Demir (2026) Multi-Depth Soil Moisture Prediction Using Machine Learning Across Türkiye's Diverse Environments

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Tarım Bilimleri Dergisi
• Year: 2026
• Date: 2026-03-24
• Authors: Muhammed Sungur Demir
• DOI: 10.15832/ankutbd.1809955

Research Groups

Not explicitly mentioned in the provided text.

Short Summary

This study developed a machine learning framework to predict soil moisture at multiple depths using environmental variables in Türkiye. The Extreme Gradient Boosting (XGBoost) model achieved strong accuracy (R² up to 0.74) and revealed depth-dependent and spatially varying controls on soil moisture dynamics.

Objective

• To develop and evaluate a machine learning framework for predicting soil moisture at multiple depths (20 cm, 40 cm, and 80 cm) using readily available environmental variables.
• To analyze the depth-dependent and spatially varying controls on soil moisture dynamics across a heterogeneous region.

Study Configuration

• Spatial Scale: Türkiye (country-wide).
• Temporal Scale: 2016–2024 (8 years).

Methodology and Data

• Models used: Seven machine learning algorithms were evaluated, with Extreme Gradient Boosting (XGBoost) identified as the best performer after hyperparameter optimization.
• Data sources: Soil moisture and environmental variables collected from 201 meteorological stations.

Main Results

• Extreme Gradient Boosting (XGBoost) demonstrated superior performance among the evaluated algorithms.
• Achieved strong prediction accuracy across all depths: R² = 0.74 at 20 cm, R² = 0.69 at 40 cm, and R² = 0.66 at 80 cm.
• Feature importance analysis revealed depth-dependent controls:
  • Shallow depth (20 cm): Primarily driven by short-term meteorological variability and seasonal dynamics.
  • Deeper layers: Increasingly regulated by stable soil hydraulic properties (e.g., clay at intermediate depths) and large-scale spatial gradients (e.g., latitude, elevation).
• The model maintained robust performance across diverse environmental conditions, with seasonal accuracy varying by approximately 10% between winter and autumn.
• Spatial analysis indicated regional variations in controlling factors: soil properties (particularly organic carbon) dominated in northern regions, while climatic and temporal variables were more influential southward.

Contributions

• Provides a scalable and cost-effective machine learning framework for multi-depth soil moisture monitoring, particularly beneficial for data-scarce regions.
• Offers novel insights into the complex, depth-dependent, and spatially heterogeneous controls on soil moisture dynamics in a climatically and pedologically diverse region.
• Supports applications in irrigation optimization, drought early warning systems, and sustainable water governance.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Demir2026MultiDepth,
  author = {Demir, Muhammed Sungur},
  title = {Multi-Depth Soil Moisture Prediction Using Machine Learning Across Türkiye's Diverse Environments},
  journal = {Tarım Bilimleri Dergisi},
  year = {2026},
  doi = {10.15832/ankutbd.1809955},
  url = {https://doi.org/10.15832/ankutbd.1809955}
}