Fig 1.
The research design.
Fig 2.
Interactions between people in the People-Space-Time (PST) model.
(a) 3D spatiotemporal diagram plotting the trajectories of citizens as location coordinates (x, y) against time (z-axis). (b) Planar representation of the 3D spatiotemporal data in a, plotted onto a street-map.
Fig 3.
Definitions of space users based on their mobility diaries.
(L refers to the location they checked in to; t denotes to the time period for defining a moving window which can be further specified according to the location i and j).
Table 1.
Centrality measures of spatial configuration.
Fig 4.
The study area.
Fig 5.
Aggregated trips between census units in Central Shanghai.
Fig 6.
Main streets (in fuchsia) in ten census areas randomly selected from Central Shanghai.
Fig 7.
Distributions of trip length and duration.
(a) Semi-log scatter plot of the distribution of trip length in the dataset. Red line shows best-fit exponential function. (b) Semi-log scatter plot of the distribution of trip duration in the dataset. Red line shows best-fit exponential function.
Fig 8.
Scatter plots of gate counts against physical co-presence intensity.
(a) Time period, 9:00–10:00; R2 of linear fit = 0.852. (b) Time period, 14:00–15:00; R2 = 0.821. (c) Time period, 21:00–22:00; R2 = 0.807.
Fig 9.
Change of the average presence/co-presence measures across time in Central Shanghai.
((a) Presence density; (b) Co-presence balance; (c) Mean cognitive distance (angular step depth); and (d) Presence/co-presence intensity).
Fig 10.
Physical co-presence intensity maps in Central Shanghai.
Fig 11.
Correlation heatmap of the physical co-presence intensity indices (a) and physical co-presence intensity maps of a selected area (b).
Fig 12.
Changing performance of different fitting models of physical co-presence intensity patterns and co-location distributions across time.
((a) change of the coverage percentage for power-law and lognormal fits over time; (b) change of shape parameter of lognormal fits over time).
Fig 13.
Detected clusters based on the modes of spatiotemporally changing physical co-presence indices in Central Shanghai.
Fig 14.
Street typology in terms of co-presence modes.
((Annotation of streets (a), mean values of the physical presence density (b), cognitive cost (c), and balance degree (d) across time for local and non-local users in each cluster in Central Shanghai).
Fig 15.
Coefficients of the significant centrality variables in the models predicting physical co-presence intensity patterns.
Table 2.
Results of the multinomial logistic regression for the impact of centrality structures on the co-presence typology of streets.