Impacts of the broadband albedo on actual evapotranspiration estimated by S-SEBI model over an agricultural area

Abstract

Surface albedo and emissivity are essential variables in surface energy balance. In recent decades, several land surface energy models have used both surface broadband albedo and emissivity in order to achieve reliable evapotranspiration retrievals on a daily basis. Despite these improvements in surface energymodels, we noticed an assumption that most studiesmakewhen using this framework. It assumes that the surface broadband albedo and emissivity can be estimated directly as aweighted average of spectral surface bi-directional reflectances, and as aweighted average of spectral surface emissivities retrieved at a given view angle, respectively. However, this approach does not take into account surface anisotropy,which is described by the Bi-directional Reflectance Distribution Function (BRDF) in the case of the surface albedo. In this paper,we analyze the influence that estimating land surface albedo directly from the surface reflectance (αREF) or through the BRDF integration (αBRDF) has on the estimation of energy balance components (net radiation, latent and sensible heat fluxes and evapotranspiration) by using the Simplified Surface Energy Balance Index (S-SEBI). To this end, in-situ data and remote sensing images acquisitioned at different view zenith angles (VZA) such as 0°, ±40° and ±57° by the Airborne Hyperspectral Scanner (AHS) over an agricultural areawere used. Results showhigh variation in αREF depending on the VZA when compared to αBRDF, with the highest difference observed in the backward scattering direction along the hot spot region (RMSE of 0.11 and relative error of 65%). Net radiation gives relative errors from 6 to 17%,with themaximumerror obtained in the images that include the hot spot effect,whereas significant changes are not observed in case of the ground heat flux and the evaporative fraction. However, sensible heat flux, latent heat flux and daily evapotranspiration show relative errors ranging between 23–39%, 6–18% and 5–15% respectively. In a future study, the influence of estimating surface emissivity directly from the average of spectral emissivities under a given view angle or using a hemispherical value will be analyzed.