sDNA+ 4.1 brings 2D and 3D spatial network analysis to GIS and CAD

Spatial network analysis, the science of measuring how streets, paths, and transit lines shape the movement of people through cities, has received its most significant upgrade in years. Version 4.1 of sDNA+, a widely used open-source toolbox for two- and three-dimensional spatial network analysis, has been formally described in the journal SoftwareX, marking the merger of the previously free sDNA package with its proprietary sister software sDNA+ and bringing an expanded suite of analytical capabilities to a single, freely licensed platform. The update, developed by Crispin H.V. Cooper, James Parrott, Jeffrey Morgan, Zitian Fan, and Alain J.F. Chiaradia, arrives at a moment when urban planners, epidemiologists, transport engineers, and computational designers are demanding ever more sophisticated tools for understanding how the geometry of the built environment shapes human behavior.

At its core, sDNA+ treats the network link as the principal unit of analysis. Rather than abstracting a city into simplified points and corridors, the software works directly with the existing network data that cities and mapping agencies already possess, computing measures of accessibility such as reach, mean distance and closeness centrality, and gravity; measures of flow through bidirectional betweenness centrality; and measures of efficiency through circuity. These statistics can be localized within radial bands bounded both below and above, allowing analysts to ask, for instance, how accessible a street is to destinations within a ten-minute walk without losing sight of the broader pattern of connectivity at larger scales. The software also computes convex hull properties, geodesics, network buffers, accessibility maps, flow bundles, and skim matrices, supported by auxiliary tools that help users prepare raw networks and calibrate models against observed data.

What distinguishes sDNA+ from many competing approaches is the extraordinary flexibility of its distance definitions and weighting schemes. Network distance can be defined as network-Euclidean, angular, or as a hybrid formula built from geometric properties and data attached to individual links. This means analysts can construct distances tailored to specific modes of travel: a pedestrian metric that balances pure distance against the angular change encountered along a route, reflecting the human preference for straight, easy-to-navigate paths; a cycling metric that combines distance, angular change, a nonlinear response to slope or gradient in either one direction or for a round trip, and a nonlinear response to vehicular traffic, which can itself be drawn from the output of another sDNA model; and a public transport metric that penalizes stops and transfers between lines based on network data tagged with bus and metro line information. Weights, meanwhile, can be specified per link, per unit length, or per polyline, drawing on data attached to links, zones, or full matrices. The hybrid metric represents the general case of all these possibilities: any user-specified formula based on network topology, geometry, and attached attribute data. Formulae are parsed at runtime with the muparser C++ library for efficient computation, and a random variance feature, combined with oversampling, can distribute predicted flows across routes of similar length rather than concentrating them on a single shortest path, producing more realistic flow patterns.

The motivation for the version 4.1 release, the authors explain, is twofold: increased interoperability with the software ecosystems that professionals actually use, and a substantially expanded feature set. On the interoperability front, the most visible change is the arrival of a plugin for Rhino and Grasshopper, the widely adopted computer-aided design platform and visual programming environment. Rhino offers advanced three-dimensional editing capabilities well beyond what the toolbox previously supported, while Grasshopper enables parametric, algorithmic, and increasingly generative AI-driven urban design without requiring users to write code. Within Grasshopper, designers can load network data, run sDNA Integral analysis, and color network links according to the results, all through a graphical programming model. The demonstration case highlighted by the developers is a three-dimensional, multi-level, indoor and outdoor pedestrian network for a transport-oriented development, illustrating how the tool can capture movement through stairways, bridges, concourses, and other vertical urban infrastructure that traditional two-dimensional analysis cannot see. The high interoperability between Rhino-Grasshopper and ArcGIS Pro opens new possibilities for urban and transport planners, urban designers, architects, and designers to analyze, coordinate, and optimize spatial configurations in both two and three dimensions. The new plugin substitutes for and expands upon the discontinued AutoCAD support, which was retired after limited uptake; users who still require AutoCAD compatibility are advised to export drawings as shapefiles via AutoCAD Map3D, process them with the free QGIS platform, and re-import, or to use Rhino’s native DWG import and export functionality.

Keeping pace with the geographic information systems world required substantial engineering. The previous version of sDNA supported QGIS version 2, but QGIS 3 introduced architectural changes so fundamental that the sDNA plugin had to be rewritten from the ground up. Jeffrey Morgan led that update, which now brings the toolbox to the current generation of the open-source GIS. On the proprietary side, ESRI replaced its legacy ArcGIS platform with the entirely new ArcGIS Pro, which supports three-dimensional network editing, visualization, and analysis. The sDNA software was updated and successfully tested with ArcGIS Pro, taking full advantage of its 3D environment. While GIS plugins have begun to introduce data-driven three-dimensional raster visualization, they do not yet offer three-dimensional network analysis natively, a gap that sDNA+ now fills across QGIS, ArcGIS Pro, the command line, Python, and Rhino-Grasshopper. The core software is written in C++ with Python and R integration, versioned with Git, and distributed under the AGPL v3.0 license, with the QGIS and Grasshopper plugins released under the more permissive MIT license.

The expanded feature set owes much to the merger with the proprietary sDNA+. All previous features of that sister software are now included in the single open-source release, most notably the hybrid radial and hybrid analytical metrics that allow users to define both locality and analysis type with complete flexibility. To reflect the union, the updated software inherits the sDNA+ name and its version numbering. The release is permanently archived with a Zenodo DOI, and the code is available through public repositories alongside developer documentation and a QGIS plugin, with support handled through the GitHub issue tracker. For a tool whose previous version has accumulated nearly 200 citations of its own methods paper, the transition from partly proprietary to fully open-source represents a meaningful democratization of advanced spatial network science.

The practical reach of sDNA+ is best appreciated through the research it has recently enabled. In the realm of three-dimensional pedestrian modeling, Zhang and Chiaradia used the software to investigate how different three-dimensional pedestrian network encodings and hybrid analyses relate to observed pedestrian volumes in Central Hong Kong, a complex, multilevel, indoor and outdoor, mixed-use, transport-oriented environment where people move through elevated walkways, shopping concourses, and transit interchanges as much as along streets. Their models achieved coefficients of determination between 0.54 and 0.72, and the modeling framework was designed to follow the early network design stage through progressive disclosure, in which information completeness increases as a design matures. This line of work demonstrated that volumetric, three-dimensional network analysis can substantially improve movement prediction in cities that extend vertically as well as horizontally.

Cycling and active travel research has similarly benefited. Earlier studies by Cooper used hybrid metrics to model pedal cycle flows, risk, and mode choice, including predictive spatial network analysis applied to cyclist flows and investment targeting, as well as longitudinal prediction of pedestrian flows following a major city center redevelopment. More recently, Schön and colleagues extended network-based analysis to active travel to school, introducing a novel measure called route betweenness, which assesses the connectivity of entire shortest routes rather than individual links and showed positive associations with children’s active school commuting in Norway’s HUNT study. The same lead author then applied these route-based metrics to evaluate how different alignment options for a proposed bridge in Trondheim would affect active-mode route and mode choice, illustrating how the tool can inform concrete infrastructure decisions before a single beam is placed.

Economic and quantitative spatial modeling represents a third major frontier. Hossain and Cooper developed a two-dimensional spatial network vehicle travel time model for the United Kingdom built purely from distance, junctions, and angular change, and found that it closely approximated Google Maps travel times, with coefficients of determination ranging from 0.919 to 0.997 depending on trip length and the inclusion of congestion. Using these travel times, they computed regional changes in network reach and showed that the relationship between accessibility and economic performance is both mediated and moderated by the proportion of knowledge-based businesses in a region, highlighting limits of transport investment as a driver of regional development. In parallel, Zhou and colleagues employed advanced quantitative spatial modeling to assess the wider economic impacts of transport infrastructure in high-density Hong Kong, demonstrating in one study that integrating complex network science with pedestrian and rail networks significantly boosts productivity estimates, and in another introducing a hybrid structural equation modeling and artificial neural network framework to evaluate the feasibility of long-term urban planning visions. Together, these studies underscore the importance of multimodal accessibility and walkable environments for economic growth.

As cities grapple with climate targets, public health crises, and the rise of generative AI in urban design, tools like sDNA+ are becoming indispensable bridges between network science and real-world decision-making. By uniting proven analytical capabilities, three-dimensional awareness, and open access across the GIS, CAD, command-line, and visual programming worlds, version 4.1 positions spatial network analysis as a shared language for everyone shaping the cities of the future.