Advancing point-to-grid scale ET mapping: comparative assessment of ECOSTRESS and MODIS LST and ET across heterogeneous landscapes

Acquisition of high-resolution land surface temperature (LST) and evapotranspiration (ET) data is critical for accurately monitoring spatio-temporal dynamics over complex terrain. This study aims to evaluate the National Aeronautics and Space Administration ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) LST and ET estimates over South Korea using observations from 13 flux towers. The uncertainty of the ECOSTRESS products was evaluated under various cloud conditions (e.g. cloudy- and clear-sky), seasonal variations (e.g. spring, summer, and autumn), land cover types (e.g. cropland and forest), and diurnal cycles. Overall, ECOSTRESS LST exhibited good agreement with surface soil temperature data under clear-sky conditions (Bias: −0.35 to 2.44°C). However, under cloudy-sky conditions, large uncertainties were observed (Bias: −18.42 to −9.18°C). ECOSTRESS LST also showed good agreement for croplands during autumn. Statistical analysis of ECOSTRESS ET revealed an overestimation (Bias: 3.33 to 9.00 mm day−1) under clear-sky conditions. The statistical performance over different land cover types indicated that topographical complexity and a lack of irrigation modeling contributed to increased uncertainty. Among the different seasons, spring yielded the lowest ET Bias, with −0.18 and 6.34 mm day−1 for the Yeoju and Paju stations, respectively. Subsequent upscaling of the flux tower observations to the grid scale confirmed that ECOSTRESS ET effectively captured the spatial heterogeneity. These findings demonstrate the strong potential of ECOSTRESS LST and ET estimates at both point and grid scales for efficient water resource management.