Measuring the relationship between morphological spatial pattern of green space and urban heat island using machine learning methods

Land use pattern can substantially shape urban thermal environment. Although previous studies have shown that urban heat island (UHI) intensity will be easily affected by the landscape pattern of green space, the relationship between the morphological spatial pattern of green space and UHI intensity remains to be discovered. Compared with landscape pattern, morphological spatial pattern analysis (MSPA) can reveal more specific details on the configuration and composition of land use. Therefore, this study aims to explore whether the morphological spatial pattern of land use matters to UHI using machine learning methods. Firstly, the morphological characteristics of green space were analyzed based on MSPA. Secondly, the linear associations between UHI intensity and a set of potential influencing factors (including morphological characteristics) were measured according to correlation coefficient. Lastly, the non-linear contribution of the morphological factors to UHI intensity was quantified based on random forest. An empirical case study in a rapidly-urbanized city has revealed the huge influence of morphological characteristics on UHI intensity with benchmark factors considered. The UHI intensity was negatively correlated with the cores, perforations, and loops of green space, but positively correlated with islets. Therefore, a few large core areas would be better than a large number of small islets when the total amount of green space is fixed. In addition, the fragmented patches of green space should be integrated or connected to enhance the cooling capacity. Our findings could offer some insights for UHI mitigation and land use planning, especially when the size of green space cannot be unlimitedly increased.