Soil organic carbon stocks and trends (1984-2019) predicted at 30m spatial resolution for topsoil in natural areas of South Africa

Link to scientific publication: https://doi.org/10.1016/j.scitotenv.2021.145384

Soil organic carbon (SOC) stocks (kg C m-2) are predicted over natural areas (excluding water, urban, and cultivated) of South Africa using a machine learning workflow driven by optical satellite data and other ancillary climatic, morphometric and biological covariates. The temporal scope covers 1984-2019. The spatial scope covers 0-30cm topsoil in South Africa natural land area (84% of the country). See methodology in linked publication for details. Data are provided here at 30m spatial resolution in GeoTIFF files. There is a dataset for the long-term average SOC and trend in SOC. Each dataset is split into four files (suffix *_1, *_2 etc.) covering separate regions of South Africa for ease of download. The raster files are:

"SOC_mean_30m..." - average of annual SOC predictions between 1984 and 2019. Values are expressed in kg C m-2

"SOC_trend_30m..." - long-term trend in SOC derived from the Sens slope (M) across annual SOC values between 1984 and 2019. Pixel values (Y) are expressed as a percentage change over the 35 years relative to the long-term mean (X). Y = M / X * 100 * 35 years

NB: All files are scaled by *100 and converted to floating data point to save space. To back-convert to original values, simply divide the raster values by 100.