Identifying the Influence of Systematic Errors in Potential Evapotranspiration on Rainfall–Runoff Models

Abstract

This study attempts to identify a threshold in the bias of potential evapotranspiration (PET) forcing data beyond which output from hydrological models will be significantly impacted. The sensitivity of a widely used conceptual rainfall–runoff model to systematic errors in PET inputs is investigated for a sample of 57 US catchments across energy- and water-limited regions. PET forcing data are biased by a constant factor ranging from ±5% to ±25%. The sensitivity of hydrologic models to PET data quality was found to be primarily driven by the long-term ratio of actual evapotranspiration (AET) to PET of catchments, which determines the energy availability of catchments. Energy-limited catchments were more sensitive to PET errors than water-limited catchments, and the PET error threshold was found to decrease along the water- to energy-limited continuum. Moreover, model performance of rainfall–runoff models was found to be more sensitive to negative PET biases than to positive PET biases. In practice, this means negatively biased PET causes catchments to artificially shift toward energy limitation, resulting in higher model sensitivity. Given the decreasing pattern of PET error threshold in catchments along the water- to energy-limited continuum, energy availability of catchments can be used as a predictor for the requirements of accuracy in PET data for conceptual hydrological models.