Global 30 m resolution distribution of organic soils (Histosols) based on soil taxonomy and peat depth training points

This repository provides predictions of global organic soil probability and peat depth (in cm) both produced using Random Forest classification/regression models. The organic soil model was trained using global soil point observations with soil types harmonized to the USDA soil classification system. These observations were converted into a binary training dataset representing organic and non-organic soil samples. Layer description: organic.soils_rf = predicted probability of organic soils (Histosols) based on the USDA Soil Taxonomy and about 447k training points (values as 0–100%: pixel values are stored as integers from 0 to 100, with a scale factor of 0.01 encoded in the Cloud Optimized GeoTIFF; a value of 255 represents NoData); peatland.depth_rf = predicted peat depth (i.e. lower boundary of the peat layer in cm) based on the cca 350k training points (peatland_pnts_global_xyt_vx.csv.gz) published in Peat-DBase v.1 and multiple national data sources; Models and predictions are based on using an extensive list of covariate layers including: peatland extent, global indundated (wetlands) extent, MODIS day-time and night-time images, CHELSA climate and bioclimatic variables, Digital Terrain variables and Landsat-derived long-term indices. From the two independent predictions, we produced an ensemble estimate showing distribution of organic soils based on the most up-to-date global data sets. For this, the peat depth values were converted to probabilities using sigmoid transformation (soils with peat depth / thickness of >40 cm can be considered Histosols); the two probabilities are then combined using an average to produce an ensemble estimate of probability of occurrence of organic soils. The 30 m spatial resolution data is available from the S3 links (note: file size is ~60GB per file): https://s3.opengeohub.org/global/soil.types/organic.soils_rf.organic_p_30m_s_20000101_20241231_go_epsg.4326_v20260506.tif https://s3.opengeohub.org/global/peatland/peatland.depth_rf_m_30m_s_20000101_20241231_go_epsg.4326_v20260513.tif https://s3.opengeohub.org/global/soil.types/organic.soils_ensemble.organic_p_30m_s_20000101_20241231_go_epsg.4326_v20260513.tif A coarser 100 m version is provided in this repository for easier access and use. Note, some islands such as Mauritius, Reunion and islands in the Pacific ocean are currently missing and will be added in the next version. For reference, please cite: Glen et al. (2026). Global disturbance and emissions from organic soils: leveraging Earth observation-based geospatial data within an IPCC framework. Under review. Hengl et al. (2026). OpenLandMap-soildb: global soil information at 30 m spatial resolution for 2000–2022+ based on spatiotemporal Machine Learning and harmonized legacy soil samples and observations. Earth Syst. Sci. Data, 18, 989–1036, https://doi.org/10.5194/essd-18-989-2026