A framework of correcting the angular effect of land surface temperature on evapotranspiration estimation in single-source energy balance models

Abstract

Evapotranspiration (ET) is an essential parameter connecting the hydrological cycle, energy balance and carbon cycle. The angular effect (AE) of remotely sensed land surface temperature (LST), one of the obstacles hindering the application of single-source energy balance models in continental and global ET estimates, is a long-standing issue that has rarely been witnessed in the last decade. In this study, we propose a general framework for correcting the angular effect that could be practically applied to reduce the uncertainties in the ET estimations with any single-source energy balance-based model, under the guidance of simulation of thermal radiation directionality through the integration of a directional module from the soil-canopy spectral radiances, photosynthesis, fluorescence, temperature, and energy balance (SCOPE) and an artificially added ET estimation module. The correction was conducted by adjusting the directional reflectance and then the reflectance-based fractional vegetation coverage to the view zenith angle of interest, together with estimating the soil and vegetation component temperatures. This framework was later tested on the widely applied surface energy balance system (SEBS) driven by the directional LST from the Moderate Resolution Imaging Spectroradiometer (MODIS) collected on 94 clear-sky days between 2009 and 2010 at the Yucheng site. The simulation results indicated that (1) the AE obviously increased as the view zenith angle increased; (2) the AE was more apparent in the backward scattering direction than in the forward scattering direction; and (3) most of the AE first increased evidently and then decreased with the increase in the leaf area index. The correction results related to the SEBS-based ET estimations, compared to the in situ ET measurements, showed that the proposed framework greatly improved the estimations of ET, with the root mean square error decreasing by 27.6 W/m2 (∼8%, relative to the average 339 W/m2) from 86.8 W/m2 (∼25%) to 59.2 W/m2 (∼17%) and the bias decreasing from 49.7 W/m2 (∼15%) to −6.5 W/m2 (∼2%). In conclusion, the proposed framework efficiently corrected the angular effect of remotely sensed LST on ET estimation. We expect more attention to be paid to the angular effect and hope our work could provide insight for solving this pendent issue.