Accuracy comparison of remotely sensed evapotranspiration products and their associated water stress footprints under different land cover types in Korean peninsula

Abstract

Robust spatial information of evapotranspiration from multiple land cover types is deemed critical for several applications in agriculture and water balance studies. Energy balance models, used in association with satellite observations, are beneficial to map spatial variability of evapotranspiration which is mainly governed by different vegetation practices and local environmental conditions. This study utilize the Surface Energy Balance System model to estimate actual evapotranspiration and water scarcity footprints under complex landscape of Korean peninsula using Moderate-Resolution Imaging Spectroradiometer satellite data for a complete hydrological year of 2012. The modeled evapotranspiration was compared with flux tower measurements obtained from a subhumid cropland and temperate forest sites for the accuracy assessment. This accuracy comparison at daily scale had good agreement yielding reasonable coefficient of determination (0.72, 0.51), bias (0.41 mm day−1, 1.01 mm day−1) and root mean squared error (0.92 mm day−1, 1.53 mm day−1) at two observation (cropland, forest) sites, respectively. Furthermore, the monthly aggregated evapotranspiration from Surface Energy Balance System showed promising results than those of obtained from Moderate-Resolution Imaging Spectroradiometer based readymade global evapotranspiration product, i.e., MOD16, when both products were compared with unclosed and closed flux tower measurements. However, the variations in monthly evapotranspiration obtained from both products were significantly controlled by several climate factors and vegetation characteristics. Water stress mapping at regional and monthly scale also revealed strong contrast between the products of two approaches. Highest mean water stress (0.74) was observed for land use areas associated with evergreen forest and under sparsely vegetation condition by using estimated evapotranspiration from Surface Energy Balance System while an extreme mean water stress value of 0.56 by using end product of MOD16 evapotranspiration was raised from cropland regions. Overall, this study revealed the performance and suitability of two distinctive remote sensing approaches for characterizing the footprints of water fluxes in the Korean peninsula and provides a baseline for the policy makers to setup the sustainable use of existing water resources in this and other similar regions.