Answers are grounded in this paper and cite their sources.
Near-surface air temperature (NSAT) is needed at both high spatial and high temporal resolution, but no single satellite sensor provides both. This study develops a clear-sky framework that estimates NSAT in near real time on a 30 m grid at 30-min intervals by fusing Geostationary Korea Multi-Purpose Satellite-2A (GK-2A) and Landsat-8 observations. A Taylor expansion-based downscaling algorithm was adapted to update its parameters at every GK-2A scan, and a thermodynamics-based formulation was simplified into an empirical regression of NSAT on land surface temperature (LST), longwave radiation and a moisture proxy derived from the normalized difference moisture index. Its three coefficients are recalibrated every hour against automatic weather station observations. Over southeastern South Korea in 2023, the root-mean-square error (RMSE) was 2.38 °C at stations withheld from calibration and 2.22 °C at the calibration stations, and it increased to 2.82 °C with previous-day coefficients. The downscaling reduced LST errors against Landsat-8 in 14 of 18 scenes but did not measurably change the station-scale NSAT accuracy. The 30 m grid therefore represents surface-type contrasts within each 2 km GK-2A pixel rather than independent thermal information at 30 m, and the product is valid only under clear-sky conditions with real-time AWS data available for hourly recalibration. Each half-hourly field was produced in about 30 s, supporting near-real-time operation.
No takes yet. Share an insight, caveat, or question.
Shin et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: