Impact of evapotranspiration process representation on runoff projections from conceptual rainfall-runoff models

Conceptual rainfall-runoff models are commonly used to estimate potential changes in runoff due to climate change. The development of these models has generally focused on reproducing runoff characteristics, with less scrutiny on other important processes such as the conversion from potential evapotranspiration (PET) to actual evapotranspiration (AET). This study uses three conceptual rainfall-runoff models (GR4J, AWBM, and IHACRES_CMD) and five catchments in climatologically different regions of Australia to explore the role of ET process representation on the sensitivity of runoff to plausible future changes in PET. The changes in PET were simulated using the Penman-Monteith model and by perturbing each of the driving variables (temperature, solar radiation, humidity, and wind) separately. Surprisingly, the results showed the potential of a more than sevenfold difference in runoff sensitivity per unit change in annual average PET, depending on both the rainfall-runoff model and the climate variable used to perturb PET. These differences were largely due to different ways used to convert PET to AET in the conceptual rainfall-runoff models, with particular dependencies on the daily wet/dry status, as well as the seasonal variations in store levels. By comparing the temporal patterns in simulated AET with eddy-covariance-based observations at two of the study locations, we highlighted some unrealistic behavior in the simulated AET from AWBM. Such process-based evaluations are useful for scrutinizing the representation of physical processes in alternative conceptual rainfall-runoff models, which can be particularly useful for selecting models for projecting runoff under a changing climate.