Improving Precipitation Nowcasting via Multiphysical Parameter Fusion in Radar Echo Extrapolation

This repository provides the official implementation of PIEE, a multimodal deep learning framework for precipitation nowcasting based on polarimetric radar observations. It contains the complete training/evaluation pipeline (Experiment/Code), the core preprocessing script for the publicly available GridRad-Severe dataset (Original data processing code/gridrad.py; https://doi.org/10.5065/2B46-1A97), the pretrained model checkpoint (Experiment/saved_model), and the corresponding runtime environment specification (environment.yml), thereby supporting reproducibility without redistributing the raw data.