Street network configuration and traffic noise patterns: a GAM-based analysis of spatial associations in Jiroft, Iran

Traffic noise is not solely a function of traffic volume but also associated with the spatial configuration of urban street networks. This study examines how key spatial features-uninterrupted road length, local alley density, and traffic-calming devices-are associated with variation in urban traffic noise levels. Event-based A-weighted noise measurements were collected at 74 stations in Jiroft, Iran, and analyzed using a generalized additive model (GAM) with smooth terms and tensor product interactions to capture nonlinear and conditional effects. Explanatory variables included road length passed (RLP), road length remaining (RLR), number of alleys (NAS), and counts of speed-reducing devices upstream and downstream (DSB, DSA). The GAM explained 77.9% of noise variability (adjusted R² = 0.699; RMSE = 4.36 dBA). Noise levels increased nonlinearly with RLR (edf = 3.178, p < 0.001) and declined sharply with increasing NAS (p < 0.001). Significant interaction effects showed that the influence of RLP on noise was contingent on local constraints, particularly alley density (ti(RLP, NAS), p < 0.001) and upstream traffic calming (ti(RLP, DSB), p = 0.003). These findings demonstrate that specific street network configurations are strongly associated with traffic noise variability. While the observed patterns are consistent with differences in driver responses to spatial conditions, vehicle behavior was not measured directly, and therefore, the proposed behavioral interpretation should be regarded as a plausible explanatory mechanism requiring further investigation.