Continental-scale evaluation of three ECOSTRESS land surface temperature products over Europe and Africa: Temperature-based validation and cross-satellite comparison
Hu T., Mallick K., Hulley G.C., Planells L.P., Göttsche F.M., Schlerf M., Hitzelberger P., Didry Y., Szantoi Z., Alonso I., Sobrino J.A., Skoković D., Roujean J.L., Boulet G., Gamet P., Hook S.
Remote Sensing of Environment, vol. 282, art. no. 113296, 2022
High spatial resolution land surface temperature (LST, <100 m) is crucial for agricultural water management, crop water stress monitoring, fire mapping, urban heat island study and volcano eruption detection. LST retrievals from the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) launched in June 2018, together with the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER, launched in 1999) and the Landsat series (since 1972), comprise the state-of-the-art high spatial resolution LST datasets publicly accessible. Recently, we generated the ECOSTRESS LST product over Europe and Africa using both the temperature and emissivity separation (TES) and split-window (SW) algorithms under the European ECOSTRESS Hub (EEH). Here, we validated the official Jet Propulsion Laboratory (JPL) TES (Collection 1), EEH TES and EEH SW ECOSTRESS LST products over Europe and Africa between August 1, 2018 and December 31, 2021 by comparing against the in-situ measurements at 9 sites over a wide variety of land cover types. Meanwhile, the validation results were compared with those obtained for ASTER and Landsat LST at the same sites for a thorough understanding of the consistency among these high spatial resolution LST products. The results reveal that the three ECOSTRESS LST products have consistent performances, with an overall RMSE around 2 K. A cold bias around 1 K exists for all three ECOSTRESS LST, which is presumably originated from the radiometric calibration of the sensor in Collection 1 data. The Landsat LST shows a similar accuracy, with an RMSE of 2.20 K and bias of 0.54 K. The EEHSW LST show the highest consistency with Landsat LST, possibly due to the identical emissivity correction process. The performance of ASTER LST is also similar, with an RMSE of 1.98 K and bias of 0.9 K. The precisions of all the LST products are around 1.5 K. Future recalibration of the ECOSTRESS Level 1 radiance data in Collection 2 is expected to further improve the accuracy of ECOSTRESS LST. Overall, this study supports the adaptation of LST retrieval algorithms for the future thermal missions.