Downscaling of GOES-16\u27s Land Surface Temperature Product Using Epitomes

Land surface temperature (LST) is an environmental variable derived from thermal infrared (TIR) imagery. Satellite platforms are a good source of TIR imagery because of their ability to provide widespread and frequent coverage of the Earthâ??s surface. It is common that a single satellite remote sensing platform is able to provide images with good spatial resolution or temporal resolution but not both. LST is an important parameter for studies on the urban heat island (UHI) effect. These studies are limited by the spatial or temporal resolutions of available LST products. This Thesis presents an algorithm to estimate land surface temperature with high spatial and high temporal resolutions by downscaling GOES16â??s LST product. This is done by extending an epitomic representations methodology, previously used to create high-resolution land cover maps, to land surface temperature data. An LST product with high spatial and high temporal resolutions will benefit UHI studies as well as any users of LST data. The accuracy and precision of our downscaled LST products are in line with the targets set by the GOES-16 satelliteâ??s LST product of 2.5 K for accuracy and 2.3 K for precision