Kai Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Kai Zhang has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Chuan Zhang
School of Agricultural Engineering, Jiangsu University, Zhenjiang 212013, China
1 shared publication - Desheng Zhang
Research Center of Fluid Machinery Engineering and Technology, Jiangsu University, Zhenjiang 212013, China
1 shared publication - Haofang Yan
School of Agricultural Engineering, Jiangsu University, Zhenjiang 212013, China
1 shared publication - Hui Wu
Research Centre of Fluid Machinery Engineering and Technology, Jiangsu University, Zhenjiang 212013, China
1 shared publication - Rongyang Wang
Research Centre of Fluid Machinery Engineering and Technology, Jiangsu University, Zhenjiang 212013, China
1 shared publication - Zhen Zheng
School of Agricultural Engineering, Jiangsu University, Zhenjiang 212013, China
1 shared publication
Latest Publications
- Open Access
- 1 August 2026
- Published by MDPI AG in Agriculture
- Vol. 16 (15) , 1660
Regionalizing Meteorological-to-Agricultural Drought Propagation for Agricultural Risk Management Using Event Metrics and Explainable Machine Learning
Developing context-specific drought regionalization is crucial for targeted risk management, as drought evolves as a cascading hazard driven by complex land–atmosphere interactions rather than isolated climatic anomalies. However, conventional regionalization frameworks remain largely static and fail to capture the dynamic propagation from meteorological forcing to agricultural impacts. To address this limitation, we developed a framework that links continuous drought dynamics to discrete drought events, enabling identification of propagation patterns and their associated environmental mechanisms across the Loess Plateau, China. By integrating run theory, dimensionality reduction, clustering, and explainable machine learning, we identified three distinct drought propagation regimes: Propagation Blocked, Disaster Amplified, and Response Desensitized zones. At the regional scale, eco-hydrological factors, particularly vegetation productivity and soil moisture dynamics, showed the strongest attribution signals for differentiating drought propagation regimes. However, regime-specific environmental associations differed substantially: (i) propagation blockage was associated with terrain–vegetation interactions; (ii) disaster amplification was associated with low ecological productivity and declining soil moisture; and (iii) response desensitization was associated with intensive agricultural activities and relatively favorable soil moisture conditions, which may partly buffer vegetation responses to thermal and meteorological stress and create apparent resilience that may mask underlying hydrological vulnerability. SHAP analysis further indicated that topography and thermal conditions were strongly associated with broad-scale differentiation, while eco-hydrological conditions showed stronger associations with local regime-specific responses. Anthropogenic activities may also be associated with altered drought propagation pathways and potential risks of unsustainable water use. These findings highlight drought as a dynamic propagation process rather than a static hazard and provide a basis for targeted drought management strategies.
- 1 August 2026
- Published by MDPI AG in Agriculture
- Vol. 16 (15) , 1660
Open AccessRegionalizing Meteorological-to-Agricultural Drought Propagation for Agricultural Risk Management Using Event Metrics and Explainable Machine Learning
Developing context-specific drought regionalization is crucial for targeted risk management, as drought evolves as a cascading hazard driven by complex land–atmosphere interactions rather than isolated climatic anomalies. However, conventional regionalization frameworks remain largely static and fail to capture the dynamic propagation from meteorological forcing to agricultural impacts. To address this limitation, we developed a framework that links continuous drought dynamics to discrete drought events, enabling identification of propagation patterns and their associated environmental mechanisms across the Loess Plateau, China. By integrating run theory, dimensionality reduction, clustering, and explainable machine learning, we identified three distinct drought propagation regimes: Propagation Blocked, Disaster Amplified, and Response Desensitized zones. At the regional scale, eco-hydrological factors, particularly vegetation productivity and soil moisture dynamics, showed the strongest attribution signals for differentiating drought propagation regimes. However, regime-specific environmental associations differed substantially: (i) propagation blockage was associated with terrain–vegetation interactions; (ii) disaster amplification was associated with low ecological productivity and declining soil moisture; and (iii) response desensitization was associated with intensive agricultural activities and relatively favorable soil moisture conditions, which may partly buffer vegetation responses to thermal and meteorological stress and create apparent resilience that may mask underlying hydrological vulnerability. SHAP analysis further indicated that topography and thermal conditions were strongly associated with broad-scale differentiation, while eco-hydrological conditions showed stronger associations with local regime-specific responses. Anthropogenic activities may also be associated with altered drought propagation pathways and potential risks of unsustainable water use. These findings highlight drought as a dynamic propagation process rather than a static hazard and provide a basis for targeted drought management strategies.