Data for project: Interpretable prediction of topsoil organic carbon content based on machine learning, deep learning and its hybrid geostatistical algorithms
Description
This dataset contains the processed data supporting the paper “Interpretable prediction of topsoil organic carbon content based on machine learning, deep learning, and its hybrid geostatistical algorithms”. It includes 317 soil sampling locations in California, USA, together with soil organic carbon (SOC) observations and remote-sensing, terrain, climate, soil, and land-cover variables. SOC observations were obtained from the USDA-NRCS Rapid Carbon Assessment (RaCA) dataset. Remote-sensing variables were derived from USGS Landsat 7 ETM+ Collection 2 Tier 1 Level-2 imagery in Google Earth Engine. Terrain variables were mainly calculated from the NASA/USGS Shuttle Radar Topography Mission digital elevation model, with MERIT Hydro data used for the wetness-related variable. Climate variables were obtained from the gridMET dataset through the Northwest Knowledge Network METDATA archive (https://www.northwestknowledge.net/metdata/data/). Numerical soil variables were obtained from SoilGrids 2.0 (https://soilgrids.org/). Soil taxonomic suborder and parent-material kind were obtained from USDA-NRCS gSSURGO (https://nrcs.app.box.com/v/soils/folder/243464012908), and land-cover classes were obtained from the 2011 National Land Cover Database through the MRLC viewer (https://www.mrlc.gov/viewer/). The definitions of the variables are as follows: Site_ID: Unique soil sampling-site identifier; Longitude/Latitude: Geographic coordinates in decimal degrees; SOC_0_30_Weighted_gkg: Depth-weighted SOC content for the 0–30 cm layer (g/kg); SR_B1–SR_B7: Landsat 7 surface-reflectance bands; LST: Land surface temperature (°C); NDVI/EVI/SAVI/MSAVI/RVI/DVI/GNDVI: Vegetation indices; BSI: Bare Soil Index; NDWI/MNDWI: Moisture- and water-related spectral indices; TCB/TCG/TCW: Tasseled Cap brightness, greenness, and wetness; NIR_Contrast/NIR_Variance/NIR_Entropy: Near-infrared texture metrics; Elevation: Elevation above sea level (m); Slope/Aspect: Terrain slope and aspect; Northness/Eastness: Cosine- and sine-transformed aspect; TPI/mTPI: Local and multi-scale topographic position indices; TWI: Wetness-related topographic index; TRI/Roughness/VRM: Terrain variation, roughness, and ruggedness metrics; Curvature: Local terrain-curvature proxy; tmmx/tmmn: Mean maximum and minimum air temperature; pr/MAT: Mean annual precipitation and temperature; RH/SRAD/VS: Relative humidity, solar radiation, and wind speed; ETR/VPD/BI: Reference evapotranspiration, vapor pressure deficit, and burning index; clay/sand/silt: Soil particle-size fractions; nitrogen: Total soil nitrogen content; phh2o: Soil pH in water; cfvo: Coarse-fragment content; cec: Cation-exchange capacity; bdod: Bulk density of the fine-earth fraction; taxsuborder: USDA Soil Taxonomy suborder; NLCD_LandCover:2011NLCD land-cover class; pmkind: Representative soil parent-material kind.