Applicability assessment of reference evapotranspiration models in Beijing based on lysimeter measurement

Abstract: Reference evapotranspiration (ET0) is the basic parameter for the vegetation evapotranspiration calculation, the regional water balance analysis, and the water resources management. The ET0 models have different adaptability for different regions because of the difference in meteorological conditions among regions. Measurement using lysimeter is the classical method for ET0 model evaluation in European and American areas, but is seldom used in China, and no research reported for North China. Assessment of four ET0 models, Penman-Monteith (FAO-56), Hargreaves-Samani, Priestley-Taylor, and Penman-van Bavel were conducted in Beijing using tall fescue (Festuca arundinacea L.) turf evapotranspiration measurement by auto-weighing lysimeter during the growing season (April - October) of 2012. Two lysimeters were established in cold-season grass tall fescue for standard ET0. The weather station (Dynamet, Dynamax Co. Ltd.) was set up in the experiment site to measure and record automatically the meteorological data of air temperature, air relative humidity, solar radiation and wind speed at 2 meter height above ground for the ET0 calculation by the models. The linear regression and the root-mean-square error (RMSE) and the index of agreement (d) were used for assessing the prediction accuracy of the different models. The results indicated that the solar radiation showed a linear relationship (R2=0.95, p=2.72×10-7) with the monthly reference evapotranspiration, indicating that the solar radiation energy was the main force to drive water from the vegetation to the atmosphere in SPAC system. The accuracy of the models declined with the time scale. The models showed different deviations in ET0 calculation because of the different parameters input. The monthly ET0 was underestimated by Priestley-Taylor model but overestimated by Penman-Monteith、Hargreaves-Samani and Penman-van Bavel models. For the daily ET0, Hargreaves-Samani and Penman-van Bavel models had a slight overestimation with the rate of 1.0167 and 1.0526, respectively, but Penman-Monteith and Priestley-Taylor models had an underestimation with the rate of 0.8204 and 0.7593, respectively. For the timely ET0, Priestley-Taylor had the lowest value among the models, while other models resulted in different values under different weather conditions. The overall estimation for the ET0 at the monthly, daily, and hourly scales showed that the Penman-van Bavel equation was the most precise method for calculating reference evapotranspiration, with a RMSE of 0.63 mm/d and a d-index of 0.96 (monthly), a RMSE of 1.43 mm/d and a d-index of 0.89 (daily), a RMSE of 0.087 mm/h and a d-index of 0.87 (hourly). The calculating accuracy of Penman-Monteith model was a little lower than that of Penman-van Bavel with a d-index of 0.73-0.93.