Shi et al. (2025) Spatial heterogeneity of agricultural drought drivers in irrigation district: A causal inference framework bridging covariation and structural equation modeling

Identification

• Journal: Agricultural Water Management
• Year: 2025
• Date: 2025-11-16
• Authors: Xiang Shi, Yubin Wang
• DOI: 10.1016/j.agwat.2025.109978

Research Groups

• Institute of Water-saving Agriculture in Arid Areas of China, Northwest A&F University, Yangling, Shaanxi 712100, China
• National Engineering Laboratory for Crop High-efficiency Water Use, Northwest A&F University, Yangling, Shaanxi 712100, China
• College of Mechanical and Electronic Engineering, Northwest A&F University, Yangling, Shaanxi 712100, China

Short Summary

This study developed a novel causal framework combining causal covariation and Structural Equation Modeling (SEM) to analyze the spatially heterogeneous drivers of agricultural drought in China's Hetao Irrigation District, revealing that temperature-related factors consistently dominate drought severity while other factors exhibit significant spatial variability influenced by elevation and drainage density.

Objective

• Develop a causal decomposition framework to identify primary drought driving factors at the sub-regional scale and their interrelationships, addressing the spatial heterogeneity problem overlooked in the overall study of the irrigation district.
• Quantify the Temperature Vegetation Dryness Index (TVDI) dynamics while revealing the effects of elevation and drainage infrastructure on TVDI, thereby demonstrating that the heterogeneity within the system is the underlying cause of the agricultural drought pattern.
• Employ Structural Equation Modeling (SEM) to verify the identified causal relationships, systematically integrating these connections with established ecological processes to construct causal pathways.

Study Configuration

• Spatial Scale: Hetao Irrigation District, Inner Mongolia, China (1.19 million hectares), divided into 23 sub-regions based on digital elevation model data. Altitude ranges from 986 m to 1292 m.
• Temporal Scale: 2001–2020 (20 years), with monthly resolution data. Non-stationarity analysis used overlapping 10-year periods (2001–2010, 2006–2015, 2011–2020).

Methodology and Data

• Models used: Causal covariation method (based on Ensemble Empirical Mode Decomposition (EEMD) and Hilbert phase coherence analysis), Structural Equation Modeling (SEM), Theil–Sen regression, Mann–Kendall test.
• Data sources: Multi-source remote sensing and reanalysis data (2001-2020, 1 km or 10 km spatial resolution, monthly temporal resolution):
  • Normalized Difference Vegetation Index (NDVI) (MOD13A2, NASA)
  • Land Surface Temperature (LST) (MOD11A2, NASA)
  • Landcover (MCD12Q1, NASA)
  • Total evapotranspiration (ET) (MOD16A2GF, NASA)
  • Precipitation (PRE) (China 1 km Monthly Precipitation Dataset, National Earth System Science Data Center)
  • Temperature (TEMP) (China 1 km Monthly Mean Temperature Dataset, National Earth System Science Data Center)
  • Potential Evapotranspiration (PET) (China 1 km Monthly Potential Evapotranspiration Dataset, National Earth System Science Data Center)
  • Soil Moisture (SM) (Global 1 km Resolution Surface Soil Moisture Dataset, Third Pole Environment Data Center)
  • Wind Speed (wind) (ERA5-Land Monthly Averaged by Hour of Day, Climate Data Store)
  • Surface Runoff (runoff) (ERA5-Land Monthly Averaged by Hour of Day, Climate Data Store)
  • Digital Elevation Model (DEM) for sub-region division.
  • Evaporative Stress Index (ESI) for validation (derived from MOD16A2GF ET and PET).

Main Results

• Drought Trends (2001–2020): 47.8 % of the Hetao Irrigation District maintained stable drought conditions, 17.4 % experienced aggravated drought, and 34.8 % saw alleviated drought. Severe drought conditions were observed in 2011, 2012, 2018, and 2019, with TVDI reaching up to 0.771.
• Spatial Heterogeneity of Drought: Elevation explained 81 % of TVDI spatial variability (r = 0.904), with higher-elevation zones (>1035 m) facing more severe drought (average TVDI 0.76–0.86). Drainage density significantly reduced drought pressure (r = −0.76), with higher drainage density (e.g., 0.61) correlating with lower TVDI (e.g., 0.53).
• Primary Driving Factors: Temperature-related factors (LST and TEMP) consistently influenced drought severity across all sub-regions. However, Potential Evapotranspiration (PET), Soil Moisture (SM), and runoff exhibited significant spatial heterogeneity in their driving strength for agricultural drought in different sub-regions.
• Causal Pathways and Strengths: The integrated causal covariation and SEM framework demonstrated excellent model fit (P > 0.05, CFI > 0.95, GFI > 0.95, RMSEA < 0.05). LST and TEMP consistently showed strong positive impacts on TVDI across all tested sub-regions (standardized path coefficients for LST often ≥ 0.4). Other factors like PET, wind, and runoff showed spatially variable impacts.
• Robustness Validation: Evaporative Stress Index (ESI) independently verified the reliability of the TVDI-derived conclusions, particularly the dominant role of LST and TEMP. It also highlighted that under strong irrigation, NDVI and PET acted as driving factors, with high vegetation coverage increasing water evaporation and high PET correlating with increased actual evapotranspiration due to crop growth.
• Non-stationarity Analysis: LST and TEMP were the most core and stable factors driving TVDI across different 10-year periods, reflecting persistent energy control. Soil Moisture (SM) showed an increasing negative impact on TVDI over time in most regions, indicating its growing importance in drought alleviation, likely due to water-saving techniques. Runoff exhibited the highest temporal and spatial variability, attributed to human regulation.

Contributions

• Developed and applied a novel causal inference framework integrating causal covariation and Structural Equation Modeling (SEM) to analyze agricultural drought drivers in human-environment coupled systems, overcoming limitations of traditional correlation-based or single-method approaches.
• Revealed significant spatial heterogeneity in agricultural drought patterns and their driving factors within an irrigation district, linking it to natural geographical factors (elevation) and drainage infrastructure.
• Demonstrated the persistent dominance of temperature-related factors (LST, TEMP) as core drivers of agricultural drought, while highlighting the spatially and temporally variable roles of other factors (PET, SM, runoff) influenced by human interventions.
• Provided a replicable model for analyzing drought mechanisms in irrigation districts, offering insights for precise prediction and sustainable water resource management tailored to sub-regional conditions.
• Validated the robustness of TVDI-derived conclusions using an independent indicator (ESI) and through non-stationarity analysis, confirming core drivers while revealing complex irrigation-driven effects on NDVI and PET.

Funding

• The National Key Research and Development Program of China (2022YFD1900802)
• Key Innovation Chain Projects of Shaanxi Province (2024NC-ZDCYL-05–01)
• Key Innovation Chain Projects of Shaanxi Province (2023-ZDLNY-58)
• Shaanxi Province Agricultural Key Core Technology Research Project (2024NYGG002)

Citation

@article{Shi2025Spatial,
  author = {Shi, Xiang and Han, Wenting and Wang, Yubin},
  title = {Spatial heterogeneity of agricultural drought drivers in irrigation district: A causal inference framework bridging covariation and structural equation modeling},
  journal = {Agricultural Water Management},
  year = {2025},
  doi = {10.1016/j.agwat.2025.109978},
  url = {https://doi.org/10.1016/j.agwat.2025.109978}
}