LAUSR

Quantifying uncertainty in runoff changes has profound implications for future runoff investigations and will support global climate model improvement. We analyze the runoff changes from model outputs of the 5th and 6th Coupled Model Intercomparison Projects (CMIP5 and CMIP6) and decompose the uncertainty of projected runoff changes into three main sources: internal variability, model uncertainty, and scenario uncertainty. The results indicate that 73% and 72% of the global land area, respectively, shows an increase in runoff for CMIP5 and CMIP6 under the high‐emissions scenarios in the long term (2070–2099) relative to 1970–1999, and runoff across all global land increases by 10.8% under the RCP8.5 scenario and 16.1% under SSP5–8.5 during 2070–2099 relative to 1970–1999. Regions with increasing runoff are mainly in Southeast Asia, eastern Africa, the Qinghai‐Tibet Plateau and the high latitudes of the Northern Hemisphere. Model agreement during 2070–2099 for runoff changes under the high‐emissions scenarios is greater than under the low‐emissions scenarios. For uncertainty of relative changes in runoff (RRC) in CMIP6, the contribution from internal variability gradually decreases over time (from 49.2% to 2.0%) while the contribution from scenario uncertainty increases (from 0.6% to 30.0%); this result is similar in CMIP5. Spatially, model uncertainty in RRC has been a major source of uncertainty for CMIP5 and CMIP6, accounting for more than 60% of total uncertainty in most regions. The results of this study help us to better understand the uncertainty in runoff changes and also provide a theoretical basis for developing mitigation measures for future changes.