Investigating the spatial effects of zonal factors on road traffic speed variability during peak hour.

Abstract

While traffic speed is a popular metric for evaluating road network efficiency and safety, its application is often limited. Studies typically focus on short stretches of road, specific design features, or particular highways, and even then, only during certain timeframes. This might be due to a lack of comprehensive traffic data. Further, analysing entire road networks and landuse patterns at a macro-level has been less common. Therefore, the study's objective is to understand the factors influencing traffic speed variations at a zonal level using crowdsourced pervasive traffic data. Crowdsourced Speed (CS) data was collected every 10 minutes for thirty consecutive days in the Central Business District (CBD) area in Sydney, Australia. Data on road network characteristics, land use, public transportation, socioeconomic variables, and travel behaviour at the zonal level were gathered from various sources for analysis. Principal Component Analysis (PCA) based local regression models were developed to understand the variables influencing speed variation at the zonal level for both morning and evening peak hours. Two Geographically Weighted Regression (GWR) models were developed to examine spatial variability in the coefficient of variation (CoV) of speed. The spatial distribution of local R² indicates that it ranges from 0.102 to 0.57 for the morning peak hour and from 0.37 to 0.54 for the evening peak hour. The variable coefficients indicate that compared to the morning peak-hour model, the evening peak-hour model exhibits greater consistency and statistical significance (at the 0.10 level) across most zones in explaining the coefficient of variation (CoV) of speed. For the evening peak hour, density, private vehicle travel, commercial activity, land use diversity and income, and road network connectivity factors are significant in explaining the variability. Analysing speed variability at the zonal level across a large network will enable planners and engineers to prioritise zones for traffic improvements.