【摘要】 基于实体地域的精准识别，正成为当前收缩城市领域的重要趋势。论文聚焦于广东省城镇实体地域的人口收缩，将城市建成区以外许多容易被忽视的城镇实体纳入研究对象中，以GlobeLand30数据集的人造地表数据识别的城镇实体单元为基础，与校正后的WorldPop人口网格数据实现空间匹配，剖析2000—2020年广东省城镇实体地域人口收缩的时空演化特征，并引入随机森林模型和Shapley加性解释结合的可解释机器学习模型，探索影响人口收缩的主要因素及其阈值效应。结果显示：(1) 2010—2020年广东省城镇实体人口收缩加剧，空间上呈现从地级市市辖区向周边下辖的县及县级市扩散趋势；(2)人口收缩具有明显的规模尺度性，即随着城镇规模尺度下降，人口收缩城镇实体数量上升；(3)影响人口收缩的因素多元，其中人口因素最为重要但随时间推移有减弱的趋势，经济和社会因素的影响呈现非线性关系，区位因素的“规模借用”效应明显，环境因素的作用较弱且表现出非线性关系；(4)各影响因素对人口收缩具有非线性的阈值效应，且对不同类型的城镇实体样本的作用关系也存在差异。研究结果为丰富人口收缩精准识别方法与深化人口收缩驱动因素认识提供参考。更多还原

【Abstract】 Accurate identification of shrinking cities based on physical urban areas represents a significant research trend in the field of urban shrinkage studies. This research focused on population shrinkage within the physical urban areas of Guangdong Province, incorporating often overlooked town-level entities. Utilizing urban physical units identified by artificial surface data from the GlobeLand30 dataset and spatially matching them with calibrated WorldPop population grid data, we analyzed the spatiotemporal change characteristics of population shrinkage in Guangdong’s physical urban areas from 2000 to 2020. A machine learning framework combining random forest modeling and Shapley additive explanations(SHAP) was employed to explore key influencing factors, including their threshold effects and interaction effects. The findings reveal that: 1) Urban population shrinkage in Guangdong Province intensified after 2010, spatially spreading from the sub-districts of prefecturelevel cities to the surrounding counties and county-level cities. 2) Population shrinkage exhibited distinct scale dependency; as the scale of urban areas decreases, the number of shrinking entities increases. 3) Demographic factors played the most significant role, economic factors exhibited nonlinear impacts, social factors exerted positive driving effects, locational factors demonstrated an "agglomeration shadow" effect, while environmental factors showed weaker influence. 4) Influencing factors exhibited nonlinear relationships with population shrinkage, alongside threshold interaction effects between factors. The research findings provide references for enriching methods of accurately identifying population shrinkage and deepening understanding of its driving factors.更多还原