Urban Network Analysis
- Introduced by
- Andres Sevtsuk & Michael Mekonnen
- Year
- 2012
- Type
- Graph-based centrality analysis of spatial urban networks
- Family
- Spatial network science / urban morphology
What does it do?
Urban network analysis treats a city as a spatial graph of streets and buildings and measures the centrality of each location — how reachable, how central, and how well-connected it is along the actual travel network.
Formalized in the Urban Network Analysis toolbox by Andres Sevtsuk and Michael Mekonnen in 2012, it differs from generic network science by weighting graph nodes with real urban data such as building floor area or population and by computing centralities within bounded search radii. The result is a set of metrics — reach, gravity, betweenness, closeness, straightness — that quantify the structural role of every building or street segment in the urban fabric.
The intuition behind the method
Imagine standing at a building's front door and asking three questions: how much of the city can I reach within a short walk, how many other people's shortest trips pass right by me, and how close am I on average to everywhere else. Urban network analysis answers all three at once by walking the street graph rather than drawing straight lines, and by counting destinations not as dots but as buildings carrying weights like floor area or jobs. A corner shop on a through-street that everyone's routes traverse scores high on betweenness; a flat deep in a cul-de-sac scores low. By repeating these counts within fixed radii, the method reveals which places are structurally advantaged for retail, footfall, or social encounter.
History
Urban network analysis grew out of two converging traditions: the configurational study of street networks in space syntax (Hillier and Hanson) and the formal centrality measures of social network analysis. Andres Sevtsuk and Michael Mekonnen consolidated these into the Urban Network Analysis (UNA) toolbox for ArcGIS in 2012, distinguished by attaching weighted buildings as the graph's nodes and by computing redundant and bounded-radius centralities. The toolbox was later reimplemented for Rhino/Grasshopper and as open libraries, and the approach has been widely applied at MIT's City Form Lab and beyond to model pedestrian movement, retail location, and the spatial structure of cities.
When to use it
Use urban network analysis when you need building- or segment-level measures of centrality and accessibility that respect the real travel network and the distribution of urban activity, for tasks such as predicting pedestrian footfall, siting retail, evaluating walkability, or describing morphological structure. It is well suited to fine-grained studies where buildings carry meaningful weights and where bounded search radii match the scale of the question. It is less appropriate when only a coarse street graph is available without building attributes, when the relevant flows are vehicular and capacity-constrained rather than topological, or when behavioural choice and congestion dominate, in which case a full travel-demand model is preferable.
Limitations and considerations
- Results depend heavily on the chosen search radius and the distance-decay parameter β.
- Topological shortest-path centralities ignore congestion, capacity, and time-of-day variation.
- Quality is bounded by the completeness and accuracy of the underlying street and building data.
- Betweenness on large weighted graphs is computationally expensive and sensitive to network simplification.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Computes centrality at the level of individual buildings or segments, not just abstract nodes.
- Weights destinations by real urban data (floor area, jobs, population) for behaviourally meaningful measures.
- Bounded search radii let the same network be analysed at walking, neighbourhood, and city scales.
- Provides a coherent family of metrics — reach, gravity, betweenness, closeness, straightness — in one framework.
Applications
- Predicting pedestrian and customer footfall along street networks for retail siting.
- Identifying structurally central streets and buildings for urban design and regeneration.
- Measuring walkable access to amenities from each residence at neighbourhood scale.
- Characterizing the morphological structure and connectivity of different city districts.
- Informing transit-station catchment and last-mile network planning.
How it works
How it works
The city is encoded as a graph G of street segments (edges) and intersections, with buildings or parcels attached to the network as origin and destination nodes. Each node carries a weight w — typically building floor area, number of dwellings, jobs, or residents — so that centrality reflects the amount of urban activity at each location rather than treating all points as equal. Edge lengths are real network distances, and the graph is the substrate on which every subsequent metric is computed.
Formula
G = (V, E),\quad w_i = \text{weight of node } i\ (\text{e.g. floor area})Reach counts the cumulative weight of all destinations within a network radius r of a building — a cumulative-opportunity measure of how much activity lies within, say, a 600 m walk. Gravity refines this by discounting each destination with an exponential distance-decay e^{-βd}, so nearer activity counts more, linking the metric directly to spatial-interaction and accessibility theory. Together they describe the local market or catchment a location commands.
Formula
\text{Reach}_i^r = \sum_{j: d(i,j)\le r} w_j,\qquad \text{Gravity}_i^r = \sum_{j: d(i,j)\le r} \frac{w_j}{e^{\beta d(i,j)}}Betweenness centrality counts how many shortest paths between other building pairs within radius r pass through a given node or segment, weighted by origin importance. High-betweenness locations sit on the through-routes of many trips and therefore capture passing footfall and traffic, which is why the measure predicts retail viability and movement potential. It is the network analogue of being on a busy corridor rather than a backwater.
Formula
\text{Betw}_i^r = \sum_{j,k\in V[r]} \frac{n_{jk}(i)}{n_{jk}}\, w_jCloseness centrality is the inverse of the mean weighted network distance from a building to all others within the radius, so central locations are those that are short trips from everything. Straightness compares the straight-line distance δ to the network distance d between locations, rewarding places whose routes to the rest of the city are direct rather than circuitous. Computing all metrics within bounded radii lets the analyst study the city at pedestrian, neighbourhood, or regional scales.
Formula
\text{Close}_i^r = \frac{1}{\sum_{j: d(i,j)\le r} d(i,j)\, w_j},\quad \text{Straight}_i^r = \sum_{j} \frac{\delta(i,j)}{d(i,j)} w_jCommon pitfalls
- Treating all buildings as equal-weight nodes when floor area or population would give very different centralities.
- Reporting a single radius without testing sensitivity to the scale at which the question operates.
- Using Euclidean distance shortcuts instead of true network distance, which defeats the purpose of the method.
- Comparing betweenness values across studies that used different radii, weights, or network cleaning rules.
Frequently asked questions
How does urban network analysis differ from ordinary social network centrality?
Ordinary network centrality treats every node as an abstract, equal point in a graph with arbitrary edges. Urban network analysis grounds the graph in geographic space: edges are real street segments with metric lengths, nodes are buildings or parcels carrying weights such as floor area or population, and centralities are computed within bounded network radii. This makes the metrics spatially meaningful — reach, gravity, and betweenness describe real catchments and movement potential rather than purely topological position.
What is the difference between reach and gravity centrality?
Both measure how much destination activity lies near a building, but reach counts all weighted destinations within a fixed network radius equally, like a cumulative-opportunity measure, while gravity discounts each destination by an exponential distance-decay so that nearer activity counts more. Reach is simpler and threshold-based; gravity is smoother and more behaviourally realistic, and it ties the analysis directly to gravity-based accessibility theory.
How does urban network analysis relate to space syntax?
Both analyse the configuration of street networks to explain movement, and urban network analysis inherits the configurational logic of space syntax. The key differences are that space syntax classically works on axial lines or segments and emphasizes angular and topological depth, whereas urban network analysis works on a metric building-weighted graph and offers an explicit menu of centralities (reach, gravity, betweenness, closeness, straightness). The two are complementary descriptions of urban structure.
Related methods
Sources
Sevtsuk, A., & Mekonnen, M. (2012). Urban network analysis: A new toolbox for ArcGIS. Revue Internationale de Géomatique, 22(2), 287–305.
Open DOI referenceCopied.