LYSIMETRIC EVALUATION OF THE APEX MODEL TO SIMULATE DAILY ET FOR IRRIGATED CROPS IN THE TEXAS HIGH PLAINS

摘要

The NTT (Nutrient Tracking Tool) was designed to provide an opportunity for all users, including producers, to run complex simulation models, such as APEX (Agricultural Policy Environmental extender), with the associated required databases. The APEX model currently nested within NTT provides estimates of the changes in nitrogen (N), phosphorus (P), and sediment losses that are associated with management practices specified by the user. Five methods (Pen- man-Monteith, Penman, Priestley-Taylor, Hargreaves-Samani, and Baier-Roberts on) for determining potential evapotranspiration (PET) are available as inputs for estimating actual ET. This study was conducted to evaluate the accuracy of the ET values obtained from the five PET equations currently available in APEX using both onsite measured climate data and data from the NTT standard databases. The mean daily, monthly, and annual ET values predicted by each of the equations in APEX for a lysimeter field at the USDA-ARS Conservation and Production Research Laboratory at Bushland, Texas, was compared to values measured for the 2001-2010 period. APEX generally underestimated ET with all PET methods (mostly during growing seasons) at both the daily and monthly levels but overpredicted for years when cotton was grown as the major cash crop due to overprediction of leaf area index during the senescing stage for cotton. The underprediction of ET in growing seasons was possibly from underprediction of rainfall due to estimation of rainfall for missing data. Overall, APEX was able to adequately (R~2 ≥ 0.82 and NSE ≥ 0.80) predict mean monthly ET for major crops grown in the semi-arid Texas High Plains region. These results should reinforce confidence in APEX's ability to simulate ET accurately for fully irrigated farms. ET predictions with the Hargreaves-Samani and Priestley-Taylor methods, which require limited data compared to the Penman and Penman-Monteith methods, were similar (p > 0.05, one-way ANOVA), with mean errors within 8.7% for measured weather data and 12.6% for NTT-generated weather data for both methods. This is encouraging because of the limited availability of measured climate data for the majority of locations in the world, including the U.S.