Zafar et al. (2026) Comparative Assessment of Machine Learning and Statistical Models for Agricultural Drought Forecasting and Driver Attribution in the North China Plain

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

Identification

• Journal: International Journal of Climatology
• Year: 2026
• Date: 2026-10-15
• Authors: Zeeshan Zafar, Jingwei Wu, Shiqiang Zhang, Guoqing Lei, Yuanyuan Zha
• DOI: 10.1002/joc.70637

Research Groups

[Information not available in the provided abstract.]

Short Summary

This study provides a spatiotemporal analysis and prediction of agricultural drought in the North China Plain (NCP) from 2001 to 2024 using the Vegetation Health Index (VHI). It identifies air temperature as a dominant negative driver of vegetation health, while soil moisture and vegetation greenness (EVI) are significant positive influences, and demonstrates that Long Short-Term Memory (LSTM) models offer the best performance for drought prediction.

Objective

• To provide a spatiotemporal analysis and prediction of agricultural drought in the North China Plain (NCP) from 2001 to 2024 using the Vegetation Health Index (VHI) derived from satellite data.

Study Configuration

• Spatial Scale: North China Plain (NCP)
• Temporal Scale: 2001 to 2024

Methodology and Data

• Models used: Trend analysis, wavelet coherence, Granger causality, structural equation modelling (SEM), Long Short‐Term Memory (LSTM), Random Forest (RF), ARIMAX.
• Data sources: Satellite data (for Vegetation Health Index (VHI), Enhanced Vegetation Index (EVI), Leaf Area Index (LAI)), meteorological variables (air temperature, precipitation), environmental variables (soil moisture, evapotranspiration).

Main Results

• Air temperature is the dominant negative driver of vegetation health (β = −0.444, p < 0.001).
• Soil moisture (β = 0.297, p < 0.001) and vegetation greenness, particularly EVI (β = 0.906, p < 0.001), exert significant positive influences on vegetation health.
• The northern and central NCP were identified as persistent agricultural drought hotspots.
• Among the prediction models evaluated, LSTM showed the best overall performance with the lowest Root Mean Square Error (RMSE) of 3.23 and the highest coefficient of determination (R²) of 0.781.
• ARIMAX exhibited relatively higher forecast uncertainty compared to LSTM and RF.
• Feature importance analysis consistently identified temperature, evapotranspiration, and vegetation as important drivers across all prediction models.

Contributions

• Provides a comprehensive spatiotemporal analysis of agricultural drought and its drivers in the North China Plain.
• Quantifies the specific influences of meteorological and environmental variables (e.g., air temperature, soil moisture, EVI) on vegetation health using a multi-method approach including SEM.
• Evaluates and identifies LSTM as a promising and superior approach for agricultural drought prediction in the NCP compared to Random Forest and ARIMAX.
• Highlights the critical roles of thermal stress and soil moisture deficit, offering insights for drought mitigation and sustainable water management strategies.

Funding

[Information not available in the provided abstract.]

Citation

@article{Zafar2026Comparative,
  author = {Zafar, Zeeshan and Wu, Jingwei and Zhang, Shiqiang and Lei, Guoqing and Zha, Yuanyuan},
  title = {Comparative Assessment of Machine Learning and Statistical Models for Agricultural Drought Forecasting and Driver Attribution in the North China Plain},
  journal = {International Journal of Climatology},
  year = {2026},
  doi = {10.1002/joc.70637},
  url = {https://doi.org/10.1002/joc.70637}
}