Spatial Design Network Analysis (sDNA)
- Introduced by
- Crispin H. V. Cooper & Alain J. F. Chiaradia
- Year
- 2020
- Type
- Link-based spatial network analysis of street and path networks
- Family
- Spatial network analysis / configurational analysis
What does it do?
Spatial Design Network Analysis (sDNA) is a toolkit for analysing street and path networks as link-based spatial graphs, measuring how individual road segments function as routes and destinations within the larger network.
Developed by Crispin Cooper and Alain Chiaradia at Cardiff University, it computes closeness- and betweenness-style measures over geometrically accurate, optionally three-dimensional networks, using hybrid distance metrics that blend metric length, angular turn cost and topological steps. By weighting links and analysing them within chosen radii, sDNA predicts pedestrian and vehicle flows, land values and accessibility, bridging the configurational tradition of space syntax with mainstream geographic-information-system network analysis.
The intuition behind the method
Picture a city's streets not as abstract points but as the actual road segments people drive and walk along, each with a real length, gradient and set of turns. Some segments are natural through-routes that lie on the shortest or straightest paths between many other places; others are quiet ends that few journeys pass through. sDNA measures exactly this for every link: how easily it reaches the rest of the network within, say, a 10-minute walk (closeness), and how often it sits on the best routes between all the other links (betweenness, or 'flow potential'). Because it works on geometrically faithful, even 3D, networks and lets you weight links by population or floorspace, the numbers it produces line up closely with where people and traffic actually go.
History
sDNA grew out of the configurational tradition of space syntax but was built to address its known weaknesses — sensitivity to how axial lines are drawn, lack of geometric accuracy, and limited integration with mainstream GIS. Crispin Cooper and Alain Chiaradia developed it at the Sustainable Places Research Institute, Cardiff University, defining analysis on geometrically faithful, link-based and three-dimensional networks with flexible hybrid distance metrics and radius-bounded, weighted closeness and betweenness. Validated against pedestrian and vehicle flow counts, land values and road-safety outcomes, the toolkit was released for ArcGIS, QGIS, AutoCAD, the command line and Python, and was documented in the 2020 SoftwareX paper that consolidated its open-source distribution.
When to use it
Use sDNA when you want to analyse a street or path network as it really is — geometrically accurate, with true lengths, turns, gradients and link weights — to predict movement, accessibility and centrality and to compare design or infrastructure scenarios. It is well suited to modelling pedestrian and vehicle flows, informing active-travel and road-safety schemes, and relating network structure to land value, retail vitality and severance. It is less appropriate when only a coarse or topologically simplified map is available, when the network is small enough that simple graph measures suffice, or when behaviour is dominated by factors outside network geometry (pricing, scheduling, individual constraints), where demand-modelling or time-geographic methods are more direct.
Limitations and considerations
- Output quality depends heavily on the geometric and topological accuracy of the input network.
- Choice of radius, distance metric and link weighting strongly shapes results and requires justification.
- Network-only measures omit signals, capacity, congestion and mode-specific behaviour unless added externally.
- Large, dense networks at many radii are computationally and memory intensive to analyse.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Works on geometrically accurate, link-based and optionally 3D networks rather than abstracted graphs.
- Hybrid metric/angular/topological distance lets the model match how networks are actually travelled.
- Radius-bounded, weighted closeness and betweenness predict pedestrian and vehicle flows accurately.
- Integrates with GIS, CAD, the command line and Python, fitting both research and professional workflows.
Applications
- Predicting pedestrian and vehicular movement from street-network configuration.
- Designing and appraising active-travel, walking and cycling networks.
- Relating network centrality to land value, retail turnover and urban vitality.
- Assessing road-safety and collision risk in relation to through-movement potential.
- Comparing masterplan and infrastructure scenarios by their accessibility and flow effects.
How it works
How it works
sDNA measures the 'distance' along a path as a flexible, hybrid cost summed over the links it traverses: a metric component (physical length, optionally adjusted for gradient in 3D), an angular component (the cumulative turn angle, penalising sharp deviations), and a topological component (number of links). By choosing the weights on these components an analyst selects a metric, angular or topological model, or any blend, to match how the network is actually travelled.
Formula
d(i,j) = \sum_{k \in p(i,j)} \big( m\, l_k + a\, \theta_k \big)Analysis is always carried out within a radius R around each link — a metric, angular, or topological bandwidth (e.g. 800 m, or 5 minutes) that defines its local network neighbourhood. The Network Quantity of Pedestrian/Destination Activity sums the weights of all links reachable within that radius, counting each link's contribution by half its length to avoid double counting, giving a weighted measure of how much network (and activity) lies within reach.
Formula
NQPDA_x = \sum_{y \in R_x} P_y\, \frac{l_y}{2}Closeness is expressed as Mean Angular (or metric) Distance: the weight-and-length-weighted average distance from a link to every other link within the radius. Low mean distance marks links that are easy to reach from their surroundings — accessible, central segments — while high mean distance marks remote ones. Inverting or rescaling this yields the familiar closeness-style accessibility used to explain destination potential and land value.
Formula
MAD_x = \frac{\sum_{y \in R_x} P_y\, d(x,y)\, \frac{l_y}{2}}{\sum_{y \in R_x} P_y\, \frac{l_y}{2}}Betweenness counts, for each link x, how often it lies on the shortest (least-cost) routes between all origin–destination pairs within radius, weighted by the origin and destination link weights and split where ties occur. High betweenness identifies through-movement routes — the segments that carry the bulk of trips passing between other places — and these flow-potential values correlate strongly with observed pedestrian and vehicular movement, making sDNA a movement-prediction tool.
Formula
Bt_x = \sum_{i}\sum_{j \in R_i} P_i P_j\, \frac{\mathbb{1}\{x \in p(i,j)\}}{\,|p(i,j)|\,}Common pitfalls
- Running sDNA on a topologically broken network (unsnapped junctions, gaps) so routes cannot connect correctly.
- Reporting a single radius when conclusions change across local, neighbourhood and city-wide bandwidths.
- Confusing closeness with betweenness — accessibility to destinations versus through-movement potential.
- Comparing raw betweenness across networks of different size or density without appropriate normalisation.
Frequently asked questions
How does sDNA differ from traditional space syntax?
Both analyse the configuration of street networks, but sDNA is link-based and geometrically accurate rather than relying on hand-drawn axial lines. It uses real segment lengths, turn angles and even 3D gradient, lets you blend metric, angular and topological distance, supports weighting links by population or floorspace, and always works within explicit radii. This makes results more reproducible and better integrated with GIS, while still producing the closeness- and betweenness-style configurational measures that space syntax pioneered.
What is the difference between closeness and betweenness in sDNA?
Closeness (reported as mean distance or related accessibility measures) describes how easily a link reaches everything else within its radius — it captures destination accessibility and centrality. Betweenness counts how often a link lies on the shortest routes between other links, capturing through-movement or flow potential. A quiet cul-de-sac near many shops can have high closeness but low betweenness, whereas a busy arterial may have high betweenness; the two together describe a segment's dual role as place and route.
Why does the analysis radius matter so much?
sDNA computes every measure within a bounded radius — a metric, angular or topological bandwidth such as 800 metres or 5 minutes — that defines each link's network neighbourhood. Small radii reveal local, walkable structure relevant to pedestrians, while large radii capture city-wide vehicular structure. Because a link can be locally central but globally peripheral (or vice versa), conclusions depend on the radius chosen, so analysts report a range of radii rather than a single value.
Related methods
Sources
Cooper, C. H. V., & Chiaradia, A. J. F. (2020). sDNA: 3-d spatial network analysis for GIS, CAD, Command Line & Python. SoftwareX, 12, 100525.
Open DOI referenceCopied.