Dual-model approaches for evapotranspiration analyses over homo- and heterogeneous land surface conditions

• Application of remote sensing based evapotranspiration models, SEBS and RSPM.
• Performance assessment of the models depending on different spatial scale.
• Evaluation of the models over two biogeophysically distinct catchments in the Korean peninsula.
• Sensitivity analyses for SEBS model based on canopy height.
• Parameterization of roughness height over areas of tall and heterogeneous vegetation.

Accurate spatio-temporal estimation of evapotranspiration (ET) is fundamental to understand land–atmosphere interactions and hydrological processes. Although various ET models based on remote sensing (RS) techniques have been presented, recent relevant studies suggest that more validation work should be conducted in regions with differences in land and meteorological properties. In this study, two RS-based models, the Surface Energy Balance System (SEBS) and Revised Remote Sensing Penman–Monteith (R-RSPM) model, were assessed over two biogeophysically different catchments on the Korean peninsula: a rice paddy agricultural field and a mixed forest area. We independently applied three approaches considering different scaled applications of a single source energy budget model (SEBS): point scale SEBS (SEBS-P), regional scale SEBS (SEBS-R), and R-RSPM. We incorporated diverse compositions of forcing datasets, Moderate Resolution Imaging Spectroradiometer (MODIS) satellite-based RS data, and Global Land Data Assimilation System (GLDAS) data and in situ meteorological data. Instantaneous flux measurements computed by the three approaches were evaluated with in situ flux measurements based on the eddy covariance (EC) system. The instantaneous net radiation estimates of all approaches were satisfactory, exhibiting strong correlations (0.68–0.99 of R2) with in situ measurements over different topographical sites. For latent heat flux (LE), all three approaches yielded similar trends of results at both cropland (CMC) and mixed forest (SMC) sites indicating that estimates of LE were overestimated. However, considering the serious lack of energy closure exhibited by the EC system in SMC, LE showed reasonable agreement (biases of 11–50 W/m2 and Root Mean Square Error (RMSE) of 78–95 W/m2) with in situ measurements adjusted by the Bowen ratio (BR) method. For daily ET analyses using the R-RSPM model, instantaneous LE was empirically scaled up to daily ET, and the estimates of ET from all three approaches had good agreement with BR-corrected in situ daily ET, even in SMC (with biases of −0.07 to 0.67 mm/day). By comparing H from two different SEBS applications (SEBS-P and SEBS-R), we concluded that GLDAS datasets exhibit robust quality as input data as well as near real time benefits. Through sensitivity analyses of a single source energy budget model (SEBS), canopy height turned out to be one of most critical parameters which can directly result in the misparameterization of roughness height for a tall canopy site. The results of this study support the applicability of RS-based ET models over homo- and heterogeneous regions, and suggest that further studies are required for exploring the accurate parameterization of roughness height over areas of tall and heterogeneous vegetation to improve the performance of SEBS.