An improved remote sensing based approach for predicting actual Evapotranspiration by integrating LiDAR

Abstract

Spatially distributed information of actual Evapotranspiration (ET) is essentially required for various applications in agriculture, as well as studies related to water balance. An Energy Balance model that uses optical and thermal remote sensing datasets is appropriate for the spatially distributed estimation of ET, which is primarily controlled by local environmental conditions. In this study, we use the Surface Energy Balance System (SEBS), which is an Energy Balance model, to estimate an improved actual ET, by integrating canopy height information obtained from ICESat/GLAS, the first space-borne LiDAR satellite. Our model estimates latent heat flux (LE) as a residual of energy balance, while sensible heat flux (H) is estimated based on Monin–Obukhov similarity theory. The spatial variability of H and LE estimated from Landsat TM/ETM+ images were compared with eddy covariance (EC) based in situ flux tower observations, in order to document the SEBS model uncertainties in three forest sites in the East and Southeast Asian ecosystem. We identified that canopy height (hc) is one of the critical parameters that can induce uncertainties in SEBS model estimations. Therefore, hc obtained from LiDAR was introduced in parameterization of the roughness length for momentum transfer (z0m). In situ based hc obtained from flux tower stations, and the normalized difference vegetation index based formulation of hc proposed by Su (2001), were also introduced in the SEBS z0m parameterization. Root mean square error of LiDAR integrated SEBS (SEBSLiDar) for estimated H values reduced by (20, 45, and 23)% compared to Landsat integrated SEBS (SEBSLandsat), and (18, 7, and 9)% compared to In situ integrated SEBS (SEBSInsitu) at the three selected forest sites (SMC, QYZ, and SMF), respectively, while the mean coefficient of determination values for the estimated LE improved by (9.09 and 5.88)% compared to SEBSLandsat and SEBSInsitu, respectively, with reference to ground-based flux tower observations. Our analysis reveals that LiDAR-based parameterization of z0m can significantly increase the accuracy of estimated roughness lengths within an ET model, which can therefore improve the accuracy of the estimated turbulent heat fluxes. SEBSLiDAR was further evaluated for the spatially distributed mapping of actual ET. We have observed that ET estimated by SEBSLandsat overestimated in forest sites compared to EC observations; however, LiDAR-integrated SEBS marginally improved the accuracy of ET, thereby leading to an improvement in the mean correlation value of 0.87 and mean normalized standard deviation value of 0.93 approaching unity based on Taylor diagram, which more closely approximates the actual ground-based measurements. The proposed methodology overcomes the current limitation in the energy balance models in estimating roughness lengths, by providing accurate formulation of hc by specifically incorporating LiDAR measurements in areas of tall vegetation canopies. The operational improvement in roughness lengths by next-generation SEBSLiDAR model provides an opportunity to improve the estimation accuracy of water and energy fluxes, specifically ET. Overall, our proposed SEBSLiDAR model provides a benchmark that is useful to policy makers and water resource managers in devising a plan for sustainable use of water resources in the East Asian region, as well as other similar ecosystems.