ABSTRACT With global climate change and human activities, environmental uncertainties are increasing, and extreme flood events are occurring more frequently. The reliability of traditional hydrological frequency analysis theories, based on the assumption of stationarity, is being increasingly questioned. This study aims to develop a non‐stationary flood frequency analysis model using the Generalized Additive Models for Location, Scale, and Shape (GAMLSS) framework, with time and climate indices as covariates. The model calculation and frequency analysis are conducted using 2000 years of climate indices reconstructed by the Paleo Hydrodynamics Data Assimilation product (PHYDA). Design floods for different return periods are then quantified based on the reconstructed data. The results show that the nonstationary model established with climate indices as covariates can accurately identify the trend of first decreasing and then increasing flood series at the FP and ZJG stations in the Daqing River Basin, achieving the best model performance. Moreover, using the PHYDA‐reconstructed climate indices from the past 2000 years to extrapolate floods and calculate design floods provides higher safety for certain return periods than observed series. However, under longer return periods, the design values are smaller than those of the existing observed series. Overall, the nonstationary model proposed in this study can serve as a tool for flood frequency analysis under climate change. Additionally, incorporating the climate indices from the past 2000 years into nonstationary flood frequency analysis provides design results that can offer valuable references for regional water infrastructure design.
Guo et al. (Mon,) studied this question.