Disentangling the non-linear impacts of natural attributes of urban built environment on life satisfaction

It is widely acknowledged that urban nature could contribute positively to people's subjective life satisfaction. However, extant literature offers limited empirical evidence about the non-linear effects of various natural attributes of urban built environment on life satisfaction. This paper applies a machine learning approach, the gradient boosting decision trees model, to fill in this knowledge gap. Using Beijing as a case study, three key natural attributes of urban built environment are considered, including exposure to greenness within immediate living neighborhood, access to the nearest non-gated community parks, and access to the nearest gated municipal parks. A dataset that combines a life satisfaction survey, GIS-based park accessibility, and neighborhood greenness extracted from georeferenced street view images is constructed and brought into modelling analysis. The findings reveal non-linear effects of all three natural attributes on residents' self-reported life satisfaction. And the association between life satisfaction and access to parks (either gated municipal parks or non-gated community parks) depends upon the level of greenness exposure within respondents' immediate living neighborhood, with significant impacts detected for those who have a relatively low supply of neighborhood greenness. This study contributes to a more nuanced understanding of the intricate association between various natural attributes of urban built environment and life satisfaction. And the empirical findings have practical implications for optimizing urban greening strategy as a cost-effective and feasible intervention to improve population life satisfaction.