Comparing field-scale eddy covariance measurements and crop coefficient estimates of alfalfa evapotranspiration to OpenET model estimates and exploring water budget implications in a dryland environment

Abstract

AbstractRemote sensing evapotranspiration (ET) models have the potential to be powerful tools for water planning and management, particularly for agriculture. OpenET is an emerging web-based tool that uses satellite imagery and climate data for calculating six distinct ET models, and an ensemble model of the six models, to provide estimates of actual ET (ETa) which is useful for field-scale irrigation management decisions. Previous studies examining the performance of individual models included in the OpenET platform showed some models used in OpenET consistently predicted lower values of ETa in dryland regions relative to in-situ measurements. The OpenET research team has made modifications to address these isues. There are few studies examining if the modified models included in OpenET sill produce lower values of ETa compared to field values in dryland environments. This study compared satellite-based OpenET estimates of ETa from three alfalfa fields in the Mesilla Valley, New Mexico, USA–one field with measurements of ETa from an eddy covariance tower and two fields with estimated crop evapotranspiration (ETc)–during the 2017 growing season to investigate if OpenET ETa estimates demonstrate an underestimation bias. OpenET ETa estimates were tested against in-situ ETa measurements and ETc estimates using two sample t-tests and Mann-Whitey U tests to determine if there were any significant differences in means between the two groups. Model seasonal percent mean bias error ranged from −33.99 to +11.37%. eeMETRIC and SIMS seasonal estimates were within ±15% of in-situ measurements at any of the three sites and within ±10% of in-situ measurements on average. SSEBop and DisALEXI produced significantly different monthly ETa estimates (p-values < 0.05) when data were extracted using the OpenET field polygons. The results of the small sample of fields suggest the OpenET models may estimate lower values of ETa relative to the field data. Future research should improve the methodology for assessing accuracy of OpenET in small agricultural fields in the western United States.