The tropical cyclone energy prediction of the North Indian Ocean in monsoon using artificial neural networks

Abstract

The North Indian Ocean (NIO), which includes the Bay of Bengal (BOB) and the Arabian Sea (AS), is highly vulnerable to tropical cyclones, emphasizing the critical importance of accurate energy predictions for effective disaster preparedness. This study focuses on predicting the Accumulated Cyclone Energy (ACE) in the NIO during monsoon using an optimized Artificial Neural Network (ANN) model. Initially, an ANN was trained with six cyclone metrics, including Velocity Flux (VF) and Power Dissipation Index (PDI), showing moderate predictive accuracy with relatively high error metrics like Mean Squared Error (MSE). The permutation feature is essential in finding the most influential features to improve model performance. This analysis identified NIO_VF, AS_PDI, and NIO_PDI as key predictors, while metrics such as BOB_PDI, BOB_VF, and AS_VF had minimal impact on NIO_ACE prediction. To improve the model's predictive accuracy while reducing complexity, the ANN model was again retrained with only the most significant features, resulting in a considerable reduction in loss (MSE). The unit of ACE is 104 knots².  In this approach, key cyclone metrics are prioritized and computationally optimized, highlighting the critical role of feature engineering in improving meteorological machine learning models. A support system based on this methodology could significantly benefit early-warning systems and strategic planning efforts in cyclone-prone regions, ultimately making these regions more resilient to extreme weather events. By analyzing the socio-economic and environmental impacts of cyclones, this study addresses 7 Sustainable Development Goals (SDGs), including SDG 13 (Climate Action), SDG 11 (Sustainable Cities), and SDG 1 (No Poverty).