Sabea et al. (2025) AI for Sustainable Development Goals: Leveraging Machine Learning for Climate-Resilient Wheat Production in Wasit, Iraq

Identification

• Journal: Engineering and Technology Journal
• Year: 2025
• Date: 2025-11-30
• Authors: Asmaa Ghali Sabea, Maryam Jawad Kadhim
• DOI: 10.47191/etj/v10i11.29

Research Groups

• Law Faculty, Sumer University, Al-Rifai, Iraq.
• Computer Science and IT Faculty, Wasit University, Al-Kut, Iraq.

Short Summary

This study develops an integrated machine-learning decision-support system to enhance wheat production resilience in the semi-arid Wasit Governorate of Iraq. By combining multi-temporal satellite data and meteorological reanalysis, the system optimizes irrigation and provides early-season yield forecasts, resulting in significant water savings and improved yield stability.

Objective

• To establish a unified, operational decision-support pipeline that bridges the gap between predictive machine learning models and practical field-level management for wheat cultivation in climate-stressed regions.

Study Configuration

• Spatial Scale: Wasit Governorate, central Iraq (regional and field-level application).
• Temporal Scale: Analysis of climate and production trends from 2010 to 2022; wheat growing cycles (typically November to May).

Methodology and Data

• Models used: A hybrid architecture featuring Long Short-Term Memory (LSTM) and XGBoost for yield forecasting; Random Forest Regression for soil moisture estimation; Gradient Boosting Classifiers for drought early warning; and Model Predictive Control (MPC) for irrigation optimization.
• Data sources:
  • Sentinel-2 Optical Imagery: 10 m resolution (NDVI/EVI) for tracking canopy vigor.
  • Sentinel-1 SAR Backscatter: 10 m resolution for near-surface soil moisture estimation.
  • ERA5 Climate Reanalysis: Daily gridded data (~0.25°) for temperature, rainfall, and solar radiation.
  • FAO SoilGrids Database: 250 m resolution for soil physical properties and water-holding capacity.

Main Results

• Model Performance: The hybrid LSTM-XGBoost model achieved a yield prediction accuracy of $R^2 = 0.82$ and an RMSE of 260 kg/ha, significantly outperforming baseline linear models ($R^2 = 0.41$).
• Resource Efficiency: Implementation of the system resulted in water savings of 18–26% and a reduction in irrigation pump runtime by approximately 21%.
• Productivity: Wheat yields increased by 7–12% due to optimized irrigation timing and reduced physiological stress.
• Drought Monitoring: The system demonstrated high reliability in drought signaling with a ROC-AUC of 0.87, allowing for interventions before visual symptoms appear.

Contributions

• System Integration: Unlike existing siloed tools, this research integrates yield forecasting, soil moisture retrieval, and irrigation control into a single operational framework tailored for semi-arid continental climates.
• Proactive Management: Shifts agricultural practices from reactive (responding to visible wilting) to anticipatory, leveraging deep learning to detect subtle phenological deviations.
• SDG Alignment: Provides a scalable framework for achieving Sustainable Development Goals related to food security (SDG 2), responsible water use (SDG 12), and climate action (SDG 13) in the Middle East.

Funding

• Not explicitly specified in the provided text; research conducted by Sumer University and Wasit University.

Citation

@article{Sabea2025AI,
  author = {Sabea, Asmaa Ghali and Kadhim, Maryam Jawad},
  title = {AI for Sustainable Development Goals: Leveraging Machine Learning for Climate-Resilient Wheat Production in Wasit, Iraq},
  journal = {Engineering and Technology Journal},
  year = {2025},
  doi = {10.47191/etj/v10i11.29},
  url = {https://doi.org/10.47191/etj/v10i11.29}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)