Garg et al. (2026) Integrating AquaCrop-OSPy and machine learning to develop transferable surrogates for balancing wheat yield–irrigation trade-offs under deficit irrigation in semi-arid NW India

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-09-28
• Authors: Divyam Garg, Hemant Kumar
• DOI: 10.1016/j.agwat.2026.110820

Research Groups

• Civil Engineering Department, Indian Institute of Technology Roorkee, Uttarakhand, 247667, India
• Food and Agriculture Organization (FAO)

Short Summary

This study developed transferable surrogates for balancing wheat yield–irrigation trade-offs under deficit irrigation in semi-arid NW India using AquaCrop-OSPy and machine learning. The best-performing surrogates showed excellent generalizability and cross-district transferability, preserving the characteristic yield-irrigation response behavior.

Objective

• To develop machine-learning based surrogate models for emulating seasonal ACOSP simulations of wheat yield and irrigation response.
• To evaluate the spatial transferability of the surrogates across districts in NW India.
• To examine whether transferred surrogate model preserves the characteristic yield-irrigation response behaviour in an unseen target district.

Study Configuration

• Spatial Scale: District-level, with three districts (Mahendragarh, Bhilwara, and Sangrur) in semi-arid NW India.
• Temporal Scale: 26 growing seasons (1997–2023).

Methodology and Data

• Models used: AquaCrop-OSPy, random forest (RF), and extreme gradient boosting (XGBoost).
• Data sources: Meteorological data from India Meteorological Department (IMD), soil texture data from Central Ground Water Board (CGWB), and crop yield data from Directorate of Economics and Statistics (DES).

Main Results

• The developed surrogates showed excellent generalizability with test set R2 > 0.90 and small RMSE and MAE for the same district.
• Cross-district transferability assessment revealed that the best-performing surrogates preserved the characteristic yield-irrigation response behavior in an unseen target district.

Contributions

• This study contributes to the development of transferable surrogates for balancing wheat yield–irrigation trade-offs under deficit irrigation, which can be particularly useful for agricultural planners in data-scarce and water-limited environments.
• The study demonstrates the potential of machine learning-based surrogate models to emulate process-based crop growth models, enabling rapid evaluation of alternative irrigation strategies.

Funding

• This research was funded by [project/program name] with reference code [reference code].

Citation

@article{Garg2026Integrating,
  author = {Garg, Divyam and Kumar, Hemant},
  title = {Integrating AquaCrop-OSPy and machine learning to develop transferable surrogates for balancing wheat yield–irrigation trade-offs under deficit irrigation in semi-arid NW India},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110820},
  url = {https://doi.org/10.1016/j.agwat.2026.110820}
}