High-resolution contemporary river temperature atlas for Central Europe

River water temperature governs aquatic metabolism, oxygen availability, and habitat suitability, yet monitoring remains sparse and uneven especially for small tributaries leaving much of the river network unobserved. Across Central Europe, long-term in situ observations are spatially heterogeneous, constraining assessments for many ecologically important rivers and tributaries that are only a few tens of meters wide. We address these limitations by developing a machine-learning framework that combines satellite thermal infrared land surface temperature (LST) with hydroclimatic, topographic, and land-cover predictors to generate high-resolution river water temperature (Tw) estimates across Central Europe, with a focus on Germany. We collocate quality-controlled in situ water temperature records with ECOSTRESS overpasses using a river-corridor sampling strategy that minimizes mixed-pixel contamination and propagates cloud and retrieval uncertainty. To ensure geographic transferability, we apply geographically structured training validation that limits spatial leakage. We then fit and compare machine-learning models to predict Tw from daily to monthly timescales. The resulting product provides spatially explicit Tw fields at ECOSTRESS native resolution (order 70 m) along the river network, resolving fine-scale thermal gradients and identifying localized hotspots associated with urbanization, flow regulation, and riparian alteration. By quantifying the combined influence of climate variability and anthropogenic modification on river heating and cooling capacity, this work supports ecological risk assessment and climate-adaptation planning, and offers a transferable template for other data-limited river systems.