Abstract :

Weather forecasting using satellite image analysis and deep learning is transforming meteorology by enabling more accurate and timely predictions. This work integrates deep learning techniques with real time satellite imagery to extract spatiotemporal atmospheric patterns for short and medium term forecasting. Satellite sensors such as geostationary imagers provide continuous global data streams, which our deep neural network models process to forecast weather variables like precipitation and cloud movement. The system includes a full stack web application that visualizes forecast results and allows interactive exploration of weather patterns. Our evaluation demonstrates substantial improvements in prediction accuracy over traditional numerical methods for short term nowcasting. The project highlights the potential of combining remote sensing, machine learning, and web technologies for robust weather forecasting solutions.