Examining non-linear relationships between campus layout, street design, and walking activity

Built environment (BE) interventions are beneficial for promoting walking activity (WA) and inhibiting sedentary behaviors and physical and mental health issues. However, few studies have investigated the non-linearity and threshold effects of the BE on WA on campus. This study uses gradient boosting decision trees to examine the non-linear relationships between campus BE features and WA, compare the relative importance of campus layout features and street design features for predicting pedestrian volume (PV), and disentangle the most effective ranges of different BE features. We measured the campus layout features of destination density, land use, street connectivity, and spatial configuration using geographic information systems and spatial design network analysis. The street design features and PV are evaluated through field audits. We observe that most campus BEs have non-linear relationships with the WA. The campus layout features have a higher collective relative importance than the street design features in predicting the PV. Different BE features have specific threshold effects on PV. Based on these findings, we propose nuanced and precise campus renewal and planning strategies for WA and walkability promotion.