SEARCH

SEARCH BY CITATION

Keywords:

  • Uranium;
  • Toxicity;
  • Dissolved organic carbon;
  • Freshwater;
  • Risk assessment

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

The present study reanalyzed 46 existing uranium (U) chronic toxicity datasets for four freshwater species to generate consistent toxicity measures and explore relationships between U toxicity and key physicochemical variables. Dissolved organic carbon (DOC) was consistently the best predictor of U toxicity based on 10% inhibitory concentration (IC10) and median inhibitory concentration (IC50) values, with water hardness also being a significant co-predictor of IC50 concentrations for one species. The influence of DOC on acute and chronic U toxicity was further characterized using existing data for five species, and was found to vary depending on species, DOC source, and exposure duration (acute vs chronic). The slopes of the relationships between DOC and (normalized) acute and chronic U toxicity were modeled using cumulative probability distributions. From these, slopes were selected for which to correct acute or chronic U toxicity values or hazard estimates based on the aquatic DOC concentration. The fifth percentiles of these cumulative probability distributions for acute and chronic exposure data were 0.064 and 0.090, respectively, corresponding to a 6.4 and 9.0% reduction in U toxicity relative to the toxicity at the base DOC concentration for each 1 mg/L increase in DOC concentration (over the DOC range 0–30 mg/L). Algorithms were developed to enable the adjustment of U toxicity values and U hazard estimates, depending on DOC concentrations. These algorithms will significantly enhance the environmental relevance of water quality/risk assessments for U in fresh surface waters. Environ. Toxicol. Chem. 2012; 31: 2606–2614. © 2012 SETAC


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Global nuclear power production is predicted to double by 2030 1. The increased demand for uranium (U) as a long-term alternative energy source to fossil fuels has focused attention on the potential environmental impacts of wastes associated with U mining and milling activities 2–5. This has prompted a marked increase over the past 10 to 15 years in the publication of U toxicity data for freshwater species (Supplemental Data, Table S1). Currently, the size of the U chronic toxicity dataset for ecologically relevant endpoints is sufficient to derive hazard estimates, including water quality guidelines or predicted no effect concentrations, under most of the major contemporary international water quality guideline/risk assessment methodologies 6–9.

However, concomitant with the expansion in U chronic toxicity research has been considerable debate in the ecotoxicology community regarding the most appropriate measure of toxicity to report, particularly for estimates of no or low/acceptable toxicity that are used to derive hazard estimates 10–14. Consequently, the published U chronic toxicity studies report a mix of no/low effects estimates, including no observed effect concentrations (NOECs), lowest observed effect concentrations, 10% bounded effect concentrations (BEC10s), minimum detectable effects concentrations, and 10% effect/inhibition concentrations (EC10s/IC10s; Supplemental Data, Table S1). Although recent methods for deriving hazard estimates have allowed for the inclusion of multiple types of toxicity measures (primarily to increase sample size and reduce uncertainty in species sensitivity distributions 9), the same toxicity measure should ideally be available for each species represented in the U toxicity dataset. With the limitations of the NOEC having been well documented 10–14, low effect concentrations (in particular, EC/IC10s; referred to herein as IC10s) estimated from concentration–response models are currently the preferred measure of toxicity 9, 12, 15. Although other measures also have merit (e.g., no effect concentrations 12, 16), these have been less widely adopted.

Of the 20 published U freshwater chronic toxicity studies listed in Supplemental Data, Table S1, 14 have been undertaken by our research group or collaborators of our research group. Of these, nine studies representing four species (the green algae, Chlorella sp. 1 and Chlorella sp. 2, the green hydra, Hydra viridissima, and the cladoceran Moinodaphnia macleayi) did not report low EC/IC values. Given our access to the raw or summarized datasets from these studies, the present study aimed to reanalyze the data (as well as data from several unpublished tests) using concentration–response modeling, and to estimate and report consistent toxicity measures, specifically IC10 and IC50 values. The 10% effect/inhibition level was selected as the low/acceptable effect level because this has typically been the most accepted and reported level in the literature 14.

In addition, the reanalysis of the U toxicity data and associated collation of accompanying test water quality data enabled relationships between the toxicity measures (IC10 and IC50) and key physicochemical variables known to influence U speciation or bioavailability to be explored. Specifically, dissolved organic carbon (DOC), pH, alkalinity, and water hardness are all known to affect U speciation or bioavailability in various ways 17 and, where possible, were included in the analyses. The aim was to determine the best predictor variable(s) for chronic toxicity and, if significant relationships were found, undertake further analyses to derive an algorithm that could be used to modify hazard estimates for U. The overall intent of the present study was to lay part of the groundwork toward the re-derivation of more robust and environmentally relevant hazard estimates for U in fresh surface waters.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Data collation

The first part of the study represented a desktop analysis of existing U chronic toxicity data from the following studies for (1) Chlorella sp. 1—Franklin et al. 18, Hogan et al. 19 and Trenfield et al. 20; (2) Chlorella sp. 2—Charles et al. 21; (3) H. viridissima—Hyne et al. 22, Markich and Camilleri 23, Riethmuller et al. 24, and Trenfield et al. 20; and (4) M. macleayi—Hyne et al. 25, Semaan et al. 26, and four unpublished tests from 1992. The Chlorella isolates were both tropical strains but were collected from different locations (1 Lake Aesake, Strickand River, Papua New Guinea, and 2 Magela Creek, Northern Territory, Australia) and considered to be different species. For H. viridissima and M. macleayi, all studies used the same tropical strains. All tests for each species were undertaken using the same protocols: 72-h cell division rate for Chlorella spp.; 96-h population growth rate for H. viridissima; and three-brood reproduction for M. macleayi, as detailed by Riethmuller et al. 27. However, depending on study objectives, sometimes differences were seen in test physicochemistry, such as pH, hardness, alkalinity, and DOC. Uranium exposure concentrations were quantified and reported for all studies. Only tests for which the control treatment met the relevant control performance acceptability criteria (defined in Riethmuller et al. 27) were reanalyzed. The collated toxicity data for each test represented either the raw data or summarized data in the form of means ± standard errors of the mean for each exposure concentration. For one of the tests (from Hyne et al. 22), standard errors of the mean values were not available for the mean response data at each of the exposure concentrations. For all of the tests, corresponding test data for pH, water hardness, alkalinity (or electrical conductivity, where hardness and alkalinity data were not reported), DOC, and water temperature were also collated and reported with the reanalyzed toxicity data.

Data analysis

Consistent with the format of the original published data, all toxicity data were presented as a function of the response of the control treatments (i.e., percentage of control response). For each dataset, nonlinear regression was undertaken (SigmaPlot Version 11.0, Systat Software) to derive a concentration–response relationship with associated 95% confidence limits, and the IC10 and IC50 concentrations for U were calculated using the model equations. Model choice was dependent on the best fit, as determined by the coefficient of determination (r2), of either a three-parameter sigmoidal or three-parameter logistic model. Assumptions of normality (Shapiro Wilk's test; α = 0.05) and homoscedasticity (Spearman rank correlation between the absolute values of the residuals and the observed value of the dependent variable; α = 0.05) were checked before model fitting.

Standard stepwise multiple linear regressions (forward and backward; Minitab 15.1.1.0, Minitab) were undertaken on the combined data for both Chlorella species (n = 20) and the data for H. viridissima (n = 16) to determine which of the key physicochemical variables were the best predictors of U toxicity, as expressed by the IC10 and IC50 values. The following independent variables were included in the stepwise regressions: Chlorella spp., pH, hardness, and DOC; and H. viridissima, hardness, alkalinity, and DOC. Variables (e.g., pH, temperature) that were excluded from the regressions were done so on the basis that their ranges were considered too small to provide adequate resolution in terms of their influence on U toxicity. Moreover, the included variables had been the subject of specific investigation in at least one of the studies for each of the species. The included studies that had assessed hardness- or alkalinity-related effects on U toxicity 21, 24 had decoupled the normally correlated nature of these two physicochemical variables in the environment and, as such, the results were not confounded by this potential correlation. A sequential Bonferroni correction was used to adjust the type I error rate (α = 0.05) to account for the effect of multiple testing on this rate 28. Assumptions of normality and homoscedasticity were tested as described. Physicochemical influences on U toxicity to M. macleayi could not be explored because of the lack of both reported physicochemical data and studies specifically addressing such effects.

Depending on the results of the stepwise regressions, further analyses were undertaken to determine an algorithm that would enable modification of U hazard estimates based on the significant physicochemical variable(s). The details of these analyses are described and discussed in the Results and Discussion section.

RESULTS AND DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Recalculated toxicity measures

The assumptions of normality and homoscedasticity were met for 45 of 46 and 39 of 46 of the datasets, respectively. Data transformations (square root and log10) of the nonconforming datasets did not rectify the assumption violations, but given the small number of violations and the excellent regression coefficients (r2 = 0.86–0.99) for all of these datasets, the original untransformed data and associated models were retained.

The recalculated toxicity measures for Chlorella sp. 1 and sp. 2, H. viridissima and M. macleayi, along with the corresponding key water physicochemical variables and original toxicity measures, are presented in Tables 1, 2, and 3 respectively, with the corresponding concentration–response plots provided in Supplemental Data, Figures S1, S2, and S3, respectively. Generally, the recalculated IC10 values were more similar to originally reported NOEC values than BEC10 values. This is illustrated in Figure 1, which shows the relationship between the original toxicity measures and the recalculated IC10 values. For the NOEC–IC10 relationship, the slope was closer to 1, and the y intercept much closer to zero than for the BEC10–IC10 relationship. The conservatism of the BEC10 compared with the IC10 is not surprising given the method of calculation of the former value, and was indeed noted by the researchers who first proposed the BEC10 as an alternative to the NOEC 29. The BEC10 has not received popular support among ecotoxicologists, possibly because of its lack of intuitiveness 14.

Table 1. Physicochemical and original and recalculated uranium toxicity data for Chlorella sp. 1 and 2 based on 72-h cell division rate
Original studypHHardness (mg/L as CaCO3)Alkalinity (mg/L as CaCO3)Dissolved organic carbon (mg/L)Temp. (°C)Original toxicity measures (µg U/L)Recalculated toxicity measures (95% CLs; µg U/L)Model typea (r2, n, p)
Nil/low effectIC50IC10IC50
  • a

    Model type fitted to data: 3-p log = 3-parameter logistic; 3-p sig = 3-parameter sigmoid.

  • b

    Based on reported nominal concentration of HCOmath image in the test medium.

  • c

    Assumption of homoscedasticity not met.

  • d

    Represents mid-point of a reported range.

  • e

    Assumptions of normality and homoscedasticity not met.

  • IC50 = median inhibitory concentration; IC10 = 10% inhibitory concentration; NR (low) = not reported, but known to be low in value; NC = not calculable; BEC10 = 10% bounded effect concentration; NOEC = no observed effect concentration; CL = confidence limits.

Chlorella sp. 1
 Franklin et al. 185.73.92.6b02721 (BEC10)7845 (35–55)87 (82–92)3-p sig (0.976, 21, <0.0001)
 6.53.92.6b02711 (BEC10)4415 (10–20)48 (41–55)3-p log (0.961, 25, <0.0001)c
 Hogan et al. 196.53.62.6b02938 (NOEC)7452 (38–64)74 (65–91)3-p log (0.920, 24, <0.0001)
 6.7d4.1114.129150 (NOEC)177135 (120–148)176 (168–185)3-p log (0.926, 24, <0.0001)c
 6.3d3.2NR (low)3.429109 (NOEC)166134 (130–140)161 (156–166)3-p log (0.995, 21, <0.0001)c
 6.5d4.778.129157 (NOEC)238176 (169–190)237 (233–242)3-p sig (0.963, 24, <0.0001)
 6.5dNR (low)<52.62972 (NOEC)137100 (89–108)134 (130–140)3-p sig (0.985, 21, <0.0001)e
 Trenfield et al. 206.23.64.1028.5NR3814 (NC–23)38 (32–43)3-p sig (0.943, 12, <0.0001)
 6.23.64.11.028.5NR12458 (NC–90)98 (68–123)3-p log (0.694, 12, 0.0020)
 6.23.64.15.128.5NR256129 (NC–173)237 (215–259)3-p sig (0.931, 12, <0.0001)
 6.23.64.110.228.5NR468196 (NC–278)396 (323–487)3-p log (0.891, 12, <0.0001)
 6.23.64.120.428.5NR744197 (NC–400)515 (310–726)3-p log (0.699, 12, 0.0018)
 6.04.64.5028.5NR133.8 (NC–7.3)11 (7.5–11)3-p log (0.921, 10, <0.0001)
 6.04.64.51.028.5NR3518 (16–20)34 (33–35)3-p log (0.997, 10, <0.0001)
 6.04.64.54.728.5NR8257 (45–66)80 (75–86)3-p log (0.977, 10, <0.0001)
 6.04.64.59.528.5NR150108 (88–127)149 (136–159)3-p log (0.978, 10, <0.0001)
Chlorella sp. 2
 Charles et al. 217.088<0.2270.7 (BEC10)569 (NC–33)66 (29–1083-p log (0.901, 10, 0.0001)
 7.0408<0.2270.7 (BEC10)7211 (NC–19)74 (55–103)3-p log (0.960, 10, <0.0001)
 7.01008<0.2272.3 (BEC10)15032 (NC–79)137 (77–205)3-p log (0.889, 10, 0.0002)
 7.04008<0.2274.5 (BEC10)27061 (NC–141)220 (125–303)3-p log (0.895, 10, <0.0001)
Table 2. Physicochemical and original and recalculated uranium toxicity data for Hydra viridissima (based on 96-h population growth rate)
Original studypHHardness (mg/L as CaCO3)Alkalinity (mg/L as CaCO3)Dissolved organic carbon (mg/L)Temp. (°C)Original toxicity measures (µg U/L)Recalculated toxicity measures (95% CLs; µg U/L)Model typea (r2, n, p)
Nil/low effectIC50IC10IC50
  • a

    Model type: 3-p log = 3-parameter logistic; 3-p sig = 3-parameter sigmoid.

  • b

    Not specifically reported in Hyne et al. 22, but obtained from historical ERISS Ecotoxicology laboratory data records.

  • c

    Represents mid-point of a reported range.

  • d

    Based on reported nominal concentration of HCOmath image in the test medium.

  • e

    Assumption of homoscedasticity not met.

  • IC50 = median inhibitory concentration; IC10 = 10% inhibitory concentration; NR (low) = not reported, but known to be low in value; NC = not calculable; BEC10 = 10% bounded effect concentration; NOEC = no observed effect concentration; CL = confidence limits.

Hyne et al. 226.5NR (low)NR (low)NR30<120 (NOEC)bNR104 (NC–168)199 (156–245)3-p sig (0.945, 6, 0.006)
 6.4cNR (low)NR (low)NR30150 (NOEC)NR170 (NC–267)>3503-p sig (0.801, 6, 0.0413)
Markich and Camilleri 2363.62.6d02756 (BEC10)10865 (58–73)106 (102–111)3-p sig (0.987, 15, <0.0001)
Riethmuller et al. 246.06.6402714 (BEC10)11449 (27–63)115 (109–122)3-p sig (0.981, 14, <0.0001)
 6.0165402781 (BEC10)177128 (111–142)181 (174–187)3-p sig (0.982, 16, <0.0001)
 6.0330402747 (BEC10)21976 (NC–128)207 (180–232)3-p sig (0.903, 16, <0.0001)
 6.016510202725 (BEC10)17157 (NC–84)164 (149–178)3-p sig (0.974, 11, <0.0001)
Trenfield et al. 206.13.64.1026NR6729 (NC–46)65 (57–74)3-p sig (0.868, 21, <0.0001)
 6.13.64.10.926NR12064 (15–82)119 (110–128)3-p sig (0.909, 21, <0.0001)
 6.13.64.14.926NR230122 (17–156)228 (213–244)3-p sig (0.861, 28, <0.0001)e
 6.13.64.19.726NR311146 (94–185)306 (288–323)3-p sig (0.944, 27, <0.0001)e
 6.13.64.119.526NR505229 (NC–313)494 (454–532)3-p sig (0.930, 21, <0.0001)
 6.14.64.5027NR5028 (23–32)48 (45–52)3-p log (0.983, 14, <0.0001)
 6.14.64.50.927NR5416 (0.3–24)54 (49–59)3-p sig (0.981, 14, <0.0001)
 6.14.64.54.827NR7922 (4.7–31)79 (73–85)3-p sig (0.982, 14, <0.0001)
 6.14.64.59.727NR11348 (26–66)113 (103–124)3-p sig (0.984, 14, <0.0001)
Table 3. Physicochemical and original and recalculated uranium toxicity data for Moinodaphnia macleayi (based on 3-brood reproduction and, for Hyne et al. 25, 3-brood parental survival).
Original studypHHardness (mg/L as CaCO3)Conductivity (µS/cm)Temp. (°C)Original toxicity measures (µg U/L)Recalculated toxicity measures (95% CLs; µg U/L)Model typea (r2, n, p)
Nil/low effectIC50IC10IC50
  • a

    Model type: 3-p log = 3-parameter logistic; 3-p sig = 3-parameter sigmoid.

  • b

    Not specifically reported in Semaan et al. 26, but calculated from Ca and Mg concentrations provided in a related internal report 34.

  • IC50 = median inhibitory concentration; IC10 = 10% inhibitory concentration; NR (low) = not reported, but known to be low in value; NC = not calculable; NOEC = no observed effect concentration; ERISS = Environmental Research Institute of the Supervising Scientist, Australia; CL = confidence limits.

ERISS unpublished data6.5NR172715 (NOEC)NR26 (NC–41)39 (27–54)3-p log (0.958, 5, 0.021)
(1992)6.7NR282716 (NOEC)NR26 (NC–42)37 (25–50)3-p log (0.940, 6, 0.007)
 7.0NR272714 (NOEC)NR12 (11–13)21 (20–22)3-p sig (0.999, 6, <0.0001)
 7.0NR292717 (NOEC)NR16 (NC–23)24 (20–29)3-p sig (0.884, 6, 0.019)
Hyne et al. 256.7NR222710 (NOEC)NR13 (NC–20)22 (15–26)3-p log (0.986, 7, <0.0001)
Semaan et al. 266.75.2b21277.8 (NOEC)NR1.5 (0.6–2.9)NC3-p log (0.994, 5, 0.003)
(Laboratory population)6.75.2b222722 (NOEC)NR16 (NC–40)32 (19–NC)3-p sig (0.765, 6, 0.04)
 7.15.2b2227>46 (NOEC)NR42 (NC–NC)NC3-p sig (0.085, 6, 0.417)
Semaan et al. 266.94.6b182725 (NOEC)NR26 (24–31)36 (33–39)3-p log (0.997, 6, <0.0001)
(Wild population)6.84.6b172729 (NOEC)NR25 (13–33)48 (42–NC)3-p log (0.942, 6, 0.0065)
thumbnail image

Figure 1. Relationships (expressed as linear regressions) between the originally reported no/low effect toxicity measures (i.e., no observed effect concentration [NOEC] = solid line; 10% bounded effect concentration [BEC10] = broken line) and the recalculated IC10 values.

Download figure to PowerPoint

The ranges within each of the IC10 and IC50 values spanned a factor of approximately 50 across the Chlorella spp. studies and 10 to 15 across the H. viridissima studies (Tables 1 and 2, respectively). However, this is not surprising given that several studies specifically investigated effects of particular physicochemical variables on U toxicity. Consequently, to assume a geometric mean of the IC10 or IC50 values for an individual species as an accurate representation of U toxicity to that species without accounting for water physicochemistry is not appropriate. For this reason, the search for quantitative relationships between key physicochemical variables and U toxicity is extremely important, as discussed in the following section.

In contrast to Chlorella spp. and H. viridissima, the toxicity data across the three M. macleayi studies were relatively consistent (Table 3). With the exception of only two tests (Semaan et al. 26), the ranges within each of the IC10 and IC50 values spanned a factor of only 2. All three of the M. macleayi studies used natural control/diluent water from sites of similar water quality within the same creek and did not specifically assess modifying factors of U toxicity. Consequently, the toxicity data (geometric means: IC10, 16 µg/L; IC50, 31 µg/L) can be considered representative of U toxicity to this species in slightly acidic, low conductivity (and hardness and alkalinity) freshwaters with DOC concentrations less than 10 mg/L (based on knowledge of typical water chemistry in this creek system). Unfortunately, DOC concentrations were not reported for any of the M. macleayi studies, precluding any correction of the data based on a U toxicity–DOC relationship (see later discussion).

Sheppard et al. 30 reviewed the U toxicity literature and attempted to estimate EC25s from the data for the purposes of deriving predicted no effect concentrations for a range of biotic groups. However, this involved simply interpolating from the data provided in the published results and thus did not represent a rigorous reanalysis of the data. Moreover, a 25% effect level is not representative of no effect.

As a final observation, the 95% confidence limits for the IC10 values were often quite wide and, in a number of cases, the lower 95% confidence limit was not calculable. This high level of uncertainty can be primarily attributed to relatively high natural variability in responses observed among concentrations in the vicinity of the low effect estimates. Additionally, a few of the original experiments (e.g., Hyne et al. 22 Supplemental Data, Fig. S2a; Semaan et al. 26 Supplemental Data, Fig. S3c) were not specifically designed for concentration–response modeling and hence included too few concentrations to enable a precise estimation of low effect concentrations.

Key variables and U toxicity (IC10s and IC50s)

The water quality dataset accompanying the toxicity data was not fully representative of the ranges of values found in natural freshwaters. In particular, the pH range was narrow and, although the hardness and alkalinity data spanned globally relevant ranges, they were skewed toward low hardness and low alkalinity waters. However, the DOC values covered a wide range and were evenly distributed across this range and, as such, can be considered globally relevant. Of the mix of four water quality variables examined, DOC was consistently the best predictor variable for the IC10 and IC50 datasets for the algae and hydra (i.e., DOC vs IC10: Chlorella spp. r2 = 0.68, p < 0.001; H. viridissima r2 = 0.48, p = 0.004; DOC vs IC50: Chlorella spp. r2 = 0.74, p < 0.001; H. viridissima r2 = 0.59, p < 0.001; Supplemental Data, Table S2). Water hardness was also a significant predictor variable for Chlorella spp. IC50 values, explaining an additional 8% of the variation in the IC50 values (r2 = 0.82; Supplemental Data, Table S2). The findings are generally consistent with the literature (both that used in the present meta-analysis as well as studies for other species); DOC has been found to significantly reduce U toxicity to all species for which it has been assessed (see DOC-based algorithms: U toxicity values and hazard estimates), whereas the degree of influence of hardness on U toxicity appears to be very species-specific (i.e., strong effect for Chlorella sp. 2 21; moderate effect for H. viridissima 24; and little to no effect for Mogurnda mogurnda 24).

Alkalinity and pH were not significant predictor variables (noting their narrow ranges), although they are considered to potentially play a role in U speciation and toxicity 17. Given the generally small sample sizes used for the meta-analyses for Chlorella spp. and H. viridissima, and the lesser amount of specific pH- and alkalinity-focused U toxicity data, the results should not be interpreted to mean that pH and alkalinity do not influence U toxicity. Although pH has been shown to affect U toxicity 18, 31, 32, the direction of this effect appears to be species-specific. This makes it difficult to appropriately account for pH when deriving water quality guidelines or predicted no effect concentrations for risk assessment. Further quantification of the influence of these variables, in particular pH, on U toxicity is warranted.

DOC-based algorithms: Adjusting U toxicity values and hazard estimates

Given the strength of the association between DOC and U toxicity demonstrated, focus was placed on the development of an algorithm to adjust U hazard estimates depending on surface water DOC concentrations. To do this, published data for as many species as possible were used. The influence of DOC on U toxicity has been specifically investigated for five species: Chlorella sp. 1 19, 20, H. viridissima 20, Velesunio angasi 31, M. mogurnda 20, and Euglena gracilis 33. The studies using V. angasi and M. mogurnda were based on acute U toxicity and were included to increase species representation and DOC–toxicity response understanding. The study using E. gracilis assessed U toxicity at only two DOC concentrations (10 and 30 mg/L) but was included because the other DOC–U toxicity relationships (based on n 3) demonstrated strong linearity across the DOC ranges (which is consistent with U.S. Environmental Protection Agency data acceptability guidelines for deriving hardness corrections 7). Linear regressions were performed (SigmaPlot Version 11.0) on the DOC versus IC50 or median lethal concentration (LC50) datasets from the above studies/species. The IC50 and LC50 values were used in preference to IC10 values because of their stronger relationship to DOC (Supplemental Data, Table S2) and because they represent more robust estimates of toxicity and, by corollary, any changes in toxicity. Uranium IC/LC50 data for each fixed DOC level were first normalized by expressing them as the proportional reduction in IC/LC50 to that IC/LC50 observed for the background (10 mg/L for E. gracilis, <0.1 mg/L for all other species) DOC level (i.e., [IC50DOCx – IC50DOCbackground]/IC50DOCbackground, where IC50DOCx = the IC50 at the fixed DOC concentrations). The resulting regression relationships are shown in Figure 2(A–E) and Table 4.

thumbnail image

Figure 2. Linear regressions of U toxicity (expressed as the proportion reduction in median inhibitory concentration/median lethal concentration [IC50/LC50] relative to the IC50/LC50 at the background dissolved organic carbon [DOC] concentration) versus DOC concentration for (A) Chlorella sp. 1; (B) Hydra viridissima; (C) Velesunio angasi; (D) Mogurnda mogurnda; and (E) Euglena gracilis. Except for the E. gracilis regression, the y intercept of all models was set to zero. The value accompanying each of the regression lines represents the slope of the relationship. Additional model details are presented in Table 4. SRFA: Suwannee River fulvic acid; SBW DOC: Sandy Billabong dissolved organic carbon.

Download figure to PowerPoint

Table 4. Regression statistics for linear regressions of dissolved organic carbon (DOC; mg/L) versus the proportion reduction in median inhibitory/median lethal concentration (IC50/LC50) valuea
SpeciesStudynSloper2p value
  • a

    Regression plots are shown Figure 2.

  • b

    Assumption of homoscedasticity not met.

  • SRFA = Suwannee River fulvic acid; SBW DOC = Sandy Billabong water dissolved organic carbon; N/A = Not applicable, due to the model being based on only two values (see text for justification of its inclusion).

Chlorella sp. 1Hogan et al. 1950.300.950.003
 

Trenfield et al. 20

–SRFA
50.690.860.014
 

Trenfield et al. 20

–SBW DOC
41.30.990.003b
Hydra viridissima

Trenfield et al. 20

–SRFA
50.350.950.003
 

Trenfield et al. 20

–SBW DOC
40.140.99<0.001b
Velesunio angasi

Markich et al. 31

–pH 5.0
30.110.870.17b
 

Markich et al. 31

–pH 5.5
30.210.890.15b
 

Markich et al. 31

–pH 6.0
30.100.970.08b
Mogurnda mogurnda

Trenfield et al. 20

–SRFA
50.190.99<0.001
 

Trenfield et al. 20

–SBW DOC
40.080.970.011b
Euglena gracilisTrenfield et al. 3320.17N/AN/A

Analysis of covariance (Minitab 15.1.1.0) indicated that the slopes of the regressions were significantly different (analysis of covariance test × DOC interaction term: df = 10, F = 22.67, p < 0.001). The same was also true when the data were separated according to acute (V. angasi, M. mogurnda; df = 4, F = 13.08, p = 0.001) and chronic (Chlorella sp. 1, H. viridissima, E. gracilis; df = 5, F = 22.08, p < 0.001) exposures. Thus, the data could not be pooled for further regression analysis and associated estimation of a single slope for the DOC–U toxicity relationship. Instead, the slopes for the acute and chronic data were plotted separately as cumulative probability distributions to create models (based on a log-normal distribution) of the distribution of slopes for the relationship between DOC and U toxicity for all species (Fig. 3). This method incorporates the natural variability in the relationship between DOC and U toxicity across species and DOC sources. A feature of this approach is that any percentile from the distribution can be selected, based on the objectives of the assessment (e.g., lower or higher protection), to represent a DOC correction factor for a hazard estimate. We selected the fifth percentile of the distribution, which is consistent with the typical derivation of 5% hazardous concentrations from species sensitivity distributions 6–9. In the case of the DOC versus U toxicity data, the fifth percentile slopes were 0.064 for acute exposures and 0.090 for chronic exposures (over the DOC range of 0–30 mg/L).

thumbnail image

Figure 3. Log-normal cumulative probability distributions of the slopes of the relationships for dissolved organic carbon versus uranium toxicity (based on normalized median inhibitory concentration/median lethal concentration [IC/LC50] values; see Table 4 for slope details) for acute and chronic exposures. The short dashed lines show the intersection of the fifth percentile with the models (acute slope fifth percentile = 0.064; chronic slope fifth percentile = 0.090). Model statistics as follows: acute slopes, n = 5, Anderson-Darling goodness-of-fit = 0.325, p = 0.357; chronic slopes, n = 6, Anderson-Darling goodness-of-fit = 0.210, p = 0.747.

Download figure to PowerPoint

These slope factors can be used to modify hazard estimates for U based on aquatic DOC concentration. First, the relevant (acceptable quality) acute or chronic U toxicity values for freshwater species require adjustment to a standard low DOC concentration. For this purpose, a DOC concentration of 1 mg/L was considered a globally applicable estimate of a low DOC concentration. Thus, the U toxicity values (e.g., IC10 or IC50) can be corrected to 1 mg/L DOC using the following equation

  • equation image(1)

where U tox1 is the U toxicity value corrected to 1 mg/L DOC, U toxi is the initial (i.e., original) toxicity value, DOCi is the DOC concentration in mg/L at which U toxi was calculated, and slope is the relevant slope factor for acute (0.064) or chronic (0.090) toxicity. Once the hazard estimate has been derived using the corrected toxicity data, it can then be adjusted based on the DOC concentration in the aquatic environment of interest, using the following equation (using water quality guidelines as an example)

  • equation image(2)

where GV1 is the default guideline value calculated at 1 mg/L DOC, DOCf is the aquatic DOC concentration of interest, and slope is the relevant slope factor for acute (0.064) or chronic (0.090) toxicity (depending on whether the hazard estimate was derived using acute or chronic toxicity data). As an example, for a U GV1 of 5 µg/L based on chronic toxicity data, the DOC modified guideline value for a surface water with a DOC concentration of 4.5 mg/L would be 6.4 µg/L (i.e., 5/[1 + 0.09] × [1 + 4.5 × 0.09]).

The above DOC correction method for U hazard estimates will significantly enhance the environmental relevance of water quality/risk assessments for U in fresh surface waters. In the near future, the method is intended to be used for revisions of a site-specific U guideline in northern Australia 19 and, potentially, the current Australian and New Zealand U water quality guideline 6. Moreover, the DOC–U toxicity slope model and associated algorithms can be readily updated and improved as additional data are published (e.g., for temperate or northern hemisphere species). Finally, further research efforts should focus on further quantifying relationships between U toxicity and other key physicochemical variables, especially pH and water hardness. If such relationships can be quantified for several species, biotic ligand models for U may represent the way forward for future derivations of U hazard estimates.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

We are grateful for the constructive advice on the manuscript by D. Jones and C. Costello of the Environmental Research Institute of the Supervising Scientist, as well as three anonymous reviewers. Advice on mathematical aspects was provided by M. Theodore (Charles Darwin University).

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information
  • 1
    International Atomic Energy Agency. 2008. Energy, electricity and nuclear power estimates for the period to 2030. Reference Data Series 1. Vienna, Austria.
  • 2
    Markich SJ, Jeffree RA. 2002. The Finniss River. A natural laboratory of mining impacts—past present and future. ANSTO E/748. Australian Nuclear Science and Technology Organisation, Sydney, Australia.
  • 3
    Pyle GG, Swanson SM, Lehmkuhl DM. 2002. Toxicity of uranium mine receiving waters to early life stage fathead minnows (Pimephales promelas) in the laboratory. Environ Pollut 116:243255.
  • 4
    van Dam RA, Humphrey CL, Martin P. 2002. Mining in the Alligator Rivers Region, northern Australia: Assessing potential and actual impacts on ecosystem and human health. Toxicology 181/182:505515.
  • 5
    Antunes SC, de Figueiredo DR, Marques SM, Castro BB, Pereira R, Gonçalves F. 2007. Evaluation of water column and sediment toxicity from an abandoned uranium mine using a battery of bioassays. Sci Total Environ 374:252259.
  • 6
    Australian and New Zealand Environment and Conservation Council and Agriculture and Resource Management Council of Australia and New Zealand. 2000. Australian and New Zealand guidelines for fresh and marine water quality. Canberra, Australia.
  • 7
    Stephan CE, Mount DI, Hansen D, Gentile JH, Chapman GA, Brungs WA. 1985. Guidelines for deriving numerical national water quality criteria for the protection of aquatic organisms and their uses. National Technical Information Service, Springfield, VA, USA.
  • 8
    European Commission. 2003. Technical guidance document on risk assessment, Part 2: Environmental risk assessment. EUR 20418 EN/2. European Chemicals Bureau, Ispra, Italy.
  • 9
    Canadian Council of Ministers of the Environment. 2007. A protocol for the derivation of water quality guidelines for the protection of aquatic life 2007. Winnipeg, Canada.
  • 10
    Chapman PM, Cardwell RS, Chapman PF. 1996. A warning: NOECs are inappropriate for regulatory use. Environ Toxicol Chem 15:7779.
  • 11
    Moore DRJ, Caux PY. 1997. Estimating low toxic effects. Environ Toxicol Chem 16:794801.
  • 12
    van der Hoeven N. 2004. Current issues in statistics and models for ecotoxicological risk assessment. Acta Biotheor 52:201217.
  • 13
    Newman MC. 2008. What exactly are you inferring? A closer look at hypothesis testing. Environ Toxicol Chem 27:10131019.
  • 14
    Warne M, van Dam R. 2008. NOEC and LOEC data should no longer be generated or used. Australas J Ecotoxicol 14:15.
  • 15
    Organisation for Economic Co-operation and Development. 1998. Report of the OECD Workshop on Statistical Analysis of Aquatic Toxicity Data. OECD Environmental Health and Safety Publication, Series on Testing and Assessment No. 10. ENV/MC/CHEM(98)18, Environment Directorate. Paris, France.
  • 16
    Fox DR. 2010. A Bayesian approach for determining the no effect concentration and hazardous concentration in ecotoxicology. Ecotoxicol Environ Saf 73:123131.
  • 17
    Markich SJ. 2002. Uranium speciation and bioavailability in aquatic systems: An overview. Sci World J 2:707729.
  • 18
    Franklin NM, Stauber JL, Markich SJ, Lim RP. 2000. pH-dependent toxicity of copper and uranium to a tropical freshwater alga (Chlorella sp.). Aquat Toxicol 48:275289.
  • 19
    Hogan AC, van Dam RA, Markich SJ, Camilleri C. 2005. Chronic toxicity of uranium to a tropical green alga (Chlorella sp.) in natural waters and the influence of dissolved organic carbon. Aquat Toxicol 75:343353.
  • 20
    Trenfield MA, Ng JC, Noller BN, Markich SJ, Van Dam RA. 2011. Dissolved organic carbon reduces uranium bioavailability and toxicity. 2. Uranium [VI] speciation and toxicity to three tropical freshwater organisms. Environ Sci Technol 45:30823089.
  • 21
    Charles A, Markich S, Stauber J, de Filippis L. 2002. The effect of water hardness on the toxicity of uranium to a tropical freshwater alga (Chlorella sp.). Aquat Toxicol 60:6173.
  • 22
    Hyne RV, Rippon GD, Ellender G. 1992. pH-dependent uranium toxicity to freshwater hydra. Sci Total Environ 125:159173.
  • 23
    Markich SJ, Camilleri C. 1997. Investigation of metal toxicity to tropical biota: Recommendations for revision of the Australian water quality guidelines. Supervising Scientist Report 127. Department of Sustainability, Environment, Water, Population and Communities, Darwin, Australia
  • 24
    Riethmuller N, Markich S, Parry D, van Dam R. 2000. The effect of true water hardness and alkalinity on the toxicity of Cu and U to two tropical Australian freshwater organisms. Supervising Scientist Report 155. Department of Sustainability, Environment, Water, Population and Communities, Darwin, Australia.
  • 25
    Hyne RV, Padovan A, Parry DL, Renaud SM. 1993. Increased fecundity of the cladoceran Moinodaphnia macleayi on a diet supplemented with a green alga, and its use in uranium toxicity tests. Aust J Mar Fresh Res 44:389399.
  • 26
    Semaan M, Holdway DA, van Dam R. 2001. Comparative sensitivity of three populations of the cladoceran Moinodaphnia macleayi to acute and chronic uranium exposure. Environ Toxicol 16:365376.
  • 27
    Riethmuller N, Camilleri C, Franklin N, Hogan A, King A, Koch A, Markich S, Turley C, van Dam R. 2003. Ecotoxicological testing protocols for Australian tropical freshwater ecosystems. Supervising Scientist Report 173. Department of Sustainability, Environment, Water, Population and Communities, Darwin, Australia.
  • 28
    Quinn GP, Keough MJ. 2002. Experimental design and data analysis for biologists. Cambridge University Press, Cambridge, UK.
  • 29
    Hoekstra JA, van Ewijk PH. 1993. The bounded-effect-concentration as an alternative to the NOEC. Sci Total Environ 705:S711.
  • 30
    Sheppard SC, Sheppard MI, Gallerand M, Sanipelli B. 2005. Derivation of ecotoxicity thresholds for uranium. J Environ Radioactiv 79:5583.
  • 31
    Markich SJ, Brown PL, Jeffree RA, Lim RP. 2000. Valve movement responses of Velesunio angasi (Bivalvia: Hyriidae) to manganese and uranium: An exception to the free ion activity model. Aquat Toxicol 51:155175.
  • 32
    Cheng KL, Hogan AC, Parry DL, Markich SJ, Harford AJ, van Dam RA. 2010. Uranium toxicity and speciation during chronic exposure to the tropical freshwater fish, Mogurnda mogurnda. Chemosphere 79:547554.
  • 33
    Trenfield MA, Ng JC, Noller BN, Markich SJ, van Dam RA. 2012. Dissolved organic carbon reduces uranium toxicity to the unicellular eukaryote Euglena gracilis. Ecotoxicology 21:10131023.
  • 34
    Semaan M. 2000. Population variability in the response of Moinodaphnia macleayi to uranium and cadmium. Internal Report 334. Department of Sustainability, Environment, Water, Population and Communities, Darwin, Australia.

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. SUPPLEMENTAL DATA
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
etc_1987_sm_SupplData.doc250KSupplementary Data

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.