Evaluating the performance of two SEB models for estimating ET based on satellite images in arid regions

Abstract

Surface energy balance (SEB) models are one of the most important methods for estimating evapotranspiration (ET). This study was aimed at assessing the applications of two SEB models under arid conditions. This was achieved by integrating in situ meteorological data and Landsat 8 satellite images for two consecutive years (2014 and 2015) for alfalfa cultivation in 2014 and maize cultivation in 2015. The performance of the two SEB models was evaluated by comparing the ET estimated by the SEB models and that measured using an Eddy covariance (EC) instrument. Mapping the ET at high resolution using internalized calibration (METRIC) displayed the best performance in 2014, with a coefficient of determination, R2 of 0.8, followed by the simplified surface energy balance index (S-SEBI) model (with R2 = 0.67). In addition, METRIC showed high efficiency, with a Nash Sutcliffe (NSE) model efficiency coefficient of 0.98, followed by S-SEBI, with an NSE of 0.93. Poor performance was observed in 2015 where the R2 for both METRIC and S-SEBI were 0.3 and 0.45, respectively. In addition, the NSE for METRIC decreased to 0.91. Conversely, the NSE for S-SEBI improved to 0.97. Therefore, the performance of the models was affected by instability due to soil water content (SWC) and low precipitation (62.2 mm in 2015). However, the performance of METRIC was better than that of S-SEBI. In general, METRIC is considered a good SEB model for arid regions. In addition, S-SEBI could be an adequate model for overcoming the difficulty associated with obtaining meteorological data.