ALBERT

Description: The behavior of extreme rainfall has become non-stationary due to climate change and global warming, leading to misleading results and putting existing structures at risk. One solution to this problem is to model distribution parameters with covariates. This study utilized a high-resolution IMD gridded dataset spanning 70 years to model extreme annual rainfall in the Indian region. The study found that NINO 3.4, dipole mode index, global and local temperature anomalies, and CO〈sub〉2〈/sub〉 are good covariates for further analysis, and including these covariates enhanced the performance of the model. The goodness test and previous studies suggested that the GEV distribution is a suitable model for such extremes. The estimated quantile uncertainties were assessed with confidence intervals (C.I.). The analysis showed that most grid points follow non-stationary patterns, with intensification in extremes. However, uncertainties in return levels due to parameter estimation and covariate selection were higher in non-stationary conditions, indicating the weaknesses of non-stationary models compared to stationary models. Nevertheless, the study revealed that the rainfall extremes follow a non-stationary pattern, highlighting the need to fit a non-stationary model with low uncertainties to provide realistic predictions. The results also showed that C.I. widened for shorter datasets for selected return periods and covariates. Overall, this study provides insights into the non-stationary dynamics of extreme rainfall and underscores the importance of using appropriate models to develop reliable predictions.