Non-stationary modeling of extreme precipitation using extreme value theory and meteorological data

Extreme precipitation estimation commonly assumes stationarity, although climate variability can modify its temporal distribution. This study assessed nonstationarity in annual maxima of daily precipitation across three climatic regions of Ecuador using extreme-value models. Hourly ERA5-Land series were processed, the Relative Oceanic Niño Index was incorporated as a covariate, and stationary and nonstationary formulations were compared through maximum likelihood, information criteria, likelihood-ratio tests, and parametric bootstrap. Guayaquil showed joint dependence on time and El Niño-Southern Oscillation variability; Puyo exhibited a temporal signal sensitive to the precipitation product; Quito remained compatible with stationarity. A CHIRPS robustness analysis showed that evidence of change is not spatially uniform and should be interpreted together with uncertainty associated with gridded precipitation products.