Variation in chronic nickel toxicity to Daphnia magna among Japanese river waters and performance evaluation of bioavailability models in predicting the toxicity.

Abstract

Nickel (Ni) ecotoxicity is dictated by water chemistry characteristics such as pH, water hardness, and amount of dissolved organic carbon. Bioavailability models have been developed to predict Ni toxicity and validated for European, Australian, and US natural waters. In this study, chronic toxicity tests in Ni-spiked Japanese river waters were conducted on a strain of Daphnia magna to test whether the chronic toxicity differs among Japanese natural waters with different water chemistries. Based on the results of chronic Ni toxicity tests, we assessed the performance of existing D. magna bioavailability models, which were developed in artificial waters (Model 1) and calibrated in European natural waters (Model 2), in terms of the accuracy and the bias of model predictions. Furthermore, we also calibrated the two models by using toxicity test results to develop a bioavailability model for Ni chronic toxicity to the strain of D. magna in Japanese river waters. The 10%, 20%, and 50% effect concentrations (EC10, EC20, and EC50) of dissolved Ni on reproduction of the D. magna strain were within ranges from 8.1 to 44.9μg/L, 9.0 to 57.1μg/L, and 10.9 to 86.1μg/L, respectively. Results indicate that differences in water chemistry among Japanese river waters influenced chronic Ni toxicity to the model organism. Model 1predicted 43% of the observed EC10, EC20, and EC50 values within a factor of 2 and 100%, 100%, and 43% within a factor of 3, respectively. Model 2 predicted 14%, 14%, and 29% of the observed EC10, EC20, and EC50 values within a factor of 2 and 43% within a factor of 3. The values of model bias based on the geometric mean of ratios of EC10, EC20 and EC50 values predicted by each of the two models and observed EC10, EC20, and EC50 values were 0.71, 0.65, and 0.62 for Model 1 and 0.27, 0.26, and 0.29 for Model 2, respectively. After calibrating two models using the results of toxicity tests, refined Model 1 predicted 71%, 57%, and 57% of observed EC10, EC20, and EC50 values within a factor of 2 and 100%, 86%, and 100% within a factor of 3; refined Model 2 predicted 71% of observed EC10, EC20, and EC50 values within a factor 2 and 100%, 86%, and 86% within a factor of 3, respectively. Our results indicate that calibrating the Ni bioavailability models in Japanese natural waters increased their predictive capacity by a factor of up to approximately five.