Evaluation and comparison of OpenET models for estimating soil water depletion of irrigated alfalfa in Arizona

Abstract

Growers in arid regions often estimate soil water depletion (D_r) of the root zone in making irrigation scheduling decisions. While growers might use soil water sensors, actual D_r in fields may be only vaguely assessed. In this study, daily crop evapotranspiration (ET_sat) data from the six OpenET satellite-based models (ALEXI/DisALEXI, eeMETRIC, geeSEBAL, PT-JPL, SIMS, SSEBop) and the OpenET ensemble were incorporated into a soil water balance (SWB) to estimate D_r (D_{r sat}) for alfalfa (Medicago sativa L.) within a 46-ha, center-pivot field in Buckeye, Arizona. A daily SWB for each model input the average ET_sat data acquired at four locations for both 30-m and 90-m pixel sizes, assuming an alfalfa root-zone of 1.8-m. Net irrigation (I_n) inputs were based on gross irrigation amounts, modified for evaporation losses using weather data from the AZMET station in Buckeye. D_{r sat} values were compared to the average observed root-zone D_r (D_{r obs}), as determined at the four locations by neutron moisture meter measurements on 21 dates between 05/23/2023 and 12/11/2023. Cumulative values were 841 mm for I_n plus effective precipitation, 776 mm (ALEXI/DisALEXI) to 1130 mm (SSEBop) for ET_sat, and 816 mm of ET obtained by a seasonal SWB (ET_swb) based on the average change in D_{r obs} (ΔD_{r obs}) from first to last measurements. Average D_{r obs} fluctuated between 59 and 133 mm, and ΔD_{r obs} was −25.0 ± 42.0 mm. Model variations in ET_sat resulted in different estimates of D_{r sat} and agreements, where mean bias error (MBE) ranged from ≈ -35 % (PT-JPL) to 195 % (SSEBop). Despite notable uncertainties in the SWB parameters and observed data, the SWB based on both the ensemble and ALEXI/DisALEXI ET_sat capably tracked D_{r obs} for most of the observations, excluding those made from late-July to mid-August. However, the best overall agreement was with the ensemble data, where differences between cumulative ET_sat and ET_swb and ΔD_{r sat} and ΔD_{r obs} were small, and MBE for D_{r sat} was less than 9 % and 15 % at 30-m and 90-m pixels, respectively. Findings suggest that using the single value OpenET ensemble daily ET could be a reliable source for estimating the D_r in arid climate alfalfa fields.