Roy et al. (2025) Smartphone-based image analysis and interpretable machine learning for soil moisture estimation across diverse Indian soils

Identification

• Journal: Remote Sensing Applications Society and Environment
• Year: 2025
• Date: 2025-07-06
• Authors: Debasish Roy, Tridiv Ghosh, Bappa Das, Raghuveer Jatav, Debashis Chakraborty
• DOI: 10.1016/j.rsase.2025.101655

Research Groups

• ICAR-Indian Agricultural Research Institute, New Delhi, India
• ICAR-Central Coastal Agricultural Research Institute, Old Goa, India
• Dr. Rajendra Prasad Central Agricultural University, Pusa, Samastipur, Bihar, India
• ICAR-National Bureau of Soil Survey and Land Use Planning, Nagpur, Maharashtra, India
• International Maize and Wheat Improvement Center, Dhaka, Bangladesh

Short Summary

The study developed a low-cost, non-destructive method for estimating soil moisture content (SMC) using smartphone imagery and interpretable machine learning across diverse Indian soil types.

Objective

• To evaluate the feasibility and accuracy of using smartphone-based image analysis and machine learning models to estimate soil moisture content across five contrasting Indian soil groups.

Study Configuration

• Spatial Scale: 14 locations across five contrasting Indian soil groups.
• Temporal Scale: Not specified (cross-sectional analysis of 238 soil samples).

Methodology and Data

• Models used: Ten machine learning (ML) models (including Random Forest), Boruta feature selection, and interpretability techniques (SHAP and ALE).
• Data sources: 238 smartphone-captured soil images used to extract 33 colour-based features.

Main Results

• The Random Forest (RF) model provided the highest predictive accuracy with an $R^2 = 0.78$ and a Root Mean Square Error ($\text{RMSE}$) of $5.98\%$.
• Key predictors for SMC estimation were identified as the Redness Index (RI), Colour Feature Index (ColFeatInd), red band (R), value (V), and X colour space.
• Boruta selection confirmed the relevance of all extracted colour features.

Contributions

• Provides a scalable, non-invasive, and cost-effective alternative to traditional invasive and expensive soil moisture measurement methods, making it potentially accessible for farmer-level implementation.

Funding

• Not specified in the provided text.

Citation

@article{Roy2025Smartphonebased,
  author = {Roy, Debasish and Ghosh, Tridiv and Das, Bappa and Jatav, Raghuveer and Chakraborty, Debashis},
  title = {Smartphone-based image analysis and interpretable machine learning for soil moisture estimation across diverse Indian soils},
  journal = {Remote Sensing Applications Society and Environment},
  year = {2025},
  doi = {10.1016/j.rsase.2025.101655},
  url = {https://doi.org/10.1016/j.rsase.2025.101655}
}