Identifying robust precursor regions of driving parameters affecting the Accumulated Cyclone Energy in the North Indian Ocean using the Causal Effect Network

Abstract

There are significant variations in tropical cyclone (TC) activity every year and related casualties. The mitigation of human and property losses brought on by TCs depends heavily on the realistic forecasting of TC intensities at longer lead times. In this regard, Accumulated Cyclone Energy (ACE) is the most appropriate measure to access cyclone energy. The study examines the association between ACE and other driving parameters influencing its variability and identifies the robust precursor regions using modern-day machine learning techniques. Initially, the correlation between ACE and driving variables (vertical wind shear, sea surface temperature, specific humidity, tropical cyclone heat potential, relative vorticity, and total column water vapor availability) is assessed. The analysis demonstrates a significant robust association between ACE and associated driving factors over the main cyclogenesis regions. Subsequently, an innovative K-means clustering process and the Causal Effect Network (CEN) based technique employing the PCMCI algorithm are utilized to investigate the relationship among the clustered regions of variables and ACE. Finally, the main contributing regions (MCR) accountable for affecting the changes in the ACE values in the basin are identified. The causal and cluster graphs illustrate the MCRs that exert direct or indirect substantial influence on ACE value. These fluctuations results are caused by the variations of atmospheric and oceanic parameters. These methods may be useful for policymakers and researchers in precise prediction, tracking cyclones, and identifying areas that affect future cyclogenesis.