Draft Ambient Water Quality Guidelines for Sulfate In British Columbia. Water Protection & Sustainability Branch October 13 th , 2011. Overview. Review process Background concentrations WQGs vs. WQOs Data requirements Process for Sulfate guideline Maximum likelihood estimates
Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.
Draft Ambient Water Quality
Guidelines for Sulfate
In British Columbia
Water Protection & Sustainability Branch
October 13th, 2011
Table 1. Summary of ambient dissolved sulfate concentrations in BC freshwaters.Sulfate background concentrations in BC
Refinement of BC-approved & working WQGs
Developed to protect the most sensitive water use, at a specific location, taking local circumstances into account
Input to permits, licenses, orders and regulations
To report to the public on water quality
To determine if remedial action is necessary
To promote water stewardship
Uncertainty factors (typically between 2-10)
Method to fit curves to data (e.g. dose-response of sulfate vs. mortality)
Quantal responses (mortality) - use Probit model
Continuous responses (e.g. growth) - Isotonic regression (ICPIN), or 3-p log-logistic
After fitting model, assess goodness-of-fit (residual plots, etc)
Use fitted curve to estimate LCxxvalues. Caution advised if extrapolating to very low effect LCxx endpoints, e.g. LC01 or LC001 values.
MLE extracts all information from data but must choose appropriate models
New chronic guideline 65 mg/L - based on 28-day LC10 of rainbow trout of 127 (47-342) mg/L with minimum uncertainty factor of 2.
Increased maximum guideline from 100 to 250 mg/L – based on LC50 data (C. dubia, D. Magna, Hyalella) with minimum uncertainty factor of 2.
Water hardness may decrease the toxicity of sulfate for some endpoints and species; however, no consistent relationship was found.
Site-specific water quality objectives using site water would be more appropriate for determining if the ion composition decreases, increases or has no-effect on sulfate toxicity to organisms in a particular water body.
The development of site-specific water quality objectives to take local conditions into consideration is done with consultation with Ministry of Environment staff.
1) “Early lifestage tests using trout
Test validity: We would like to discuss control performance of the
early life stage rainbow trout tests. We do not believe that the
control in the soft water test passed a reasonable test performance
criterion for this type of test. Furthermore, we are concerned that
the poor survival in the soft water, and the high degree of
variability in this test is indicative of a stressed population of
test organisms, and that these tests should be rejected because of QA/QC concerns.”
The acceptable cumulative control mortality cannot be > 35% (65% survival or better is OK.)
For the R. trout eyed egg test there was 27% cumulative mortality or 73% survival which passes the validity criteria.
Environment Canada 1998. Biological Test Method: Toxicity Tests Using Early Life Stages of Salmonid Fish (Rainbow Trout), EPS 1/RM/28 second edition.
1) “Early lifestage tests using trout
Statistical analysis: We would like to discuss the statistical power associated with the early life stage tests using rainbow trout. Specifically, we do not believe that the test had sufficient power to detect a 10% deviation from the control with a reasonable degree of confidence, and that the LC10 value reported does not make sense in the context of the dataset.”
The sulfate levels vary considerably from control levels, e,g. up to around 2000. Over the entire range of measured sulphate level, the probit model (allowing for overdispersion) had a statistically significant slope (p=.0016 for hardness 50; p=.0252 for hardness 100; p=.0247 for hardness 200) so an effect of sulfate level over the range of doses was detected.
Once model is fit using ENTIRE dataset, you can extrapolate back to LC10 values despite there being large variability in raw data.
3) “Tests using the freshwater mussel
Statistical analysis: We would like to discuss the effect levels reported for the soft water test using the freshwater mussel. Specifically, we do not believe that the LC10 reported in the Draft document is supported by the data because the test does not have sufficient power to detect that level of effect with a reasonable degree of confidence, and the statistics used did not account for background mortality in the control from that test.
The LC10 value reported does not make sense in the context of the dataset, and we believe that the LOEC, MATC, LC25 or LC50 might be a more appropriate value.”
4) Tests using the Pacific Tree frog
Statistical analysis: We would like to discuss the effect levels reported for moderate hardness test (80 mg/L) using the Pacific Tree frog. Specifically, the analyses used did not account for the background mortality in the control from that test and the LC10 value reported does not make sense in the context of the dataset. We believe that the point estimates reported by Elphick et al. (2011) for this test were calculated appropriately for this test.
Only 2 hardness levels used. At hardness 15, the control sulfate is given as "1" - was this real or merely a placeholder for data entry?
No observed mortality observed (0 out of 15 on test). At hardness 80, the control sulfate level is 93 and 2/15 mortalities observed.
We fit a model where no threshold effect was observed at either hardness level, and the 2/15 mortalities at sulfate level 93 is not unreasonable with the fitted dose response curve. A model with a control threshold was not fit because of the sparseness of the data.
If you look at the raw mortality numbers, the effect of sulfate at hardness 80 appears to be "worse" than at hardness 15 as the total mortalities tend to be higher in general at comparable sulfate levels. This might be the result of a threshold taking effect at the higher hardness levels, but is difficult to discern because the control doses are too different between the 2 studies.
This is also the species where the CETIS printouts use a control threshold in 1 hardness level and not the other hardness level.
Need to consider all hardness levels simultaneously. For example, is it biologically reasonable to have no natural response at hardness 15 and a natural response at hardness 80?
5) “Tests using Hyalellaazteca
Statistical analysis: We would like to discuss the effect levels reported for growth of Hyalellaazteca. We do not believe that these tests are sufficiently robust to calculate an IC10 with a reasonable degree of confidence. It should be noted that the data are contradictory to information from other tests performed by Nautilus, which indicate a lower sensitivity to sulphate in higher hardness using both growth and survival endpoints.”
In these studies the observed response "increased" from baseline and then decreased. We tried a variety of models, but the most suitable was separate Isotonic Regression model for each hardness level which gave estimates of 1326 at hardness 50, 645 at hardness 100; and 333 at hardness 200.
CETIS fit a 3P log-gompertz (IC10=683) at hardness 200; a ICPIN=Isotonic Regression (IC10=638) at hardness 100; a ICPIN=Isotonic Regression (IC10=1321) at hardness 50.
Our results are identical to CETIS except for the very hard water, but all our estimates have very wide confidence limits which are reported in the guideline.
We didn't have access to this other dataset, but it could be integrated into the analysis if available.
6) “Tests using fathead minnows (Nautilus data)
Statistical analysis: We would like to discuss the recalculated LC10 value for 80 mg/L water hardness (LC10 of 426 mg/L sulphate); this value makes no sense in the context of the dataset, in which there is no deviation from the control response at up to 1300 mg/L sulphate.”
Raw data has control dose of 37 for sulfate at hardness 40 with 1/30 mortality; control dose of 74 for sulphate at hardness 80 with 3/30 mortality; control dose of 130 for sulfate at hardness 160 with 1/30 mortality and control dose of 300 for sulphate at hardness 320 with 0/30 mortality and EC10 1555;
Two top models are the separate probit model (model weight 0.53) and the monotonic effect of hardness model (model weight of 0.47). Neither has a natural response.
Estimated LC10 were 301 for hardness 40; 426 for hardness 80; 1074 for hardness 160; 2318 for hardness 320.
CETIS EC10 were
2450 for hardness 320 (with a 0% threshold);
3231 for hardness 160 (with a 10% threshold)
1555 for hardness 80 (with a 10% threshold)
558 for hardness 40 (with a 3% threshold);
The key differences was the use of the threshold by CETIS and no threshold by us. The CETIS thresholds don't vary in a consistent fashion, i.e. threshold goes from 3% to 10% and down to 0%. Is this a sensible thing for thresholds? Given the non-zero sulphate levels at "control" doses the observed mortality is consistent with an effect of sulphate rather than a threshold.
7) “Selection of point estimates - General
We would like to discuss problems with calculation and use of tenth percentile effect levels from tests that allow 10 or 20 percent effect in the control as acceptable, and in which the minimum significant difference that is statistically detectable is typically in the range of 20 to 30%. In such tests, variability precludes calculation of 10th percentile estimates with a reasonable degree of confidence.”
8) “We would also like to discuss inconsistencies between MoE guidance on deriving water quality
guidelines, which suggest that "the lowest observed-effect concentration (LOEC) or EC(low-effect generally thought to be EC15-20) from a reliable chronic exposure study, preferably on sensitive native BC species, are selected.".
9) “Selection of statistical methods - General
We would like to discuss the use of linear interpolation for determination of point estimates from data sets with continuous data, since this approach is considered less appropriate than non-linear regression by Environment Canada".
10) “Role of water hardness in modifying acute toxicity: We would like to discuss the role that water hardness plays in altering the acute toxicity of sulphate. Toxicity test results from acute toxicity tests are consistent with the conclusion that water hardness reduces toxicity of sulphate (with the sole exception being Chironomids, which are insensitive to sulphate), and we would like to discuss why water hardness was not incorporated into the maximum guideline for sulphate".
BC Water Quality Guidelines intended for generic provincial application
Protect most sensitive species and life stage during indefinite exposure
We did not find a consistent relationship with water hardness and sulfate toxicity
11) “Role of water hardness in modifying chronic toxicity: We would like to discuss why it is necessary that decreasing toxicity with increasing water hardness is observed with all test species. As long as sensitive test organisms are protected across the full range of hardnesses, it should not matter is some less sensitive species do not show less sensitivity to sulphate at higher hardness.".
12) “Role of total dissolved solids in toxicity to Ceriodaphnia: We would like to discuss adverse effects observed with Ceriodaphnia in 320 mg/L water hardness, and the relevance of this datapoint to setting water quality guidelines for sulphate".