Short-Term Weather Forecasting Using Deep Learning Techniques

Abstract

Weather nowcasting is the process of predicting the weather for a period of o to 6 hours.  Advanced deep learning models for weather nowcasting, emphasizes precise prediction of factors such as cumulative precipitation, humidity, wind direction, etc.Deep Learning models such as LSTMs and a CNN-LSTM Hybrid and stacked LSTM were applied to the AgriMet dataset. Stacked LSTM demonstrates notable performance with low Mean Squared Error (MSE) and Mean Absolute Error (MAE), indicating effective pattern capture. These results underscore the potential of deep learning for substantial improvement in short-term weather forecasting, providing pragmatic insights for decision-making in dynamic weather conditions.