Assessing the impact of PET estimation methods on hydrologic model performance

Evapotranspiration is a necessary input and one of the most uncertain hydrologic variables for quantifying the water balance. Key to accurately predicting hydrologic processes, particularly under data scarcity, is the development of an understanding of the regional variation of the impact of potential evapotranspiration (PET) data inputs on model performance and parametrization. This study explores this impact using four different potential evapotranspiration products (of varying quality). For each data product, a lumped conceptual rainfall-runoff model (GR4J) is tested on a sample of 57 catchments included in the MOPEX data set. Monte Carlo sampling is performed, and the resulting parameter sets are analyzed to understand how the model responds to differences in the forcings. Test catchments are classified as energy- or water-limited using the Budyko framework and by eco-region, and the results are further analyzed. While model performance (and parameterization) in water-limited sites was found to be largely unaffected by the differences in the evapotranspiration inputs, in energy-limited sites model performance was impacted as model parameterizations were clearly sensitive to evapotranspiration inputs. The quality/reliability of PET data required to avoid negatively impacting rainfall-runoff model performance was found to vary primarily based on the water and energy availability of catchments.

Automated summary

This study investigates the impact of potential evapotranspiration (PET) data inputs on model performance and parameterization using different quality PET products and a rainfall-runoff model. It was found that differences in evapotranspiration inputs largely did not affect model performance in water-limited sites, but did impact performance in energy-limited sites. The quality/reliability of PET data required for avoiding negative impacts on model performance varied based on water and energy availability of catchments.