High-resolution digital soil mapping of amorphous iron- and aluminium-(hydr)oxides to guide sustainable phosphorus and carbon management

Summary

This study developed a high-resolution digital soil map predicting oxalate-extractable iron and aluminium oxide contents across the Netherlands at 25 m spatial resolution using 12,110 wet-chemical and 102,393 NIR spectroscopy observations combined with over 150 spatial covariates. Map quality varied by target variable and depth (MEC 0.19–0.80), with better performance for topsoil than subsoil and for aluminium than iron oxides; prediction uncertainties were generally reliable. The resulting maps can inform spatial management of phosphorus retention and carbon storage to optimise crop production, water quality and carbon sequestration at the field and regional scale.

Regional applicability

This study was conducted in the Netherlands and provides methodological precedent for high-resolution soil property mapping applicable to United Kingdom soils. Transfer to UK conditions would require development of analogous national datasets and validation against UK soil sample networks, though the quantile regression forest approach and uncertainty quantification framework are directly transferable. Given the UK's different soil parent materials, climate and soil typology, a country-specific calibration would be necessary before operational deployment.

Key measures

Model Efficiency Coefficient (MEC: 0.19–0.80), Root Mean Square Error (RMSE: 13.5–56.9 mmol kg−1), Mean Error (ME: −6.8 to 6.8 mmol kg−1), Prediction Interval Coverage Probability for 90% prediction intervals

Outcomes reported

The study spatially predicted oxalate-extractable iron and aluminium oxide contents across the Netherlands at 25 m resolution across six soil depth layers (0–200 cm) using quantile regression forest modelling. Map quality was assessed using Model Efficiency Coefficient, Root Mean Square Error, and Mean Error metrics, with evaluation of prediction interval coverage probability.

Topic tags