Enhancing Cyclone Precipitation Nowcasting with Deep Learning: A Convolutional LSTM Approach Using INSAT 3D Data

Due to the intricate interplay between several atmospheric and oceanic elements, as well as the potential for abrupt changes in strength and trajectory, predicting the severity and intensity of precipitation during a cyclone may be difficult. Because of this unpredictability, it is extremely difficult to estimate cyclone consequences correctly and to prepare populations for possible calamities.There are several limits to traditional weather prediction techniques, including those related to accuracy, resolution, data quality, and processing capacity. These techniques include Numerical Weather Prediction (NWP), Weather Observations, Climate Models, Satellite Imagery, and Radar Technology. When it comes to using the Convolutional Long Short-Term Memory (ConvLSTM) model to forecast the intensity of cyclone precipitation, particularly during Cyclone Michaung in the Indian subcontinent, this research sets the standard. With a 4 -hour lead time, the model forecasts precipitation at 30 -minute intervals and initially achieves a high correlation value of 0.9514, which later drops to 0.8579 as the lead time increased.