A novel scheme for seamless global mapping of daily mean air temperature (SGM_DMAT) at 1-km spatial resolution using satellite and auxiliary data

The daily mean air temperature (DMAT) is an essential descriptor of climate change. Seamless global DMAT maps will significantly improve our knowledge of terrestrial meteorological and climatic conditions. This study proposes a novel scheme, Seamless Global Mapping of Daily Mean Air Temperatures (SGM_DMAT). The SGM_DMAT scheme comprises two key phases: Estimating DMAT under clear-sky conditions, and reconstructing missing values under cloudy conditions using data from 2020 to 2022 as the calibration dataset and data in 2023 as the validation dataset. The results demonstrate that combining all valid Moderate Resolution Imaging Spectroradiometer (MODIS) TERRA/AQUA daytime and nighttime land surface temperature (LST) observations under clear-sky conditions, and applying spatial temporal analysis techniques with reference images for cloudy days, ensures robust and seamless DMAT estimation. Specifically, the Extreme Gradient Boosting (XGBoost) was selected as the optimal model of DMAT estimation. The optimal feature dataset includes satellite-derived LSTs, latitude, longitude, elevation above sea level, month, and day of year. The optimal calibration dataset comprises all valid calibration data (AVCD). Additionally, the priority order of DMAT clear-sky estimation models was established using different LST combinations. Finally, robust and seamless global maps of DMAT were generated for the period 2020–2023. For globally seamless mapping products, the R2 was 0.956, with an RMSE of 2.825 °C and a MAE of 1.985 °C. The proposed SGM_DMAT scheme may aid DMAT estimation in regions that lack sufficient meteorological stations. The seamless global DMAT products have broad applicability including in trend analysis, urban heat island research, and assessment of crop stress due to temperature extremes.