Summary
BIS-4D is a machine learning-informed soil mapping platform for the Netherlands that produces high-resolution (25 m) three-dimensional maps of seven key soil properties and their associated uncertainties to 2 m depth, informed by up to 950,000 soil observations and 366 environmental covariates. The platform additionally maps temporal changes in soil organic matter from 1953 to 2023. Prediction accuracy was property-dependent, with clay, sand and pH performing well (MEC ≥0.6) but phosphorus remaining difficult to predict reliably, highlighting both the utility and limitations of national-scale digital soil mapping.
Regional applicability
This study was conducted in the Netherlands and therefore does not directly address United Kingdom soil conditions or policy, though the methodological approach and machine learning framework may be transferable to UK soil mapping initiatives. The spatial resolution (25 m) and depth range (0–2 m) provide a benchmark for comparable national soil mapping efforts in the UK context.
Key measures
Soil texture (clay, silt, sand content), bulk density, pH, total nitrogen, oxalate-extractable phosphorus, cation exchange capacity, soil organic matter; model efficiency coefficient (MEC); prediction interval coverage probability; 10-fold cross-validation results
Outcomes reported
The study delivered high-resolution maps (25 m resolution) of eight key soil properties and their uncertainties across 0–2 m depth in three-dimensional space, plus temporal maps of soil organic matter between 1953 and 2023. Prediction accuracy varied by property, with clay, sand and pH showing highest accuracy (MEC 0.6–0.92), whilst oxalate-extractable phosphorus was most difficult to predict (MEC −0.11 to 0.38).
Topic tags
Dig deeper with Pulse AI.
Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.