Abstract: Severe convective weather is a critical meteorological factor that affects the safety of space launches, particularly in tropical coastal regions characterized by frequent and highly variable atmospheric conditions. To enhance the meteorological risk prevention and control capabilities of launch sites, a convective weather early warning method based on the AeolusStorm model was proposed. The model integrates multi-source meteorological data, including wind field data from the ESA Aeolus satellite, radiosonde observations, and ground-based radar measurements. A random forest regression algorithm was employed to establish a dual risk assessment framework consisting of a risk index (RiskIndex), which quantifies the current atmospheric instability, and a future index (FutureIndex), which predicts forthcoming meteorological trends. Validation using multi-source observations from the Wenchang Space Launch Site (2020−2023) showed that the AeolusStorm model outperformed traditional statistical methods in nonlinear fitting and temperature prediction. The model effectively identified severe convective weather under complex tropical coastal conditions and provided quantitative decision support for launch safety.