Shao et al. (2026) A knowledge-guided machine learning framework for cross-scale wheat harvest monitoring via sample augmentation

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-09-15
• Authors: Mingchao Shao, Chongya Jiang, Jingwei An, Haokai Zhu, Yue Li, Xia Yao, Tao Cheng, Hengbiao Zheng, Weixing Cao, Yan Zhu
• DOI: 10.1016/j.rse.2026.115671

Research Groups

National Engineering and Technology Center for Information Agriculture, MOE Engineering Research Center of Smart Agriculture, MARA Key Laboratory of Crop System Analysis and Decision Making, Jiangsu Key Laboratory for Information Agriculture, Nanjing Agricultural University.

Short Summary

This study proposes a Knowledge-Guided Machine Learning (KGML) framework that integrates multi-satellite Earth observation data to monitor wheat harvest from the field to regional scales with high accuracy. The framework combines spectral knowledge rules with machine learning models to generate high-confidence augmented samples and estimate harvest dates.

Objective

• Investigate the feasibility of using a Knowledge-Guided Machine Learning (KGML) framework for cross-scale wheat harvest monitoring via sample augmentation.

Study Configuration

• Spatial Scale: Field-level, regional scales.
• Temporal Scale: 2023 and 2024 wheat harvest periods.

Methodology and Data

• Models used: Random Forest model, Hybrid CNN-Transformer-LSTM (HCTL) model.
• Data sources: PlanetScope, Sentinel-2, MODIS satellite Earth observation data, vehicle-mounted cameras, smartphones.

Main Results

• The KGML framework generated numerous high-confidence augmented samples from PlanetScope imagery with regional accuracy >0.80.
• The HCTL model achieved overall accuracy = 0.93 for field-level harvest mapping and R2 = 0.97, RMSE = 0.07, rRMSE = 0.15 for sub-pixel harvest fraction estimation.

Contributions

• This study provides an effective solution for wheat harvest monitoring by bridging the gap between limited ground-truth data and multi-scale satellite observations.
• The proposed KGML framework supports food security assessments and informed agricultural management decisions.

Funding

• National Engineering and Technology Center for Information Agriculture, MOE Engineering Research Center of Smart Agriculture, MARA Key Laboratory of Crop System Analysis and Decision Making, Jiangsu Key Laboratory for Information Agriculture.

Citation

@article{Shao2026knowledgeguided,
  author = {Shao, Mingchao and Jiang, Chongya and An, Jingwei and Zhu, Haokai and Li, Yue and Yao, Xia and Cheng, Tao and Zheng, Hengbiao and Cao, Weixing and Zhu, Yan},
  title = {A knowledge-guided machine learning framework for cross-scale wheat harvest monitoring via sample augmentation},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115671},
  url = {https://doi.org/10.1016/j.rse.2026.115671}
}