Please enable JavaScript to use this site.

Powered by Esri
An assessment of pedestrian accessibility in the world's main urban centers, aggregated at country and city level. Indicators include the average walking time to different categories of destinations, as well as the proportion of inhabitants that can access each category of services within a 15 minutes walk.
This is an assessment of pedestrian accessibility in the world's main urban centers, aggregated at country and city level. Indicators include the average walking time to different categories of destinations, as well as the proportion of inhabitants that can access each category of services within a 15 minutes walk. The data is produced and maintained by the UN's Sustainable Development Solutions Network (SDSN) as part of the SDG Transformation Center.

Pedestrian accessibility is the extent to which the built environment supports walking access to destinations of interest. This measure is particularly useful for assessing spatial justice in cities, usually represented by underpriviledged communities which are pushed to live in deteriorated urban areas receiving a minor share of public investments and thus low levels of accessibility. Monitoring spatial indicators of pedestrian accessibility helps planners and policymakers evaluate the impacts of urban design and transport interventions and guides targeted interventions towards creating healthy, sustainable cities, and achieving the United Nations (UN) Sustainable Development Goals (SDGs).

Data Sources
Two main sources of data are behind this study.  OpenStreetMap  is used to collect data on pedestrian infrastructure and geographically allocated places of interest (POI): hospitals, schools, supermarkets, restaurants, schools, etc. Pedestrian infrastructure networks are returned by the OpenStreetMap API as networks of nodes and edges, where each node represents a street intersection and each edge represents a segment of road with walkable features. Data on population density for every city is retrieved from the European Commission's 2020  Global Human Settlement Layer (GHSL) . This data is retrieved in the form of a grid of 100m by 100m squares and their associated population density values covering the entire world.

Geographical extent
The geographical extent of a particular city or region often varies according to different authorities and interpretations. Novel projects, such as the Global Human Settlements (GHS) Urban Centres Database (UCDB), seek to establish a consistent, shared geographic definition of “urban centres” globally. This study does not consider municipal boundaries for defining city borders. Rather, it considers "Functional Urban Areas"  as defined by the OECD and the European Commission . The boundaries of Functional Urban Areas consider urbanization factors such as commuting flows and population density, and are less arbitrary than municipal boundaries. For this reason, cities presented here may have a different (and often bigger) shape expected.

Accessibility analysis
To measure accessibility to services for each city, we perform a network analysis on the pedestrian street networks and POIs data to quantify and map accessibility to urban infrastructure at the street intersection level. For each 100m cell from the population grid data, the resulting "walking time" reflects the time that a person residing inside that cell would have to walk for, using the existing pedestrian infrastructure, to reach the first amenity from a given category of services. The analysis was performed using geopandas and pandana python packages.

These calculations were performed for all cities where at least one POI could be identified for each square kilometer. This threshold is applied in order to enforce representativity and accuracy. These scores were then be generalized for each country, by taking the population weighted average of the accessibility score for each point in the population grid. Countries where less than 40% of the urban population is represented after applying the aforementioned thresholds were excluded from the final dataset.