fitted using the global curve fit method, blue lines

represent the 95% confidence bands and the red lines represent the 95% prediction bands. Arrows in (A)

represent points of inflection.

Table 5.4 Logistic cumulative frequency curve parameters (α and β) and their fit (R2 value) to ecotoxicity data

of different groupings of the PSII herbicides and using the global fit and Traas et al. (2002) methods.

Method for parameter estimation

No. of PSII herbicides (n)

β R2

Traas et al (2002) 13 0.387 0.9070

Global Curve Fit - All PSII herbicides

13 0.4414 0.9113

Global Curve fit - Group A

2 0.7267 0.9451

Global Curve Fit - Group B

11 0.3387 0.9870

5.5.5 Comparison with Guideline Values

In water quality management, it is important that the single toxicant pesticide DGVs and the msPAF

method give consistent results, given that management efforts are likely to use either or both

methods to evaluate and communicate risk. However, a number of differences in the methods used

to calculate the two sets of values have the potential to generate different results. Firstly, in bringing

together the local SSDs to create global SSDs to calculate the msPAF, an inherent level of uncertainty

accumulates (Figure 5.4 and Figure 5.5). Thus, when we feed concentration data from individual

herbicides into the global SSDs, this uncertainty carries through to the risk estimates pertaining to

98

the individual herbicides. Secondly, the distribution models used to estimate the proposed DGVs in

King et al. (2017a, b) were chosen using a best-fit approach for each herbicide. Hence, three

different models were used (log-logistic, Burr Type III and Weibull), in comparison to the single

model used in the msPAF method (i.e. logistic). Lastly, as previously mentioned, the King et al.

(2017a, b) proposed DGVs generally have separate values for freshwater and marine ecosystems,

and therefore the species that populate the SSDs may differ for some herbicides compared to the

msPAF SSDs.

Based on these factors, it was important to examine what, if any, differences there were in the

estimates of risk (i.e. the PC99 and PC95s) generated by the msPAF approach and the proposed

DGVs. The PC99 and PC95 for each herbicide were estimated using the corresponding Group A or

Group B global SSD (based on the results from the previous section in this Chapter). For example, the

PC values for atrazine were calculated from the Group A global SSD and the diuron PC values were

calculated from the Group B global SSD. In addition, the PC values for each herbicide were calculated

from their corresponding local SSDs. The PC99 values were plotted against the King et al. (2017)

proposed DGVs for freshwater and marine (Figure 5.6). As similar relationships were observed with

the PC95 comparisons (Figure B.1, Supplementary Material), the discussion below will focus just on

the PC99 results.

For many of the PSII herbicides, there were differences observed between the PC values calculated

from the global SSDs and the proposed DGVs (Figure 5.6). To decipher the root cause of the

observed differences, a process of elimination of possible sources of the differences was undertaken

(Table 5.5). This included observing the following:

(i) The difference between the local and global PC values – this would indicate the variation

in results due to the process of pooling the local SSDs to create global SSDs.

(ii) The difference between the local PC value and the proposed DGVs - this would indicate

that either different distribution models or the differences in the data set of the SSDs

were the source of the observed differences.

(iii) In the cases where the data sets were the same (see below for an explanation), but

there were still observed differences between the local PC values and the proposed

DGVs, we could conclude that the likely source of variation was from the distribution

model.

As previously mentioned, the freshwater and marine species were collated to generate the local and

global SSDs for calculating the msPAF. However, in most cases (eight herbicides) fresh and marine

species had already been collated for estimating the proposed DGVs (Table S1). This was due to the

99

lack of available toxicity data to populate an SSD based solely on just marine or freshwater species

(King et al., 2017a, b; Warne et al., 2018).

The results in Table 5.5 suggested there was no consistent source of variation across the 13

herbicides between the globally derived PC99 and the proposed DGVs. In a few cases (ametryn and

prometryn) there was good agreement between all values. The pooling of local SSDs to create a

global SSD did seem to cause some observed differences in the results, e.g. fluometuron, metribuzin,

propazine, terbuthylazine and terbutryn (Table 5.5). Better agreement in parallelism of SSDs would

reduce this variation. The use of different distribution models for the SSDs was also a likely source of

variation for hexazinone, metribuzin, propazine, tebuthiuron, terbuthylazine and terbutryn (Table

5.5). Using a Burr III model for the global curve fit might alleviate some of these differences,

although, it may not reduce the inconsistencies for those SSDs fitted with log-logistic and inverse

Weibull models (Table S1). Lastly, in four cases it was likely that at least some of the observed

variation was due to combinig the data sets of freshwater and marine species (atrazine, diuron,

bromacil and simazine).

Bringing greater consistency between the msPAF method and the proposed DGVs for the individual

PSII herbicides will require a more thorough investigation of the true sources of the observed

differences, which was beyond the scope of this study. It is likely that, through careful consideration

of the ecotoxicity data used to generate SSDs, there would be greater agreement between risk

estimates of toxicant mixtures with individual toxicant DGVs. This may include preferentially using

more robust toxicity values, e.g. chronic EC50 data over EC10 or NOEC data, using a standard

distribution model for contaminants with the same MoA, and/or using the same data sets of species

to calculate the msPAF as those used to generate DGVs.

Even with these changes, the limited availability of robust toxicity data to populate many SSDs for

pesticides is still an issue (and probably many other toxicant types), leading to SSDs with high

degrees of uncertainty, deviations in parallelism and ultimately differences between msPAF

estimations and individual DGVs.

100

Figure 5.6: Comparison of the protective concentration for 99% of species (PC99) estimated using the global

curve fits to herbicides in Groups A and B (black X), local curve fits (grey X), with the proposed default guideline

values (King et al., 2017) for freshwater species (blue dot with blue 95% confidence intervals) and marine

species (green dot with green 95% confidence intervals).

0.0001

0.001

0.01

0.1

1

10

100

1000C

on

cen

trat

ion

(m

g/L)

101

Table 5.5: Observed differences between the local and global protective concentration (PC) values and the marine and freshwater (FW) default guidelines values, the

possible causes of the observed differences and, based on these, the concluding reason for the observed differences.

Possible cause of observed differences:

Effect of pooling data from multiple

herbicides

Effect from different distribution models and/or combining freshwater and marine species

Concluding reason for observed difference

Observations:

Was there a difference between local and global PC

values?

Was the same data set used in the local SSD

and the proposed DGV SSDs†?

Was there a difference between proposed DGVs and

local PC values?

Ametryn Minimal Yes (FW) Minimal Good agreement between all values

Atrazine Small No Large Difference due to distribution model and/or combining FW and marine species

Diuron Minimal No Large difference with Marine proposed DGV

Difference is between marine DGV and all others. The FW species drive the msPAF estimations.

Bromacil Small Yes (Marine) Large difference with FW proposed DGV.

Good agreement between models and local and global fits. Difference is due to species in SSDs.

Fluometuron Large Yes (FW and Marine) Minimal difference Difference due to global curve fit.

Hexazinone Minimal Yes (Marine) Moderate difference with marine proposed DGV

Difference due to distribution model.

Metribuzin Moderate Yes (Marine) Moderate difference with marine and FW proposed DGV.

Difference due to the distribution model and global curve fit.

Prometryn Minimal Yes (Marine) Minimal Good agreement between all.

Propazine Moderate Yes (Marine) Moderate difference with marine proposed DGVs


Simazine Small No Large difference with Marine proposed DGV

Difference is between marine DGV and all others. The FW species drive the msPAF estimations. Possible effect from marine distribution model.

Tebuthiuron Small Yes (Marine) Small difference with marine proposed DGVs


Terbuthylazine Moderate Yes (Marine) Moderate difference with marine and FW proposed DGVs

Difference due to the distribution model, however the global curve fit adjusts for it.

Terbutryn Moderate Yes (FW and Marine) Large difference with proposed DGVs

Difference due to the distribution model however the global curve fit reduces the difference.

† ‘Marine’ and ‘Freshwater’ in parentheses refer to the SSDs used to generate the freshwater and marine proposed DGVs (respectively) reported in King et al. (2017a, b).

102

5.6 Conclusions

In examining the suitability of the msPAF method published by Traas et al (2002) to estimate the risk

(percent of species affected) of mixtures of PSII herbicides, we used an iterative approach for testing

and adapting the method to best suit the limitations of the PSII herbicide data. The msPAF

concentration addition method (Traas et al., 2002), for estimating the risk of mixtures of chemicals

with the same MoA, relies on the assumptions that the ecotoxicity data within each SSD follow a

normal distribution and the SSDs are parallel. This study, however, identified that both these

assumptions were violated for 13 PSII herbicide SSDs tested here and therefore we adapted the

Traas et al. (2002) method to account for the violations. As a result, it is recommended that global

PSII herbicide SSDs are used for calculating the msPAF for environmental samples that contain any

mixture combination of the 13 PSII herbicides tested here. Based on the data that populate the SSDs

published in King et al. (2017a, b), we found that the ms-PAF estimations were more robust when

the 13 SSDs were separated into two groups based on having parallel SSDs (Group A and B) and then

fitted with a global logistic distribution function. These global PSII herbicide SSDs should be used

irrespective of the combination of PSII herbicides present within the environmental sample to be

tested.

Even the method developed here is not ideal and other improvements could be examined. Ideally,

one global PSII herbicide that represented all PSII herbcides would be most favourable. Theoretically,

the SSDs of chemicals with a similar MoA should be parallel and therefore, the process of converting

the x-axis to a relative scale acts to convert the SSDs to pseudo-replicates for that MoA. For example,

the SSDs for diuron, atrazine and hexazinone, once converted to HUs would each become a SSD that

represents the PSII herbicide MoA. By that theory, you could calculate one set of global parameters

that represented all mixture combinations of PSII herbicides, irrespective of which PSII herbicides

you were specifically investigating. In this sense, we could estimate the ‘true’ parameters of the PSII

herbicide SSD from a sample population of PSII herbicide’s. Thus, the global parameters calculated

from multiple PSII herbicides should theoretically be closer to the true parameters from which the

true population is derived. A similar concept was suggested by De Zwart (2002) who suggested that

each MoA could be represented by one value of β.

De Zwart (2002) also observed deviations in parallelism between SSDs, which they concluded were a

result of insufficient observations. In this study, we suggest that there are other potential sources

including the use of NOEC data, data converted to chronic EC10 values, and using a range of test

endpoints. Faust et al. (2001) and Chevre et al. (2006) also found that PSII herbicide response curves

103

did not always conform to parallelism. Both agreed also that it was not a reason to continue to use

the concentration addition model when estimating the risk of PSII herbicide mixtures.

The results from this study will mean that the risk (percent of species affected) from mixtures of PSII

herbicide can be estimated for 13 PSII herbicides commonly found in ecosystems of the GBR. This

will help management programs achieve greater understanding of the full impact pesticides are

having to Reef ecosystems. With further development, pesticides with different MoA can also be

included in the assessment.

104

5.7 References










determination of water quality criteria for herbicides in surface water. Environ Sci Tech, 40(2): 426-

435.

De Zwart D. 2002. Observed regularities in species sensitivity distributions for aquatic species. (2002).

In: Posthuma L, Suter GW II, Traas TP, editors. Species Sensitivity Distributions in Ecotoxicology.

Boca Raton (FL), USA: Lewis Publishers.


methodologies. Environ Toxicol Chem 24: 2665-2676.



Barrier Reef 2011-2015. A report for the Queensland Department of Environment and Heritage










1628–1631.

105

Garzon-Garcia A, Wallace R, Huggins R, Turner RDR, Smith RA, Orr D, Ferguson B, Gardiner R, Thomson

B and Warne MStJ. 2015. Total suspended solids, nutrients and pesticide loads (2013–2014) for

rivers that discharge to the Great Barrier Reef – Great Barrier Reef Catchment Loads Monitoring

Program 2013–2014. Brisbane (QLD), Australia: Department of Science, Information Technology,

Innovation and the Arts, 108p.



Huggins R, Wallace R, Orr DN, Thomson B, Smith RA, Taylor C, King O, Gardiner R, Wallace S, Ferguson

B, Preston S, Simpson S, Shanks J, Warne MStJ, Turner RDR, Mann RM. 2017. Total suspended

solids, nutrient and pesticide loads (2015–2016) for rivers that discharge to the Great Barrier Reef

– Great Barrier Reef Catchment Loads Monitoring Program. Brisbane (QLD), Australia: Department

of Environment and Science, 139p.


Ser 261: 149-159.


values for pesticides commonly used in the Great Barrier Reef catchment area: Part 1 - 2,4-D,






Part 2 - Bromacil, Chlorothalonil, Fipronil, Fluometuron, Fluroxypyr, Haloxyfop, MCPA,







Raton (FL), USA: Lewis Publishers. p 3–11.

106






Queensland, Australia. Integ Environ Manag Assess 13(4):746–753.

Traas TP, Van de Meent D, Posthuma L, Hamers T, Kater BJ, De Zwart, D, Aldenberg, T. 2002. The

potentially affected fraction as a measure of ecological risk. In: Posthuma L, Suter GW II, Traas TP,


p 315-344.

Turner R, Huggins R, Wallace R, Smith R, Vardy S, Warne MStJ. 2012. Sediment, nutrient and pesticide

loads: Great Barrier Reef Loads Monitoring 2009–2010. Water Sciences Technical Report, Volume

2012, Number 14. Brisbane (QLD), Australia: Department of Science, Information Technology,










Wallace, R, Huggins, R, Smith, RA, Turner, RDR, Vardy, S, Warne, MStJ. 2015. Total suspended solids,









107













University of Hertfordshire. 2013. The Pesticide Properties Data Base (PPDB). Developed by the




108

Chapter 6: Discussion and Conclusions

There is a common theme throughout this thesis; to develop quantitative methods that incorporate

the cumulative toxic affects of pesticides in order to measure their risk, and the progress

management practices have made towards reducing that risk on Great Barrier Reef ecosystems.

Three research objectives were set:

1. Develop a quantitative method to calculate a total photosystem II (PSII) herbicide load from

the loads of individual PSII herbicides that accounts for their differences in toxicity to measure

progress towards the 2013 pesticide target.

2. Validate this method, using monitoring data from the GBR, to prove the most robust and

environmentally relevant method has been developed.

3. Develop a quantitative method to determine the percent of species affected from mixtures of

PSII herbicides in GBR ecosystems, to measure progress towards the 2017 pesticide target.

Presented below is a summary of how the objectives were achieved and the observed and expected

impact of the outcomes of this research. The chapter is concluded with recommendations for

continuing and advancing this research in the future.

6.1 Summary of results and research impact

When research for this thesis began, the pesticide target for the Reef Water Quality Protection Plan

(Reef Plan) 2013 (Australian Government and Queensland Government, 2013a) was a load-based

target, that is: ‘at least a 60 per cent reduction in the end-of-catchment pesticide loads in priority

areas’. At that time, there was a need to determine a method for measuring the progress towards

achieving this target that could incorporate the relative toxicities of the five priority photosystem II

(PSII) herbicides that posed the greatest risk to the GBR. The Reef Plan 2013 targets were a revision of

the previous pesticide target (Reef Plan, 2009) and was advocated as an interim measure (to ensure

continuity with earlier reporting) towards setting more ecologically relevant targets. The revision of

the pesticide targets for Reef Plan 2013 was the impetus for the research of this thesis and resulted in

setting the first objective.

To address the first objective, the toxic load method was developed to convert a pollutant load,

comprised of multiple chemicals, to a toxicity-based load (toxic load), using a modified toxic

equivalency factor (TEF) approach (Chapter 3). The method concatenated the relative potencies (ReP)

of multiple species to estimate a weighting factor (the TEF) that could easily be applied to convert a

mass load of a pesticide to a toxic load. The issue going forth lay in determining what value from the

109

distribution of RePs should be used to represent the TEF. Two novel additions to the previously

published TEF approach were introduced that determined which percentile of the ReP distribution

would generate the most environmentally relevant and robust toxic loads. In doing so, the estimated

toxic loads would provide a more realistic evaluation of where management efforts should be focused

to reduce pesticide impacts on GBR ecosystems.

Chapter 4 followed on to validate the above method (second objective) using monitoring data

collected from GBR catchments. In this case study, it was determined that the most environmentally

relevant and robust toxic loads for the priority PSII herbicides were generated using the 75th percentile

of the ReP cumulative distribution. The results also demonstrated that it was essential to use the toxic

load method (rather than the mass loads as used in previous years) to measure progress towards the

2013 targets, and doing so meant that measured reductions in the toxic load would signify a

commensurate reduction in impacts on GBR ecosystems.

The results presented in Chapters 3 and 4 addressed a requirement of the Australian and Queensland

governments’ commitment for measuring the progress towards achieving the water quality

improvement targets. The chapters were published as sister papers in a peer-reviewed journal in the

scientific literature and the outcomes adopted by the Paddock to Reef Integrated Monitoring,

Modelling and Reporting Program4 as the method for measuring progress towards the 2013 target.

While the move from measuring the mass loads of pesticides to a toxic load was an important interim

step in making the pesticide reduction targets more ecologically relevant, some issues remained:

1. Toxicity and risk of chemicals are determined by the concentration of the chemical(s) and the

duration organisms are exposed to the chemical(s), however toxic loads do not provide this

information.

2. While we knew that a reduction in the toxic load would likely lead to a commensurate

reduction in ecosystem impact, the toxic load provided no indication of what the impact from

the load was initially or what it was reduced to and was inconsistent with the Australian and

New Zealand Guidelines for Fresh and Marine Water Quality (ANZECC and ARMCANZ, 2000;

Warne et al., 2018; King et al., 2017 a, b).

3. The method for deriving toxic loads could only combine pesticides that have the same mode

of action (MoA) (e.g. PSII herbicides). It could not be used to combine pesticides with different

4 The Paddock to Reef Integrated Monitoring, Modelling and Reporting Program is jointly funded by the Australian and Queensland government. The program collects and integrates data and information on agricultural management practices, catchment indicators, catchment loads and the health of the Great Barrier Reef. The primary objective of the program is to measure and report on progress towards Reef Plan's goal and targets through annual Report Cards (Australian Government and Queensland Government, 2013b).

https://www.reefplan.qld.gov.au/about/

https://www.reefplan.qld.gov.au/about/

https://www.reefplan.qld.gov.au/measuring-success/report-cards/

110

MoA. This was a major limitation as there are numerous pesticides other than PSII inhibiting

herbicides detected in GBR ecosystems (Turner et al., 2012, 2013; Wallace et al., 2014, 2015,

2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017; Devlin et al., 2015).

An important consideration of the most recent update to the Reef Plan (renamed as the Reef 2050

Water Quality Improvement Plan) was to ensure the water quality improvement targets were more

focused and environmentally relevant (Brodie et al., 2017; Australian Government and Queensland

Government, 2017). The pesticide targets switched from a loads reduction approach to a risk-based

target – the ‘protection of at least 99% of species at end-of-catchments’. The target change meant

there was also a need for a change in the methods for measuring progress towards the new targets

(i.e. the third objective). Chapter 5 investigated the use of a method, previously published in the

literature, that could estimate the percent of species affected by mixtures of chemicals; the

multisubstance Potentially Affected Fraction (msPAF) method. Running in parallel with the

development of this method was the development of species sensitivity distributions (SSDs) of 28

pesticides common to the GBR ecosystems (King et al., 2017a, b). With these SSDs, the msPAF method

was investigated for 13 PSII herbicides found in GBR ecosystems. Features of the data underpinning

the PSII herbicide SSDs violated the assumptions of the published msPAF method and, hence, the

method was modified to account for these violations. The developments to the method improved the

robustness and reliability of the estimates of risk (percent of species affected) from mixtures of PSII

herbicides. The results from Chapter 5 have been incorporated into calculations of pesticide risk for

the 2017 Scientific Consensus Statement (Waterhouse et al., 2017) and the Mackay Whitsunday and

Wet Tropics regional report cards.

6.2 Recommendations for Future Research

Developing the msPAF method was a big step forward in measuring the risk of pesticides to GBR

ecosystems. However, the methods presented in Chapter 5 only measure one element of risk – the

magnitude of exposure (i.e. the pesticide concentrations). Other factors also determine the

ecological risk a chemical has on an ecosystem, i.e. the temporal and spatial exposure. Investigating

a quantitative approach for incorporating temporal and spatial exposure with the msPAF metric was

beyond the scope of this thesis. However, it is crucial that the msPAF metric presented here does

incorporate these additional factors into the method for measuring progress towards the new

pesticide targets.

Temporal exposure will be important to quantify as progress towards the targets are reported on an

annual basis. Temporal exposure to pesticides is highly seasonal and occurs in pulses as pesticides

are transported in runoff from rainfall events (Smith et al., 2012; Turner et al., 2012, 2013; Wallace

111

et al., 2014, 2015, 2016; Garzon‐Garcia et al., 2015; Huggins et al., 2017). Quantifying the ecological

risk of pulsed exposures will require methods that can account for acute exposures followed by

recovery periods, which can become a complex process (Vallotton et al., 2008). As the pesticide

targets are set at the end of the catchment, quantifying the spatial exposure may be less relevant

when assessing individual catchments. This will become more important when assessing progress

towards the target is scaled up from each catchment to the basin, regional and whole of GBR level.

In addition, there are a number of other pesticides detected frequently in GBR ecosystems that have

not been considered in this thesis. To measure the risk of all pesticides, the methods presented in

Chapter 5 will need to be expanded to other types of pesticides. The msPAF method was selected in

particular because of this need as there is a capacity to aggregate the effects of multiple chemicals

with different MoA using the response addition approach (Traas et al., 2002). It is recommended

that a similar approach to that presented in Chapter 5 is used to generate a global SSD for each

pesticide MoA detected in GBR ecosystems.

6.3 Reference

Australian Government and Queensland Government. 2013a. Reef Water Quality Protection Plan

2013. Securing the health and resilience of Great Barrier Reef World Heritage Area and adjacent


Queensland.

Australian Government and Queensland Government. 2013b. Paddock to Reef Integrated Monitoring,

Modelling and Reporting Program; Reef Water Quality Protection Plan 2013–2018. Brisbane (QLD),

Australia: The State of Queensland. Available from https://www.reefplan.qld.gov.au/measuring-

success/paddock-to-reef/

Brodie J, Baird M, Waterhouse J, Mongin M, Skerratt J, Robillot C, Smith R, Mann R, Warne MStJ, 2017.



Queensland, 68 pp.






https://www.reefplan.qld.gov.au/measuring-success/paddock-to-reef/


112






Vallotton N, Eggen RI, Escher BI, Krayenbühl J, Chèvre N. 2008. Effect of pulse herbicidal exposure on

Scenedesmus vacuolatus: a comparison of two photosystem II inhibitors. Environ Toxicol Chem

27(6): 1399-407.

114

Appendix A. Supplementary Material for Chapter 3

Table A.1 Sources of herbicide ecotoxicity data used in the calculation of relative potencies.

Reference Extracted herbicide data

Abou-Waly, H., Abou-Setta, M. M., Nigg, H. N., and Mallory, L. L. 1991. Growth response of freshwater algae, Anabaena flos-aquae andSelenastrum capricornutum to atrazine and hexazinone herbicides. Bulletin of Environmental Contamination and Toxicology, 46(2), 223–229.

Atrazine and hexazinone

Cain, J. R., and Cain, R. K. 1983. The effects of selected herbicides on zygospore germination and growth of Chlamydomonas moewusii (chlorophyceae, volvocales). Journal of Phycology, 19(3), 301–305.

Ametryn, atrazine and diuron

Drost, W., Backhaus, T., Vassilakaki, M., and Horst Grimme, L. 2003. Mixture toxicity of s-triazines to Lemna minor under conditions of simultaneous and sequential exposure. Fresenius Environmental Bulletin, 12(6), 601–607.

Ametryn and atrazine

Fai, P. B., Grant, A., and Reid, B. (2007). Chlorophyll a fluorescence as a biomarker for rapid toxicity assessment. Environmental Toxicology and Chemistry, 26(7), 1520–1531.

Atrazine and diuron

Flores, F., Collier, C. J., Mercurio, P., and Negri, A. P. 2013. Phytotoxicity of four photosystem II herbicides to tropical seagrasses. PloS one, 8(9), e75798.

Atrazine, diuron, hexazinone and tebuthiuron

Gaggi, C., Duccini, M., Bacci, E., Sbrilli, G., Bucci, M., and Naby, A. M. 1995. Toxicity and hazard ranking of s-triazine herbicides using microtox® two green algal species and a marine crustacean. Environmental Toxicology and Chemistry, 14(6), 1065–1069.

Ametryn and atrazine

Garrett, D. C. 2004. Effects of methanol, atrazine, and copper on the ultrastructure of Pseudokirchneriella subcapitata (Selenastrum capricornutum).

Atrazine

Grossmann, K., Berghaus, R., and Retzlaff, G. 1992. Heterotrophic plant cell suspension cultures for monitoring biological activity in agrochemical research. Comparison with screens using algae, germinating seeds and whole plants. Pesticide Science, 35(3), 283–289.

Atrazine and diuron

Hickey, C. W., Blaise, C., and Costan, G. 1991. Microtesting appraisal of ATP and cell recovery toxicity end points after acute exposure of Selenastrum capricornutum to selected chemicals. Environmental Toxicology and Water Quality, 6(4), 383–403.

Tebuthiuron

Jones, R. J., and Kerswell, A. P. 2003. Phytotoxicity of Photosystem II(PSII) herbicides to coral. Marine Ecology Progress Series, 261, 149–159.

Ametryn, hexazinone and tebuthiuron

Ma, J. 2002. Differential sensitivity to 30 herbicides among populations of two green algae Scenedesmus obliquus and Chlorella pyrenoidosa. Bulletin of Environmental Contamination and Toxicology, 68(2), 275–281.


Ma, J., Lin, F., Wang, S., and Xu, L. 2003. Toxicity of 21 herbicides to the green alga Scenedesmus quadricauda. Bulletin of Environmental Contamination and Toxicology, 71(3), 0594–0601.


Ma, J., Xu, L., Wang, S., Zheng, R., Jin, S., Huang, S., and Huang, Y. 2002. Toxicity of 40 herbicides to the green alga Chlorella vulgaris. Ecotoxicology and Environmental Safety, 51(2), 128–132.

Atrazine and diuron

115

Reference Extracted herbicide data

Ma, J., Wang, S., Wang, P., Ma, L., Chen, X., and Xu, R. 2006. Toxicity assessment of 40 herbicides to the green alga Raphidocelis subcapitata. Ecotoxicology and Environmental Safety, 63(3), 456–462.


Ma, J., Liang, W., Xu, L., Wang, S., Wei, Y., and Lu, J. 2001. Acute toxicity of 33 herbicides to the green alga Chlorella pyrenoidosa. Bulletin of Environmental Contamination and Toxicology, 66(4), 536–541.


Magnusson, M., Heimann, K., and Negri, A. P. 2008. Comparative effects of herbicides on photosynthesis and growth of tropical estuarine microalgae. Marine Pollution Bulletin, 56(9), 1545–1552.

Atrazine, diuron and hexazinone

Masojídek, J., Souček, P., Máchová, J., Frolík, J., Klem, K., and Malý, J. 2011. Detection of photosynthetic herbicides: Algal growth inhibition test vs. electrochemical photosystem II biosensor. Ecotoxicology and Environmental Safety, 74(1), 117–122.

Atrazine and diuron

Negri A. P., Flores F., Röthig T., & Uthicke S. 2011. Herbicides increase the vulnerability of corals to rising sea surface temperature. Limnology and Oceanography, 56(2): 471–485.

Atrazine and hexazinone

Peterson, H. G., Boutin, C., Freemark, K. E., and Martin, P. A. 1997. Toxicity of hexazinone and diquat to green algae, diatoms, cyanobacteria and duckweed. Aquatic Toxicology, 39(2), 111–134.

Hexazinone

Schafer, H., Hettler, H., Fritsche, U., Pitzen, G., Roderer, G., and Wenzel, A. 1994. Biotests using unicellular algae and ciliates for predicting long-term effects of toxicants. Ecotoxicology and Environmental Safety, 27(1), 64–81.

Atrazine and diuron

Schrader, K. K., de Regt, M. Q., Tidwell, P. D., Tucker, C. S., and Duke, S. O. 1998. Compounds with selective toxicity towards the off-flavor metabolite-producing cyanobacterium Oscillatoria cf. chalybea. Aquaculture, 163(1), 85–99.

Atrazine and diuron

Seery, C. (unpublished) Ametryn, atrazine, diuron, hexazinone and tebuthiuron

Singh, S., Datta, P., and Tirkey, A. 2011. Response of multiple herbicide resistant strain of diazotrophic cyanobacterium, Anabaena variabilis, exposed to atrazine and DCMU. Indian Journal of Experimental Biology, 49: 298–303.

Atrazine and diuron

St Laurent, D., Blaise, C., MacQuarrie, P., Scroggins, R., and Trottier, B. 1992. Comparative assessment of herbicide phytotoxicity to Selenastrum capricornutum using microplate and flask bioassay procedures. Environmental Toxicology and Water Quality, 7(1), 35–48.

Hexazinone

Teisseire, H., Couderchet, M., and Vernet, G. 1999. Phytotoxicity of diuron alone and in combination with copper or folpet on duckweed (Lemna minor). Environmental Pollution, 106(1), 39–45.

Diuron

116

Figure A.1 Cumulative distribution functions of the relative potency values for ametryn and atrazine,

hexazinone and atrazine, hexazinone and ametryn, tebuthiuron and diuron, tebuthiuron and ametryn, and

tebuthiuron and hexazinone.

117

Table A.2 P-values from Kruskal Wallis tests applied to the relative potency values of species belonging to

freshwater, marine and estuarine ecosystems for all combinations of test chemicals with atrazine and diuron

as reference chemicals.

Reference Chemical Test Chemical n p-value

Atrazine Ametryn 11 0.256

Diuron 18 0.46

Hexazinone 8 0.119

Tebuthiuron 5 0.564

Diuron Ametryn 10 0.669

Hexazinone 8 0.486

Tebuthiuron 5 0.083

118

Appendix B. Supplementary Material for Chapter 5

Table B.1 Proposed freshwater and marine pesticide default guideline values for the protection of at least

99% of species, their reliability and the distribution model used to generate the values.

PSII Herbicide Freshwater Marine

Reliability n Model PC99 Reliability n Model PC99

Ametryn M 17* Burr III 0.074 M 9 Burr III 0.1

Atrazine VH 50 Burr III 0.17 M 17 Log-Logistic 1.2

Bromacil L 5 Log-Logistic 1.6 M 7* Log-Logistic 0.23

Diuron VH 26 Burr III 0.08 VH 20 Burr III 0.43

Fluometuron* M 7* Log-Logistic 20 M 7* Log-Logistic 20

Hexazinone L 5 Log-Logistic 0.31 L 8* Inverse Weibull

1.8

Metribuzin VH 18 Inverse Weibull

2 M 19* Inverse Weibull

2

Prometryn L 7 Log-Logistic 0.094 M 9* Burr III 0.11

Propazine L 5 Log-Logistic 1.3 L 5 Log-Logistic 2.2

Simazine H 17 Burr III 3.2 L 6 Log-Logistic 28

Tebuthiuron M 5 Log-Logistic 4.8 M 7* Log-Logistic 4.7

Terbuthylazine VH 16 Burr III 0.43 M 18* Burr III 0.4

Terbutryn M 19* Burr III 0.079 M 19* Burr III 0.079

*SSD is based on both marine and freshwater species and makes up the full set of species used to generate the

msPAF SSD, i.e. n = number species in (Table 5.1). All other msPAF SSDs were generated from a combination of

the species that populate the individual freshwater and marine SSDs used to generate the proposed DGVs.

119

Table B.2 Proposed freshwater and marine pesticide default guideline values for the protection of at least 95%

of species, their reliability and the distribution model used to generate the values.

PSII Herbicide Freshwater Marine

Reliability n Model PC95 Reliability n Model PC95

Ametryn M 17* Burr III 0.33 M 9 Burr III 0.61

Atrazine VH 50 Burr III 0.98 Log-Logistic 3.1

Bromacil L 5 Log-Logistic 3.6 M 7* Log-Logistic 1.1

Diuron VH 26 Burr III 0.23 VH 20 Burr III 0.67

Fluometuron M 7* Log-Logistic 40 M 7* Log-Logistic 40

Hexazinone L 5 Log-Logistic 1.1

L 8* Inverse

Weibull 2.5

Metribuzin VH 18 Inverse

Weibull 2.6

M 19* Inverse

Weibull 2.7

Prometryn L 7 Log-Logistic 0.49 M 9* Burr III 0.52

Propazine L 5 Log-Logistic 3.1 L 5 Log-Logistic 4.6

Simazine H 17 Burr III 10 L 6 Log-Logistic 63

Tebuthiuron M 5 Log-Logistic 13 M 7* Log-Logistic 11

Terbuthylazine VH 16 Burr III 1.2 M 18* Burr III 0.97

Terbutryn M 19* Burr III 0.26 M 19* Burr III 0.26

Figure B.1: Comparison of the protective concentration for 95% of species (PC95) estimated using the global

logistic curve fits – GC11 and GC2 (black X), local logistic curve fits (grey X), and those estimated by King et al

0.01

0.1

1

10

100

1000

Co

nce

ntr

atio

n (

g/L)

120

2017a,b for freshwater species (blue dot with blue 95% confidence intervals) and marine species (green dot

with green 95% confidence intervals).

121

Bibliography

This bibliography compiles the list of all references cited in this thesis. As the thesis has been

structured for publication of individual chapters, a reference list has also been provided at the end of

each chapter. It should be noted that, in a few cases, citations with a common first author and year

of publication were distinguished with ‘a’ and ‘b’ suffixes to the publication year. However, in some

chapters only one of the two citations were used and therefore no suffix was required. In these

cases, the suffix has been included in the list below as ‘(a)’ and ‘(b)’ to let the reader know that the

citation exists in the main body of the thesis with and/or without the suffix.

Ågerstrand M, Edvardsson L, Rudén C. 2014. Bad reporting or bad science? Systematic data evaluation

as a means to improve the use of peer-reviewed studies in risk assessments of chemicals. Human

and Ecological Risk Assessment: An International Journal, 20(6), 1427-1445.

Aldenberg T, Jaworska JS, Traas TP. 2002. Normal species sensitivity distributions and probabilistic

ecological risk assessment. In: Posthuma L, Suter II GW, Traas TP, editors. Species Sensitivity

Distributions in Ecotoxicology. Boca Raton (FL), USA: Lewis Publishers. pp 49–102.









Zealand.







122

Australian Government and Queensland Government. 2013(a). Reef Water Quality Protection Plan

2013. Securing the health and resilience of Great Barrier Reef World Heritage Area and adjacent


Queensland.

Australian Government and Queensland Government. 2013b. Paddock to Reef Integrated Monitoring,

Modelling and Reporting Program; Reef Water Quality Protection Plan 2013–2018. Brisbane (QLD),

Australia: The State of Queensland. Available from https://www.reefplan.qld.gov.au/measuring-

success/paddock-to-reef/



Backhaus T, Faust M, Scholze M, Gramatica P, Vighi M, Grimme LH. 2004. Joint algal toxicity of

phenylurea herbicides is equally predictable by concentration addition and independent action.

Environ Toxicol Chem 23(2): 258-264.




Bartkow M, Dunn A, Komarova, T, Paxman, C, Mueller J. 2008. Monitoring of organic chemicals in the

Great Barrier Reef Marine Park and selected tributaries using time integrated monitoring tools:

October 2008. Report to the Great Barrier Reef Marine Park Authority (Marine Monitoring

Programme), National Research Centre for Environmental Toxicology (EnTox).

Bentley C, Devlin M, Paxman C, Chue KL, Mueller J. 2012. Pesticide monitoring in inshore waters of

the Great Barrier Reef using both time‐integrated and event monitoring techniques (2011 ‐ 2012).

The University of Queensland, The National Research Centre for Environmental Toxicology (Entox).

Berenbaum MC. 1985. The expected effect of a combination of agents: the general solution. Journal

of Theoretical Biology 114, 413‐431.










123



Brodie J, Baird M, Waterhouse J, Mongin M, Skerratt J, Robillot C, Smith R, Mann R, Warne MStJ. 2017.



Queensland, 68 pp.

Cappo M, Kelley R. 2000. Connectivity in the Great Barrier Reef World Heritage Area: an overview of

pathways and processes. In: Wolanski (editor). Oceanographic Processes of Coral Reefs: Physical

and Biological Links in the Great Barrier Reef. Boca Raton (FL), USA: CRC Press. pp.161–187.


determination of water quality criteria for herbicides in surface water. Environ Sci Tech, 40(2): 426–

435.


Strengths and limitations. Hum Ecol Risk Assess, 5: 33–42.

Davis AM, Lewis SE, Bainbridge ZT, Glendenning L, Turner R, Brodie JE. 2012. Dynamics of herbicide

transport and partitioning under event flow conditions in the lower Burdekin region, Australia. Mar

Poll Bull, 65: 182‐193.

Davis AM, Thorburn PJ, Lewis SE, Bainbridge ZT, Attard SJ, Milla R, Brodie JE. 2013. Environmental

impacts of irrigated sugarcane production: Herbicide run‐off dynamics from farms and associated

drainage systems. Agriculture, Ecosystems and the Environment, 180: 123‐135.

Davis AM, Lewis SE, O’Brien DS, Bainbridge ZT, Bentley C, Mueller JF, Brodie JE. 2014. Water resource

development and high value coastal wetlands on the lower Burdekin floodplain, Australia. In:

Wolanski, E, editor. Estuaries of Australia in 2050 and beyond. Dordrecht, Germany: Springer.

pp. 223–245

De Zwart D. 2002. Observed regularities in species sensitivity distributions for aquatic species. (2002).

In: Posthuma L, Suter GW II, Traas TP, editors. Species Sensitivity Distributions in Ecotoxicology.

Boca Raton (FL), USA: Lewis Publishers.


methodologies. Environ Toxicol Chem 24: 2665–2676.

124

DEHP [Department of Environment and Heritage Protection]. 2009. Queensland Water Quality

Guidelines, Version 3. Brisbane (QLD), Australia: Department of Environment and Heritage

Protection, 184pp.

Deloitte Access Economics. 2013. Economic contribution to the Great Barrier Reef. Townsville (QLD),

Australia: Great Barrier Reef Marine Park Authority.



Barrier Reef 2011–2015. A report for the Queensland Department of Environment and Heritage



Durda JL, Preziosi DV. 2000. Data quality evaluation of toxicological studies used to derive

ecotoxicological benchmarks. Hum Ecol Risk Assess 6: 747-765.

EC [European Commission]. 2011. Common implementation strategy for the Water Framework

Directive (2006/60/EC). Guidance document no. 27. Technical guidance for deriving environmental

quality standards. Brussels, Belgium: European Commission.

Environment Canada and Health Canada. 2001. Priority substances list assessment report:

Nonylphenol and its ethoxylates. Canada: Environment Canada and Health Canada, Minister of

Public Works and Government Services.

eWater Cooperative Research Centre. 2015. Water Quality Analyser v2.1.2.4. Canberra (ACT),

Australia: eWater Cooperative Research Centre.





assessments for dioxins, furans, and PCBs. J Toxicol Environ Health Part A 66: 533–550.

Finney D. 1979. Bioassay and the Practice of Statistical Inference. International Statistical Review /

Revue Internationale de Statistique 47(1): 1–12.



125

Flores F, Collier CJ, Mercurio P, Negri AP. 2013. Phytotoxicity of four photosystem II herbicides to

tropical seagrasses. PloS One 8: e75798.


Great Barrier Reef using both time-integrated and event monitoring techniques (2012–2013).









1628–1631.

Garzon-Garcia A, Wallace R, Huggins R, Turner RDR, Smith RA, Orr D, Ferguson B, Gardiner R, Thomson

B and Warne MStJ. 2015. Total suspended solids, nutrients and pesticide loads (2013–2014) for

rivers that discharge to the Great Barrier Reef – Great Barrier Reef Catchment Loads Monitoring

Program 2013–2014. Brisbane (QLD), Australia: Department of Science, Information Technology,


GBRMPA [Great Barrier Reef Marine Park Authority]. 2010. Water Quality Guidelines for the Great

Barrier Reef Marine Park. Townsville (QLD), Australia: Great Barrier Reef Marine Park Authority,

109pp.

Gutendorf B, Westendorf J. 2001. Comparison of an array of in vitro assays for the assessment of the

estrogenic potential of natural and synthetic estrogens, phytoestrogens and xenoestrogens.

Toxicol 166, 79-89.


programs across the US states. Environ Manage 4: 301-310.

Harrington L, Fabricius K, Eaglesham G, Negri A. 2005. Synergistic effects of diuron and sedimentation

on photosynthesis and survival of crustose coralline algae. Mar Poll Bull 51: 415‐427.

Harrison R. 1999. Endocrine disrupting chemicals. Cambridge: RSC.

126










toxicity data. Integr Environ Assess Manage 1: 174–180.

Huggins R, Wallace R, Orr DN, Thomson B, Smith RA, Taylor C, King O, Gardiner R, Wallace S, Ferguson

B, Preston S, Simpson S, Shanks J, Warne MStJ, Turner RDR, Mann RM. 2017. Total suspended

solids, nutrient and pesticide loads (2015–2016) for rivers that discharge to the Great Barrier Reef

– Great Barrier Reef Catchment Loads Monitoring Program. Brisbane (QLD), Australia: Department

of Environment and Science, 139p.

Hutchinson TH, Scholz N, Guhl W. 1998. Analysis of the ECETOC aquatic toxicity (EAT) database IV –

Comparative toxicity of chemical substances to freshwater versus saltwater organisms.

Chemosphere 36(1): 143–153.

Jones R. 2005. The ecotoxicological effects of Photosystem II herbicides on corals. Mar Poll Bull 51(5):

495–506.


Ser 261: 149–159.

Kapernick A, Shaw M, Dunn A, Komarova T, Mueller J, Carter S, Eaglesham G, Schaffelke B, Bass D,

Haynes D. 2006. River pesticide loads and GBR lagoon pesticide data. Attachment A, Task 2.6.

Report to the Great Barrier Reef Marine Park Authority (Marine Monitoring Programme). Brisbane

(QLD), Australia: National Research Centre for Environmental Toxicology (EnTox).

Kapernick A, Shaw C, Dunn A, Komarova T, Mueller J, Carter S, Eaglesham G, Alberts V, Masters B,

Rohde K, Packett R, Prange J, Haynes D. 2007. River pesticide loads and GBR lagoon pesticide data

(2006–2007). Report to the Great Barrier Reef Marine Park Authority (Marine Monitoring

Programme). Brisbane (QLD), Australia: National Research Centre for Environmental Toxicology

(EnTox).

127

Kennedy K, Paxman C, Dunn A, O'Brien J, Mueller JF. 2010. Monitoring of organic chemicals in the

Great Barrier Reef Marine Park and selected tributaries using time integrated monitoring tools

(2008-2009). Brisbane (QLD), Australia: National Research Centre for Environmental Toxicology

University of Queensland (Entox), University of Queensland.

Kennedy K, Devlin M, Bentley C, Lee‐Chue K, Paxman C, Carter S, Lewis SE, Brodie J, Guy E, Vardy S.

2012. The influence of a season of extreme wet weather events on exposure of the World

Heritage Area Great Barrier Reef to pesticides. Mar Poll Bull 64: 1495–1507.











Klimisch HJ, Andreae M, Tillmann U. (1997). A systematic approach for evaluating the quality of

experimental toxicological and ecotoxicological data. Regul Toxicol Pharmacol 25: 1-5.

Leung KMY, Morritt D, Wheeler JR, Whitehouse P, Sorokin N, Toy R, Holt M, Crane M. 2001. Can

saltwater toxicity be predicted from freshwater data? Mar Poll Bull 41(11): 1007–1013.




Magnusson M, Heimann K, Negri AP. 2008. Comparative Effects of Herbicides on Photosynthesis and

Growth of Tropical Estuarine Microalgae. Mar Poll Bull 56: 1545–1552.





128

Negri AP, Flores F, Röthig T, Uthicke S. 2011. Herbicides increase the vulnerability of corals to rising

sea surface temperature. Limnol Oceanogr 56: 471–485.

O’Brien DS, Lewis SE, Davis A, Gallen C, Paxman C, Smith R, Warne MStJ, Mueller J, Brodie JE. 2013.

Pesticide exposure within the Barratta Creek Catchment. Cairns (QLD), Australia: Report to the Reef

Rescue Water Quality Research and Development Program. Reef and Rainforest Research Centre

Limited, 58pp.

OME [Ontario Ministry of the Environment] 1984. Scientific criteria document for standard

development. No. 4–84. Polychlorinated Dibenzop- dioxins (PCDDs) and polychlorinated

dibenzofurans (PCDFs). Ontario, Canada: Intergovernmental Relations and Hazardous

Contaminants Coordination Branch, Ontario Ministry of the Environment.

Pedersen JA, Yeager MA, Suffet IH (Mel). 2006. Organophosphorus insecticides in agricultural and

residential runoff: Field observations and implications for total maximum daily load development.

Environ Sci and Technol 40:2120-2127.



Raton (FL), USA: Lewis Publishers. p 3-11.




Resour 11: 67-78.

Safe S. 1998. Development validation and problems with the toxic equivalency factor approach for risk

assessment of dioxins and related compounds. J Animal Sci 76(1): 134–141.

Schäfer RB, Caquet T, Siimes K, Mueller R, Lagadic L, Liess M. 2007. Effects of pesticides on community

structure and ecosystem functions in agricultural streams of three biogeographical regions in

Europe. Sci Tot Environ 382(2–3): 272–85.

Schaffelke B, Collier C, Kroon F, Lough J, McKenzie L, Ronan M, Uthicke S, Brodie J. 2017. Scientific

Consensus Statement 2017. Scientific Consensus Statement 2017: A synthesis of the science of

land-based water quality impacts on the Great Barrier Reef, Chapter 1: The condition of coastal

and marine ecosystems of the Great Barrier Reef and their responses to water quality and

disturbances. Brisbane (QLD), Australia: State of Queensland.

129

Schneider K, Schwarz M, Burkholder I, Kopp-Schneider A, Edler L, Kinsner-Ovaskainen A, Hartung T,

Hoffmann S. 2009. “ToxRTool”, a new tool to assess the reliability of toxicological data. Toxicol Lett

189: 138-144.




Smith RA, Warne MStJ, Mengersen K and Turner RDR. 2017(a). An improved method for calculating


746-753.

Smith RA, Warne MStJ, Mengersen K, Turner RDR. 2017(b). An improved method for calculating

toxicity‐based pollutant loads: Part 2. Application to contaminants discharged to the Great Barrier

Reef, Queensland, Australia. Integ Environ Manag Assess 13(4): 754–764.

Stephan CE, Mount DI, Hansen DJ, Gentile JH, Chapman GA, Brungs WA. 1985. Guidelines for deriving

numerical national water quality criteria for the protection of aquatic organisms and their uses.

Washington DC, USA:US EPA. US EPA Report No. PB-85-227049.

Traas TP, Van de Meent D, Posthuma L, Hamers T, Kater BJ, De Zwart D, Aldenberg, T. 2002. The



p 315–344.

Turner R, Huggins R, Wallace R, Smith R, Vardy S, Warne MStJ. 2012. Sediment, nutrient and pesticide

loads: Great Barrier Reef Loads Monitoring 2009–2010. Water Sciences Technical Report, Volume

2012, Number 14. Brisbane (QLD), Australia: Department of Science, Information Technology,






University of Hertfordshire (2013). The Pesticide Properties Data Base (PPDB). Developed by the




130



US EPA [United States Environmental Protection Agency]. 2000. Revisions to the Water Quality

Planning and Management Regulation and Revisions to the National Pollutant Discharge

Elimination System Program in Support of Revisions to the Water Quality Planning and

Management Regulation. Washington DC, USA: US EPA. 65 Federal Register, 43: 586.




US EPA [United States Environmental Protection Agency]. 2012. Species Sensitivity Distribution

Generator, Ver 1. Cincinnati (OH), USA: US EPA, Office of Research and Development, National

Center for Environmental Assessment. Available from:


US EPA [United States Environmental Protection Agency]. 2015. ECOTOX User Guide: ECOTOXicology

Database System. Version 4.0. [cited 2012/09/09]. Available: http://www.epa.gov/ecotox/

Vallotton N, Eggen RI, Escher BI, Krayenbühl J, Chèvre N. 2008. Effect of pulse herbicidal exposure on

Scenedesmus vacuolatus: a comparison of two photosystem II inhibitors. Environ Toxicol Chem

27(6): 1399-407.

van de Plassche EJ, Polder MD, and Canton JH. 1993. Derivation of maximum permissible

concentrations for several volatile compounds for water and soil. Bilthoven, The Netherlands:

National Institute of Public Health and Environment Protection. Report No. 679101 008.



Walker N, Peterson RE. 2006. The 2005 World Health Organization re-evaluation of human and


241.






http://www.epa.gov/ecotox/

131










Walsh, C. 2004. Pharmacology Drug Actions and Reactions. (7. ed.). New York (NY), USA: Informa

Healthcare.

Warne MStJ, Hawker DW. 1995. The number of components in a mixture determines whether

synergistic and antagonistic or additive toxicity predominate: the funnel hypothesis. Ecotoxicol

Environ Saf 31: 23–28.





Australia: Department of Science, Information Technology, Innovation and the Arts.











Waters DK, Carroll C, Ellis R, Hateley L, McCloskey GL, Packett R, Dougall C, Fentie B. 2014. Modelling

reductions of pollutant loads due to improved management practices in the Great Barrier Reef

132

catchments – Whole of GBR. Toowoomba (QLD), Australia: Queensland Department of Natural

Resources and Mines. Technical Report 1.

Wheeler JR, Leung KMY, Morritt D, Sorokin N, Rogers H, Toy R, Holt M, Whitehouse P, Crane M. 2002.

Freshwater to saltwater toxicity extrapolation using species sensitivity distributions. Env Tox Chem

21(11):2459–2467.



e ologi ally relevant quantitative ethods for ...the gbr. as a result of the governments evolving...

Documents