Data analysis choices influence the relative importance of multiple stressors on macroinvertebrates in agricultural streams

Hanh H. Nguyen , Verena C. Schreiner, Ralf B. Schäfer

Robustness of multiple stressor rankings is essential for credible ecotoxicological assessments and policy guidance, yet how data aggregation and modelling choices shape conclusions about stressor importance remains poorly resolved. We addressed this gap using a fully reproducible approach on a widely cited dataset of 101 small agricultural streams across Germany, where earlier data analyses have been subject to discussion, applying linear and penalized regression with bootstrap stability analysis. Pesticide toxicity remained consistently identified as a key stressor regardless of modelling approach. Summed toxic units combining event and grab sampling strengthened the pesticide-macroinvertebrate association compared to maximum toxic units or single sampling methods, suggesting that mixture-level exposure mechanisms and substance-specific toxicokinetics matter.  Beyond pesticides, data aggregation choices influenced the relative importance of multiple stressors, with agricultural land use, nutrients, and hydromorphological degradation showing stronger effects than previously reported and no single stressor dominating ecological responses. Different ecological metrics responded to distinct stressor sets, highlighting metric choice as a relevant consideration in interpreting multiple stressor effects. Our findings reveal that conclusions about dominant stressors can be sensitive to analytical decisions, calling for transparent, multi-metric, multi-model approaches to enable more defensible evaluations of chemical mixture and multiple stressor effects on freshwater biodiversity.