Revisiting the Impact of the Built Environment on Noise Exposure: From Static Spaces to Individually Tracked Mobility

Although many studies have examined how the built environment influences neighborhood noise levels, few have examined individual-level exposure across indoor, outdoor and in-vehicle contexts. This limitation in research design could lead to the uncertain geographic context problem (UGCoP), further suggesting that the influence of built environment features could vary across different contexts. Moreover, built environment–noise relationships might be nonstationary, contradicting the linearity assumption in prior research. To address these gaps, this study integrates Global Positioning System trackers with wearable noise sensors to capture individual trajectories and corresponding noise exposure at one-minute resolution in Hong Kong. Using 1.91 million observations, this study developed a multiple eXtreme Gradient Boosting machine learning framework in combination with SHapley Additive exPlanations analysis to compare the importance and nonstationary effects of built environment factors on individual noise exposure across three contexts. The main findings are as follows. First, the key built environment factors influencing individual noise exposure differ across contexts. Population and intersection density are most influential in indoor contexts; green space and transportation land use are predominant in outdoor contexts; and population density and transportation land use remain critical in in-vehicle contexts. Second, many built environment–noise relationships exhibit nonstationary effects, often following patterns of diminishing marginal returns or U-shaped curves. Third, individual noise exposure varies by hours and seasons, and age, education, and household size significantly influence exposure levels. This study advances the understanding of how the built environment shapes noise exposure at the individual level and provides practical insights for designing more sustainable and livable neighborhoods.