Faced with intensifying heatwaves in the Béni Mellal-Khénifra region, local authorities lacked the tools to anticipate climate risks or organize rapid logistical responses for the most vulnerable populations. This study presents BeniMeteo, a comprehensive decision-support architecture that transforms weather forecasts into operational response plans. The approach rests on four pillars: forecasting future temperatures using a hybrid SARIMA-XGBoost model trained on twelve years of historical data; calculating a territorial risk score (Ri) that integrates heat hazard, demographic exposure, social vulnerability, and medical isolation for each *douar* (rural settlement); automatically generating four-level alerts based on hybrid seasonal thresholds calibrated against official climate norms; and optimizing logistics via the HRW-LRP model, which plans the location of response centers and mobile team routes by prioritizing critical areas. Results show forecast accuracy reaching an R² of 0.98 for short-term horizons, a 100% validation rate for alert thresholds in test cases, and the identification of 78 *douars* at a critical risk level during a heatwave simulation. BeniMeteo demonstrates that it is possible—using open data and free tools—to build a rigorous, geographically targeted, and immediately operational decision-support system to protect vulnerable populations from climate risks. 

Faced with intensifying heatwaves in the Béni Mellal-Khénifra region, local authorities lacked the tools to anticipate climate risks or organize rapid logistical responses for the most vulnerable populations. This study presents BeniMeteo, a comprehensive decision-support architecture that transforms weather forecasts into operational response plans. The approach rests on four pillars: forecasting future temperatures using a hybrid SARIMA-XGBoost model trained on twelve years of historical data; calculating a territorial risk score (Ri) that integrates heat hazard, demographic exposure, social vulnerability, and medical isolation for each *douar* (rural settlement); automatically generating four-level alerts based on hybrid seasonal thresholds calibrated against official climate norms; and optimizing logistics via the HRW-LRP model, which plans the location of response centers and mobile team routes by prioritizing critical areas. Results show forecast accuracy reaching an R² of 0.98 for short-term horizons, a 100% validation rate for alert thresholds in test cases, and the identification of 78 *douars* at a critical risk level during a heatwave simulation. BeniMeteo demonstrates that it is possible—using open data and free tools—to build a rigorous, geographically targeted, and immediately operational decision-support system to protect vulnerable populations from climate risks.