You are here:
USE OF MODELS TO SUPPORT WATER QUALTIY CRITERIA - A CASE STUDY
Citation:
Munns Jr., W R. AND A Kuhn. USE OF MODELS TO SUPPORT WATER QUALTIY CRITERIA - A CASE STUDY. Presented at European Science Foundation Workshop: Development of a Guideline to Select Applicable Mechanistics Models Aiming at Assessment of Impact of Pollutants in Food Chains, Prague, Czech Republic, April 15, 2004.
Description:
In the United States, current methods for deriving chemical criteria protective of aquatic life depend on acute and chronic toxicity test results involving several species. These results are analyzed statistically to identify chemical concentrations that protect the majority of aquatic species a majority of the time. An assumption underlying this approach is that criteria that minimize adverse affects (mortality, reproduction and individual growth) on individual organisms will also prevent effects on populations and communities. The veracity of this assumption has been evaluated using microcosm and field experiments, which have yielded somewhat mixed results.
Mechanistic population models provide a means to integrate individual-level effects as measured by toxicity tests into projections of population-level effects due to pollutant exposure, and therefore assess the degree of protection afford to aquatic populations by the U.S. criterion approach. In a series of investigations [1,2,3], we evaluated these relationships for an estuarine species of mysid shrimp, Americamysis bahia, which is used commonly in U.S. toxicity testing and criteria development. Data from lethal and sublethal toxicity tests using A. bahia (96-hr acute, 7-d rapid chronic, and 28-d chronic) were available to us for more than 20 chemicals for which these tests were used to develop protective criteria. We developed a deterministic, density-independent, age-classified matrix model based on the survivorship and fecundity data obtained daily in the 28-d tests, estimating population growth rate ( ) as the measure of population-level effect [1]. Concentration-based thresholds of population-level effect (where = 1.0), denoted C*, were compared to the acute LC50, and a statistic calculated during water quality criterion derivation, the criterion continuous concentration (CCC). C* was well correlated with the CCC and 28-day chronic statistic (reflecting the lower of the mortality or reproduction endpoint), and surprising, with the LC50. Weaker correlation was observed with the 7-d rapid chronic value. For 94% of the chemicals evaluated, the CCC was lower than C*, indicating the criterion to be reasonably protective of population-level effects in A. bahia.
But, was the model a reasonable predictor of actual population dynamics? To answer this, we compared projections of population effects based on toxicity test results with observations from populations maintained under continuous exposure conditions for three generations. Specifically, we parameterized the model with data from a 28-day test of nonylphenol toxicity, using nine exposure concentrations (including two control treatments) [2]. Treatment-wise projections were compared with the outcomes of a 55-d multigenerational experiment of A. bahia dynamics when exposed to similar concentrations in the laboratory. The model performed well both qualitatively and quantitatively. In a final analysis of different modeling approaches, data from the 28-d chronic were reevaluated using three additional formulations: a delay-differential equation model (DDE), an age-truncated DDE which assumed that individuals died after age 28, and a partial-differential equation model [3]. Comparisons of model predictions to the 55-d data set suggested model outcomes to be relatively robust to the details of model formulation, but that certain assumptions were important and the veracity of all models suffered by not accounting for density effects. These results indicate that, while certain aspects of model selection may be important in assessing pollutant impacts on populations, even simply-constructed models can provide information useful in pollutant criteria derivation and protection against population-level effects.