Evolution and Prediction of Rainstorm Disasters Based on ARIMA-BP Neural Network

As global warming intensifies, extreme weather events like heavy rainfall become more frequent and severe, posing significant challenges to society. Accurate precipitation prediction is crucial for mitigating disaster risks. This study combines the ARIMA(1,1) and BP neural network models to improve forecasting accuracy for predicting precipitation from 2026 to 2035. The ARIMA model captures linear trends, while the BP network addresses nonlinear patterns. The hybrid model is validated using k-fold cross-validation and sensitivity analysis. Results show that accuracy varies under different climate scenarios, with increased uncertainty in high-emission cases.