Resumen
Assessing the impact of climate change on streamflow is critical to understanding the changes to water resources and to improve water resource management. The use of hydrological models is a common practice to quantify and assess water resources in such situations. In this study, two hydrological models with different structures, e.g., a physically-based distributed model Liuxihe (LXH) and a lumped conceptual model Xinanjiang (XAJ) are employed to simulate the daily runoff in the Xijiang basin in South China, under historical (1964?2013) and future (2014?2099) climate conditions. The future climate series are downscaled from a global climate model (Beijing Climate Centre-Climate System Model, BCC-CSM version 1.1) by a high-resolution regional climate model under two representative concentration pathways?RCP4.5 and RCP8.5. The hydrological responses to climate change via the two rainfall?runoff models with different mathematical structures are compared, in relation to the uncertainties in hydrology and meteorology. It is found that the two rainfall?runoff models successfully simulate the historical runoff for the Xijiang basin, with a daily runoff Nash?Sutcliffe Efficiency of 0.80 for the LXH model and 0.89 for the XAJ model. The characteristics of high flow in the future are also analysed including their frequency (magnitude?return-period relationship). It shows that the distributed model could produce more streamflow and peak flow than the lumped model under the climate change scenarios. However the difference of the impact from the two climate scenarios is marginal on median monthly streamflow. The flood frequency analysis under climate change suggests that flood magnitudes in the future will be more severe than the historical floods with the same return period. Overall, the study reveals how uncertain it can be to quantify water resources with two different but well calibrated hydrological models.