Non-Stationarity flood frequency analysis in the Lancang-Mekong River Basin under Climate Change

Extreme floods exceeding historical records have become more frequent globally in recent years due to climate change, signaling an increasing non-stationarity in flood patterns. Traditional design floods, based on the assumption of stationarity, are no longer sufficient to ensure engineering safety and human welfare, necessitating a re-evaluation and revision of design flood standards. The Lancang-Mekong River Basin (LMRB) is both climate-sensitive and a high-risk area for flood disasters. To better manage future flooding in the LMRB, six hydrological stations along the mainstream are focused to analysis flood non-stationarity. In this study, a GAMLSS model based on temporal covariates is developed and nine global climate models and two SSPs-RCPs scenarios are designed for flood peaks frequency analysis. The results show that annual maximum flood peak series exhibit significant non-stationarity, with a noticeable increasing trend across the entire basin under the BCC, CCCMa, and MIRCO climate models. In contrast, the remaining models show an increasing trend in the upstream and a decreasing trend in the downstream. When non-stationary models are constructed, the flood peak series at most stations follow log-normal and gamma distributions under different future scenarios, with both the mean and variance showing a strong linear relationship with time. Compared with traditional stationary models, future design floods present heterogeneous deviations from upstream to downstream. At the upstream Chiang Saen station, flood estimates shift from overestimation to underestimation over time, with a 5% underestimation of the 100-year flood by 2065. This suggests that additional flood control infrastructure will be needed to withstand more frequent extreme floods. Conversely, at the downstream Kratie station, an opposite trend is observed, with a 7% overestimation of the 100-year flood by 2065, suggesting that some existing infrastructure may become redundant in the future. This study providing a more accurate scientific basis for flood risk forecasting and offering new support for flood management and disaster risk reduction in the basin.