全球变化知识资源中心

In this study, we assess whether dynamical downscaling using a regional climate model (RCM) would provide added value against the driving global climate model (GCM) and whether a statistically bias-corrected GCM could be used as an alternative to a bias-corrected RCM for assessing the hydrological response to climate change. The target region is the Chungju Basin in Korea, which is characterized by complex geographical features. The analysis focuses on runoff simulations driven by raw outputs and bias-corrected outputs from the HadGEM2-AO and RegCM4, which are used to generate national standard climate change scenarios throughout the Korean Peninsula. Statistical methods, such as linear scaling and quantile mapping, improve the performance in precipitation simulations and effectively correct the systematic biases seen with the climatological means of the HadGEM2-AO and RegCM4. Although there is no relevant difference between the bias-corrected HadGEM2-AO and RegCM4 in terms of long-term averaged climatology, the bias-corrected RegCM4 performs better in capturing the dominant mode of runoff variability and hydrological extremes (e.g., floods measured by 1-day maximum runoff and droughts measured by the standardized runoff index). Because the magnitude of the bias correction factor tends to influence the preservation of future change signals derived from raw and bias-corrected outputs, the HadGEM2-AO, which requires a large bias correction factor to compensate for poor performance, produces future changes in high flows with low coherence between the raw and bias-corrected projections.