A XGBoost-based composite drought index combining multi-source remote sensing data for drought monitoring in China

作者：Xing Huang, Xianghu Li, Yani Song, Zhenhe Lv, Ligang Xu, Dan Zhang (51283) · 发表于：Journal of Hydrology Regional Studies · 年份：2026 · DOI：10.1016/j.ejrh.2026.103623 · 研究领域：Hydrology and Drought Analysis、Remote Sensing in Agriculture、Soil Moisture and Remote Sensing

Study region China. Study focus The composite drought index (CDI) is crucial for monitoring and evaluating drought, which can overcome the limitations of single indicators and provide a more comprehensive depiction of drought evolution. This study integrated multi-source data based on the XGBoost to construct a monthly-scale CDI at the national scale and analyzed the spatiotemporal patterns of drought in China from 2001 to 2020. New hydrological insights for the region The CDI showed stable performance across contrasting climatic and land-surface conditions and was able to reflect both meteorological drought signals and soil moisture stress. In representative drought events, it provided a more coherent depiction of drought evolution than single-source indices. At the national scale, the CDI revealed marked spatiotemporal heterogeneity of drought in China: drought was generally most severe in autumn, with the highest severity (1.67) and the longest duration (3.44 months), whereas spring showed the lowest drought severity (1.17). Spatially, Inner Mongolia (IM) and Northeast China (NEC) emerged as the main drought-prone regions, with persistently high severity in IM (1.57–1.99). These findings indicate that integrating multi-source drought signals within a machine-learning framework can improve national-scale drought monitoring in China and provide a stronger basis for region-specific drought-risk assessment and early warning.