Fig 1.
Workflow diagram illustrating the sequential stages of the research methodology.
Data from various sources were integrated to analyze service accessibility and connectivity, visualized to explore spatial patterns across urban areas.
Fig 2.
Example of a node n and its closest Points of Interest (PoIs).
Three example service categories are shown: purple triangles, blue diamonds, and green squares. The closest PoIs in each category are respectively i (at walkable distance of 5 minutes), a (at 7 minutes), and x (at 10 minutes). In this example, the PoI-proximity because residents in n can walk to a PoI for each of the given categories within 10 minutes. Equivalently, the largest isochrone for n that includes at least one service in each category is
. Base map and data from OpenStreetMap and OpenStreetMap Foundation.
Table 1.
Summary of the accessibility and connectivity metrics used in the study.
Table 2.
Summary of nodes, edges, POIs, and categories across different cities. Bold values indicate global max values.
Fig 3.
Cities world-wide ranked by average PoI-proximity(), weighted by population.
Each city is represented by a box-plot showing the distribution of values across its area. Cities are sorted from lowest (left) to highest (right) weighted average proximity. Box-plots colors indicate the continent a city belongs to.
Fig 4.
Cumulative PoI-proximity curve for six representative cities.
Each curve shows the the percentage of the population within a given PoI-proximity threshold. The six cities include those with the highest and lowest area under the curve (AUC) values, plus one city at each of the 20th, 40th, 60th, and 80th percentiles.
Fig 5.
Distributions of the accessibility metrics for six representative cities.
Shown are (PoI-proximity),
(PoI-density within a 15-minute walk),
(PoI-entropy within a 15-minute walk), and
(PoI-accessibility within a 15-minute walk) across all nodes in each city.
Fig 6.
Correlation heatmaps for key urban metrics across 81 cities.
The matrices show pairwise correlations between PoI-proximity, PoI-density, PoI-entropy, PoI-accessibility, closeness, and population using (a) Pearson correlation coefficients and (b) Kendall‘s tau. Colors range from –1 (blue, strong negative correlation) to 1 (red, strong positive correlation), with gray indicating near-zero correlation. Reported values indicate the corresponding correlations and p-values.
Fig 7.
PoI-accessibility ranking and its spatial patterns for the six representative cities.
(a) Cities world-wide sorted in descending order of average PoI-accessibility weighted by population, with
. (b-g) Heatmaps of
values for the six representative cities at the intersection level. Points heatmaps are color coded by population-weighted
(scale shown on the right): darker areas indicate better access to services, white areas have no (or a negligible number of) residents, and red areas represent nodes with no access to services. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Fig 8.
Comparison between PoIs within 15-minute isochrones for two example intersections in Paris.
(Left) An intersection with good PoI-proximity but poor PoI-entropy, and PoI-density. (Right) An intersection with poor PoI-proximity but good PoI-entropy and PoI-density. Colored areas refer to the presence of PoIs in a certain category. Base map and data from OpenStreetMap and OpenStreetMap Foundation.
Fig 9.
Closeness ranking and its spatial patterns for the six representative cities.
(a) Cities world-wide sorted in descending order of average closeness , weighted by population. (b-g) Heatmaps of
values for the six representative cities at the intersection level. Points are color coded by population-weighted
(scale shown on the right): darker areas indicate higher closeness, while white areas correspond to nodes with no (or a negligible number of) residents, no connection to the city network, or both. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Fig 10.
Comparison of PoI-accessibility and normalized closeness across cities.
Bubble chart showing cities’ PoI-accessibility and normalized closeness. Marker size is proportional to city population, and marker color refers to the geographical region a city belongs to. Outliers, identified using an elliptic envelope (with contamination parameter of 0.22), are textured. Text annotations’ backgrounds are color-coded by geographical region, except for the six representative cities (light gray) and outliers discussed in the text (white).
Fig 11.
PoI-accessibility vs. normalized closeness for Paris (a, b), Turin (c, d) and Vancouver (e, f).
Left panels: each bubble corresponds to a neighborhood detected by the Infomap algorithm and color-coded accordingly. Axes are ranged differently, to simplify the comparison between neighborhoods within a city. Right panels: nodes represent intersections, with sizes proportional to the number of residents. Population distributions are shown as histograms along the axes. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Fig 12.
PoI-accessibility vs normalized closeness for Ottawa (a, b), Melbourne (c, d), and Houston (e, f).
Left panels: each bubble corresponds to a neighborhood detected by the Infomap algorithm and color-coded accordingly. Axes are ranged differently, to simplify the comparison between neighborhoods within a city. Right panels: nodes represent intersections, with sizes proportional to the number of residents. Population distributions are shown as histograms along the axes. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Table 3.
Summary statistics of the six representative cities. Mean and the standard deviation of the PoI-accessibility and normalized closeness are aggregated by Infomap clusters. The Pearson r and Kendall’s τ coefficients are calculated between the two variables.
Fig 13.
PoI-accessibility vs. normalized closeness for the nine most populous Italian cities.
Two bubble charts comparing cities by PoI-accessibility (y-axis) and normalized closeness (x-axis) with marker size proportional to city population (left) and to average income (right). Markers are color-coded according to three geographical macro-regions (North, Center, South). A central map of Italy displays the location of the nine cities, with sizes proportional to city population, providing geographical context for the comparison.
Table 4.
Summary statistics of the top nine Italian cities by population. The mean and standard deviation of the PoI-accessibility, normalized closeness, and the average income are aggregated by Infomap clusters. The Pearson r and Kendall’s τ coefficients are calculated between the first two variables.
Fig 14.
PoI-accessibility vs. closeness by neighborhoods in Bari (a), Bologna (b), and Florence (c).
Markers are color-coded according to the administrative neighborhoods (as shown in the embedded city map), and sized by population (left) and average income (right). Axes are ranged differently, to simplify the comparison between neighborhoods within a city. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Fig 15.
PoI-accessibility vs closeness by neighborhoods in Genoa (a), Milan (b), and Naples (c).
Markers are color-coded according to the administrative neighborhoods (as shown in the embedded city map), and sized by population (left) and average income (right). Axes are ranged differently, to simplify the comparison between neighborhoods within a city. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Fig 16.
PoI-accessibility vs closeness by neighborhoods in Palermo (a), Rome (b), and Turin (c).
Markers are color-coded according to the administrative neighborhoods (as shown in the embedded city map), and sized by population (left) and average income (right). Axes are ranged differently, to simplify the comparison between neighborhoods within a city. Maps contain information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.
Fig 17.
Kendall’s Tau correlation between different city rankings across studies.
Correlation between city rankings based on the accessibility measures presented in this paper, to Bruno et al. [25], and Nicoletti et al. [11]. The comparison is restricted to the subset of cities included in all three studies. Colors range from −1 (blue, strong negative correlation) to 1 (red, strong positive correlation), with grey indicating near-zero correlation. Reported values indicate the corresponding correlations and p-values.