Consistency of stream insect trait responses to instream pesticide exposure across five U.S. regions

Trait-environment relationships are a promising tool to understand how environmental factors shape freshwater communities, particularly on large scales. However, inconsistent responses of the community trait composition to environmental gradients have been reported for freshwater invertebrates. One reason for these discrepancies may be the predominant use of single traits without accounting for intercorrelations among traits. In this study, we evaluated whether grouping taxa into trait profile groups (TPGs), i.e., groups with similar trait combinations, yields more consistent relationships with estimated pesticide toxicity across large spatial scales compared to single traits. We analyzed invertebrate assemblages and environmental data from five U.S. regions compiled within the United States Geological Surveys (USGS) regional stream-quality assessments (RSQA). We used boosted regression trees (BRTs) to assess the ability of single traits and TPGs to predict estimated pesticide toxicity. TPGs responded slightly more consistently than single traits across regions, but effective predictive models could be established for only one region. Then, we identified the most responsive traits and TPGs with the feature importance of the BRTs and used them as ecological indicators in a comparison with the fraction of EPT taxa and the SPEARpesticides indicator. We compared indicators by evaluating, for each, the performance, consistency, and specificity of its relationship to pesticide toxicity. The marginal effects of estimated pesticide toxicity were slightly higher and more consistent across regions for TPGs than single traits. However, in terms of performance, consistency, and specificity the SPEARpesticides indicator surpassed both TPGs and single traits. Given the limitations of current trait data, expert-defined indicators may be more effective than a data-driven approach using TPGs for assessing the impacts of pesticides on stream insect communities. Nonetheless, future developments of the TPG approach—particularly through the integration of physiological and genomic traits—hold promise for improving its predictive power.