Generating 60–100 m, hourly, all-weather land surface temperatures based on the Landsat, ECOSTRESS, and reanalysis temperature combination (LERC)

Satellite-derived land surface temperatures (LSTs) often encounter a tradeoff between spatial and temporal resolutions, as well as severe cloud contamination. While extensive efforts have focused on resolution enhancement and under-cloud reconstruction, generating fine-resolution (≤100 m) diurnal LSTs under all-weather conditions remains a challenge, which hampers fine-scale monitoring of climatological, hydrological, and ecological processes. The latest 70-m ECOSTRESS observations at varying times of day provide an unprecedented opportunity for detailed mapping of diurnal LST dynamics, and reanalysis products with spatiotemporal continuity offer promising references for all-weather thermal dynamics. However, these advantages have rarely been integrated to concurrently achieve high spatiotemporal resolution and completeness. Here, we present a simple yet effective framework for reconstructing 60–100 m, hourly, all-weather LSTs based on the Landsat, ECOSTRESS and Reanalysis temperature Combination (termed LERC). The framework involves three steps: (i) preliminary under-cloud estimations within annual cycles of several times by fitting an enhanced annual temperature cycle (EATC) model to clear Landsat/ECOSTRESS scenes and China Land Data Assimilation System (CLDAS) LST fluctuations; (ii) optimized daily estimations at each selected time by correcting biases of the preliminary under-cloud estimations and re-modeling the EATC with temporally densified samples; and (iii) hourly seamless estimations by interpolating the two nearest daily estimations with reference to the weighted diurnal changes in CLDAS. LERC was evaluated in an urban-dominated region throughout 2020, resulting in an average root-mean-squared-error of 2.0 K (3.0 K) against 50 Landsat and ECOSTRESS images (hourly ground measurements at 13 sites). Compared to the enhanced spatial and temporal adaptive reflectance fusion model, classic ATC model, diurnal temperature cycle model, and three sophisticated all-weather products, LERC demonstrates outperformance in terms of general accuracy, spatiotemporal variability, and robustness against sparse input. LERC has great potentials for generating long-term reliable all-weather LST records with a high spatiotemporal resolution to promote broad applications.