Khomri et al. (2025) Optimizing Sugar Beet Irrigation in Arid Regions: A Machine Learning Approach to Soil Moisture Prediction

Identification

• Journal: Sugar Tech
• Year: 2025
• Date: 2025-11-28
• Authors: Zine-eddine Khomri, Samir Boudibi, A. Aissaoui, Abdelhamid Foughalia, Mohamed Kamel Bensalah
• DOI: 10.1007/s12355-025-01698-9

Research Groups

• Centre de Recherche Scientifique et Technique sur les Régions Arides (CRSTRA), Biskra, Algeria

Short Summary

This study evaluated eight machine learning models for soil moisture prediction in sugar beet cultivation in southern Algeria to optimize irrigation. The deep learning models, LSTM and GRU, demonstrated superior accuracy, leading to an estimated 15–25% reduction in water usage and a 5–10% increase in crop yield potential.

Objective

• To evaluate the performance of eight machine learning models (Linear Regression, Random Forest, XGBoost, Support Vector Regression, K-Nearest Neighbors, ARIMA, Long Short-Term Memory, and Gated Recurrent Unit) for predicting soil moisture for sugar beet cultivation in southern Algeria.
• To establish a relationship between sensor-based soil moisture measurements and irrigation schedules generated with the Penman–Monteith equation to optimize irrigation efficiency.

Study Configuration

• Spatial Scale: Southern Algeria, specifically the ITIDAS Institute in Biskra (latitude 34.850° N, longitude 5.733° E, elevation 87 m above sea level). Climatic data from TerraClimate dataset with approximately 4 km spatial resolution.
• Temporal Scale: Two years of soil moisture data (over 5,000 data points) recorded every 5 minutes for sugar beet cultivation from 2022 to 2024. Monthly climatic data from TerraClimate (1958 to present, used for the study period).

Methodology and Data

• Models used: Linear Regression (LR), Random Forest (RF), XGBoost, Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU). The Penman–Monteith (PM) equation was used for reference evapotranspiration calculation and irrigation scheduling.
• Data sources:
  • Climatic data: TerraClimate dataset (maximum and minimum temperatures, relative humidity, wind speed).
  • Soil moisture data: Readings from a patented smart irrigation system (Patent No. 12574) collected every 5 minutes.
  • Additional inputs: Watering start times and frequencies.

Main Results

• LSTM and GRU models achieved the highest accuracy in soil moisture prediction.
  • GRU model: Correlation coefficient (R) = 0.82, Root Mean Squared Error (RMSE) = 0.987, Mean Absolute Error (MAE) = 0.747.
  • LSTM model: R = 0.77, RMSE = 0.998, MAE = 0.826.
• Traditional models (LR and KNN) performed poorly, with R values of 0.30 and 0.13, respectively.
• Integrating ML predictions with the Penman–Monteith equation for irrigation scheduling resulted in an estimated 15–25% reduction in water usage (approximately 1,500–2,500 m³/ha/year savings) and a 5–10% increase in crop yield potential for sugar beet cultivation.

Contributions

• First study to assess multiple machine learning models for soil moisture prediction in agricultural fields characterized by shallow, sandy soils, and highly variable weather conditions in southern Algeria.
• Demonstrated the transformative potential of deep learning techniques (LSTM and GRU) for optimizing irrigation systems and advancing sustainable agricultural practices in water-scarce arid regions.
• Provided a hybrid approach combining theoretical evapotranspiration estimates (Penman–Monteith) with observed field data and ML predictions for more reliable and context-specific irrigation schedules.

Funding

• This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Citation

@article{Khomri2025Optimizing,
  author = {Khomri, Zine-eddine and Boudibi, Samir and Aissaoui, A. and Foughalia, Abdelhamid and Bensalah, Mohamed Kamel},
  title = {Optimizing Sugar Beet Irrigation in Arid Regions: A Machine Learning Approach to Soil Moisture Prediction},
  journal = {Sugar Tech},
  year = {2025},
  doi = {10.1007/s12355-025-01698-9},
  url = {https://doi.org/10.1007/s12355-025-01698-9}
}