Addressing Nonstationarity in Extreme Rainfall Patterns: A Case Study on Indian Cities

The evolving landscape of extreme rainfall patterns, triggered by climate change and global warming, introduces nonstationary behavior, challenging conventional hydrologic design assumptions rooted in stationarity. This study addresses this paradigm shift by modeling distribution parameters with covariates, utilizing a 70-year high-resolution IMD gridded dataset to extract and model extreme annual rainfall across diverse Indian cities. Drawing on previous research and goodness-of-fit tests that favor the Generalized Extreme Value (GEV) distribution for modeling extremes, the study incorporates various indices, including Nino3.4, dipole mode index, global and local temperature and time, to characterize nonstationarity in extreme annual rainfall, leveraging climate cycles and global warming trends. Performance assessment utilizes the Akaike information criterion and Likelihood ratio test, while quantile reliability is scrutinized through confidence intervals (CIs). The findings uncover widespread nonstationary trends in most grid points, resulting in broader CIs for estimated quantiles, return periods, and covariates in fitted models. Despite the broader confidence bands associated with nonstationary conditions, indicating higher uncertainty, the results affirm a nonstationary pattern in rainfall extremes. Consequently, the study underscores the imperative to develop nonstationary models that effectively capture these dynamic trends with reduced uncertainty.