Evaluating evapotranspiration estimation methods in APEX model for dryland cropping systems in a semi-arid region

Evapotranspiration (ET), is a major component of the hydrologic budget and therefore it requires accurate estimation. The Agricultural Policy/Environmental eXtender (APEX), a hydrologic and water quality model developed for evaluating the effect of agricultural production management practices on the environment. It has five different methods to simulate ET. The objectives of this study were to: determine the impact of using different ET methods in APEX on sensitive ET parameters in semi-arid environments, determine reasonable range of values for sensitive parameters, and compare performance of these methods using multiple criteria. ET and crop yield data measured in the Lysimeter fields located in the USDA-ARS Conservation and Production Research Laboratory, Bushland, Texas were used to calibrate and validate the model. Results indicated that selection of statistical model performance measure and ET method affects the number of sensitive parameters and parameter rank. The number of sensitive parameters remained relatively stable among ET simulation methods but there was large variability in parameter sensitivity ranks. It is also important that users carefully choose appropriate statistical measure to use depending on the goals of their study. With the exception of maximum rainfall intercept, exponent coefficient rainfall, SCS index coefficient, rain intercept coefficient, and root growth soil strength, all methods had similar ranges of values for the sensitive parameters. In general there were no large differences in the performance of ET methods. However, Penman-Monteith method simulated ET relatively better than the other methods, which may be explained by the fact that it is a physically-based method with many weather variables whose data was available for the study area. However, the study findings indicate that the other ET methods can provide satisfactory results in regions with limited weather data.