Comparison of different predictors of exposure for modeling impacts of metal mixtures on macroinvertebrates in stream microcosms

► Responses of macroinvertebrates to metals in microcosms were statistically modeled. ► The amount of metals binding to humic acid was the best predictor for the modeling. ► Our results support the use of the predictor for metal effects on macroinvertebrates. Knowledge about which predictors of metal exposure are best to model the impacts of metal mixtures on river macroinvertebrates remains uncertain. A new predictor based on the amount of metals binding to humic acid, which is assumed to be a proxy of non-specific biotic ligand sites, has been proposed. The amount can be calculated using Windermere Humic Aqueous Model (WHAM), which we will refer to as the WHAM-HA approach. Here, we tested the hypothesis that the predictor based on the WHAM-HA approach provided a better estimate of metal effects observed in microcosm experiments than three other measures: total metal concentrations, free metal ion concentrations, and the cumulative criterion unit (CCU) which is a measure of the ratios of measured metal concentrations relative to the U.S. Environmental Protection Agency hardness adjusted criterion values. For this evaluation, we used nine macroinvertebrate metrics of abundance and richness obtained from microcosm experiments conducted with metal mixtures (Zn alone, Zn+Cd, and Zn+Cd+Cu). For each of the four predictors, we performed multiple linear regression with variables corresponding to the three metal concentrations or CCU and selected the best model based on Akaike's information criterion corrected for small sample sizes. For all of the macroinvertebrate metrics affected by metals, the WHAM-HA approach was selected as the best among the four predictors, followed by the model with total metal concentration. In most of best models, Zn and Cu or Cu alone was responsible for reductions in invertebrate metrics, even though the highest concentrations of Cd exceeded 100 times the hardness-adjusted criterion value. Either of the models with free metal ion concentration and CCU was the third ranked model. Our results suggest that the estimated amount of metals binding to humic acid is a better predictor for the effects on macroinvertebrate richness and abundance observed in microcosm experiments than total or free ion concentrations of metals and CCU.