Evaluation of surface energy balance models for mapping evapotranspiration using very high resolution airborne remote sensing data

Abstract

Agriculture is the largest (90%) consumer of all fresh water in the world. The consumptive use of water by vegetation represented by the process evapotranspiration (ET) has a vital role in the dynamics of water, carbon and energy fluxes of the biosphere. Consequently, mapping ET is essential for making water a sustainable resource and also for monitoring ecosystem response to water stress and changing climate. Over the past three decades, numerous thermal remote sensing based ET mapping algorithms were developed and these have brought a significant theoretical and technical advancement in the spatial modeling of ET. Though these algorithms provided a robust, economical, and efficient tool for ET estimations at field and regional scales, yet the uncertainties in flux estimations were large, making evaluation a difficult task. The main objective of this study was to evaluate and improve the performance of widely used remote sensing based energy balance models, namely: the Surface Energy Balance Algorithm for Land (SEBAL), Mapping Evapotranspiration at high Resolution and with Internalized Calibration (METRIC), and Surface Energy Balance System (SEBS). Data used in this study was collected as part of a multi-disciplinary and multi-institutional field campaign BEAREX (Bushland Evapotranspiration and Agricultural Remote Sensing Experiment) that was conducted during 2007 and 2008 summer cropping seasons at the USDA-ARS Conservation and Production Research Laboratory (CPRL) in Bushland, Texas. Seventeen high resolution remote sensing images taken from multispectral sensors onboard aircraft and field measurements of the agro-meteorological variables from the campaign were used for model evaluation and improvement. Overall relative error measured in terms of mean absolute percent difference (MAPD) for instantaneous ET (mm h[superscript]-[superscript]1) were 22.7%, 23.2%, and 12.6% for SEBAL, METRIC, and SEBS, respectively. SEBAL and METRIC performances for irrigated fields representing higher ET with limited or no water stress and complete ground cover surfaces were markedly better than that for dryland fields representing lesser ET and greater soil water deficits with sparser vegetation cover. SEBS algorithm performed equally well for both irrigated and dryland conditions but required accurate air temperature data. Overall, this study provides new insights into the performance of three widely used thermal remote sensing based algorithms for estimating ET and proposed modifications to improve the accuracy of estimated ET for efficient management of water resources.