Data supplement: Optimisation of sampling design for spatial soil mapping with machine learning

Data supplement for the paper: Optimisation of sampling design for spatial soil mapping with machine learning. This repository contains the core datasets needed to reproduce the analyses reported in the manuscript (AOA computation, k-means clustering, SOC prediction, and model evaluation). Contents Training datasets (point samples) original_training_data.gpkg: Original SOC training dataset (n = 904). Includes SOC observations and extracted predictor values. updated_training_data.gpkg: Updated SOC training dataset (n = 1004), comprising the original dataset (n = 904) plus additional field samples (n = 100) collected using the proposed sampling strategy. Raster datasets estonia_merged_raster.tif: National covariate stack used for AOA computation and SOC prediction. Predictor layers include: `clay`, `silt`, `sand`, `rock`, `slope`, `lsf`, `tri`, `twi`, `ndvi_spring`, `ndvi_summer`, `ndvi_autumn`, `bsi_spring`, `bsi_summer`, `bsi_autumn`, `drained`, `landuse_type`. Water bodies and artificial land-use classes were excluded during preprocessing. merged_public_land_candidate_area.tif: Covariate stack restricted to the candidate sampling domain (e.g., publicly owned land and feasibility-filtered areas) used for k-means clustering. estonia_reference_grid.tif: Reference grid (10 m, EPSG:3301) used to align and rasterise outputs (e.g., k-means clustering) to a consistent national extent, resolution, and pixel grid. baseline_aoa.tif: A binary classification map representing the AOA across Estonia. This identifies regions where the baseline Random Forest (RF) model (trained on original data) can provide reliable predictions. basline_dissimilarity_index.tif: A continuous raster showing the Dissimilarity Index (DI) for the baseline RF model. updated_aoa.tif, A binary classification map representing the AOA across Estonia. This identifies regions where the updated RF model (trained on updated data) can provide reliable predictions. updated_dissimilarity_index.tif: A continuous raster showing the Dissimilarity Index (DI) for the updated RF model. Raster metadata Spatial resolution: 10 m CRS: EPSG:3301 (Estonia national grid) NoData value: -9999 Full bibliographic details are provided in the manuscript.