Suleymanov et al. (2026) Three-dimensional mapping of key soil properties with multi-stage validation and big data

Identification

• Journal: CATENA
• Year: 2026
• Date: 2026-02-26
• Authors: Azamat Suleymanov, Gaziz Valiev, Ruslan Shagaliev, Ruslan Suleymanov, Larisa Belan
• DOI: 10.1016/j.catena.2026.109949

Research Groups

• Laboratory of Artificial Intelligence in Environmental Research, Decarbonisation Technologies Center, Ufa State Petroleum Technological University, Ufa, Russia
• Laboratory of Soil Science, Ufa Institute of Biology, Ufa Federal Research Centre, Russian Academy of Sciences, Ufa, Russia
• Department of Geodesy, Cartography and Geographic Information Systems, Ufa University of Science and Technology, Ufa, Russia
• UNESCO Department of Geoparks and Territories of Sustainable Development, Ufa University of Science and Technology, Ufa, Russia

Short Summary

This study generated high-resolution three-dimensional digital soil maps for soil organic carbon (SOC) and pH in the Republic of Bashkortostan, Russia, across five depth intervals up to 1 meter using a machine learning approach. The maps, validated through a multi-stage process, showed reliable predictions in plain regions but highlighted data limitations in mountainous areas.

Objective

• To generate updated, high-resolution three-dimensional digital maps of soil organic carbon (SOC) and pH KCl for the Republic of Bashkortostan (Russia) across five depth intervals (up to 1 meter) using a machine learning algorithm.

Study Configuration

• Spatial Scale: Republic of Bashkortostan, Russia.
• Temporal Scale: Current state mapping based on existing soil measurements.

Methodology and Data

• Models used: Random forest (machine learning algorithm).
• Data sources:
  • Training data: 29,402 soil organic carbon (SOC) measurements and 22,301 pH measurements.
  • Covariates: A wide range of soil-forming factors.
  • Validation: Profile-based cross-validation, uncertainty assessment, area of applicability (AOA), expert-based pedological knowledge, and comparison with existing digital and conventional maps.

Main Results

• Predictive models explained 50% of the variance in SOC and 64% of the variance in pH.
• Climate variables were identified as the key predictors for both soil properties.
• The Area of Applicability (AOA) assessment revealed that the models were unable to provide reliable predictions across all depths in mountainous areas due to insufficient data and unique environmental conditions.
• Predictions for both soil properties are generally reliable in plain regions, where the model benefits from robust representation of soil-landscape relationships and sufficient training data.
• Comparison with other maps showed both differences and similarities depending on the specific map.

Contributions

• Generation of the first high-resolution three-dimensional digital soil maps for SOC and pH for the Republic of Bashkortostan, Russia, up to 1 meter depth.
• Application of a comprehensive multi-stage validation approach, including Area of Applicability (AOA) and expert-based pedological knowledge.
• Identification of critical data gaps in mountainous ecosystems, providing guidance for future soil data collection efforts.
• Provision of valuable 3D-DSM products to support local agricultural, environmental, and socio-economic needs, enhancing regional and national understanding of soil resources.

Funding

• Not specified in the provided text.

Citation

@article{Suleymanov2026Threedimensional,
  author = {Suleymanov, Azamat and Valiev, Gaziz and Shagaliev, Ruslan and Suleymanov, Ruslan and Belan, Larisa},
  title = {Three-dimensional mapping of key soil properties with multi-stage validation and big data},
  journal = {CATENA},
  year = {2026},
  doi = {10.1016/j.catena.2026.109949},
  url = {https://doi.org/10.1016/j.catena.2026.109949}
}