Kharraz et al. (2026) Intelligent Division of Agricultural Farms into Homogeneous Management Zones for Precision Irrigation Using Remote Sensing and Artificial Intelligence

Identification

• Journal: Smart Agricultural Technology
• Year: 2026
• Date: 2026-04-01
• Authors: Siham Kharraz, Okacha Amraouy, Mohammed Nabil Kabbaj, Mohammed Benbrahim
• DOI: 10.1016/j.atech.2026.102070

Research Groups

• Engineering, Modeling and Analysis of Systems Laboratory, Sidi Mohamed Ben Abdellah University, Fez, Morocco

Short Summary

This study developed a hierarchical framework integrating multi-source remote sensing data, topographic information, and soil properties with machine learning (LightGBM) to delineate homogeneous management zones for precision irrigation, achieving 94.1% accuracy in agricultural land discrimination and providing a physically interpretable basis for water management.

Objective

• To identify agricultural areas using object-level machine learning.
• To delineate environmentally consistent homogeneous management zones (HMZs) reflecting combined moisture, thermal, and topographic controls.
• To incorporate soil hydraulic constraints from diverse data sources.
• To provide a reproducible framework to support precision irrigation planning in semi-arid environments.

Study Configuration

• Spatial Scale: A privately owned agricultural farm of 210,000 m² (21 hectares) in the Fès-Meknès area of north-central Morocco, located between latitudes 33.946°–33.954° N and longitudes 5.313°–5.307° W.
• Temporal Scale: Data acquisition and analysis focused on the 2024–2025 agricultural season (May to October), with specific periods for early-growth (April–May 2024) and peak-stress (July–August 2024) conditions. The study was conducted for a single year.

Methodology and Data

• Models used:
  • K-means clustering (for unsupervised segmentation and soil zoning)
  • Random Forest (RF)
  • Extreme Gradient Boosting (XGBoost)
  • Light Gradient Boosting Machine (LightGBM) (selected as the final classifier)
• Data sources:
  • Satellite:
    • Sentinel-2 (optical imagery, Level-2A, 10 m and 20 m spatial resolution) for vegetation monitoring, surface characterization, and spectral indices (NDVI, EVI, NDWI, NDMI, BSI, NDBI).
    • Sentinel-1 (Synthetic Aperture Radar (SAR) data, Ground Range Detected (GRD) products, 10 m spatial resolution) for VV and VH backscatter coefficients as soil moisture proxies.
    • Landsat-8 (Thermal Infrared Sensor (TIRS) imagery, 100 m resampled spatial resolution) for Land Surface Temperature (LST) as an indicator of crop water stress.
  • Observation/Reanalysis:
    • USGS SRTMGL1_003 Digital Elevation Model (DEM, 30 m spatial resolution) for topographic variables (Elevation, Slope, Topographic Wetness Index (TWI)).
    • SoilGrids database (ISRIC, ~250 m spatial resolution) for soil attributes (clay, sand, silt, bulk density, organic carbon) averaged over 0–60 cm depth.

Main Results

• LightGBM demonstrated the best performance for agricultural/non-agricultural discrimination, achieving an overall accuracy of 94.1% and an F1-score of 0.928 for the agricultural class.
• The generated agricultural mask showed high spatial consistency, with the estimated agricultural area (206,853.30 m²) differing by only -0.6% from the reference area (208,172.11 m²).
• Topography-based zoning, using an optimal number of 4 clusters determined by the elbow method, delineated zones categorized as water loss, balanced, and water storage based on a composite Hydric Stress Index (HSI).
• Soil-based zoning, derived from normalized soil indices (Clay Index, Drainage Index, Soil Texture Index, Water Holding Index), further refined management units, highlighting substantial intra-field soil heterogeneity.
• The final multi-layer fusion integrated agricultural mask, topography-based zones, and soil-based zones to produce 7 distinct Homogeneous Management Zones (HMZs), representing combined influences of vegetation, terrain, and soil for site-specific irrigation control.

Contributions

• Proposes an integrated and scalable framework for delineating homogeneous management zones by combining multi-source remote sensing data, geospatial analysis, and machine learning techniques.
• Introduces a three-level hierarchical zoning strategy: (i) separation of agricultural and non-agricultural areas using supervised artificial intelligence, (ii) topographic-based subdivision, and (iii) soil-based zoning using either in situ data or global SoilGrids products.
• Shifts from traditional pixel-level mapping to an object-oriented approach, generating spatially coherent polygons that are directly usable as decision-support layers for precision irrigation planning.
• Develops a physically interpretable, operationally practical, and transferable framework, demonstrated in a semi-arid agricultural region of north-central Morocco, designed to operate under limited field-data availability.
• Formulates a decision-oriented composite Hydric Stress Index (HSI) to integrate thermal, vegetation, and terrain-related drivers of hydric stress for irrigation zoning and management prioritization.

Funding

Funding information for this research was not explicitly provided in the paper.

Citation

@article{Kharraz2026Intelligent,
  author = {Kharraz, Siham and Amraouy, Okacha and Kabbaj, Mohammed Nabil and Benbrahim, Mohammed},
  title = {Intelligent Division of Agricultural Farms into Homogeneous Management Zones for Precision Irrigation Using Remote Sensing and Artificial Intelligence},
  journal = {Smart Agricultural Technology},
  year = {2026},
  doi = {10.1016/j.atech.2026.102070},
  url = {https://doi.org/10.1016/j.atech.2026.102070}
}