AI-Driven Hyper-Local Early Warning System for Severe Weather Nowcasting

Source summary

The proposed early-warning prototype uses a shared spatiotemporal multi-task model to estimate thunderstorm, cloudburst and flash-flood risks two to six hours ahead. It combines IMDAA baselines, INSAT-3D/3DR moisture and cloud-top observations and terrain data, preserving moisture, instability and lift signals. A GIS dashboard should show distinct risk maps and explain meteorological triggers, with categorized API alerts. Forecast accuracy and useful warning lead times require validation; they are not established outcomes.