Measuring Urban Green Space Exposure Based on Street View Images and Machine Learning

Abstract

Exposure to green spaces (GSs) has been perceived as a natural and sustainable solution to urban challenges, playing a vital role in rapid urbanization. Previous studies, due to their lack of direct spatial alignment and attention to a human-scale perspective, struggled to comprehensively measure urban GS exposure. To address this gap, our study introduces a novel GS exposure assessment framework, employing machine learning and street view images. We conducted a large-scale, fine-grained empirical study focused on downtown Shanghai. Our findings indicate a pronounced hierarchical structure in the distribution of GS exposure, which initially increases and subsequently decreases as one moves outward from the city center. Further, from both the micro and macro perspectives, we employed structural equation modeling and Geodetector to investigate the impact of the urban built environment on GS exposure. Our results highlight that maintaining an appropriate level of architectural density, enhancing the combination of sidewalks with GSs, emphasizing the diversity of regional characteristics, and avoiding excessive concentration of functions are effective approaches for increasing urban GS exposure and promoting human wellbeing. Our study offers scientific insights for urban planners and administrators, holding significant implications for achieving sustainable urban development.