Skip to contents

Mapping, classifying, and integrating diverse street network datasets: new methods and open source tools for active travel planning

Robin Lovelace

Leeds Institute for Transport Studies, University of Leeds, UK

Zhao Wang

Leeds Institute for Transport Studies, University of Leeds, UK

Hussein Mahfouz

Leeds Institute for Transport Studies, University of Leeds, UK

Juan Pablo Fonseca Zamora

Leeds Institute for Transport Studies, University of Leeds, UK

Angus Calder

Sustrans Scotland

Martin Lucas-Smith

CycleStreets Ltd, Cambridge, UK

Dustin Carlino

Alan Turing Institute, London, UK

Josiah Parry

Environmental Systems Research Institute (Esri), Redlands, CA, USA

Rosa Félix

University of Lisbon
Source: vignettes/gisruk2025.Rmd
gisruk2025.Rmd

Introduction

Active travel is an accessible and cost-effective way to replace short car trips and make transport systems more sustainable. The Network Planning Tool (NPT) for Scotland is a new web-based strategic network planning tool that estimates cycling potential down to the street level. The NPT builds on the functionality of the Propensity to Cycle Tool (PCT) and related tools [@lovelace2017; @goodman2019; @lovelace2024; @félix2025], offering a detailed nationwide cycling potential analysis for Scotland. The NPT is funded by Transport Scotland and developed by the University of Leeds in collaboration with Sustrans Scotland. The tool is open-source and hosted at github.come/nptscot, enabling others to learn from and contribute to the underlying methods to help make transport planning more open and participatory [@lovelace2020a].

A unique feature of the NPT is its integration of multiple layers into a single tool, overcoming limitations of previous strategic network planning tools, which generally focus either on behaviour data (such as the PCT) or physical infrastructure [@vybornova2024; @vierø2024]. The NPT brings together more than a dozen datasets, including Ordnance Survey Mastermap and OpenStreetmap data products to provide comprehensive data on the potential for mode shift, infrastructure, modelled motor traffic levels, and street space at the network level. This paper presents new geographic methods we developed to support this work, with reference to reproducible code that use the new osmactive R package for classifying national-scale OpenStreetMap (OSM) datasets based on attributes and geographic relationships, and the anime Rust crate for astonishingly fast and accurate route network data integration.

Datasets and methods

The NPT uses datasets from diverse sources including from the DfT’s network of traffic sensors, motor traffic counts based a survey evaluating Edinburgh’s roll-out of 20 mph speed limits, the National Travel Survey, and the Scottish Household Travel Survey. We used four key datasets representing the road network for this project:

  • Ordnance Survey OpenRoads, an open access and simplified representation of the road network that is ideal for visualisation
  • OS MasterMap Highways, a more detailed dataset that includes information on road widths and other features
  • OS Mastermap Topography, the most detailed vector dataset that includes detailed information on the geometry of many features of the man-made environment, including curb lines. In the OS Topo layer roads and other pieces of transport infrastructure such as footways (pavements) are represented not as corridor or lane centerlines, but as polygons.
  • OpenStreetMap, a crowdsourced dataset that is continuously updated by community volunteers. The dataset is rich, with tags for width and smoothness of infrastructure, but is not as consistent as the OS datasets.

Ordnance Survey’s MasterMap Highways and Topography datasets provide unparalleled accuracy and detail, but are a struggle to import using consumer-grade hardware and standard tools such as QGIS. To overcome this issue we developed the mastermapr package in collaboration with the government agency Active Travel England (ATE), to efficiently and flexibly imports MasterMap datasets (50 GB compressed). After importing the four route network datasets outlined above, we integrated them using the following steps:

  • Pavement widths were calculated from the OS MasterMap Topography dataset using the function get_pavement_widths() from the osmactive package. This function calculates the width of manmade roadside features associated with each road segment by dividing the area of matching polygons associated with each road segment by the length of the road segment.
  • The get_bus_routes() function from the osmactive package was used to determine the number of bus lanes on each road segment.
  • We used the anime Rust crate to efficiently join the networks, based on alignment and proximity. The use of spatial indexes makes the implementation highly efficient compared with the rnet_join() function in the R package stplanr that we were using previously. See the josiahparry/ANIME codebase on GitHub for details of the new algorithm and its implementation as an R (and planned Python) interface to the new anime Rust crate.

Classifying OpenStreetNetwork ways

Functions including get_travel_network() and classify_cycle_infrastructure() from the osmactive package were used to classify cycle infrastructure types based on the presence of cycle lanes, tracks, and other features (see the code snippet which generates @fig-bristol below).

osm = get_travel_network("bristol")
cycle_net = get_cycling_network(osm)
drive_net = get_driving_network(osm)
cycle_net = distance_to_road(cycle_net, drive_net)
cycle_net = classify_cycle_infrastructure(cycle_net)
plot_osm_tmap(cycle_net)
Cycle infrastructre and painted cycle lanes in Bristol, generated with 5 lines of code using the osmactive package. See interactive version online at github.com/nptscot/osmactive/releases.
Cycle infrastructre and painted cycle lanes in Bristol, generated with 5 lines of code using the osmactive package. See interactive version online at github.com/nptscot/osmactive/releases.

Road width measurements

Two key measurements are needed to assess whether existing roads can accommodate cycle infrastructure: carriageway width and corridor width, as defined below.

Accurate and available carriageway width measurements are important because they determine if proposed infrastructure can fit solely within the carriageway, without the need for moving curbs or other roadside features. The Scottish Cycling by Design (CbD) guidance document outlines three types of cycle infrastructure that run along the carriageway, building on Department for Transport guidance [@departmentfortransport2020]:

  • Cycle track at carriageway level
  • Stepped cycle track
  • Cycle track at footway level

The first option, cycle track that is at carriage level, has additional advantages over stepped cycleway, being often cheaper to construct with no new tarmac required in many cases but requires sufficient width on the existing carriageway to accommodate the cycle infrastructure.

Corridor width captures the carriageway plus any built roadside features, such as footways. The following dplyr query was used to extract the width of manmade roadside features: filter(descriptive_group == "Roadside", make == "Manmade"). From that point, we calculated the width of pavements associated with each road segment by dividing the area of matching polygons associated with each road segment by the length of the road segment, as implemented in the get_pavement_widths() function in the osmactive package.

Minimum cycle track and buffer widths

The corridor width is important because it determines whether, in cases where there is insufficient space on the carriageway, part of the footway or other manmade roadside features may be reallocated for cycle infrastructure, while still maintaining recommended minimum widths for pedestrians. CbD guidance on widths is summarised in table @tbl-cbd.

Table illustrating the minimum width requirements for cycle infrastructure according to the Cycling by Design (CbD) guidance. {#tbl-cbd}
Cycle infrastructure type Absolute minimum width Desirable minimum width
Unidirectional cycle track 1.5 m 2.0 m
Bidirectional cycle track 2.0 m 3.0 m

CbD also specifies buffers that must be accounted for when calculating the effective available width for cycle infrastructure (@tbl-buffers).

Table illustrating the buffer widths for cycle infrastructure based on road type and speed limit according to the Cycling by Design (CbD) guidance. {#tbl-buffers}
Road type / Speed limit Buffer width
30 mph 0.5 m
40 mph 1.0 m
50 mph 2.0 m
60 mph 2.5 m
70 mph 3.5 m

Speed limit data taken from OSM was used to determine the buffer width for each road segment.

Bus routes and road traffic assumptions

The minimum space requirements for motor traffic depend on the uses of the road, including whether it is a bus route and whether there are dedicated bus lanes. Based on the Active Travel England cross section check tool, part of Active Travel England’s open access Excel-based design tools (see acteng.github.io/inspectorate_tools/ for web-based versions), we assumed the following widths for motor traffic:

  • Non-bus routes: 2 × 2.75 m
  • Bus routes without dedicated bus lanes: 2 × 3.2 m
  • Bus routes with dedicated bus lanes: 2 × 3.2 m plus an additional space of n_bus_lanes × 3.2 m for the dedicated bus lanes.

The number of bus lanes was determined from the OSM data with the function get_bus_routes() in the R package osmactive that we developed for this project. The results are illustrated in @fig-edinburgh-bus-lanes, which can also be reproduced with the following Overpass Turbo query:

relation["route"="bus"]({{bbox}});

Results

The result of the data engineering, data integration, modelling and visualisation steps described above is datasets that are visualised interactively in the NPT web application, using pmtiles for efficient rendering of large datasets [@gonçalves2023]. The key components of the NPT are:

  • The “Route network” layer, presenting data on cycling potential on OSM-based and simplified [@lovelace_reproducible_2024] network layers.
  • The “Infrastructure and traffic” layer, which includes data on motor traffic levels and cycle infrastructure.
  • The “Street space” layer, which categorises roads in accordance with the Cycling by Design guidance, specifying the methodology for classifying road spaces and cycle infrastructure, presented in @fig-street-space.
  • The “Core network” layer, representing a cohesive network that could be a priority when planning new infrastructure, building on previous work [@szell2021].
  • The Network Planning Workspace (NPW) web application, which allows users to explore the data and create custom scenarios of infrastructure change.

Conclusion

The Network Planning Tool for Scotland is a cutting-edge web application designed for strategic cycle network planning. A unique feature of the NPT is its integration of multiple layers into a single tool, overcoming limitations of previous strategic network planning tools, which generally focus either on behaviour data or physical infrastructure. The new street space layer represents a step change in access to combined carriageway and corridor widths, the first time this data has been made available at a national level, to the best of our knowledge.

In future work we plan to improve the NPT by incorporating new datasets from a variety of sources, including Scotland’s Spatial Hub. We would like to develop context-specific classifications in the osmactive package to better address the diverse needs of different urban environments and support the roll-out the methods in new places.

References

Acknowledgements

Thanks to Transport Scotland for funding the development of the Network Planning Tool for Scotland, and to the many users who have provided feedback on the tool. Thanks to Sustrans Scotland, and Matt Davis and Angus Calder in particular, for collaborating on the project.

Thanks to the OpenStreetMap community for creating and maintaining the data that underpins the NPT. Thanks to Ordnance Survey for providing the MasterMap data that underpins the NPT (contains OS data © Crown copyright and database rights 2025 OS licence number 100046668).

Biographies

Robin Lovelace is Professor of Transport Data Science at the Leeds Institute for Transport Studies (ITS) and specialises in data science and geocomputation, with a focus on modeling transport systems, active travel, and decarbonisation.

Zhao Wang is a researcher at the Leeds Institute for Transport Studies (ITS). Zhao specializes in machine learning, data science and geocomputation for transport planning and engineering.

Hussein Mahfouz is a PhD student at the Leeds Institute for Transport Studies (ITS) and specialises in data science and geocomputation for transport planning and engineering, with a focus on Demand Responsive Transport (DRT).

Juan Pablo Fonseca Zamora is a PhD student at the Leeds Institute for Transport Studies (ITS) and specialises in new methods for traffic modelling.

Angus Calder is a Senior Mobility Planner at Sustrans Scotland and specialises in active travel planning and infrastructure design.

Martin Lucas-Smith is a Director at CycleStreets Ltd and specialises in open data and open source software for cycling.

Dustin Carlino is an independent researcher and software developer, specialising in open source software development and geographic data science.

Josiah Parry is a Senior Product Engineer at Environmental Systems Research Institute, Inc (Esri) and an open source software developer.

Rosa Félix is a researcher at the University of Lisbon and specialises geographic methods for transport planning, with a focus on active travel and decarbonisation.