Toxicity thresholds of nine herbicides to coral symbionts (Symbiodiniaceae)

Over 30 herbicides have been detected in catchments and waters of the Great Barrier Reef (GBR) and their toxicity to key tropical species, including the coral endosymbiotic algae Symbiodiniaceae, is not generally considered in current water quality guideline values (WQGVs). Mutualistic symbionts of the family Symbiodiniaceae are essential for the survival of scleractinian corals. We tested the effects of nine GBR-relevant herbicides on photosynthetic efficiency (ΔF/Fm′) and specific growth rate (SGR) over 14 days of cultured coral endosymbiont Cladocopium goreaui (formerly Symbiodinium clade C1). All seven Photosystem II (PSII) herbicides tested inhibited ΔF/Fm′ and SGR, with toxicity thresholds for SGR ranging between 2.75 and 320 µg L−1 (no effect concentration) and 2.54–257 µg L−1 (EC10). There was a strong correlation between EC50s for ΔF/Fm′ and SGR for all PSII herbicides indicating that inhibition of ΔF/Fm′ can be considered a biologically relevant toxicity endpoint for PSII herbicides to this species. The non-PSII herbicides haloxyfop and imazapic did not affect ΔF/Fm′ or SGR at the highest concentrations tested. The inclusion of this toxicity data for Symbiodiniaceae will contribute to improving WQGVs to adequately inform risk assessments and the management of herbicides in tropical marine ecosystems.

Improving the applicability of water quality guideline values for herbicides in tropical marine ecosystems. Reliable water quality guideline values (WQGVs) are required to assess the potential risks posed by herbicides and mixtures of herbicides, frequently detected in tropical marine waters including the GBR 8,12 . Conventionally, concentrations of individual herbicides detected in the field are compared against the relevant WQGV for that herbicide. For example, time averaged concentrations of up to 0.778 µg L −1 diuron, obtained over a month-long passive sampler deployment, have been detected in the latest GBR monitoring program 12 , a value that exceeds the proposed PC99 WQGV for diuron (0.43 µg L −1 ; protection concentration for 99% of species) 39 . Importantly, measured concentrations of herbicide mixtures can be assessed against a risk metric which predicts the proportion of species potentially affected by that mixture. The metric applied in GBR herbicide monitoring is the multi-substance Potentially Affected Fraction (ms-PAF) 40 , and more exceedances of WQGVs are reported using this approach which accounts for all herbicides present in the environment 12 . The ms-PAF method requires high quality WQGVs for all herbicides detected in the environment; however, several alternative herbicides do not have reliable WQGVs and their contribution to risk can be difficult to assess. The Australian and New Zealand guidelines currently include WQGVs for nine herbicides: the five priority PSII herbicides and four alternative herbicides (2,4-D, bromacil, MCPA, and simazine), but many of these are of low reliability due to lack of toxicity data 41 . Their relevance to tropical marine habitats is also uncertain as the majority of these WQGVs have been derived from toxicity data from freshwater temperate species 23,41 . Marine proposed guideline values (PGVs) have recently been proposed for 20 herbicides: 2,4-D, ametryn, bromacil, diuron, fluometuron, fluroxypyr, haloxyfop, hexazinone, imazapic, isoxaflutole, metribuzin, MCPA, simazine, tebuthiuron, pendimethalin, prometryn, propazine, terbutryn, triclopyr, terbuthylazine 39,42,43 but the majority are of low to moderate reliability due to lack of suitable marine toxicity data.
More toxicity data are required to improve WQGVs for alternative herbicides detected in tropical marine environments, especially for species of critical ecological value such as corals and their symbionts. Here, we developed a 14-day chronic growth inhibition test for cultures of the free-living coral symbiont Cladocopium goreaui to determine the toxicity thresholds for nine herbicides detected in the GBR and the GBRCA. C. goreaui was chosen as a suitable test species as it is common in tropical benthic environments and has a relatively rapid growth rate in comparison to many other Symbiodiniaceae 35, 44 . The PSII herbicides diuron, bromacil, hexazinone, metribuzin, propazine, simazine and tebuthiuron, as well as the non-PSII herbicides haloxyfop and imazapic were chosen in consultation with the Water Quality and Investigation Team at the Queensland Department of Environment and Science and indicate current toxicity data gaps for the improvement of WQGVs. The effects on specific growth rate (SGR) and ΔF/F m ′ were investigated 23,37 . The chronic SGR toxicity thresholds (10% effect concentrations (EC 10 ) and no effect concentrations (NEC)) represent ecologically relevant endpoints required for inclusion in deriving WQGVs 45 . Correlations between ΔF/F m ′ and SGR effect concentrations (EC 50 ) would further validate the application of PAM fluorometry as a sensitive tool to measure sub-lethal stress in this ecologically important coral symbiont.
Bioassay performance. Cladocopium goreaui exhibited consistent exponential growth in control treatments ranging from 0.0877 to 0.163 SGR day −1 among experiments (which were all conducted on separate occasions) ( Table 1). There was no effect of carrier solvent (DMSO and ethanol) on SGR (t-test: F IMK-DMSO (1, 3) = 1.185; p = 0.356 and F IMK-EtOH (1, 4) = 0.529; p = 0.507). The effective quantum yield ΔF/F m ′ was also consistent across control treatments (ΔF/F m ′ = 0.35 ± 0.04). The reference toxicant diuron (6 μg L −1 ) applied in each toxicity test inhibited SGR by 64.9% ± 3.6% (mean ± SD, n = 44) and ΔF/F m ′ by 92.8% ± 1.1% (mean ± SD, n = 22). Toxicity of PSII herbicides: specific growth rate. The inhibition of SGR by PSII herbicides increased with concentration (Fig. 1). Concentrations of each herbicide that inhibited 10% and 50% of SGR (EC 10 and EC 50 , respectively), as well as no effect concentrations (NEC) in C. goreaui are listed in Table 2. All PSII herbicides were toxic but exhibited a broad range of potencies. C. goreaui was most sensitive to the reference herbicide diuron (EC 50 = 4.45 μg L −1 ) and least sensitive to simazine (EC 50 = 387 μg L −1 ) ( Table 2). All relative potencies were proportionally lower than diuron ranging from bromacil (ReP = 0.16) to simazine (ReP = 0.012). Slopes for fitted concentration-response curves were similar for the PSII herbicides ranging between 2.5 and 5.4 (R 2 ≥ 0.83), while hexazinone had a much greater slope (Slope = 9.1, R 2 = 0.97) (Fig. 1). The order of toxicity based on EC 50 values were: diuron > bromacil > metribuzin > propazine > hexazinone > tebuthiuron > simazine ( Table 2). The EC 10 and NEC toxicity thresholds for SGR followed a similar order of toxicity ( Toxicity of PSII herbicides: effective quantum yield. The order of toxicity of PSII herbicides on ΔF/ F m ′ in C. goreaui was the same as per SGR (Table 2). However, ΔF/F m ′ was a more sensitive endpoint: diuron EC 50 = 1.20 μg L −1 and simazine EC 50 = 93.3 μg L −1 ( Table 2). Fitted concentration-response curves ( Fig. 1) had similar shapes, with slopes for all PSII herbicides ranging from 1.2 to 1.9 (R 2 ≥ 0.86). Relative potencies of each herbicide relative to diuron ranged from 0.013 to 0.14 ( Table 2). The EC 10 and NEC thresholds for ΔF/F m ′ inhibition followed a similar order of toxicity to inhibition of SGR ( Toxicity of non-PSII herbicides. Both non-PSII herbicides, imazapic and haloxyfop, failed to inhibit growth or ΔF/F m ′ at the highest concentrations of 165,000 μg L −1 and 2980 μg L −1 , respectively (Table 2; Fig. 3). Higher concentrations of imazapic and haloxyfop were unable to be tested as they altered the pH of the IMK media below an acceptable range.

Discussion
Effects of PSII herbicides on growth. Each of the PSII herbicides inhibited SGR and ΔF/F m ′ in the cultures of the coral symbiont C. goreaui over 14 days. This was expected as the PSII herbicides all bind to the D1 protein site in PSII which is common across phototrophs 46 . Despite the shared mode of action there was a > 80-fold difference in potency between the most toxic herbicide diuron and the least toxic herbicide simazine (comparison of SGR EC 50 s). There was no relationship between toxicity and chemical class. For example, diuron and tebuthiuron are both phenylureas but the toxicity of diuron was > 70-fold greater than tebuthiuron. Substantial differences in toxicity were also evident among the triazines (simazine and propazine) and triazinones (metribuzin and hexazinone). There was also no relationship between EC 50 values ( Table 2) and water solubility or log K OW (Table S1), indicating that transport across cell walls and membranes and/or accumulation within the cells did not have a major influence on relative potency. Instead, the toxicity of each herbicide is most likely related to the specific affinity (steric compatibility and binding properties) between each herbicide and the Q B binding site on the D1 protein in PSII 47 .
Direct comparisons of toxicity between species should be made with caution due to differences in experimental conditions; however, comparing the EC 50 s against datasets with several species' studies carried out under similar conditions can provide insights into the relative sensitivity of C. goreaui compared with other species. Diuron, applied in this study as a reference toxicant, is the most studied PSII herbicide with respect to effects on marine microalgal growth. The growth inhibition EC 50 values for 17 species range between 3.4 and 110 μg L −1 diuron 23,38,39,43,48 , indicating cultured C. goreaui is among the most sensitive. The toxicity of herbicides to two other microalgal species Rhodomonas salina 23 and Chaetoceros muelleri 38 were tested in the same laboratory under similar conditions (Table 3). Comparisons between EC 50 values reveal broadly similar sensitivities www.nature.com/scientificreports/  www.nature.com/scientificreports/ to each PSII herbicide (within threefold differences) except for hexazinone, which was > tenfold more toxic to the cryptophyte R. salina than C. goreaui.

Toxicity of non-PSII herbicides on growth. The two non-PSII herbicides haloxyfop and imazapic
had no observable effect on SGR in C. goreaui. This insensitivity to the non-PSII herbicides is consistent with other marine microalgae R. salina 23 , C. muelleri 38 (Table 3) and Navicula spp. 49 . The phenoxy herbicide haloxyfop inhibits the enzyme acetyl-CoA carboxylase (ACCase inhibitor) and blocks production of lipids and fatty acids 42,50 . ACCase inhibitors such as haloxyfop act on homomeric (eukaryotic) ACCases and sequences for homomeric ACCases have been reported in dinoflagellates 51 , indicating the potential for activity. However, the herbicidal activity of haloxyfop-p-methyl also depends on its hydrolysis within a plant and this may not occur within C. goreaui 42 . Furthermore, Thomas et al. 23 proposed that the active hydrolyzed form of haloxyfop which contains a carboxyl moiety may bind calcium (Ca 2+ ) and/or magnesium (Mg 2+ ) ions in seawater 52 which could affect activity directly or help to stabilize the herbicide at the surface water:air interface 53 . Like haloxyfop, imazapic also has a carboxyl group which may explain its low toxicity to C. goreaui cultured in seawater. Further- www.nature.com/scientificreports/ more, acetohydroxy acid synthase (AHAS or ALS), the target enzyme for imazapic, has not yet been reported in dinoflagellates 54 .  www.nature.com/scientificreports/ Effects of herbicides on photosynthetic efficiency. All PSII herbicides inhibited effective quantum yield (ΔF/F m ′), which is proportional to photosynthetic efficiency at a given light intensity 55 , in cultured Symbiodiniaceae. Inhibition of ΔF/F m ′ by the seven PSII herbicides was on average 4.4 times more sensitive than SGR based on EC 50 ratios (Table 2) and the linear regression slope (Fig. 4). The ratio of inhibition of SGR: ΔF/ F m ′ is comparable to other tropical marine microalgae: R. salina (average SGR: ΔF/F m ′ = 4.3 for 7 PSII herbicides) 23 ; Chaetoceros muelleri (3.0 for 3 PSII herbicides) 38 ; Navicula spp. (1.5 for 3 herbicides) and Nephroselmis    49 . Although growth in microalgae is directly dependent on photosynthesis the relationship with inhibition of ΔF/F m ′ by PSII herbicides may not necessarily be 1:1 since microalgae may draw upon their current resources to continue growth, and some (including Symbiodiniaceae) can adjust to reduced photosynthesis by increasing reliance on heterotrophy 56 . Differences in ΔF/F m ′ : SGR between studies and species may also be due to light intensity which can affect both ΔF/F m ′ and SGR 57 , and the nutritional status of cells which is dependent on the composition of the growth media and may change depending on the duration of exposure 58 . Interestingly, when based on EC 10 s the average ratio of SGR: ΔF/F m ′ was 10.7, indicating an even greater sensitivity of ΔF/F m ′ to herbicides relative to SGR at lower herbicide concentrations (also reflected by the differences in slopes of the concentration response curves for both parameters (Fig. 1)). The reason for this is unclear; however, relative sensitivities between these endpoints may change with exposure duration. For example, inhibition of ΔF/F m ′ by PSII herbicides is relatively consistent after it reaches a rapid maxima 27 , while effects of PSII herbicides on SGR may continue to decline over the course of the exposure period (and the nature of this relationship is likely to change with light intensity 57 ). The insensitivity of ΔF/F m ′ in C. goreaui to the non-PSII herbicides haloxyfop and imazapic was expected as their modes of action do not block electron transport in PSII, hence there would be no increase in chlorophyll a fluorescence 57 . This result was consistent with other tropical marine microalgae R. salina 23 , C. muelleri 38 (Table 3) and Navicula spp. 49 .
The effects of PSII herbicides on ΔF/F m ′ have been assessed for both cultured and symbiotic (in hospite) forms of Symbiodiniaceae. EC 50 values have been reported for Symbiodiniaceae in hospite with corals for the PSII herbicides: ametryn, diuron, hexazinone, atrazine, simazine, tebuthiuron and irgarol 1051 20,27,28,59 , diuron and hexazinone in a jellyfish 60 and diuron in an anemone 61 . However, the species of Symbiodiniaceae in hospite was only identified in one instance 20 . A comparison of the EC 50 s for diuron among cultured and in hospite Symbiodiniaceae (Table 4) shows relatively consistent values among and between these groups. Given that experimental conditions, especially light intensity 22,57 , can affect inhibition of ΔF/F m ′ in microalgae, direct comparisons are difficult between studies. However, one study found Durusdinium trenchii to be twice as sensitive to C. goreaui under identical culture and measurement conditions 35 . It was hypothesized that differences in sensitivity to diuron among Symbiodiniaceae types may be due to the reported differences in capacity to repair damaged PSII reaction centres affected by herbicide-mediated photoinhibition 62 . The diuron EC 50 for C. goreaui in the current study was the lowest reported for this species and, while this may be due to differences in experimental conditions, the comparatively long herbicide exposures (14 days) may also play a role, allowing for a build-up of damage to PSII (chronic photoinhibition) and leading to a depression of ΔF/F m ′ 63 . Nevertheless, the strong and consistent relationships between ΔF/F m ′ and SGR and between Symbiodiniaceae species for multiple PSII herbicides supports the measurement of ΔF/F m ′ inhibition in marine microalgae as a valuable and biologically relevant toxicity endpoint for PSII herbicides (but should not be applied to herbicides with other modes of action).
Relevance of herbicide sensitivity of free-living Symbiodiniaceae. The apparent lack of differences in sensitivity to diuron between cultured, freshly isolated and in hospite Symbiodiniaceae, along with the rapid onset of ΔF/F m ′ inhibition reported for symbiotic corals 27 , indicates the multiple membrane layers of the coral host provide little barrier to diuron reaching intracellular Symbiodiniaceae. There were also similarities in the sensitivity of cultured C. goreaui to EC 50 s reported for coral-hosted Symbiodiniaceae: hexazinone 8.8 µg L −128 and 14 µg L −120 ; simazine 150 µg L −1 and tebuthiuron 175 µg L −127 . These comparisons demonstrate a consistency of access and binding of PSII herbicides to the D1 protein within Symbiodiniaceae, regardless of whether the dinoflagellate is in its symbiotic or free-living form and supports the application of cultured C. goreaui for assessing the toxicity of PSII herbicides.
The exposure of PSII herbicides to Symbiodiniaceae within corals can lead to coral bleaching (breakdown of symbiosis), reduced translocation of autotrophically-derived nutrients to the host, reduced reproduction and mortality 29,30,65 , and similar responses are possible in symbiotic foraminifera 22 , jellyfish 64 and ascidians 61 . Table 3. Comparison among toxicity thresholds of Cladocopium goreaui, Rhodomonas salina and Chaetoceros muelleri. Effect concentration values (SGR EC 50 ) with 95% confidence intervals derived for Cladocopium goreaui, Rhodomonas salina and Chaetoceros muelleri exposed to herbicides. All experiments were conducted under similar conditions in the same laboratory. Concentrations are reported in µg L −1 . EC 50 values for Rhodomonas salina obtained from Thomas et al. 23 and Chaetoceros muelleri from Thomas et al. 38 .

Species
Cladocopium www.nature.com/scientificreports/ The free-living Symbiodiniaceae in culture and in the environment alternate between motile and non-motile forms 66 . This mobility facilitates dispersal and infection of host recruits, which are initially symbiont free for most coral species 67 . Impacts of PSII herbicides on populations of free-living Symbiodiniaceae could limit the onset of mutualistic endosymbiosis in corals and other symbiotic invertebrates that have a critical reliance on their symbionts for autotrophic energy acquisition. Free-living Symbiodiniaceae also represent a reservoir of symbionts for bleached coral hosts that are needed for rapid recovery and survival following thermal bleaching events 31,68 . Indeed, free-living Symbiodiniaceae are widespread in the ocean, with a recent study detecting this family in over 90% of sampling sites and making up 0.1% of total eukaryotic reads in tropical and sub-tropical waters 31 . Impacts by herbicides on these highly diverse and abundant free-living populations of Symbiodiniaceae may therefore have ecological consequences beyond the risk to their role in invertebrate symbiosis.
Implications for water quality guideline values. The Australian and New Zealand marine WQGVs for all herbicides tested here (apart from the reference herbicide diuron) are of low reliability due to lack of appropriate marine toxicity threshold data 39,[41][42][43] . The SGR toxicity thresholds for cultured C. goreaui represent the first toxicity dataset for Symbiodiniaceae that is suitable for application to water quality guideline derivation. The comparatively slow growth rate of C. goreaui (common for all Symbiodiniaceae) meant that the chronic growth assay for this species needed to be longer than the standard 3-or 4-day exposure applied for most algal species. Consequently, the toxicity thresholds identified here could be considered relatively conservative and appropriate for comparison against long in situ exposure durations. NEC and EC 10 toxicity thresholds are the preferred data for inclusion in Species Sensitivity Distributions (SSDs) used to derive WQGVs 45 and these are compared in Table 5 against existing and proposed WQGVs for each of the herbicides. Australian and New Zealand WQGVs are currently derived to protect 99, 95, 90 and 80% (PC99, 95, 90 and 80, respectively) of species in marine and freshwater ecosystems 41 , and Table 5 compares SGR endpoints rather than those for ΔF/F m ′, as inhibition of growth is currently considered to better represent an ecologically relevant impact 45 . The SGR NECs and EC 10 s for all nine herbicides were greater than the proposed WQGVs (Table 5), indicating the proposed WQGVs would be protective of C. goreaui growth. The current WQGVs 41 were proposed three decades ago and current guideline values for bromacil and hexazinone would not be protective of C. goreaui. However, the SGR toxicity thresholds for C. goreaui, in combination with other similar data for the tropical marine cryptophyte R. salina 23 , the diatom C. muelleri 38 and the coral Acropora tenuis 69 will contribute to improving the reliability of WQGVs and their relevance to tropical marine ecosystems such as the GBR. The SGR toxicity thresholds reported here for C. goreaui were all higher than concentrations detected in the GBR of up to 0.778 µg diuron L −1 over time averaged month-long passive sampler deployments 12 . Additionally, the ΔF/Fm′ toxicity thresholds for C. goreaui, except for diuron, were also higher than concentrations reported in the GBR 12 . However, improving WQGVs for alternative pesticides (such as the herbicides in this study) will improve confidence in the application of ms-PAF to predict the total toxicity of all pesticides detected in water quality monitoring programs. For example, over 80% of water samples taken in the GBRCA between 2011 and 2015 contained between 2 and 20 quantifiable pesticides for samples collected in the GBRCA 8 . Including the contribution of all pesticides to ecological risk using ms-PAF results in more reported exceedances of WQGVs 12 , and is the preferred method in monitoring programs that inform agricultural management practices adjacent to sensitive marine habitats such as the GBR 6 . www.nature.com/scientificreports/

Conclusion
The toxicity growth thresholds for alternative herbicides to the free-living form of the common coral symbiont C. goreaui were reported for the first time. The seven PSII herbicides all inhibited photosynthetic efficiency (ΔF/ F m ′), leading to reduced SGR over the 14-d exposure period. The sensitivity of ΔF/F m ′ to PSII herbicides was on average 4.4-fold greater than for SGR, but the clear mechanistic link and consistent relationship between inhibition of ΔF/F m ′ and SGR indicates that inhibition of ΔF/F m ′ should be considered a biologically relevant toxicity endpoint for PSII herbicides to marine microalgae 23,37,38 . The non-PSII herbicides haloxyfop and imazapic did not affect SGR in C. goreaui at very high concentrations indicating these individual herbicides do not pose a significant risk to this species. The effects of PSII herbicides on free-living C. goreaui occurred at similar concentrations as shown for Symbiodiniaceae in hospite, indicating cultures of this species are relevant proxies for both free-living and symbiotic forms of this species. As mutualistic endosymbiotic partners with scleractinian corals, inclusion of toxicity data for Symbiodiniaceae to relevant contaminants such as alternative herbicides will improve our confidence that future WQGVs are adequate to inform risk assessments for tropical marine ecosystems. While this study targeted some of the most frequently detected alternative herbicides in GBR waters, there remains several pesticides, including insecticides and fungicides, with no current WQGVs and further testing is needed to address this.

Materials and methods
Test species and culture conditions. A monoclonal strain of Cladocopium goreaui (formerly Symbiodinium clade C1 70   Specific growth rates (SGR) were expressed as the logarithmic increase in cell density from day i (t i ) to day j (t j ) as per Eq. (1), where SGR i-j is the specific growth rate from time i to j, X j is the cell density at day j and X i is the cell density at day i 58 72 and Karim et al. 77 .
Percent inhibition was calculated relative to controls as per Eq. (3). Inhibition of ΔF/F m ′ suggests reduced energy conversion in PSII of the organism under a known light intensity. These measurements were used to compare photoinhibition between the controls and respective treatments. Three to six replicates at 6 μg L −1 diuron were included in each bioassay to estimate the consistency of inhibitory responses of cells. Microscopy iPAM settings were: actinic light = 1, measuring light = 10-12, gain = 3, damp = 2, actinic light width = 180 s and saturation pulse intensity = 2.
Physico-chemical and herbicide analyses. Physical and chemical characteristics of each treatment were measured at 0 days and 14 days including pH and salinity (LAQUAact-PC110 Meter, HORIBA Scientific) and dissolved oxygen (HQ30D Portable Meter, HACH) as per Thomas et al. 23 . Temperature was logged in 10-min intervals over the duration of the test (HOBO, Onset). Herbicide analytical samples were taken at 0 day and 14 days. Aliquots (1 mL) for chemical analysis were transferred into 1.5 liquid chromatography amber glass vials and spiked with surrogate standards (i.e. diuron-d6, hexazinone-d6, metribuzin-d3, simazine-D10, propazine-D6, bromacil-D3, haloxyfop-D4, and imazapic D7) with a 10 ng mL −1 final concentration of the surrogate standard. The herbicide and degradation product concentrations were determined by HPLC-MS/MS using an SCIEX Triple Quad 6500 QTRAP mass spectrometer (SCIEX, Concord, Ontario, Canada) equipped with an IonDrive Turbo V ion source using a TurboIonSpray probe 23,78,79 . Data acquisition and quantification were performed on MultiQuant 3.0 software by Sciex. Compound identification and confirmation were achieved using retention times and via comparison of SRM transition intensity ratios between the sample and isotopically labelled standard in the same run. The geometric mean from start and end concentrations (time-weighted average) was assigned as the 'measured' concentration in that sample. The average loss from these measured concentrations was then applied to all nominal concentrations as per Thomas et al. 23 .
Data analyses. Measured concentrations were used for all concentration-response modelling and toxicity threshold estimates (Table S1). Specific growth rate and photosynthetic inhibition were calculated as percent inhibition relative to the solvent control or IMK media control (in bioassays when no solvent carrier was used). Herbicide concentrations that inhibited 10 and 50% of growth or ΔF/F m ′ relative to controls (EC 10 and EC 50, respectively) were interpolated from modelled mean values using GraphPad Prism V 7.0 (GraphPad Software Inc., San Diego, USA). Relative potencies (ReP) were calculated for each herbicide (ReP = EC 50 diuron/EC 50 herbicide) against the reference toxicant diuron. The estimation of no effect concentrations (NEC) for SGR was calculated in R (V 3.6.1) as per Thomas et al. 23 . Proportional decline in SGR (1-inhibition) was modelled as a function of log concentration of each herbicide using a Bayesian non-linear gaussian model with the R package jagsNEC 80 . This model has been specifically developed to derive no effect concentrations (NECs) but also allows (1) SGR i -j = ln X j − ln X i t j − t i (day −1 )