River system thermal dynamics under Dual pressures of climate change and cascade reservoir operations

Understanding the cumulative impacts of climate change and cascade reservoir operations on river thermal regimes is essential for sustainable water and ecosystem management. We integrate multi-sensor Landsat observations (2000-2024) in Google Earth Engine, reconstructing daily river temperatures via sinusoidal dynamic regression. Three physically interpretable metrics-mean thermal offset (Δa), amplitude dampening (R), and phase lag (Δτ)-quantify reservoir-induced thermal alterations. A Bayesian-optimized CatBoost model with SHapley Additive exPlanations (SHAP) identifies shifts in climatic response patterns of river temperature under cascade regulation. Application to the lower Jinsha River shows a mean warming rate of 0.12 °C/year, peaking at 0.23 °C/year in summer. Reservoir operations cause persistent Δa increases, smoothed seasonal extremes, and progressive Δτ delays. SHAP analysis reveals declining air temperature influence (from 32.7 % to 24.6 %) and rising contributions from solar radiation (26.8 %) and wind (18.3 %), based on standardized Shapley Contribution Index (SCI) values. These results demonstrate that climate change and reservoir regulation jointly reshape river thermal regimes. The integrated approach-combining remote sensing, physically based metrics, and uncertainty-quantified machine learning-offers a transferable framework for fine-scale monitoring and adaptive scheduling in regulated river systems.