Non-stationary dynamics of compound climate extremes: a WRF-CMIP6-GAMLSS framework for southeastern China

Abstract

Understanding future changes in compound climate extremes (CCEs) is critical for climate risk assessment. Existing research, however, has largely relied on stationary assumptions, overlooking the dynamic evolution of CCEs under non-stationary climate change. To address this gap, this study employs an enhanced Generalized Additive Model for Location, Scale, and Shape (GAMLSS) framework to provide novel insights into the non-stationary characteristics of hot-wet (HW), hot-dry (HD), cold-wet (CW), and cold-dry (CD) extremes under future climate scenarios. We focus on the Minjiang River Basin (MRB) in Southeast China. A high-resolution dataset for CEE detection was generated by dynamical downscaling a bias-corrected CMIP6 dataset, using the Weather Research and Forecasting (WRF) model. Our results indicate that (1) CCEs increase significantly at a rate of 3.55 d per decade under the SSP5-8.5 scenario, with hot extremes (HW and HD) being the dominant contributors. Spatially, the increases exhibit a distinct west to east gradient, peaking in the downstream areas of the MRB. (2) Under the SSP5-8.5 scenario, CCEs exhibit a marked shift from stationary to non-stationary characteristics, with non-stationarity detected in 95.20 % of grid cells. This transition is primarily driven by mean warming, which explains 80.81 % of the change, rather than by variability. (3) The non-stationary results demonstrate that the severity and recurrence frequency of CCEs are systematically underestimated under stationary assumptions. Most CCE types (except for CD) show an increasing recurrence frequency under the SSP5-8.5 scenario. For instance, the frequency of events with a 100-year return period increases at a stronger trend of 3.12 d per decade. This study emphasizes the necessity of updating the frequency changes of CCEs under a non-stationary framework.