As urban green spaces have significant cooling effects on the urban heat island (UHI), a precise understanding of these effects is necessary to devise precise greenspace strategies for abating the UHI. This paper explores the impacts of different greenspace (trees, grass, and water) patterns on the UHI in Beijing's Olympic Area, using different grid cell sizes and spatial statistical models. Greenspace pattern metrics include percent cover, mean patch size (MPS), mean patch shape index (MSI), edge density (ED), and largest percent index (LPI). The results show that different greenspace metrics have varying effects on surface temperature. The spatial error model (SEM) turns out to be a good choice for estimating the relationship between Land Surface Temperature (LST) and the greenspace metrics. The regression coefficients of these metrics vary with grid cell size. Tree and grass edge densities have opposite effects, which suggest that trees should be planted in smaller clusters, whereas grass should be planted in larger and continuous patches in order to reach maximum LST cooling. The optimal grid cell size is in the [120-240 m] range. These findings can help urban planners mitigate the UHI in a city with limited green space availability.
Keywords: Grass; Spatial autocorrelation; Spatial metrics; Trees; Urban heat island; Water.