Jump to Main Content
Evaluation of two end-member-based models for regional land surface evapotranspiration estimation from MODIS data
- Tang, Ronglin, Li, Zhao-Liang
- Agricultural and forest meteorology 2015 v.202 pp. 69-82
- algorithms, climate change, energy balance, evapotranspiration, heat transfer, meteorology, models, moderate resolution imaging spectroradiometer, momentum, roughness, soil water, soil water content, spatial data, surface temperature, uncertainty, vegetation cover, water management, China
- The importance of accurate estimations of regional and global evapotranspiration (ET) has been recognized worldwide in the fields of hydrology, water resource management, meteorology, and global climate change. Given the different structures of the end-member-based surface energy balance algorithm for land (SEBAL) and surface temperature–vegetation index (Ts–VI) triangle models, which nonetheless use essentially the same definitions of dry and wet pixels, questions arise regarding the nature of the differences between the two models’ estimated sensible heat fluxes and latent heat fluxes and what controls the heat flux differences. This study aims to investigate how the SEBAL model and the Ts–VI triangle method differ from each other in the regional evaporative fraction (EF) and ET estimates through analytical deduction and model applications. Because both the SEBAL model and the Ts–VI triangle model are of limited use when deep-layer soil moisture is moderately to significantly stressed, MODIS remote sensing data from 23 clear-sky overpass times between January 2010 and late October 2011, covering a wide range of soil moisture content and fractional vegetation cover conditions, are acquired to assess the two models over a non-water stressed study area on the North China plain. The results show that the SEBAL model is able to produce satisfactory sensible heat flux (H) and latent heat flux (LE) estimates compared with ground-based large aperture scintillometer measurements at the Yucheng station, with small biases of 4.1W/m2 and 2.3W/m2 and root mean square errors (RMSEs) of 46.4W/m2 and 48.6W/m2, respectively. However, the Ts–VI triangle model produces much worse H and LE estimates, with biases of 98.5W/m2 and −92.2W/m2 and RMSEs of 119.3W/m2 and 115.5W/m2, respectively. Variations of pixel-by-pixel surface available energy and momentum roughness length over the study area are responsible for the differences of the H and LE estimates between the two models. The SEBAL model produces larger EF and ET for most pixels than the Ts–VI triangle method when the same group of dry and wet pixels is applied, especially for pixels characterized by medium to high vegetation fractions or surface soil moisture contents. The Ts–VI triangle method is more sensitive to the surface temperatures of the dry and wet pixels than the SEBAL model. The findings from this study benefit the proper selection of end-member-based models for regional ET estimation and aid in quantifying the resultant uncertainties in the model-derived surface energy components.