An evapotranspiration product for arid regions based on the three-temperature model and thermal remote sensing

Abstract

An accurate estimation of evapotranspiration (ET) is crucial to better understand the water budget and improve related studies. Satellite remote sensing provides an unprecedented opportunity to map the spatiotemporal distribution of ET. However, ET values from barren or sparsely vegetated areas in arid regions are often assumed to be zero in typical ET products because of their low values. In addition, separating ET into soil evaporation (E-s) and vegetation transpiration (E-c) is difficult. To address these challenges, we developed an ET product (MOD3T) based on a three-temperature model and thermal remote sensing, specifically Moderate Resolution Imaging Spectroradiometer (MODIS) data. MOD3T has a spatial resolution of 1 km and a temporal resolution of 8 days. All input parameters except air temperature were obtained from MODIS datasets. Validation in two adjacent arid river basins in northwestern China showed that the mean absolute errors (mean absolute percent errors) between the MOD3T and flux tower ET were 0.71 mm d(-1) (18.5%) and 0.16 mm d(-1) (24.9%) for a densely vegetated area and sparsely vegetated sandy desert, respectively. The error between the MOD3T and water balance ET was 24 mm y(-1) (8.1%). The E-c/ET or E-s/ET of MOD3T was comparable to the observed stable oxygen and hydrogen isotopes. Unlike the MODIS ET (MOD16), MOD3T could not provide continuous ET values (as 70% of the MOD16 area lacked data) but exhibited relatively low uncertainty, particularly in cold seasons. Therefore, MOD3T can provide ET, E-s and E-c estimates for arid regions within acceptable ranges. (C) 2015 Elsevier B.V. All rights reserved.