Estimation of all-sky 1 km land surface temperature over the conterminous United States

Land surface temperature (LST) is a crucial parameter for hydrology, climate monitoring, and ecological environmental research. LST products from thermal infrared (TIR) satellite data have been widely used that. However, TIR information cannot provide under cloudy-sky conditions. All-sky can be estimated microwave measurements, but their coarse spatial resolution, narrow swaths, short temporal range make it impossible to generate long-term, high-resolution, accurate global all-sky global. This study proposes methodology generating the product by combining multiple Moderate Resolution Imaging Spectroradiometer (MODIS), reanalysis, ground in situ measurements using random forest. Field AmeriFlux Surface Radiation Budget (SURFRAD) networks were model training validation. Cloudy-sky clear-sky models developed separately. To further improve accuracy of model, conventional RF was extended incorporate information. The validated 2010, 2011, 2017 that not training. For models, root-mean-square-error (RMSE) = 2.767 2.756 K, R2 0.943 0.963, bias −0.143 − 0.138 respectively. same validation samples validate both MODIS conditions Global Data Assimilation System (GLDAS) at 0.25° with RMSE 3.033 4.157 −0.362 0.224 0.904 0.955, Additionally, 10-fold cross-validation results all datasets indicate stability. applied 2000 2015 over conterminous United States (CONUS). Our shows similar patterns GLDAS products, more accurate. Both comparisons demonstrated robustness our proposed product.