Singha et al. (2025) Hybrid framework of physics-inspired optimization and explainable ensemble learning for irrigation classification mapping in Morocco

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Environmental Research Communications
• Year: 2025
• Date: 2025-11-25
• Authors: Chiranjit Singha, Satiprasad Sahoo, Ajit Govind
• DOI: 10.1088/2515-7620/ae2449

Research Groups

Not specified in the provided abstract.

Short Summary

This study developed a novel integrated ensemble classification framework, leveraging remote sensing data, field surveys, and advanced machine learning, to generate high-resolution irrigation maps for the Moroccan region. The framework achieved high accuracy, particularly for drip irrigation, identified key hydro-meteorological features influencing classification, and revealed significant upstream irrigation expansion with implications for water sustainability.

Objective

• To present a novel integrated ensemble classification framework to generate high-resolution irrigation maps for the Moroccan region in Africa.

Study Configuration

• Spatial Scale: Regional (Moroccan region) to basin scale (Upper Rabat-River Basin).
• Temporal Scale: Ground truth datasets for 2024.

Methodology and Data

• Models used: Hybrid ensemble approach combining physics-inspired machine learning algorithms (Photon Search Algorithm - PSA, Quantum-Behaved Avian Navigation Optimizer - QANO, Kepler Optimization Algorithm - KOA, Nuclear Reaction Optimization - NRO) with Extreme Gradient Boosting (XGB). Synthetic Minority Over-sampling Technique (SMOTE) for class imbalance. Graph centrality Laplacian score for feature selection. SHAP analysis for feature importance.
• Data sources: Remote sensing (RS) data, field surveys, ground truth datasets (754 training, 324 testing for drip; 288 training, 120 testing for flood; 953 training, 411 testing for sprinkler). Input variables included topographic, hydro-meteorological, and RS indicators. Specific influential variables identified were vapor pressure deficit (VPD), land surface temperature (LST), and actual evapotranspiration (AET).

Main Results

• Vapor pressure deficit (VPD) and land surface temperature (LST) were identified as the top two most influential variables for classification performance.
• Drip irrigation was the most accurately classified type across all models, with XGB-QANO achieving the best overall performance (F1-score: 0.9639, AUC: 0.99).
• Flooding irrigation was the most challenging to classify, with XGB-NRO achieving the highest F1-score (0.7029) and AUC (0.934).
• Sprinkler irrigation classification was highly consistent, with XGB-NRO achieving top results (F1-score: 0.9071, AUC: 0.967).
• SHAP analysis identified VPD and AET as two key features influencing irrigation classification.
• Spatially, irrigation has notably expanded upstream in the Upper Rabat-River Basin, especially near tributaries, where new croplands remained irrigated during droughts, unlike downstream areas.

Contributions

• Development of a novel integrated ensemble classification framework for high-resolution irrigation mapping, combining physics-inspired ML algorithms with XGB.
• Successful application of advanced machine learning techniques and remote sensing for accurate irrigation type mapping in a water-stressed region.
• Identification of key hydro-meteorological variables (VPD, LST, AET) critical for distinguishing irrigation types.
• Detailed spatial analysis revealing significant upstream irrigation expansion in the Upper Rabat-River Basin, highlighting long-term water sustainability concerns for Moroccan agriculture.

Funding

Not specified in the provided abstract.

Citation

@article{Singha2025Hybrid,
  author = {Singha, Chiranjit and Sahoo, Satiprasad and Govind, Ajit},
  title = {Hybrid framework of physics-inspired optimization and explainable ensemble learning for irrigation classification mapping in Morocco},
  journal = {Environmental Research Communications},
  year = {2025},
  doi = {10.1088/2515-7620/ae2449},
  url = {https://doi.org/10.1088/2515-7620/ae2449}
}