Multi-city comparison of the relationships between spatial pattern and cooling effect of urban green spaces in four major Asian cities

Abstract

Studies on the role of urban green spaces (UGS) in mitigating the urban heat island (UHI) effect demonstrate the influence of two related, but distinct attributes of UGS: the composition (amount), and configuration (spatial attributes and distribution) of UGS patches. While the positive role of UGS amount in providing cooling effect seems unequivocal, it is still not possible to develop a consensus on the role of UGS configuration from the small number of studies that have been conducted to-date. The latter observation could be due to different methodologies, such as choice of landscape metrics and statistical methods used in different studies, as well as differences in urban form of cities. This study addresses the current knowledge gap by using a same set of methodologies applied to four cities of different urban forms: Singapore and Hong Kong as compact cities, and Jakarta, Kuala Lumpur as sprawling cities. Landsat imageries were used to derive land cover and land surface temperature (LST) maps. Different statistical and spatial analysis techniques were also used. We report several novel findings from this study: (1) only four metrics of percentage of landscape (PLAND), area-weighted mean shape index (SHAPE_AM), patch density (PD), and mean Euclidean nearest neighbor distance (ENN_MN) were adequate to explain UGS pattern-LST relationships in all four cities; (2) the relative importance of composition versus configuration of UGS in determining LST seems to be a function of the existing UGS pattern: what the average per unit area of the city is like in terms of patch size (area-weighted mean patch size (AREA_AM)), shape complexity (SHAPE_AM), and distribution (skewness and kurtosis) of patches; and (3) decision tree classifier is a novel and effective method to unravel hidden patterns in complex UGS pattern-LST relationships. Our results also provide insights on UGS management in cities for higher cooling effects. Greening priorities may differ among cities. In cities where configuration is not a determinant of LST (e.g. Kuala Lumpur and Hong Kong), greening may focus on providing conditions necessary for configuration to influence LST and adding greenery, and where configuration already affects LST (e.g. Jakarta and Singapore), the focus may be placed on optimization of UGS by providing larger, simpler in shape, more aggregated and connected, and less fragmented UGS patches.