A cloud-regulated land surface warming model to reconstruct daytime surface temperatures under cloudy conditions

Daytime land surface temperature (D-LST) plays a pivotal role in regulating net ecosystem exchanges and is characterized by rapid fluctuations. Thermal infrared satellite remote sensing (TIRS) is widely used to acquire D-LST data owing to its global coverage and high-frequency observations. However, the presence of cloud cover impedes the TIRS technique by obstructing ground thermal emissions. A prevalent solution to this challenge involves correcting clear-sky surface temperatures using the cloud effect which is derived from surface energy balance (SEB) models representing distinct weather conditions. Yet, conventional methods might encounter substantial uncertainties primarily due to the oversimplified SEB modeling, which exacerbates the temperature estimation errors caused by the biases in their employed data products. This study introduces a novel SEB model termed ‘C-SWARM’, designed to reconstruct D-LST under cloudy conditions. The C-SWARM model characterizes D-LST as the result of a cloud-moderated surface warming process, with coefficients indicating the complementary mechanism for solar heating and atmospheric insulation driving surface warming throughout the day. The new model was implemented to fill missing data caused by cloud cover in the LST product of NOAA's Geostationary Operational Environmental Satellite (GOES-R), demonstrating a mean absolute error of 2.57 K and accuracy improvements of 0.38 to 1.89 K over benchmark methods at 49 flux tower sites across the contiguous United States. The explicit physical mechanisms make the C-SWARM model a generalized solution for all-weather remote sensing across spatial and temporal scales.