High-resolution processing of Himawari meteorological satellite data

SARNet, a deep-learning model, was used to generate high-resolution (>32×) NDVI, RGB, and LST images based on Himawari meteorological satellite data, targeting application in the agricultural field. High-quality high resolution was achieved using NDVI data. However, limited resolution was achieved using RGB and LST data, owing to the extreme upscaling. Geographic factors such as cloud coverage, water coverage, and climatic zone significantly affected prediction accuracy. Retraining using local data is necessary when the climatic conditions are highly variable. The proposed approach exhibits the potential for achieving high-resolution satellite-based images for agricultural applications. However, further improvements in resolution and accuracy will require careful consideration of environmental conditions as well as further enhancements.