Wu et al. (2026) Multi-Dimensional Monitoring of Agricultural Drought at the Field Scale

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

Identification

• Journal: Agriculture
• Year: 2026
• Authors: Yehao Wu, Liming Zhu, Maohua Ding, Lijie Shi
• DOI: 10.3390/agriculture16020227

Research Groups

• Remote sensing and agricultural monitoring research units (specific departments not listed in the provided text, focused on Hebi City, Henan Province).

Short Summary

This study develops a high-resolution, field-scale agricultural drought monitoring model for Hebi City using multi-source satellite data and machine learning, identifying XGBoost as the most effective algorithm with 89% accuracy. The research demonstrates that integrating radar, optical, and topographic data significantly improves the detection of rapid-onset, small-scale drought events.

Objective

• To construct and evaluate a multi-dimensional agricultural drought monitoring model at the field scale by integrating multi-source environmental variables and machine learning algorithms to overcome the limitations of single-variable monitoring and low-resolution data.

Study Configuration

• Spatial Scale: Field scale; Hebi City, Henan Province, China (including mountainous and plain regions).
• Temporal Scale: 2019 to 2024.

Methodology and Data

• Models used: eXtreme Gradient Boosting (XGBoost), Random Forest (RF), and Random Forest-Recursive Feature Elimination (RF-RFE); Mann–Kendall test for time series trend analysis.
• Data sources: Sentinel-1 (HV, VV polarizations), Sentinel-2 (NDVI, B2, B11 bands), Digital Elevation Model (DEM: elevation, slope, aspect), and Global Precipitation Measurement (GPM) precipitation data.

Main Results

• Model Performance: The XGBoost model outperformed others with a Correlation Coefficient (R) of 0.79, a Relative Root Mean Square Error (RRMSE) of 0.45, and an overall accuracy of 89%.
• Comparative Accuracy: RF and RF-RFE achieved accuracies of 88% and 86%, respectively.
• Feature Importance: Full-factor input (combining radar, optical, precipitation, and topography) yielded the best results. The exclusion of DEM data significantly reduced predictive accuracy and the ability to capture key feature patterns.
• Spatiotemporal Analysis: Successfully mapped a 2023 drought event in Hebi City, identifying abrupt changes in drought status and noting a lower probability of drought in western mountainous areas.
• Sample Sensitivity: A significant correlation was observed between the size of the training sample set and the resulting model performance.

Contributions

• Multi-dimensional Integration: Combines high-resolution radar (Sentinel-1) and optical (Sentinel-2) data with topographic and meteorological variables to address the spatial-temporal trade-offs in drought monitoring.
• Field-Scale Precision: Advances the ability to capture small-scale, rapid-onset drought events that are typically missed by low-resolution satellite products.
• Validation Innovation: Discusses the feasibility of using irrigation signals as a verification tool for drought occurrence, providing a new perspective for ground-truthing remote sensing models.

Funding

• Not specified in the provided text.

Citation

@article{Wu2026MultiDimensional,
  author = {Wu, Yehao and Zhu, Liming and Ding, Maohua and Shi, Lijie},
  title = {Multi-Dimensional Monitoring of Agricultural Drought at the Field Scale},
  journal = {Agriculture},
  year = {2026},
  doi = {10.3390/agriculture16020227},
  url = {https://doi.org/10.3390/agriculture16020227}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)