Modeling aesthetic ecosystem services in megacity streetscapes

Abstract Urban streetscapes serve as essential public domains shaping community well-being and identity. Our study investigates the nuanced factors influencing the provision of urban aesthetic ecosystem services — a key non-material benefit with implications for mental well-being and urban quality of life. Employing online surveys, deep learning analyses, and spatial modeling, we bridge ground-level perception with landscape-level features, exploring the intricate interplay between green and built areas in shaping aesthetic preferences in São Paulo’s streets — the largest megacity in the Southern Hemisphere and a highly diverse urban environment. We found that the perceived beauty of streets is positively affected by the heterogeneous arrangement of vegetation and built-up areas and by the three-dimensionality of trees — and not solely by the quantity of greenery. Surprisingly, socioeconomic profiles of respondents exhibit no discernible impact on aesthetic evaluations, suggesting consensus across people with diverse social characteristics. Using convolutional neural networks trained on our survey, we predicted aesthetic scores for over 350,000 street images, yielding for the first time a map of the scenic beauty ecosystem service of an entire megacity. This aesthetic map uncovers significant mismatches between supply and demand for aesthetic services, exposing urban inequalities. By revealing these drivers and spatial patterns, our framework provides actionable insights for policymakers — linking perception and landscape-level planning — and offers a pathway to cultivate more socially equitable and aesthetically meaningful urban environments.