Estimating the spatial and model uncertainties in yielding extreme rainfall return levels across India
Year: 2025
Type: article
Abstract: Frequency and magnitude of extreme rainfall are greatly influenced by anthropogenic climate change and natural climate variability. Modeling these events is crucial for assessing risks and vulnerability of ecosystems in a changing climate. Former studies modify the traditional General Extreme Value (GEV) distributions that associate distribution parameters with large-scale oceanic and atmospheric circulations. However, the spatial and model uncertainties in modeling extremes have yet to be addressed. To address this, current study develops 64 Non-stationary GEV (NSGEV) distributions to model the Annual Maximum Rainfall (AMR) across India. NSGEV distributions consider a pool of feasible functional relationships between distribution parameters and climate drivers. Preliminary analysis suggests that although AMR across India exhibits substantial non-stationarity, a statistically significant linear and non-linear trend is rare. We find that hybrid NSGEVs, which consider multiple climate drivers and time as covariates to the location parameter, best explain the AMR variance. Moreover, the study finds that the frequency of rare rainfall events occurring once every 71 years is reduced to less than ten years in many parts of the country. Uncertainty analysis confirms that modeling the extreme events yields higher uncertainty over northeastern, central and coastal parts of the country; however, in a few zones, the spatial variance in rainfall return levels remains constant as the return periods increase, indicating that the model structure is inconsequential for low-frequency high-magnitude rainfall events.
Language: English
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: gold
Estimated APC paid: $2,600
OpenAlex ID: W4410083363