Comparing evapotranspiration estimations using crop model-data fusion and satellite data-based models with lysimetric observations: Implications for irrigation scheduling

Authors: STOCKLE, CLAUDIO - Washington State University; LIU, MINGLIANG - Washington State University; KADAM, SUNIL - Washington State University; Evett, Steven; Marek, Gary; Colaizzi, Paul

Submitted to: Agricultural and Forest Meteorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/12/2025
Publication Date: N/A
Citation: N/A

Interpretive Summary: The scarcity of water resources in the U.S. Southern High Plains is of regional, national and even international concern because the region acts as a breadbasket for the nation and world. The majority of agricultural production in this region depends on irrigation, largely dependent on pumping from the Ogallala or High Plains Aquifer, which are yielding less water every year. Accurate computer estimates of field scale crop water use can improve irrigation management and crop yield per unit of water applied. But computer models of crop water use vary widely in their accuracy. Scientists at the USDA ARS Conservation & Production Research Laboratory at Bushland, Texas, teamed with scientists in Washington State and in India to test three important crop models against three years of accurately measured daily crop water use data from corn grown at Bushland. The two models that were based on remote sensing (satellite images) overestimated crop water use later in the season and would have caused overirrigation if used to guide irrigation. One model, called CropSyst-W, based its estimates on easily obtained weather data and had the best agreement with directly measured crop water use.

Technical Abstract: Accurate estimation of field scale evapotranspiration (ET) is the foundation for efficient on-farm water management. In this study, three models to estimate ET were compared with measurements collected in four lysimeters at Bushland, Texas, USA. The lysimetric data was for maize (Zea mays L) during the seasons 2013, 2016, and 2018, irrigated using a MESA (Mid Elevation Irrigation System) linear-move system applying 100% ET (SW lysimeter) and 75% ET (NW lysimeter), and SDI (Subsurface Drip Irrigation) applying 100% ET (NE and SE lysimeters). The estimations were performed with CropSyst-W, a crop model integrating NDVI data to derive green canopy cover, and two remote sensing-based energy balance models: EEFlux and OpenET. Two statistical indices (Willmott index of agreement, d, and the normalized root mean square deviation, NRMSD) were used to evaluate the performance of the models. The average d values of the 12 combinations of 3 years and four lysimeters were 0.93 and 0.77 for CropSyst-W and EEFlux, respectively. Estimations from OpenET were only available in 2016 and 2018, and the average d values for those years were 0.95, 0.85, and 0.89 for CropSyst-W, EEFlux, and OpenET, respectively. The average NRMSD values of the 12 combinations were 0.31 and 0.47 for CropSyst-W and EEFlux, while the average NRMSD for the two years that included OpenET was 0.28 (CropSyst-W), 0.43 (EEFlux), and 0.32 (OpenET). Overall, CropSyst-W had the best performance (larger d and lower NRMSD), also showing the lowest fraction of systematic RMSD. The agreement between lysimeter and models ET fluctuated for early growth, midseason, and senescence periods, with challenges for EEFlux and OpenET during senescence due to the consistent lag of the NDVI decline leading to overestimation. Uncertainty factors affecting the ET estimation of the models and the lysimetric measurements, the adequacy of models to support irrigation scheduling, and avenues for improvement of the models are discussed.