Clustering of green space qualities: evidence from three Australian cities

Background: Health-relevant features of green space are often referred to under the catch-all term ‘quality’, but this masks how some qualities may cluster geographically and manifest contrasting patterns with socioeconomic circumstances. We measured multiple green space qualities to derive cluster typologies and their socioeconomic patterning. Methods: Novel green space qualities clustering was developed using K-means clustering for “Amenity and Environment” and “Percentage-based Land Cover” schemes. Amenity and Environment qualities spanned access/topography, amenities (positive and negative), biodiversity, safety, surrounding trees, total green space and beaches/coastline. A one-way parametric Analysis of Variance (ANOVA) was conducted to analyse associations between these clusters and Australian Bureau of Statistics Index of Relative Socio-economic Disadvantage (IRSD). A quality check exercise was conducted with team members examining screenshots of cluster results and answering a 4-point Likert scale. Results: Leafy clusters were more likely to be found in areas of lower disadvantage (when IRSD was split into quintiles and reversed so that a value close to 1 is low disadvantage and value close to 5 is high, mean = 1.80, SE = 0.011 at 1600 m). Grassy clusters were more likely to be found in areas of higher disadvantage (mean = 3.44, SE = 0.012). Many inland areas of the three cities were typically low in all domains except safety. The quality check exercise showed that the clustering matched lived experiences between 70.6%–88.2% depending on the scheme and scale. Discussion: Many green space quality domains are co-located. A typology of clusters was shown to relate to socioeconomic circumstances with potential implications for different health benefits from green space exposure.