Accurate weather forecasting is vital for safety but faces challenges with calculation errors. We developed a new predictive model that enhances accuracy by reducing data redundancy and optimizing information flow. Experiments with real radar data show our approach significantly outperforms existing methods, particularly for heavy rainfall. This model maintains clear details over time, offering a robust tool for timely severe weather warnings and effective disaster prevention.

This work is the first study to simultaneously analyze the number concentrations and species of insoluble particles in hailstones. The size distribution of insoluble particles for each species vary greatly in different hailstorms but little in shells. Two classic size distribution modes of organics and dust were fitted for the description of insoluble particles in deep convection. Combining this study with future experiments will lead to refinement of weather and climate models.