LAUSR

Accurate estimation of reference crop evapotranspiration (ET0) is important due to its crucial role in determining crop water requirement in irrigated agriculture. Though a great number of models have been developed, their rigorous evaluation with measurements is still lacking, leading to confusion and arbitrariness in model selection. In this paper daily estimates of 16 ET0 models, including five combination-, six radiation and five temperature-based ones, were compared with weighing lysimeter measurements during crop growing season (April through October) in 2012 at a semiarid site in China. Daily ET0 was measured by two weighing lysimeters (area 1.3m×1.3m, depth 2.3m) located in a fescue grass (Festuca arundinacea Schreb) plot (100m×100m) surrounded by a 167ha crop, winter wheat rotated with summer maize. We found the models were ranked decreasingly as: FAO-ppp-17 Penman>1963 Penman>FAO-24 Blaney-Criddle (BC)>1996 Kimberly Penman>FAO-24 radiation>FAO-56 Penman-Monteith (PM)>FAO-24 Penman>Turc>DeBruin-Keijman>Jensen-Haise>Priestley-Taylor>Hargreaves>Makkink>Hamon>Blaney- Criddle>Mcloud on basis of RMSE (root mean square error). Overall, the combination models performed best with RMSE ranging from 0.93 to 1.32mmd−1, followed by the radiation models with RMSE from 1.28 to 1.79mmd−1, and the temperature models with RMSE from 1.09 to 2.48mmd−1. The best combination model (FAO-ppp-17 Penman) was respectively 29% and 17% more accurate than the best radiation (FAO-24 radiation) and temperature (FAO-24 BC) models. Better performance of the combination and radiation models resulted because they explicitly contain the dominant factors influencing ET0. All models tended to overestimate under low evaporative demand while underestimating the measured values under high demand, but on average the combination and radiation methods underestimated by 0.46mmd−1 and 0.60mmd−1, respectively, whereas the temperature method overestimated by 0.21mmd−1. All combination and radiation models, and the Hargreaves and FAO-24 BC in temperature method showed robust structure. To improve them future efforts should be on local calibration, but for temperature models showing structure failure focus should be on its optimization. The coefficients of commonly used models were calibrated and related to meteorological variables. Particularly, those of the Priestley-Taylor, Makkink, Turc and the Hamon were enhanced, while those of the Hargreaves and BC were decreased. In climate similar to the current site in China we suggest continued use of the older Penman equations for combination method and the FAO-24 radiation or Turc for radiation method. Meanwhile, two questions need to be addressed in future studies: i) adoption of the FAO-56 PM equation as the sole standard for computing ET0 and the proper value for surface resistance; and ii) the effectiveness of the later modifications to the original wind function in the Penman equation.