资源环境科技发展态势分析平台

Projections of runoff from global multi-model ensembles provide a valuable basis for the estimation of future hydrological extremes. However, projections suffer from uncertainty that originates from different error sources along the modeling chain. Hydrological impact studies have generally partitioned these error sources into global impact and global climate model (GIM and GCM, respectively) uncertainties, neglecting other sources, including scenarios and internal variability. Using a set of GIMs driven by GCMs under different representative concentration pathways (RCPs), this study aims to partition the uncertainty of future flows coming from GIMs, GCMs, RCPs, and internal variability over the CONterminous United States (CONUS). We focus on annual maximum, median, and minimum runoff, analyzed decadally over the twenty-first century. Results indicate that GCMs and GIMs are responsible for the largest fraction of uncertainty over most of the study area, followed by internal variability and to a smaller extent RCPs. To investigate the influence of the ensemble setup on uncertainty, in addition to the full ensemble, three ensemble configurations are studied using fewer GIMs (excluding least credible GIMs in runoff representation and GIMs accounting for vegetation and CO2 dynamics), and excluding intermediate RCPs. Overall, the use of fewer GIMs has a minor impact on uncertainty for low and medium flows, but a substantial impact for high flows. Regardless of the number of pathways considered, RCPs always play a very small role, suggesting that improvement of GCMs and GIMs and more informed ensemble selections can yield a reduction of projected uncertainties.