Modeling engagement-mediated associations between urban green space quality and well-being in a Global South metropolis

Rapid urban expansion has progressively decoupled daily life from natural environments, intensifying concerns about psychological well-being in densely populated metropolitan areas. As a result, urban green spaces (UGS) have increasingly been recognized as critical nature-based infrastructure for mitigating mental distress. Although the mental health benefits of nature are well documented, empirical evidence remains disproportionately concentrated in the Global North, limiting understanding of how perceived environmental quality is associated with psychological well-being through patterns of use and engagement in Latin American metropolitan contexts. This study examines the associations between engagement with UGS and psychological well-being, explicitly testing whether perceived environmental quality (naturalness, infrastructure, and accessibility) is statistically associated with well-being through engagement-mediated pathways. Drawing on survey data from 2,108 residents of the São Paulo Metropolitan Region, Brazil, structural equation modeling showed that higher levels of engagement were positively associated with better well-being (β = 0.352; p < 0.001). Perceived environmental quality, especially naturalness, was associated with well-being primarily through indirect, engagement-mediated pathways rather than through direct associations. Although the strength of these associations varied across population groups, the findings highlight the central role of engagement as an associative pathway linking environmental attributes to psychological well-being. From an urban planning perspective, these results suggest that policies should move beyond the quantitative provision of green areas and prioritize qualitative improvements that foster safe, frequent, and meaningful engagement with urban nature. Future research adopting analytical frameworks explicitly centered on social differentiation may further elucidate how these engagement-mediated associations vary across sociodemographic contexts.