Abstract
<title>Abstract</title> <p>Climate change is reshaping extreme-rainfall behaviour, making station-scale risk assessment increasingly dependent on how climatic information is incorporated. Yet nonstationary extreme-value studies often treat rainfall extremes as a pooled process, using similar covariates to explain both how often extremes occur and how large they become. In this study, station-level extreme rainfall across diverse hydroclimatic regimes in India is analyzed within a peaks-over-threshold framework to identify climatic and atmospheric covariates separately associated with event frequency and exceedance severity, providing a basis for representing these risk-relevant pathways independently. The results reveal a clear component-specific structure. Occurrence is most consistently associated with atmospheric moisture transport, with Integrated Vapor Transport emerging as the dominant occurrence covariate across most stations, while Total Column Water Vapor is dominant in the moisture-limited arid northwest. Unlike occurrence, severity does not follow a uniform national-scale covariate structure. Instead, the covariates associated with exceedance magnitude are highly localized and regime dependent, involving different combinations of moisture transport, oceanic variability, convective instability, local thermodynamic anomalies, and vertical uplift. These contrasts vary across coastal and peninsular monsoon environments, inland plains, arid zones, and mountainous regions, indicating that rainfall extremes cannot be adequately represented by a single covariate-response structure. The proposed diagnostic framework captures this nuance by screening covariates separately for frequency and severity, offering a practical prior-screening step for selecting component-appropriate covariates in nonstationary rainfall-risk models and improving the physical interpretability of station-scale risk assessment across diverse water-resources settings.</p>