High Spatiotemporal Rugged Land Surface Temperature Downscaling over Saihanba Forest Park, China

Satellite-derived rugged land surface temperature (LST) is an important parameter indicating the status of Earth’s energy budget and its seasonal/temporal dynamic change. However, existing LST products from areas are more prone to error when supporting applications in mountainous Earth processes that occur at high spatial temporal resolutions. This research aimed develop a method for generating with resolution by using improved ensemble model combining three regressors, including random forest, ridge, support vector machine. Different combinations high-resolution input parameters were also considered this study. The datasets included Moderate Resolution Imaging Spectroradiometer (MODIS) (MxD11A1) nighttime, Sentinel-2 Multispectral Instrument (MSI) datasets, digital elevation (DEM) datasets. 30 m derived compared against situ dataset obtained Saihanba Forest Park (SFP) sites ASTER-derived 90 LST, respectively. results measurements demonstrated significant details, R2 higher than 0.95 RMSE around 3.00 K both Terra/MOD- Aqua/MYD-based slightly better being Terra/MOD. inter-comparison ASTER showed over 80% pixels difference image two within 2 K. In light complex topography distinct atmospheric conditions, these comparison encouraging. proposed study depicts seasonality surfaces.