A Dynamical Framework to Understand and Predict the Indian Summer Monsoon Low Pressure Systems

Abstract

About 60% of the rainfall during the Indian Summer Monsoon (ISM) is manifested by the synoptic-scale storms form over North Bay of Bengal (BoB) and the adjacent land area known as “Low-Pressure Systems” (LPS). Unlike tropical cyclones, the storms during this season (LPSs) are embedded in the background monsoon flow, which makes them difficult to predict, considering the chaotic nature of the monsoon. Nearly one-third of these synoptic-scale storms are formed due to the amplification of disturbance which is propagating from the Western North Pacific (WNP) (categorized as “downstream LPS”). We observed an association of tropical cyclones (TCs) originating over WNP with the genesis mechanisms of downstream LPS over the BoB. The TCs over the WNP are classified into different clusters based on different features like length, genesis location, landfall, etc., using the gaussian mixture models. We found that four major clusters of WNP TCs are responsible for triggering 83% of the downstream LPS genesis. We established a causality using the transfer entropy analysis between the fluctuations in mean sea-level pressure over BoB and the Rossby wave activity over the WNP prior to the initiation of an LPS.Our results suggest a plausible prediction of downstream LPS at least a week ahead. The current generation of climate models has low skill in simulating the LPS; understanding the dynamics behind the genesis of LPS is the way to improve the LPS-related precipitation in climate models. The recent advancement in using AI/ML in predicting various weather and climate phenomena, including our recent study in predicting the synoptic-scale sea-level pressure using the ConvLSTM model explains the importance of dynamics-based data-driven ML models to predict complex weather patterns. Understanding the dynamics of such physical phenomena will help in identifying the appropriate predictors for the data-driven ML models.