The Propensity to Cycle Tool: An open source online system for sustainable transport planning

1 Introduction

• Growing evidence links cycling infrastructure to higher cycling rates, while interest and investment in infrastructure are increasing.This context makes the question of where to intervene salient to transport planning.Source: §1 · p. 2
• The paper centers on where to build cycling infrastructure and highlights locally specific evidence about where cycling could grow.It presents online, interactive, publicly accessible tools as a way to support this planning question.Source: §1 · p. 3
• The Propensity to Cycle Tool is an online planning support system for mapping cycling potential.It was funded by the UK Department for Transport and is open source.Source: §1 · p. 3
• The paper connects the tool’s open-source design to wider questions about how models should inform transport decision-making.It frames the PCT as a distinct approach among tools for assessing cycling potential.Source: §1 · p. 3

2.1 The policy context

• Cycle paths can follow shorter, flexible routes through parks or disused railways, increasing the importance of high-resolution map-based tools; local budgets also create demand for tools enabling local cycling targets.Source: §2.1 · pp. 3–4
• The PCT addresses the underdeveloped question of where to build high-quality cycling infrastructure, rather than what to build, and provides evidence on potential route-network capacity needs.Source: §2.1 · p. 4

2.2 Research into cycling potential

• Unlike most reviewed methods, the PCT combines route-based, systematic, quantitative, and scalable analysis, features shared by only notable exceptions Larsen et al. (2013) and Zhang et al. (2014).Source: §2.2 · p. 4
• The PCT uses an origin-destination-level mode-choice model of shifts to cycling and presents its outputs in an online, interactive map, at scales relevant to local planning control.Source: §2.2 · p. 4

2.3 Planning support systems

• Unlike cycling-potential tools designed for a single study region, the PCT aims to be a generalisable and scalable planning support system.Source: §2.3 · p. 8
• The PCT’s open data, open-source code and public online interface encourage transparency and independent participation, but effective use depends partly on deployment resources and user training.Source: §2.3 · p. 9, §2.4 · p. 10

2.4 National context and features of the Propensity to Cycle Tool

• Commissioned to identify where cycling propensity is greatest, the PCT provides evidence to prioritise cycling infrastructure rather than model full transport demand or land use.Source: §2.4 · p. 9
• The PCT estimates cycling potential at area, desire-line and route-network levels, using an interactive map to view results from a few streets to a metropolitan region.Source: §2.4 · pp. 9–10
• The tool route-allocates origin-destination pairs with a cycling-specific routing algorithm and presents estimated health-economic and carbon impacts under future scenarios for investment cases.Source: §2.4 · pp. 9–10

3.1 Processing OD data

• The core input is a 2011 Census table of commuter flows between MSOAs by travel mode, augmented with cyclist gender, area mortality, route distance, and hilliness data.Routes were assigned to the transport network using stplanr.Source: §3.1 · p. 11

3.2 Modelling baseline propensity to cycle

• Baseline cycling propensity was estimated with individual-level logistic regression using route distance, hilliness, nonlinear distance terms, and distance–hilliness interactions.Additional equations estimate mode share for intrazonal commuters and people without a fixed workplace; this baseline underpins Government Target, Go Dutch, and Ebikes.Source: §3.2 · pp. 11, 13

3.3 Scenarios of cycling uptake

• Government Target doubles cycling in line with the 2025 target, while Gender Equality sets women’s cycling share on each OD pair equal to men’s current share.The Government Target scenario is generated by adding observed and modelled cyclists; Gender Equality has its greatest relative impact where cycling is highly gender-unequal.Source: §3.3 · pp. 14–15
• National cycling doubles under the Government Target scenario, but areas with many short, flat trips and below-average cycling more than double while areas with above-average cycling and many long, hilly routes grow less.The scenario combines observed cyclists with modelled counts, producing uneven local growth in both absolute and relative terms.Source: §3.3 · pp. 14–15
• Gender Equality has its greatest relative impact in areas where cycling is highly gender-unequal.Source: §3.3 · p. 15
• Go Dutch applies Dutch-derived scaling factors to baseline propensity, making cycling propensity depend on each OD pair’s distance and hilliness rather than its current cycling level.The Dutch multiplier is larger for shorter trips and was estimated from British and Dutch National Travel Surveys.Source: §3.3 · p. 15
• Ebikes extends Go Dutch by applying additional scaling factors estimated from UK, Dutch, and Swiss travel surveys, with effects varying by trip distance and hilliness.The factors estimate how much more likely a commute is to be cycled by ebike owners than by cyclists generally.Source: §3.3 · p. 16
• The Ebikes scenario extends Go Dutch and applies survey-derived scaling factors that vary with trip distance and hilliness.The factors reflect how much more likely owners were to cycle commute by Ebike than cyclists generally, accounting for longer journeys and hills.Source: §3.3 · p. 16

3.4 Estimation of health and carbon impacts

• The PCT estimates net health effects with HEAT, accounting for deaths avoided through cycling and deaths associated with reduced walking.It assumes non-cycling modes within an OD pair are equally likely to be replaced; for Ebikes it estimates faster speeds and lower health benefit per minute than for conventional cycling.Source: §3.4 · p. 17
• The carbon assessment estimates emissions reductions from decreased car driving, using 0.186 kg CO2-equivalent per kilometre as the 2015 average-car value.It assumes all modes are equally likely to be replaced by cycling.Source: §3.4 · p. 18

3.5 Visualisation, route allocation and network generation

• The PCT displays regional desire lines as centroid-to-centroid flows aggregated across both directions, and allocates them to fast or quiet routes.The quiet option favors smaller, quieter roads and off-road paths; the fastest option was selected, informed by recommendations favoring direct routes.Source: §3.5 · pp. 18, 21
• The Route Network aggregates overlapping fast routes, but excludes within-zone commuters, commuters without fixed workplaces, and commuters working in another region.Its values also depend on adjustable OD-pair distance and trip-count thresholds; generated routes do not necessarily match cyclists’ current paths.Source: §3.5 · p. 21
• On Otley Road, Go Dutch raises cycling from 73 to 296 commuters on one route and from 546 to 1780 in the Route Network; nationally, that layer covers around two thirds of cycle commuters.The authors advise greater confidence in relative than absolute Route Network values.Source: §3.5 · p. 21, Figure 5 · p. 22

4.1 Model output tabs

• The interactive map displays cycling potential at area, desire-line, route, and route-network levels; popups give element-level figures, while Lines provides sample OD-pair data.The Areas tab provides area-level data, and Model output supplies regional statistics, diagnostic plots, and model results.Source: §4.1 · p. 22

4.2 Trip distance distributions

• Oxfordshire has more short trips under 5 km, while West Yorkshire has more longer-distance commutes and lower cycling; Go Dutch and Ebikes narrow regional cycling-rate differences.The paper relates Oxfordshire’s short-trip profile to its relatively high cycling level.Source: §4.2 · p. 23, Figure 6 · p. 24

pacts

• Under Go Dutch, Leeds cycling patterns shift substantially south and more closely represent short-distance trips across the city than under the Government Target scenario.Both scenarios’ desire lines focus on the city centre.Source: pacts · p. 24
• 88 commuters use York Road in the baseline, rising to 318 under Government Target and 1,426 under Go Dutch; Otley Road only triples under Go Dutch.The route-network outputs illustrate how estimated commuter cyclist flows on road segments change across scenarios.Source: pacts · p. 24, Figure 8 · p. 27
• OpenCycleMap shows existing cycle provision, the Index of Deprivation maps social inequalities, and Satellite imagery helps assess road space for paths.The Satellite basemap shows road widths and land uses along desire lines.Source: pacts · p. 26, Figure 10 · p. 29
• The PCT lets users rank routes by health or carbon impacts: health criteria favor OD pairs with low current walking, while emissions criteria favor those with high baseline car use.Health benefits do not rise simply in proportion to the number of people cycling, because longer trips bring greater health benefits than short ones.Source: pacts · pp. 28–29

5 Discussion

• The PCT reports health-related benefits alongside trip counts because health benefits do not always rise simply in proportion to additional trips.Source: §5 · p. 30
• Illustrative uses show that cycling demand’s spatial distribution may shift as cycling grows, demonstrating the tool’s potential utility.The discussion describes settings in which demand is likely to shift as cycling grows.Source: §5 · p. 30
• For planning, it can translate national targets into local aspirations, compare regional potential, support route business cases, and inform network capacity planning.Source: §5 · p. 30
• The tool relies on 2011 Census origin-destination data, so results are limited to commuting and may miss recent travel-behavior shifts; its interface offers only discrete scenarios.Source: §5 · p. 30
• Open-source development could add scenarios, including medium-term and local targets, while allowing stakeholders to modify scenarios for their own purposes.Source: §5 · pp. 30–31
• Potential extensions include individual-level estimates and additional trip purposes, such as an education layer to prioritize safe routes to school.Source: §5 · p. 31
• Deployment in new cities, regions, or countries depends on suitable origin-destination data and routing services that estimate cyclable routes.Source: §5 · p. 31
• Authorities, campaign groups, and the public can access the same evidence base, which the authors hypothesize may encourage informed, rigorous debate.Source: §5 · pp. 31–32

• JW and RL originated the project concept, and RL led manuscript writing with input from all authors.Source: Author contribution statement · p. 32
• AG led the PCT model and the generation of cycling-level, health-gain, and carbon-impact estimates for each scenario.Source: Author contribution statement · p. 32
• JW and AG led methods for calculating health impacts, while JW, RL, AG, and RA contributed to cycling-uptake rules.Source: Author contribution statement · p. 32
• RL led spatial visualisation of model estimates and coordinated online-tool development; RA led policy implementation and practitioner feedback, and AA and NB led the user interface.Source: Author contribution statement · p. 32

• RL’s contribution was supported by the Consumer Data Research Centre, and JW’s by an MRC Population Health Scientist Fellowship.Source: Funding · p. 33
• JW’s and AA’s contributions were supported by CEDAR, a UKCRC Public Health Research Centre of Excellence funded by several research and health charities and councils.Source: Funding · p. 33
• AG’s contribution was supported by an NIHR post-doctoral fellowship; the authors state that the paper’s views do not necessarily represent those of the listed funders and organisations.Source: Funding · p. 33