Assessing future runoff changes with different potential evapotranspiration inputs based on multi-model ensemble of CMIP5 projections

Runoff projection under future climate scenarios has been widely studied to investigate the impacts of climate change on regional water availability. However, uncertainty in runoff projection due to different ETp inputs has not been fully assessed. This study firstly adopted the physically-based Penman model, temperature-based Hargreaves model, and radiation-based Abtew, Jensen-Haise, and modified Makkink models to drive Xinanjiang (XAJ) model, thus investigating the influence of different potential evapotranspiration (ETp) inputs on runoff simulation. Then, we used the validated XAJ model to project runoff in North Johnstone catchment, northeast Australia. Lastly, we quantified the uncertainty caused by 34 global climate models (GCMs), different representative concentrative pathway (RCP) scenarios (RCP4.5 & RCP8.5), and different ETp models with the technique of three-way analysis of variance (ANOVA). We found that XAJ model performed well (R² ≥ 0.88, NSE ≥ 0.86) and showed low sensitivity to different ETp inputs in runoff simulation and projection. Under future climate scenarios, spring and winter runoff had a large decrease, which was mainly caused by the decrease in rainfall. The mean decreases in spring and winter runoff were 14.6% – 20.1% and 10.3% – 15.2% respectively by 2090s under RCP8.5. GCMs (50.9% – 67.4%) and their interaction with RCPs (35.4% – 46.6%) were the dominant factors resulting in uncertainty in runoff projection. Our study not only advanced the understanding of the impacts of different ETp inputs on runoff projection but also offered insights on the understanding of the roles different factors played in the uncertainty in runoff projection. We expect such knowledge can provide a way forward to narrow down the uncertainty in runoff projection, thus providing more robust information for policy makers in water management.