Sahoo et al. (2026) Soil Moisture Prediction Using a Scalable and Validated Random Forest Regressor Model

Identification

• Journal: Lecture notes in networks and systems
• Year: 2026
• Date: 2026-08-01
• Authors: Asmet Ranjan Sahoo, Krish Agrawal, Debrup Ghosh, Hilal Abdullah Younis
• DOI: 10.1007/978-3-032-33535-7_43

Research Groups

• Kalinga Institute of Industrial Technology, Bhubaneswar, India
• Al-Esraa University, Baghdad, Iraq

Short Summary

The study proposes a scalable and cost-effective soil moisture prediction system using a Random Forest Regressor (RFR) optimized for deployment on edge devices to improve irrigation efficiency.

Objective

• To develop a deployable machine learning-based framework for predicting soil moisture that overcomes the high costs and scalability limitations of in-situ sensors and remote sensing satellites, specifically for smallholder farmers in developing regions.

Study Configuration

• Spatial Scale: Site-specific application (designed for deployment in actual agricultural environments, with a focus on regions like India).
• Temporal Scale: Temporal dynamics are captured using lag features spanning the previous 7 to 30 days.

Methodology and Data

• Models used: Random Forest Regression (RFR), validated via k-fold cross-validation.
• Data sources:
  • Meteorological data (precipitation, temperature).
  • Static geophysical data (soil type).
  • Remote sensing indicators (Normalized Difference Vegetation Index [NDVI], Land Surface Temperature [LST]).

Main Results

• The model's performance was validated using standard statistical indicators: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared ($R^2$).
• The system was successfully optimized for deployment on edge computing hardware (Raspberry Pi), enabling real-time, offline soil moisture forecasting and site-specific irrigation guidance.

Contributions

• Integration of diverse data streams (meteorological, geophysical, and remote sensing) with temporal lag features to improve prediction accuracy.
• Transition of a complex ML model to a scalable, low-cost edge device implementation, reducing dependency on constant connectivity and expensive infrastructure.
• Provision of a practical tool for precision farming tailored to the economic constraints of smallholder farmers.

Funding

• Not specified in the provided text.

Citation

@article{Sahoo2026Soil,
  author = {Sahoo, Asmet Ranjan and Agrawal, Krish and Ghosh, Debrup and Younis, Hilal Abdullah},
  title = {Soil Moisture Prediction Using a Scalable and Validated Random Forest Regressor Model},
  journal = {Lecture notes in networks and systems},
  year = {2026},
  doi = {10.1007/978-3-032-33535-7_43},
  url = {https://doi.org/10.1007/978-3-032-33535-7_43}
}