CYCLONE INTENSITY ESTIMATION USING INSAT-3D IR IMAGERY AND DEEP LEARNING

Abstract

The objective is to pioneer an innovative Deep Learning Convolutional Neural Network (CNN) model, acknowledging the inherent intricacies of cyclone intensity estimation and emphasizing the imperative for exacting assessments. This model integrates imagery visualization and prediction of Tropical Cyclone intensity. The comprehensive approach utilizes half-hourly INSAT-3D IR Images to improve the accuracy and reliability of cyclone intensity predictions. In pursuit of this objective, the proposed novel approach leverages Convolutional Neural Networks (CNNs) for estimating cyclone intensity. CNNs excel in extracting intricate patterns from imagery data, making them ideal for analyzing satellite images of cyclones. The model utilizes established metrics and observational data to train the CNN, ensuring robust performance in cyclone intensity estimation. The experimental results demonstrate the efficacy of the CNN-based approach, achieving a significant accuracy of 93.45% & 92.7% in predicting cyclone intensity values. This capability extends to tracking the evolution, intensification, and landfall of cyclones. Surpassing traditional methods, this highlights its potential to enhance meteorological forecasting and disaster management efforts. Furthermore, the model provides both original and predicted intensity values ranging from Cyclonic Storm (34-47 KT) to Severe Cyclonic Storm (48-63 KT), accompanied by graphical representations. This comprehensive approach offers insights into cyclone dynamics, facilitating informed decision-making and preparedness strategies. Novelty: This research pioneers the application of CNNs for cyclone intensity estimation, marking a significant contribution to the field. Leveraging deep learning capabilities, it introduces a paradigm shift in how cyclone intensity is assessed, promising enhanced precision and reliability in meteorological predictions and disaster preparedness strategies. This innovative approach signifies a transformative advancement in cyclone intensity estimation methodologies, offering new insights and avenues for improving forecasting accuracy and mitigating the impacts of cyclonic events.