Downscaling the GOES ABI Data in Support of High-Resolution Wildfire Mapping

Due to the limited spatial or temporal resolution, current satellite-based products from VIIRS (Visible In- frared Imaging Radiometer Suite), MODIS (Moderate Resolution Imaging Spectroradiometer), and GOES ABI (Geostationary Operational Environmental Satellite Advanced Baseline Imager) are often inadequate for real-time applications such as wildfire detection. For example, although the VIIRS data exhibit superior spatial resolution, they suffer from infrequent revisits and limited scan swath. The GOES-R series provide more regular and frequent revisits, resulting in improved coverage and temporal resolution. But the 2-km spatial resolution of the infrared channels on the GOES-R series is insufficient to provide detailed information of the atmospheric conditions. The overarching goal of this research is to employ deep learning to create high spatial and temporal resolution fire products needed by fire managers for real-time wildfire monitoring and perimeter mapping. In particular, an enhanced super-resolution generative-adversarial network is devised and implemented to enhance the 2-km resolution of the GOES ABI data (channels 7 and14 in particular) to a higher resolution at 375-m. To achieve this, the concurrent overlapping higher spatial resolution VIIRS data from Suomi NPP, NOAA-20, and NOAA-21 will be used as references in training the super-resolution model for downscaling GOES ABI data. The results show that the super-resolution GOES/ABI data can provide more detailed spatial information than the original imagery and maintain its better temporal resolution at the same time. The outcomes of this approach can be further utilized to enhance the spatial resolution of operational application products derived from the ABI measurements, e.g., the active fire detection product.