Nationwide 30 m maps of forest composition, biomass, and diversity in South Korea (2021–2025) from direct prediction and plot-index imputation

This dataset provides nationwide 30 m maps of four forest attributes in South Korea: coniferous fraction, dominant tree species, aboveground biomass, and tree-species diversity. The maps were developed using data from the eighth Korean National Forest Inventory (2021–2025) and predictors derived from Sentinel-2 imagery, climate data, and topography. Two mapping approaches are provided for each attribute. The direct-prediction products were generated using a separate random forest model for each attribute and include a cell-level relative uncertainty layer. The plot-index imputation products assign one matched inventory plot to each forest cell. The attributes assigned to a cell were therefore observed together at the same field plot. These products also include a layer describing the distance to the matched reference plot. Both approaches are available with two predictor sets. The base set consists of satellite, climate, and topographic variables. The FTM set additionally includes the national forest-type map. The same products are published as a public Google Earth Engine image collection with a browser viewer, so they can be examined before or instead of downloading them: https://ee-korea-forest-maps.projects.earthengine.app/view/korea-forest-maps (collection projects/ee-korea-forest-maps/assets/korea-forest-attributes). The analysis code is at https://github.com/kim-seunguk/korea-forest-attribute-maps. All raster products have a spatial resolution of 30 m and use EPSG:5179. Model performance was evaluated using 50 km spatial-block cross-validation and a spatially disjoint test set. Attribute-specific accuracy estimates and limitations are described in README.md. File contents and field definitions are listed in data_records.csv. Inventory plot coordinates are not included. The lookup tables retain the official inventory plot numbers, which can be used to obtain additional attributes from the public National Forest Inventory database.