Research papersEvaluation of six potential evapotranspiration models for estimating crop potential and actual evapotranspiration in arid regions

- AET showed significantly linear and positive correlation with PET on field scale.
- Crop coefficient existed remarkably hyperbolic function with canopy conductance.
- A hyperbolic model was used to parameterize crop coefficient.
- SW, PE and PT models combined with crop coefficient model estimated AET accurately.

Using potential evapotranspiration (PET) to estimate crop actual evapotranspiration (AET) is a critical approach in hydrological models. However, which PET model performs best and can be used to predict crop AET over the entire growth season in arid regions still remains unclear. The six frequently-used PET models, i.e. Blaney-Criddle (BC), Hargreaves (HA), Priestley-Taylor (PT), Dalton (DA), Penman (PE) and Shuttleworth (SW) models were considered and evaluated in the study. Five-year eddy covariance data over the maize field and vineyard in arid northwest China were used to examine the accuracy of PET models in estimating daily crop AET.Results indicate that the PE, SW and PT models underestimated daily ET by less than 6% with RMSE lower than 35 W m−2 during the four years, while the BC, HA and DA models under-predicted daily ET approximately by 10% with RMSE higher than 40 W m−2. Compared to BC, HA and DA models, PE, SW and PT models were more reliable and accurate for estimating crop PET and AET in arid regions. Thus the PE, SW and PT models were recommended for predicting crop evapotranspiration in hydrological models in arid regions.