Gao et al. (2026) A parallel evaluation framework integrating a modified SWAP model and AutoML for simulating soil salinity under mulched drip irrigation

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-06-30
• Authors: Fukui Gao, Yadan Wang, Ni Song, Huifeng Ning, Peiwen Xu, Shenglin Li, Hao Liu
• DOI: 10.1016/j.agwat.2026.110591

Research Groups

• Institute of Farmland Irrigation, Chinese Academy of Agricultural Sciences/Key Laboratory of Crop Water Use and Regulation, Ministry of Agriculture and Rural Affairs, Xinxiang, Henan 453002, PR China
• Shangqiu, Agro-ecological System, National Observation and Research Station, Shangqiu, Henan 476000, PR China

Short Summary

This study developed a parallel evaluation framework integrating a modified SWAP model and H2O AutoML for simulating soil salinity under mulched drip irrigation. The results show that the modified SWAP model improved the physical simulation of soil salinity, while H2O AutoML outperformed in handling plot-scale spatial heterogeneity.

Objective

• To reconstruct the upper boundary flux mechanism of the SWAP model by introducing a film coverage factor and to evaluate its performance in simulating soil salt transport under mulched drip irrigation.
• To achieve adaptive optimization of model structures and hyperparameters using H2O AutoML, identify the dominant factors driving soil salinity through variable importance analysis, and improve deep soil salinity prediction through a recursive strategy.

Study Configuration

• Spatial Scale: Field scale in arid regions of Northwest China.
• Temporal Scale: Long-term simulation covering the entire cotton growing seasons from 2021 to 2024.

Methodology and Data

• Models used:
  • Modified SWAP model with a film coverage factor for simulating soil salt transport under mulched drip irrigation.
  • H2O AutoML for adaptive optimization of model structures and hyperparameters, variable importance analysis, and recursive strategy for deep soil salinity prediction.
• Data sources: Field measurements from six independent field experiments conducted between 2021 and 2024.

Main Results

• The modified SWAP model improved the physical simulation of soil salinity compared to the standard SWAP model.
• H2O AutoML outperformed in handling plot-scale spatial heterogeneity, with a reduction in NRMSE by 46.79% relative to the modified SWAP model.
• A recursive strategy constructed a vertical information chain that alleviated the attenuation of surface information during downward transmission.

Contributions

• This study provides a parallel evaluation framework integrating physical and data-driven models for simulating soil salinity under mulched drip irrigation.
• The results highlight the value of process-based models for long-term salinity extrapolation and risk assessment, and the advantage of AutoML in capturing spatial heterogeneity and reconstructing salinity profiles.

Funding

• Not specified.

Citation

@article{Gao2026parallel,
  author = {Gao, Fukui and Wang, Yadan and Wang, Jinglei and Song, Ni and Ning, Huifeng and Xu, Peiwen and Li, Shenglin and Liu, Hao},
  title = {A parallel evaluation framework integrating a modified SWAP model and AutoML for simulating soil salinity under mulched drip irrigation},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110591},
  url = {https://doi.org/10.1016/j.agwat.2026.110591}
}