Downscaling the GOES-R Land Surface Temperature

The GOES-R series, a constellation of geostationary satellites operated by NOAA, provides critical environmental data, including Land Surface Temperature (LST). The LST product is derived from the Advanced Baseline Imager (ABI) instrument onboard these satellites. Its frequent observations allow for monitoring of rapid LST changes every hour. It provides comprehensive LST data across the Americas. The LST product supports various applications in meteorology, hydrology, agriculture, and environmental monitoring.

Despite its rich temporal resolution, the coarse spatial resolution of GOES-R LST limits its application in certain domains. Challenges arise in accurately capturing temperature variations within urban areas, small water bodies, and complex terrain. Mixed pixel effects, where a single pixel encompasses multiple land cover types with varying temperatures, further complicate analysis. Coastal and mountainous regions, characterized by rapid temperature gradients, also pose challenges for the GOES-R LST product. To fully leverage the valuable temporal information provided by GOES-R, downscaling techniques are essential for addressing these spatial limitations and expanding the product's applicability to a wider range of applications.

A novel downscaling technique is proposed to enhance the spatial resolution of GOES-R LST data. This method leverages the complementary strengths of Landsat's high spatial resolution and GOES-R's rich temporal coverage. By utilizing five spectrally matched bands between the two sensors, a downscaling model is developed to generate high-resolution LST estimates. The high-spatial-resolution information is obtained by comparing Landsat's high-resolution pixels with the corresponding coarser resolution aggregated to GOES-R pixel size. This model minimizes the difference between averaged Landsat observations and corresponding GOES-R pixels, incorporating in-situ measurements for improved accuracy. The resulting downscaled LST product achieves an approximate spatial resolution similar to Landsat's at nadir.