In the UK, the National Cycle Network (NCN) provides a major backbone for cycling and wheeling across both urban and rural areas. Covering over 16,000 miles of signed and managed routes, the NCN is used for both leisure and utility cycling, significantly with 5000 miles of the network completely traffic free. It comprises shared use paths, segregated cycle lanes, streets with low traffic, and off-road tracks, all with signage. The “on-road” sections typically are found in rural areas, whilst the traffic free routes are found in urban areas, but there is much variation in the quality and type of provision (Sustrans, 2023). However, contrary to modern design principles (Department for Transport, 2020), much of both the NCN and wider UK cycle network contain barriers. These barriers, often in the form of physical obstructions, are discriminatory against a host of active modes, such as wheelchairs, tri-cycles and many adapted cycles (Sustrans, 2020; Wheels for Wellbeing, 2022).

Historically, barriers have been implemented on active travel corridors due to planning stipulations at the time of construction (BikeBiz, 2016), often as a result of perceived issues of crime, such as trespass by motorcycle riders, as well to prevent cyclists from approaching intersections or tight passages on the network at a high speed. The design of such barriers, however, often focuses on reducing the passable width of the infrastructure or even requiring the need to dismount a cycle. This has led to the unintended consequence of rendering the routes inaccessible for users who cannot pass through the barrier due to their choice of cycle, be that through requirements as a disabled person, or simply a choice based on utility and function.

Currently, over 7000 barriers have been identified on the network in the UK, roughly equating to a barrier every 0.4 miles. These barriers have been identified as part of a volunteered geographic information (VGI) initiative curated by Sustrans, now renamed The Walk Wheel Cycle Trust, an active travel and sustainable mobility charity organisation, who are custodians of the network (Sustrans, 2018). Sustrans aims to remove all barriers on the NCN by 2040, with the target of making the network accessible and useable for all. Progress remains slow, however, with no established methods for determining which of these barriers are the highest priority for removal. This has led to sporadic and often uncoordinated work on network improvements, and therefore likely has not obtained the highest potential gains in access for users. To tackle this, this paper presents a decision support tool to assist the prioritisation of the barrier identification and removal process, with the goal of enabling a greater rate of change in the ability for non-standard cycle users to access and benefit from the existing NCN in the UK.

Burns, Oram and Claris (2019) state that 31% of disabled people would like to start cycling but potential users are faced with limitations in accessing active travel infrastructure, such as barriers on routes. Transport for London (2012) asked survey respondents whether they could ride a bike: 70% of disabled people responded “Yes”, highlighting that the issue lies not with ability to move via cycling but with the opportunity to use cycling as a mode of transport. In the UK, there is a low rate of observed cycling per week for people with disabilities of 9%, compared with 17% for non-disabled people (Cycling UK, 2025). Moreover, despite cycling forming a key part of physical activity, transport and health promotion for disabled people, it is largely not catered for in active travel planning in the UK (Inckle, 2020). Cycling can be easier for disabled people than other modes such as walking or using public transport, and thus accessing cycling infrastructure is an important component of daily living (Inckle, 2022; Andrews and Clement, 2021). However, cycling continues to be difficult or unfeasible as a result of challenging and unaccommodating infrastructure.

These issues commonly relate to the prevalence of physical barriers that do not allow a user to continue onto the designated foot/cycle path (Figure 1), particularly if using a non-standard cycle. Roadblocks, barriers and gates are frequently too narrow for use by many users (Orving et al., 2019; Clayton and Parkin, 2016; Cox and Bartle, 2020) and Wheels for Wellbeing found inaccessible infrastructure is the greatest barrier to cycling for disabled cyclists (Wheels for Wellbeing, 2021). Design standards from the Department for Transport, the UK government department responsible for transport policy, mandate that corrections must be made where effective widths are below 1.5m, however audit data from Sustrans (Figure 2) indicates that over 5000 barriers still exist across the UK with a critical width lower than this (Department for Transport, 2020).

Barriers were originally implemented predominantly as a measure against anti-social behaviour, such as the illegal use of motorbikes on traffic-free routes. Crime is often the source of conflict in the context of barriers on cycle networks, as the perceived benefit of keeping crime away is seen to outweigh the benefits of allowing all users of the network to use the routes fully (Brice, 2020). Intervention and monitoring studies have highlighted, however, that the perceived level of crime is far higher than the reality on the ground. In a barrier redesign/removal trial in York over the period of two months, for every illegal motorised trip on the network, 22,212 legitimate trips occurred (Sustrans, 2024). Through video recording at the site of barrier removal, a total of four illegal motorised trips were recorded, while 45 additional trips per day by active modes were recorded that could not have taken place prior to the changes (due to the barriers now being accessible for a wider number of users, allowing riders of adapted cycles to pass through). This study also interviewed users and found that roughly half of all users knew of people who could now use the routes who would have previously been excluded due to their use of non-standard cycle or other mobility aid, meaning access would not be possible.

Barriers vary in their impacts as a result of their spatial location, their inherent design and the user interacting with them. The form of adapted cycles or mobility aids can have impacts on which barriers are passable or not (Cox and Bartle, 2020). For example, a tri-cycle may struggle to pass through a narrow “A-frame” barrier but be able to pass through a chicane. Where the barrier occurs can also be of importance. As highlighted by Larrington-Spencer (2025), barriers can restrict access to what would be good quality infrastructure, if the user was able to pass through them. Living within close vicinity to such a situation prevents cycling to a much wider area. Users are essentially in a geographic lottery as to if they can access cycling or not locally, regardless of how committed to cycling they are.

When measuring access within a cycle network, a typical analysis will look to assess the quantity and quality of the network and the destinations available (Lowry and Loh, 2017). These analyses, however, do not tend to consider that some portions of network may not be accessible by some user groups, instead calculating access only for the average user.

Some studies look to address this by measuring the impact of barriers on networks. Direct barrier effects have been defined as the impacts created by transport infrastructure on physical access to activities for people (and opportunities for business) (van Eldijk et al., 2020). The effects of a barrier on a given area have been quantified by many methods using geospatial analysis, commonly including at least one of four measurements: travel time or distance, amenity choice, catchment size, and service efficiency (van Eldijk et al., 2020; Reyier, 1987; Jarlebring et al., 2002; Clark, 1991; Anciaes, 2013).

Travel time or distance requires a cut-off point to be defined. Both et al. (2022) observed cycling for 6 kilometres at a speed of 13 km/h to be a distance citizens are willing to cycle comfortably. This figure, however, only relates to regularly commuting adults; a reduction in both average speed and distance is to be expected for people with additional needs with respect to cycling.

Studies of barrier impacts are usually applied to dense street networks with high interconnectivity (Anciaes, Jones and Mindell, 2016). In contrast, the NCN typically functions as linear connections between locations. Without any alternative routes, the impact of a barrier is more absolute: a single impassable barrier can render an entire section of the network inaccessible to certain users. Whilst real-world journeys may be able to deviate from the NCN route, the intention of a national cycle network is to enable safe and convenient travel by cycle along its entirety, thus travel along the NCN between barriers is the focus of this paper.

Furthermore, there are different factors that are important to disabled cyclists, compared with the factors used in previous barrier impact studies. Within prioritisation studies in which multiple factors must be considered, the relative importance of each factor is unlikely to be equal. The process of determining the importance of each factor originates from the research area of Multi-Criteria Decision Making (MCDM). MCDM evaluates the relative importance of each factor in a pairwise matrix to provide a weighting by which each factor is multiplied (Stojčić et al., 2019). Many MCDM approaches are also prevalent in studies relating to geospatial problems. Most commonly, analytical hierarchy process (AHP) is used, particularly in studies of suitability analysis and GIS (Ohta et al., 2007; Panahi, Rezaie and Meshkani, 2013; Sałabun, Wątróbski and Shekhovtsov, 2020). Conceived by Saaty (1987), AHP is a commonly implemented MCDM approach to reduce multi-factor metrics to a set of pairwise comparisons. In this pairwise comparison, metrics are traded off against each other and their relative importance is judged using a numerical scale that indicates how much more important one factor is over another. The AHP method has been used effectively in GIS studies, as shown in the research from Ohta et al., (2007) and Panahi, Rezaie and Meshkani, (2013).

Based on the above literature, it is evident that research is yet to provide a comprehensive bridge between the “physical” accessibility relating to ability to move through a space, and the “transport” accessibility relating to the ability to reach opportunities. There is a particular gap in considering the impact of physical barriers in active travel networks on the accessibility of those networks to different users. This research aims to fill this gap by using both types of accessibility to score the impacts of physical barriers on the ability of users to reach a given opportunity.

Furthermore, whilst some studies have shown the impact of barriers (van Eldijk et al., 2020), to the authors’ knowledge no study has attempted to rank barriers in terms of removal priority. Given the potential benefits of unlocking key sections of active travel infrastructure for all, it is surprising that little research has attempted to quantify where barriers have the biggest impact on accessibility. Considering the lack of knowledge regarding the prioritisation of barrier removal, the overarching research question for this paper can therefore be defined as: How can barriers be evaluated for removal priority within the context of accessibility for all?

Research has been conducted in the city region of York, and rural county region of Pembrokeshire. York and Pembrokeshire were chosen for the high prevalence of barriers within their respective areas, and for their distinct characteristics (Table 1). York is a populated city region with a relatively high cycling rate and represents an urban cycling context. Pembrokeshire presents the rural case as the area is sparsely populated and use of cycling for transport is low. The regions covered by both study areas can be seen in Figure 3.

Central to this analysis is a dataset of barriers collected by Sustrans volunteers between 2015 and 2018 across the entire NCN. Each barrier contains data on the barrier’s critical width (the maximum width to pass through a given barrier), the restriction type, such as gate, bollard or pinch point, along with a spatial location and a photo of the restriction.

The methodology aims to consolidate the factors discussed within in the Background section of this paper into a simple overall score for each barrier on the cycle network. The proposed score is formed of five overall metrics with six further sub-metrics (Figure 4), which are weighted and normalised to provide an overall score.

Processing and analysis use an open-source framework (QGIS.org, 2025; Van Rossum and Drake, 1995), with all code and processing methods available on a publicly available GitHub repository (Appendix 6.1). Workflows can be found in Appendix 6.3 – 6.5.

From each AHP questionnaire, the consistency ratio is checked to ensure valid answers. This is a measure of how consistent decision-makers pairwise comparisons are, where a ratio less than or equal to 0.1 is an acceptable level of consistency (Saaty, 1987). Of the completed questionaries, two responses gave consistency ratios greater than 0.1. These questionnaires were removed from the overall results accordingly. From the remaining responses, the linguistic scores were translated to numerical ratios (Table 3) and weightings created (Table 7).

The consistency ratio, 0.026, is below the threshold value so therefore the determined weights can be accepted, while the overall consensus value is 90.4%. The barrier critical width has been determined to have the highest priority in overall scores, with the importance of the route on which the barrier is located determined to be of lowest importance. The largest variance of opinion is with barrier critical width, whilst potential distance gained and importance of the route saw the lowest variance of opinion.

Shown in Table 8 are the derived weights from the AHP process. Sub-metric weights are calculated as a percentage of their respective metric’s weighting, and can be found in detail in Appendix 6.7.

The weightings are used to score the barriers within the two case study areas (Figure 13). For York, high scoring barriers are found in clusters along sections of cycle network that provide access from the periphery of the city into the centre. In Pembrokeshire, the high priority barriers are found within the urban areas in the region, where there is both a higher density of barriers and a higher impact from individual barriers.

Through visual analysis of the scores, clear spatial patterns emerge. Clusters appear in the north-east and south-west of York, and in Pembrokeshire in south-west. Such clusters may indicate a high density of barriers that present a significant obstruction to access and use of the network. These spatial clusters can be confirmed or rejected through use of the Local Moran’s I statistic of spatial association (Anselin, 1995) for both areas. Local Moran’s I is a test for local spatial autocorrelation, to measure how similar of different values are from neighbouring locations. Clusters are found where (95% confidence interval), with the -value indicating whether the observed pattern around the location is unlikely to have occurred at random.

Significant clusters of high values (High-High Clusters), are most prevalent in Pembrokeshire, where distinct areas of high removal priority take up 11.7% of all barriers (Figure 14). These are predominately found near urban regions, where the highest proportion of potential users are found. Comparatively, low-low clusters, low values with low significance, are generally found in rural regions. Four outliers are found in total, with three low-high outliers found in and around the town of Pembroke. The single high-low outlier is found the coastal town of Tenby in the south-east.

York has no significant outliers but does show signs of clustering. A section of seven low-low clusters is found near the city centre, representing 21.6% of all barriers within this area. Three barriers within high-high clusters are found, two within the north-east and one to the south-west of the region.

The methodology has produced an overall score to assess the removal priority of barriers across the NCN to help improve access to disabled people or those using non-standard cycles. Both study sites used to test the methodology display separate spatial characteristics but the overall pattern of high priority barriers remains similar. Barriers that block movement in and out of zones such as town or city centres are predominately the highest ranked as might be expected, with these giving a greater benefit potential to users. This is shown as linking areas of the city within York or linking towns together within Pembrokeshire (Figure 15), and therefore opening up access for non-standard cycle users to a greater number of services, economic opportunities and a greater length of protected cycling infrastructure with associated health benefits. The potential reduction in community severance will provide better opportunities in terms of accessibility to those who live around these highest ranked barriers.

Combining into an overall score provides an easily implementable tool for decision makers to use, regardless of their geospatial knowledge. There is sufficient variability between scores to differentiate between barriers but scores are not so widely spread that groups of barriers cannot be visually identified (Figures 13 and 14). The clustering of highly-scored barriers provides a practical solution for decision makers in being able to more easily identify clusters that can be targeted for removal, which is often more economically effective than focusing on isolated elements. The clusters indicate optimal areas for barrier removal as one overall removal project due to their proximity. This approach would offer economic savings and achieve maximum societal improvement for disabled cyclists and other potential beneficiaries.

Many barriers within Pembrokeshire had zero change in accessibility scores (Table 6). This could be due to the remoteness of much of the area, as many barriers are simply not in a realistic cycling range of any amenities. All barriers within York show a change in amenity accessibility, suggesting all barriers within York are currently discriminatory to anyone who cannot pass through them. This highlights the importance of barrier removal schemes, with these barriers causing reduced accessibility to basic amenities.

As highlighted by the difference in the count of amenities accessible within the walk/cycle range of a barrier between York and Pembrokeshire, access to amenities varies dependent on location (Table 6). This is to be expected as rural areas have lower provisions of amenities and services compared with built-up areas (Vitale Brovarone and Cotella, 2020). However, it has been well documented that not all areas of OpenStreetMap are displayed to the same depth or detail (Haklay, 2010). Thus, it is to be expected that York contains better documentation of amenities than Pembrokeshire because of its urban density but also because that urban density leads to more potential OpenStreetMap users. Thus, while the results obtained in section 4.1.5 were expected, these results should not be taken as perfect findings.

The results from the AHP have led to significant differences in the weighting of factors. The critical width and potential accessibility gained metrics account for the majority of the overall score with a combined weighting of 77.1%, resulting in scores being most influenced by their critical width and accessibility scores. However, critical width saw the greatest variation in opinions from the respondents to the AHP, varying from a weighting of 50.2% to as low as 28.3%. This is intuitive as barriers are physical factors that affect the usability of the cycle path for disabled people or those in larger than standard bicycles. A very wide critical width may not affect the majority users, thus if removal is to occur, the gains will be minimal compared with a largely impassable barrier. Feedback from the consulted stakeholders indicated that accessibility change was deemed to best measure latent demand and therefore was equally rated as a high factor.

The lowest weighted factor is the route usage (1.1%), due to the belief from topic experts that as this only factored existing usage, being based on travel to work data from the UK census, it did not capture usage for other activities or latent demand and so should not have a significant influence on barrier scoring. While no metrics are fully able to measure potential latent demand from the removal of a barrier, the route usage is deemed to be less capable of measuring this than edge betweenness centrality with a 100% consensus on the weighting of this metric from all questionnaires received. Vehicle ownership in an area was also deemed to have very low importance (1.7%), and similarly the minimum distance gained metric also received a low weighting (1.5%). Given the low influence of these metrics, it is feasible that similar outputs could be achieved without them, though this would reduce the holistic nature of the score. Further sensitivity analysis could be undertaken to explore the effect of such metrics and if the processing time associated with these is worthwhile.

Weightings are inherently a reflection of what a given person perceives the value of active travel networks to be, and who/what the network is for. The weightings found with this study indicate a consensus that the primary use for the network is to reach amenities, as indicated by the high critical width and accessibility weightings already discussed. However, other stakeholder groups may have different perceptions and thus weightings may differ, with those interested in active travel for leisure likely to value distance gained more strongly, for instance, as a measure of the ability to undertake longer trips. The questions asked of the stakeholders could also be adapted for use with other stakeholder groups, as the questions used in the MCDM process of this study do not cover aspects that could be considered, such as leisure, health or tourism. Alternatively, local authorities or administrators may value the route importance more highly as they aim to improve commuting by active travel. The methodology presented here allows for the alteration of weights at the end of the processing pipeline, and therefore can accommodate different scenarios.

Whilst the approaches presented aim to use quantifiable and measurable metrics, ultimately the true hinderance of a barrier can only be known through more qualitative measures. Extensive consultation with the people who currently use the active travel network and the people who may use the active travel network if changes were implemented would bring viewpoints and opinions that an index may not be able to convert to a number. It is these viewpoints, however, that would paint the true picture of which barriers needed to be removed first, and this tool enables a broader assessment prior to more local focused work to identify those barriers whose removal would bring the greatest benefit.

Although wider progress is being made to remove barriers on active travel corridors, the perceived risks of removing them, such as increase in crime, still remain. This friction will persist until a larger evidence base is created for pre- and post-removal studies. Data driven tools, such as the one developed in this paper, begin to provide decision makers an evidence base for removing barriers. Taking objective measurements regarding the characteristics of each barrier within the local authority district allows—for the first time—to document multiple measured benefits of removing each barrier. This is in contrast to previous schemes, which have tended to focus on community reporting, perceptions of importance and local knowledge of the active travel network. Our data driven approach can be used to complement these, not replace them, as these less quantifiable factors are nonetheless important.