Non-stationary frequency analysis of extreme rainfall events on the east coast of KwaZulu-Natal, South Africa

This study examines changes in the frequency and magnitude of extreme rainfall events in South Africa's KwaZulu-Natal region. Traditionally hydrological design assumed a stationary climate, but recent extreme rainfall events have prompted investigation into the presence of change in observed data series and its potential drivers. Analysing rainfall data from 39 stations, the study finds weak evidence of increasing annual maximum daily rainfalls over time, with about 40% of sites showing positive trends, though only one is significant. Nonstationary extreme value models incorporating climate drivers as covariates (Southern Oscillation Index, Dipole Mode Index, CO2 and global mean temperature) alongside time are explored, revealing CO2 as a significant influencer. However, stationary models outperform nonstationary models at 56% and 36% of stations, based on Akaike information criterion and Bayesian information criterion measures, respectively.