Runoff Prediction and Uncertainty Analysis for Xijiang River Basin Based on CMIP6 Climate Scenarios

Due to the combined effects of global climate change and strong human activities, extreme floods have become frequent and widespread, with significant changes in runoff sequences. Predicting future runoff changes in flood-prone areas under the influence of climate change and human activities is of great significance for regional water disaster prevention and rational water resource utilization. Therefore, by taking the Xijiang River Basin, a region with frequent floods, as the research object, this paper adopts the Mann-Kendall mutation test and univariate linear regression methods to reveal the non-uniform characteristics of the basin's runoff sequences. Based on this, the Xin'anjiang hydrological model (XAJ) is built, and the particle swarm optimization (PSO) algorithm is employed to calibrate and validate the model parameters. Furthermore, by utilizing data from 15 climate models under CMIP6, the bias correction and spatial disaggregation (BCSD) downscaling method is applied to downscale the data to the Xijiang River Basin. Additionally, the successfully built XAJ model is then driven by the data, and indicators such as the coefficient of variation and relative runoff changes are adopted to reveal the characteristics of future water resource variability in the basin. The results show that the annual runoff of the basin decreases year by year at a rate of 17.26 m³/s, with the abrupt change occurring in 2002. The built XAJ model can be more effectively applied to the Xijiang River Basin, and its ability to capture peak flow is superior. In the near, medium, and long term, the runoff during the non-flood season will increase significantly, while that in flood seasons will decrease. The results can provide an important scientific basis for water management agencies to plan, allocate, and utilize water resources reasonably in the basin, thereby reducing the influence of GCMs and SSPs uncertainty on runoff prediction uncertainty.