Researchers at UNIST developed an AI tool called UNIQUE that combines Landsat and MODIS satellite data to estimate daily plant carbon uptake at 30-meter resolution. It uses two satellite-specific models plus a generative AI fusion step, U-SCALER, to fill gaps caused by clouds and produce continuous maps of gross primary production. The approach was tested in South Korea and Wisconsin, where it picked up differences between forest types and mixed farmland that coarser 500-meter data missed. The team says the method cuts daily estimation errors by 13 to 28 percent versus existing global datasets, and the merged map beat the ESTARFM fusion technique by about 8 percent when checked against 30 ground stations. UNIQUE is now being applied to the land carbon uptake section of the Korea Carbon Project. That makes it a practical tool for forest and farmland monitoring, climate damage assessment, and carbon neutrality planning.
