Evaluation of evapotranspiration models for estimating daily reference evapotranspiration in arid and semiarid environments

Abstract

Daily outputs from eight evapotranspiration models were tested against reference evapotranspiration (ETo) data computed by FAO56PM to assess the accuracy of each model in estimating ETo. Models were compared at eight stations across Jordan. Results show that Hargreaves modified models were the best in light of mean biased error (MBE), root mean square error (RMSE) and mean absolute error (MAE). The MBE, RMSE, and MAE values ranged from -1.47 to 0.81, 3.87 to 1.14 and 0.87 to 3.15 mm/day for HarM1, and from -1.45 to 0.89, 1.08 to 3.91, and 0.85 to 3.16 mm/day for HarM2, respectively, which would make it the best models in light of the MBE, RMSE and MAE ranging from -6.18 to 2.79, 6.90 to 1.08 and 4.74 to 0.85 mm/day for all models and stations. Comparisons were also made using three composite regions: countrywide, semiarid, and arid regions. In conclusion, local calibration is needed for the whole models or the linear regression can be used to calculate the ETo.