Climate Change Induced Extreme Rainfall and Its Impacts on Large Reservoir Systems: A Non-Stationarity Perspective

Design Flood (DF) inputs for large reservoir systems, such as Intensity-Duration-Frequency (IDF) curves and Probable Maximum Precipitation (PMP), are traditionally derived under stationarity assumptions, which are increasingly challenged under a changing climate. The current study examines changes in extreme rainfall characteristics across ten important dams in the Godavari River Basin (GRB), India, under three climate scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5). Daily rainfall projections from the NEX-GDDP-CMIP6 dataset are evaluated against gridded observations of the India Meteorological Department (IMD) for the historical period (1951-2014). Nine statistical performance metrics, combined with five Multi-Criteria Decision-Making (MCDM) methods and a Group Decision-Making (GDM) framework, are used to identify the best-performing (top-five) Global Climate Models (GCMs). Based on this evaluation, five GCMs – BCC-CSM2-MR, CMCC-ESM2, MPI-ESM1-2-HR, MPI-ESM1-2-LR, and NESM3 are selected for GRB. Next, non-stationarity in extreme rainfall is assessed using epoch-wise analysis, trend detection methods (Mann-Kendall test and Sen’s slope estimator), and a change-point detection technique (Pettitt’s Test). The results of statistical analyses show significant increases in short-duration rainfall extremes in recent decades. Subsequently, IDF curves are developed for multiple return periods (100-, 200-, 500, and 1000-year) using the Gumbel distribution (GEV-1). The results revealed a robust intensification of short-duration rainfall extremes under future climate scenarios, with SSP5-8.5 exhibiting the largest increases, implying that stationary design assumptions may underestimate future dam safety risks. Furthermore, PMP is estimated using the Hershfield method, and the results indicated increases ranging from 8.55% to 44.11% across the selected dam locations. Overall, the study underscores the necessity of revisiting stationary design assumptions and offers a scalable framework for climate-resilient design storm estimation for large reservoir systems. While increases in PMP are evident, their direct application without field-level validation may lead to over- or under-conservative design decisions. Hence, future work should focus on reconciling model-based PMP estimates with observed extreme events, local meteorological records, and dam-specific field conditions, alongside hydrological and reservoir routing analyses, to support robust and reliable dam safety assessments.

Keywords: Climate Change, Non-stationarity, Intensity-Duration-Frequency (IDF), Probable Maximum Precipitation (PMP), NEX-GDDP-CMIP6 models