Sign in Subscribe
Periodical publishing

Machine Learning Algorithm Accurately Maps Humus Horizon Thickness in Russian Soils

Strategic Insights

— 1 min read

A recent study published in the Journal of Engineering has showcased the efficacy of a machine learning algorithm in mapping the humus horizon thickness (HHT) in Russian soils. The research, conducted by the Russian Academy of Sciences, utilized the Random Forest algorithm implemented on the Google Earth Engine cloud online platform to analyze 92 predictors, including climate, relief, vegetation, spatial position, and soil properties. The study covered an area of 92,400 square kilometers in the Cis-Salair Plain, with financial support provided by the Ministry of Science and Higher Education of the Russian Federation.

Key Takeaways:

  • The Random Forest algorithm demonstrated high accuracy in mapping HHT, with a coefficient of determination (R²) of 0.88 for the training dataset and 0.12 for the validation dataset.
  • The root mean square error (RMSE) for the validation dataset was 9.7 cm, while the mean absolute percentage error (MAPE) and mean absolute error (MAE) were 24.3% and 6.5 cm, respectively.
  • The modeling accuracy was deemed satisfactory, with actual HHT values ranging from 3 to 110 cm and a trend of decreasing HHT from northwest to southeast.
  • The study found that the lowest average HHT values were typical of meadow-chernozemic solonetz (Solonetz (Salic)), while the highest values were observed in ordinary meadow soils (Mollic Gleysols).
  • The research has been peer-reviewed and published in the journal Eurasian Soil Science.

Statistics:

  • 92 predictors used to characterize soil formation factors, including climate, relief, vegetation, spatial position, and soil properties.
  • 92,400 square kilometers of area analyzed in the Cis-Salair Plain.
  • 718 training dataset samples constructed based on archive materials from 1974 to 1984.
  • 130 validation dataset samples used to evaluate the algorithm's performance.
  • RMSE of 9.7 cm for the validation dataset.
  • MAPE of 24.3% and MAE of 6.5 cm for the validation dataset.

Sources:

  • NewsRx. Findings from Russian Academy of Sciences Has Provided New Data on Machine Learning (Digital Mapping of the Humus Horizon Thickness In Soils of the Cis-salair Plain Using the Random Forest Machine Learning Algorithm). Journal of Engineering. October 13, 2025; p 779.
  • Eurasian Soil Science. Digital Mapping of the Humus Horizon Thickness In Soils of the Cis-salair Plain Using the Random Forest Machine Learning Algorithm. Eurasian Soil Science, 2025;58(11).

Read more