Abstract. Understanding the heterogeneity of climate change and its impacts on annual and seasonal discharge and the difference between median flow and extreme flow in different climate regions is of utmost importance to successful water management. To quantify the spatial and temporal heterogeneity of climate change impacts on hydrological processes, this study simulated river discharge in the River Huangfuchuan in semi-arid northern China and in the River Xiangxi in humid southern China. The study assessed the uncertainty in projected discharge for three time periods (2020s, 2050s and 2080s) using seven equally weighted GCMs (global climate models) for the SRES (Special Reports on Emissions Scenarios) A1B scenario.

Climate projections that were applied to semi-distributed hydrological models (Soil Water Assessment Tools, SWAT) in both catchments showed trends toward warmer and wetter conditions, particularly for the River Huangfuchuan. Results based on seven GCMs' projections indicated changes from −1.1 to 8.6 °C and 0.3 to 7.0 °C in seasonal temperature and changes from −29 to 139 % and −32 to 85 % in seasonal precipitation in the rivers Huangfuchuan and Xiangxi, respectively. The largest increases in temperature and precipitation in both catchments were projected in the spring and winter seasons. The main projected hydrologic impact was a more pronounced increase in annual discharge in the River Huangfuchuan than in the River Xiangxi. Most of the GCMs projected increased discharge in all seasons, especially in spring, although the magnitude of these increases varied between GCMs. The peak flows were projected to appear earlier than usual in the River Huangfuchuan and later than usual in the River Xiangxi, while the GCMs were fairly consistent in projecting increased extreme flows in both catchments with varying magnitude compared to median flows. For the River Huangfuchuan in the 2080s, median flow changed from −2 to 304 %, compared to a −1 to 145 % change in high flow (Q05 exceedance threshold). For the River Xiangxi, low flow (Q95 exceedance threshold) changed from −1 to 77 % and high flow changed from −1 to 62 %, while median flow changed from −4 to 23 %. The uncertainty analysis provided an improved understanding of future hydrologic behavior in the watershed. Furthermore, this study indicated that the uncertainty constrained by GCMs was critical and should always be considered in analysis of climate change impacts and adaptation.