A Comprehensive Analysis of Agricultural Drought in the North China Plain: Evaluating Contributing Factors with Drought Indices

Agricultural drought threatens crop production and regional food security, increasing pressure on water-resource management. Although remote sensing drought indices are widely used, their applicability varies, and the dominant drivers and causal interactions remain unclear. Focusing on the North China Plain (NCP), we evaluated different drought indices, validated their performance using the July 2014 and July 2019 drought events, quantified spatial heterogeneity with Geodetector, and identified causal relationships via the Peter–Clark Momentary Conditional Independence Plus (PCMCI+) by constructing causal networks for Henan Province. The results found that, except for the Precipitation Condition Index (PCI), single-variable drought indices generally showed weaker correlations with the Standardized Precipitation Evapotranspiration Index (SPEI) and exhibited poorer monitoring performance than comprehensive indices. Among the comprehensive indices, the Microwave Integrated Drought Index (MIDI) performed best and showed the strongest correlation with SPEI (r = 0.52). Moreover, the results from the Geodetector showed that Precipitation (PRE), Soil Moisture (SM), and Land Surface Temperature (LST) each demonstrated the highest explanatory power in different years, with the highest precipitation cumulative q value was 7.23 for the past 21 years, the main driving factor for the spatial differentiation of agricultural drought. In addition, interactions among the driving factors generally enhanced their explanatory power for the spatial heterogeneity of agricultural drought. Across the five regions of the NCP, causal links between LST and Potential Evapotranspiration (PET) and between PRE and the drought indices were consistently identified, highlighting the joint roles of surface energy conditions and water supply in agricultural drought formation. The Henan causal networks further confirmed this pattern and revealed positive lagged effects of LST on SM.