生态学报2026，Vol.46Issue(15)：8094-8110,17.DOI:10.20103/j.stxb.202509062307
基于XGBoost机器学习的公园降温效应及影响因素研究
A study on the cooling effects of urban parks and their influencing factors using XGBoost machine learning
摘要
Abstract
The cooling effect of parks is a crucial approach for mitigating the urban heat island effect.While a substantial body of existing research has predominantly focused on plain cities,the understanding of the cooling effects exerted by different park types within the complex context of mountainous cities,along with a comprehensive elucidation of their multidimensional influencing mechanisms,remains notably insufficient and requires further in-depth investigation.Addressing this research gap,the present study takes 219 parks located in the central urban area of Chongqing,a prototypical mountainous city in China,as its research objects.It innovatively undertakes a comparative analysis of the differences in cooling effects among four distinct park categories:green parks(primarily vegetated),blue parks(featuring significant water bodies),comprehensive parks(integrating green,blue,and other features),and gray parks(dominated by impervious surfaces such as plazas).By integrating two-dimensional(2D)and three-dimensional(3D)influencing factors and employing the XGBoost machine learning model combined with the SHAP(Shapley Additive exPlanations)interpretation method,this study reveals the dominant factors influencing the park cooling effect and their nonlinear interaction mechanisms.The results indicate that:(1)The average cooling intensity was highest in blue parks(2.32℃),followed by comprehensive parks and green parks(1.60℃ and 1.58℃,respectively),and was lowest in gray parks(0.88℃).The average cooling range was largest for blue parks(401.47m),followed by green parks(288.75m).Gray parks exhibited the highest average cooling efficiency(0.27,dimensionless),yet their absolute cooling intensity and range were limited.(2)The overall influence of 2D factors on the cooling effect was superior to that of 3D factors.Park area,perimeter,and water body area were the dominant 2D factors,while building height was the primary 3D influencing factor,showing a significant nonlinear relationship with cooling intensity.Influenced by the data spatial resolution,the effect of vegetation canopy height on the cooling intensity of parks exhibited significant type-specific differences:it showed significant effects in green and blue parks but weaker effects in gray and comprehensive parks.(3)Regarding factor interactions,the interaction between park perimeter and the Normalized Difference Built-up Index(NDBI)had the most significant impact on cooling intensity.This study reveals the unique patterns of park cooling effects in mountainous urban environments,addresses the gap in understanding the influencing mechanisms from a multidimensional perspective,and provides a scientific basis for the precise planning of parks and enhanced thermal environment resilience in mountainous cities.关键词
降温效应/影响因素/城市公园/机器学习/重庆市Key words
cooling effect/influencing factors/urban parks/machine learning/Chongqing引用本文复制引用
龙银珠,戴技才,谭耀湛,郑启月..基于XGBoost机器学习的公园降温效应及影响因素研究[J].生态学报,2026,46(15):8094-8110,17.基金项目
教育部人文社科规划基金项目(20XJAZH002) （20XJAZH002）
