基于机器学习量化对城市热岛的形态学影响

Quantitative morphological impact of urban heat island based on machine learning

文章以广州珠江新城为例,利用XGBoost结合Shapley值(SHAP)和部分依赖图(PDP)评估城市形态对城市热岛的影响。结果强调了城市形态属性对热岛的差异性影响,优化建筑配置可降低热岛风险。城市布局因素对热岛表现出阈值影响,要在有限时间内控制建筑表面分数和建筑高度,以减少负面影响。该研究模型关注适应空间规划,抵御城市热岛极端风暴,通过可解释的机器学习,为城市形态与热岛风险的关系提供建议。

Taking the Pearl River New Town in Guangzhou as an example,this paper uses XGBoost,Shapley Value(SHAP)and Partial Dependency Map(PDP)to evaluate the impact of urban morphology on urban heat island.The results emphasize the differential impact of urban form attributes on heat islands,and optimizing building configurations can reduce heat island risks.Urban layout factors have a threshold impact on heat islands,and it is necessary to control the surface fraction and height of buildings within a limited time to reduce negative impacts.This research model focuses on adapting to spatial planning,resisting extreme urban heat island storms,and providing recommendations for the relationship between urban form and heat island risk through interpretable machine learning.