Nonlinear relationship between perceived residential environment and neighborhood park visitation: An analysis of mobile data and street view imagery in Tokyo

Keywords: Urban green spaces, Urban perception, Park visitation, Urban big data, Mobile phone data, Street view imagery (SVI)
Abstract Type: Virtual Paper Abstract

Abstract

Urban green spaces are vital for improving the well-being of residents in densely populated cities. While the benefits of urban parks are well-known, the influence of perceived residential environments on park visitation, particularly through nonlinear relationships, remains underexplored. This study addresses this gap by examining the nonlinear relationship with perceived residential environment on park visitation in Tokyo’s 23 special wards. Using street view imagery (SVI) to assess residents' perceptions, we combined this data with mobile phone records of park visits. Gradient Boosted Decision Trees (GBDT) were used to model the relationships. The results indicate nonlinear relationships between the perceptions and park visitation, with perception metrics accounting for over 41% of the feature importance for visitation frequency and diversity. 'Depressing' and 'beautiful' perceptions were linked to lower visitation, while 'wealthy' and 'lively' perceptions showed threshold effects, with notable reversals at specific points. 'Safety' was found to be relatively less influential. By identifying key perceptual factors and their nonlinear impacts, this study provides actionable insights for urban planners to design data-driven strategies that enhance neighborhood liveliness, foster social interaction, and promote more localized and sustained park use.