Predicting future meteorological drought risk in mainland China using random forest model

作者：Chenyao Huang, Dunxian She, Xinbo Liu, Yanjun Zhang, Yingying Wu, Xinrong Shi, Yiheng Wei · 发表于：Journal of Hydrology Regional Studies · 年份：2025 · DOI：10.1016/j.ejrh.2025.102633 · 被引用次数：3 · 研究领域：Hydrology and Drought Analysis、Climate variability and models、Hydrology and Watershed Management Studies

Study region The study focuses on mainland China. Study focus As climate change exacerbates drought in China, accurate drought risk assessment is crucial for formulating adaptation strategies. The study established a comprehensive machine learning (ML)-based framework for meteorological drought risk assessment, integrating 13 indicators across hazard, exposure, and vulnerability categories. We employed Artificial Neural Networks (ANN) and Random Forest (RF) to evalute historical drought risk. The best-performing model was applied to project future drought risks from 2021 to 2100 under Shared Socioeconomic Pathways (SSP) scenarios (SSP1–2.6, SSP2–4.5, SSP3–7.0, and SSP5–8.5). New hydrological insights for the region The projected future scenarios indicate a significant increase in meteorological drought risk, particularly in western China (QH, GS, NX, SC, and some regions of XJ), with high drought risk under scenarios of higher emissions. Notably, under the SSP3–7.0 scenario, rapid population growth exacerbates exposure, leading to 14.1 % of areas presenting very high drought risk. This ML-based drought assessment framework not only identifies high-risk drought areas effectively but also provides essential insights into the implications of various emission scenarios on drought severity and frequency. These insights are critical for enhancing regional drought resilience and forumulating drought risk management strategies under climate change.