Vutukuru Sathya Sai Harshit

Accurate short-term atmospheric forecasting is essential for climate monitoring, disaster management, and environmental planning. Traditional numerical weather prediction models require significant computational resources and may struggle to capture complex nonlinear atmospheric patterns. This study presents ClimaVision, a deep learning-based framework for short-term atmospheric prediction using ERA5 reanalysis data. The proposed system leverages historical atmospheric variables including temperature, sea level pressure, wind components, and cloud cover to forecast future atmospheric conditions. Multiple deep learning models, including Long Short-Term Memory (LSTM), Graph Neural Networks (GNN), Transformer, and NBEATS, are implemented to capture spatiotemporal dependencies in atmospheric data. The models are trained using historical sequences of atmospheric observations and evaluated using standard regression metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and R² score. Experimental results demonstrate that deep learning models can effectively learn atmospheric patterns and provide reliable short-term forecasts. The proposed ClimaVision framework highlights the potential of AI-driven approaches for improving atmospheric prediction systems.