Rising temperatures and an increasing frequency of extreme events, such as heatwaves pose major threats to ecosystems, impacting ectothermic organisms including fish. Elevated water temperatures, coupled with reduced oxygen levels, intensify stress and mortality among fish fauna, necessitating urgent action to address the impacts of heatwaves on biodiversity. Here, we propose to combine sub-seasonal to seasonal (S2S) forecasting methodologies with ecological modeling integrated with physiological parameters. Specifically, we developed a sub-seasonal forecast model designed to assess fish stress during heatwaves in Swiss rivers. We first compiled from the literature physiological parameters related to fish thermal stress to construct a thermal stress index. The developed model integrates deep-learning forecasts of water temperature with fish physiological data and species distribution modelling to offer a process-driven prediction of heat-stress responses. To validate its efficacy, we retrospectively applied the model to the 2018 heatwave across 20 measurement stations on lowland Swiss rivers, comparing its results with assessments from experts and practitioners. We found that the model successfully forecast stress peaks 2–3 weeks in advance at two validated key sites, where high mortality was reported by practitioners. By classifying fish species based on their thermal sensitivity, we identified a high vulnerability of salmonids. Applications of our model intend to alleviate climate change impacts, resulting in targeted actions to reduce local fish mortality.