The non-stationary additive regional frequency analysis method

Abstract The Regional Frequency Analysis (RFA) method is widely used to improve probabilistic assessments of extreme hydrometeorological events. However, its classical formulation assumes stationarity - an assumption often violated under climate change. To overcome this limitation, we propose the Non-Stationary Additive Regional Frequency Analysis (NS-Add-RFA), an extension of the additive RFA that integrates nonstationary probabilistic models to represent temporal changes in air temperature frequency distributions at the regional scale. Application of the NS-Add-RFA in controlled simulation experiments and to daily extreme maximum air temperature data from São Paulo State, Brazil (1991–2024), showed that it outperforms nonstationary at-site approaches in capturing the probabilistic structure of extremes under diverse climate and climate change scenarios. The new regionalization method revealed widespread increases in extreme air temperatures across 93% of the state, capturing changes in both the central tendency and dispersion of the series. Furthermore, it provides a more detailed and reliable trend assessment than traditional approaches such as the regional Mann–Kendall trend test. To facilitate broader adoption, we developed the R package NSTempRFA ( https://github.com/gabrielblain/NSTempRFA ), which computes all NS-Add-RFA statistics. In conclusion, the NS-Add-RFA offers a scientifically grounded framework for nonstationary regional frequency analysis, enhancing the capacity to assess and interpret extreme air temperature events under changing climate conditions.