Coastal ecosystems are exposed to multiple anthropogenic stressors. Effective management actions would be better informed from generalized predictions of the individual, combined and interactive effects of multiple stressors; however, few generalities are shared across different meta-analyses. Using an experimental study, we present an approach for analysing regression-based designs with generalized additive models that allowed us to capture nonlinear effects of exposure duration and stressor intensity and access interactions among stressors. We tested the approach on a globally distributed marine diatom, using 72 h photosynthesis and growth assays to quantify the individual and combined effects of three common water quality stressors; photosystem II-inhibiting herbicide exposure, dissolved inorganic nitrogen (DIN) enrichment and reduced light (due to excess suspended sediment). Exposure to DIN and reduced light generally resulted in additivity, while exposure to diuron and reduced light resulted in additive, antagonistic or synergistic interactions, depending on the stressor intensity, exposure period and biological response. We thus find the context of experimental studies to be a primary driver of interactions. The experimental and modelling approaches used here bridge the gap between two-way designs and regression-based studies, which provides a way forward to identify generalities in multiple stressor interactions.

This raw dataset was collected through a series of laboratory exposure experiments - see Supplementary Material, section 'S1. Methods for laboratory-based experimental work'.

The data was processed using Microsoft Excel and R statistical software (versionn 4.1.0).

Scripts and data to perform the analyses can also be found at: https://github.com/oliviakpj/Changing_multistressor_interactions