A Hybrid CNN-LSTM Approach for Precipitation Forecasting under Climate Change Scenarios

Accurate precipitation forecasting is vital for water resource management and climate change mitigation. This study proposes a hybrid approach combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to improve precipitation predictions using CMIP6 projections. The model integrates spatial features extracted by CNN from multi-modal data sources, while LSTM captures the temporal dependencies of precipitation and climate variables. By leveraging CMIP6's high-resolution outputs and combining them with real-time observational data, the model learns complex spatial-temporal patterns to enhance forecast accuracy. Performance is evaluated using metrics like Root Mean Square Error (RMSE) and correlation coefficients, showing substantial improvements over traditional methods. This approach provides an effective framework for multi-source data integration, offering strong potential for climate adaptation and water resource management.