Identification, Characterization and Correction of Directional Anisotropy in High-Resolution Land Surface Temperature through a Comparison of Sharpened Sentinel-3 with ECOSTRESS Data.

This study identifies, characterizes and corrects for directional effects in high-resolution (HR) Land Surface Temperature (LST) maps based on a comparison of sharpened Sentinel-3 (S3) and ECOSTRESS data at 70-meter resolution. Firstly, the two HR LST datasets were assembled. A Data Mining Sharpener (DMS) derived HR LST to overcome the lack of HR data. This technique sharpened 1 kilometer spatial resolution LST from the S3 Sea and Land Surface Temperature Radiometer (SLSTR), for the Flanders region from 2019 to 2024. The DMS uses a bagging ensemble, that incorporates Sentinel-2 (S2) Multi-Spectral Instrument (MSI) reflectance bands and Copernicus Digital Elevation Model (DEM) data, to enhance the spatial resolution of S3 SLSTR LST. Quasi-simultaneous ECOSTRESS LST acquisitions, for the same region, were acquired to serve as a secondary data source. Secondly, the combined dataset, containing multi-angular observation pairs, was used for analysing directional effects. After cross-calibrating the observation pairs, directional effects are identified through pixel-by-pixel comparisons. A Vinnikov-Roujean-Lagouarde (RL) parametric model simulated directional effects up to 1.6 K. Additionally, case-specific parametric models have been established, characterizing specific morning, night, seasonal, land cover and vegetation density effects. A validation study demonstrates that the proposed correction methodology decreases LST differences of the observation pairs, with tailored strategies for specific conditions yielding the best results. Our study underscores the importance of considering directional effects in HR LST retrievals and provides a global correction framework for upcoming HR thermal missions, including LSTM, TRISHNA and SBG.