Warming driven intensification of Heat–Wet Alternations over South Indian Peninsula (1981–2024): Diagnosis and Machine Learning based Predictive Modeling

<title>Abstract</title> <p>Compound extremes events are emerging as a defining feature of the changing climate over sensitive tropical monsoon regions. Heat - Wet Alternations (HWA) defined as intense heat spells followed by heavy rainfall episodes are such consequential compound extremes whose evolution, temporal organization and drivers remain poorly understood over tropical regions like South Indian Peninsula (SIP). This study characterizes dry–wet spell alternation (DWA) and HWA occurrences over SIP during 1981–2024, using daily scale spell run decomposition, trend analysis and Empirical Orthogonal Function (EOF) supported by continuous wavelet transforms. Mann–Kendall trend across three successive ~ 40-year epochs (1901–2024), revealed a domain-wide significantly increasing trend of DWA in the 1981–2024 epoch. This epoch exhibited domain-wide increase of 8.84 (p &lt; 0.05) HWA events per year. The wavelet of the EOF1 of HWA demonstrated non-stationarity and statistically significant intensification of power in the 2–4 year band post 2010, coherent with the reorganized periodicities of climatic modes such as ENSO and IOD periodicities, directly attributing HWA intensification to planetary warming. Collectively, the findings establish that SIP is transitioning toward a more variable, extreme-prone regime for compound heat–wet extremes, with the intensification of variability itself emerging as a robust feature of this transition. Building on these diagnostics, a validated gradient-boosted (RUSBoost) classifier predicts HWA onset at daily, grid-scale resolution, achieving ROC-AUC of 0.911 and Brier Skill Score of 0.31 with feature attribution confirming antecedent local heat as the dominant HWA driver.</p>