To know where to provide for cycling, and what type of provision, it is important to have good information – such as how many people cycle or wish to cycle, where they wish to ride, for what purpose they ride, and how competent they are to handle a variety of conditions.
To help build this picture, this section describes:
the origins and destinations of cycling trips
methods for identifying the routes used by people on bicycles
the types and numbers of people who use these cycling routes or who may use them in the future.
The methods used will depend on the particular project focus (eg developing a totally new cycle network versus adding to an existing network), and what information is already available.
These items can be covered by questionnaires or focus groups.
City/district planning information
District planning documents map the existing land use and the hierarchy of roads. They also contain information about land use zones and growth areas, major residential subdivisions or commercial or community developments. They are a most useful source of primary data about likely origins and destinations of cycling trips. A higher concentration of people cycling can be expected near popular cycling destinations.
This information is readily available and helps identify likely origins and destinations for people who wish to travel by cycle. It is useful for planning for cycling to future developments.
This method provides no information about numbers of people or the routesthey cycle on.
Identify where cycle traffic could be expected by plotting significant trip origins and destinations on a map, alongside any existing cycle facilities and the road hierarchy.
The five-yearly census includes questions about the mode of travel to work on census day and the locations of the respondent’s residence and workplace.
This data can identify the number and distribution of residents and employees in various age brackets and those who cycled to work on census day.
This data provides reliable numbers. It can be used to plot graphically the significance of areas as origins and destinations for cycling trips to work and by connecting them, the desire lines for commuting to work.
Plotting family size or population density in school-age or the 30–45-year-old bracket may allow a comparison of the likely uptake of cycling in different parts of cities. Larger families and these age groups are likely to involve more people who cycle.
There are disadvantages in time and cost. It duplicates some of the qualitative information available from land use, which may sometimes be sufficient for the purpose.
Census data does not reveal the routes chosen by people cycling.
The census trip-to-work data are a snapshot of one day. This can be affected by weather and any other factors peculiar to the particular day on which the census was held. The census provides no data about cycling trips made to locations other than places of employment.
If using this method, be aware of its limitations. Be mindful also to state the true typical percentage of travellers who cycle to work, by excluding in the total number of respondents those who did not travel to work that day or who worked at home.
School cycle traffic
School cycle traffic is localised and likely to be a significant proportion of the total cycling occurring in many areas. If there is a low rate of school children cycling, this could suggest that the cycling environment is not suitable for the interested but concerned group.
Questionnaires and counting parked cycles are commonly used to assess cycle demand at schools.
By obtaining the number of students attending school on a survey day, the percentage of students cycling to school can be calculated (Note: bear in mind how many staff also cycle).
A school represents a concentration of cycle users who are relatively easy to survey. They are also a target group with great potential to grow cycling numbers with travel behaviour promotion and local network changes.
Additional localised information can be obtained about popular routes or problem locations.
Surveying school cycle traffic:
requires school approval
has a time and cost factor, especially when questionnaires are used
is limited as some areas have few children cycling to school.
During network planning, count parked cycles to quantify existing school cycle use. To get a representative value, ensure that counting does not take place on ‘unusual’ days, eg cycle skills training week or weekly school sports day.
During route planning, survey people to obtain detailed information on route choice and problem areas. Where possible, incorporate this survey into a neighbourhood accessibility plan or school travel plan.
This method uses the total number of visitors to particular locations, attractions or facilities to indicate their likely significance as cycling destinations, as a proportion of visitors will arrive by bike. These data could be obtained for example by the number of ticket sales at a venue.
As long as the information is readily available, this is a quick method for prioritising sites for more detailed investigation, such as counting cycle traffic or parked cycles.
There are disadvantages in time and cost if the information is not readily available.
Only accounts for people currently cycling.
Use this method where the information is readily available. Note that cycle facilities such as urban bike parks are likely to attract a much higher proportion of visits by bike than non-cycling attractions.
Counting parked bicycles
Counting the number of bicycles parked at particular locations on a typical day can help determine the significance of those places as cyclist destinations.
Counts of parked cycles are particularly useful for places with defined cycle parking places such as schools, public transport stations and workplace bike lockups. They are quick and simple to perform.
Without numbers of actual travellers to the destination, only absolute cycle numbers, not proportions, can be calculated.
Counts at private venues require permission from the facility owner to undertake the count.
A snapshot (especially if the weather is poor) may not get a true representation of cycling numbers.
Develop a programme for counting parked bicycles at key destinations.
The Ministry of Transport Household Travel Survey identifies the general characteristics of cycle trips and the people who cycle. Research on this topic has been undertaken by Abley et al (2008) and Milne et al (2011).
Some councils undertake travel surveys which provide information on current and potential cycling.
The Ministry of Transport Household Travel Survey is a national survey with a low sample size for cycling in each area each year. However, for larger centres and those with a higher cycling mode share there is enough data available by combining surveys over more than one year to obtain a useful overall picture of cycling usage. While origins and destinations are available and likely routes mapped, there is not enough data to identify use on individual routes.
Council surveys can have a small sample size, or a biased (eg self-selected) sample.
Use the household travel data to monitor overall changes in cycling, an indication of who is cycling, and a general picture about the cycling purposes other than the trips to employment, which are better from the census data.
Work with council research teams to design effective surveys. These are powerful tools for planning and communicating cycleway projects at a local scale.
District plans usually include maps of the road hierarchy in their areas (typically arterial, collector and local roads).
A first assumption could be that the number of people wishing to cycle on a particular link in the road network will be in direct proportion to those using motor traffic on that link. So highly trafficked roads could be expected to carry relatively high volumes of cycle traffic, given appropriate cycling conditions.
This method gives the simplest and quickest indication of potential cycle demand across the whole area.
This method relies on the appropriate form of provision being implemented. People may avoid cycling on sections of arterial roads that they perceive as hazardous or unpleasant for cycling, and unless the appropriate provision is provided, the latent demand (as calculated by this method) will remain suppressed.
This method does not take into account the short cuts that might be available to people cycling but not to motor traffic, eg through parks.
Some arterial roads carry long-distance traffic and therefore there is a lower level of associated demand for cycling trips.
Comparing cycling trips with motor vehicle trips probably creates a trip type biased towards utilitarian trips rather than leisure trips (ie where the journey itself is part of the point of making the trip). Arterial routes have the primary function of moving traffic and generally have less attractive surroundings, more noise and more traffic fumes; factors that make these locations less conducive for leisure cycling trips.
This method is a good way to begin assessing utilitarian cycle demand, especially for longer cycling trips. Other cycle demand assessment methods should be used to refine further the understanding of cycle travel patterns in an area.
Cycle crash data covering a long period of time can indicate those routes that people have difficulty negotiating safely by cycle. Because crash numbers are dependent on exposure rates, they can also provide a proxy indicator of cycling usage.
Useful crash data can be obtained from the NZ Transport Agency’s Crash Analysis System (CAS), ambulance services and individual road controlling authorities’(RCA) databases of locally reported crashes.
This data is readily available and is also needed for determining appropriate facility types and evaluating cycle route options.
It can also help to prioritise cycle safety treatments based on crash numbers alone.
While CAS can include cyclist crashes that do not involve a motor vehicle, reporting rates for these are very low and off-road crashes rarer still (see safety issues for people who cycle).
Ambulance data has good location information but is biased towards the more serious injuries.
This method will be poor at identifying:
sections of the road network that carry significant numbers of cyclists and are relatively safe for cyclists
off-road routes (ambulance data are useful here).
People may avoid cycling on hazardous sections of otherwise desirable cycle routes and therefore the crash data will not work as a proxy to identify current or desired cycling routes.
Use this method as a supplement to volume-based methods, but be aware of its limitations.
Start with CAS data. For a more complete picture, supplement this with ambulance and RCA data, and possibly even crowdsourced crash information, but remove any duplicate data from the combined database to avoid double counting.
This method involves plotting on a map the location of any existing cycle facilities along routes, at intersections and for trip ends. This may indicate where cycle demand is, or has been considered, significant.
This information has more than one use, as it is part of the base inventory required before cycle route options are evaluated.
The existence of cycle facilities does not always indicate significant cycle traffic, as they may have been poorly located in the first place. The quality of current infrastructure must also be taken into account.
Many favoured cycling routes may simply be convenient quiet streets with no specific cycle facilities.
Use this method as it provides information needed for other purposes, including cycling promotion.
Counting the people currently cycling on a particular route may give an indication for the level of demand for that route. The monitoring and reporting sectionoutlines the type of programme that should already be in place to monitor activity on a cycle network. But the specific route in question may not be already covered by the monitoring plan and therefore it may be necessary to add count stations, or to use data from other sites to estimate the level of cycling at the route in question.
Manual cycle counts
In this method, people at cycling sites record the numbers of people on bikes, their travel direction, and possibly their gender and/or whether each person cycling is assumed to be a primary school pupil, a secondary school student or an adult.
At busy sites people on bikes should be counted separately rather than as part of a general traffic count, as they are easily overlooked. Counting is usually done during the morning or afternoon peak, but counts undertaken at other times can also be scaled up to give a representative peak period or average daily volume (see information on calibration and scaling cycle counts in Monitoring cycle throughput .
The methods already mentioned above may provide a good indication of where to start counting to assess demand on particular cycle routes.
Manual counts have the potential advantage of also collecting more detailed cycle user demographic/behaviour data, but they are also subject to more human error, especially at busy sites.
Automated cycle counts
Automatic mechanical counters can be used to count bicycles, even in conjunction with counting other traffic, however specialised cycle counters may be more accurate (ViaStrada, 2009).
Bicycle detectors at traffic signals can also be used to regularly monitor the number and time pattern of people cycling. Beware of false counts that could be generated by cars driving in adjacent lanes or straying into the cycle lane.
A number of different automatic counting technologies are available; a summary of cycle counting technologiesis given in the section on automatic count technologies in Monitoring and reporting. Each of the technologies has different operational constraints and advantages.
Cycle traffic counts provide hard, conclusive evidence of existing cycle demand.
Automatic counters can provide continuous/ongoing count data over long periods of time (useful for also assessing seasonal changes in demand).
The method only has time and cost disadvantages, particularly for manual counts.
It is still difficult for most automated count methods to collect intersection turning count data automatically.
Manual counts can only give a ‘snapshot in time’ for a particular location.
Counting existing cycle volumes does not help to assess latent demand, or identify what future demand will be.
Each local authority should carry out an annual programme of cycle counts to monitor cycle use trends and provide data to support funding applications.
In addition to counting cycles using sections of routes soon to be investigated or designed in detail, it is recommended that some strategic counts be repeated annually. This could include counting cycles crossing a cordon around the central business district and or other key destinations, as well as on some outlying arterial routes.
Cycle counting should not be performed only as a means of assessing demand for future provisions, but also as a way of monitoring existing and new facilities to gain a better understanding of cycle travel patterns and seasonal trends.
‘Crowdsourcing’ entails gathering cycle use data from a large group of people, generally through use of GPS tracking of their routes on the internet. The rise of smartphones facilitates a greater availability of data. Applications are available that produce ‘heat maps’ of cycling activity – ie where people cycle within the network. Some applications allow users to provide additional data, such as demographics and trip purposes. It is also possible to filter data, for example it is possible to look for typical commuting periods and distances to filter out ‘training rides’ and the like.
Generally, users of smartphone cycling applications subscribe voluntarily, therefore such applications generally only capture a certain sub-set of total cyclists and is considered biased. Data will only be gathered from cyclists who have a smartphone, the motivation to use the required application, and the dedication to do so consistently.
However, this technology could be employed to gauge a specific subset of targeted users who have been solicited for a particular study, thus addressing the potential data bias.
Crowdsourcing applications give access to a large data set of cycling use patterns across the whole network in a way that no other source can currently provide.
The technology is readily available – many people own smartphones and no extra equipment is necessary.
Publicly available heat map data is ‘self-selected’ ie it is provided voluntarily by users, who have chosen to adopt the tracking technology. The data is inevitably biased towards a subset of people who cycle – those who enjoy technologically tracking their rides.
People without smartphones are unable to use the technology, therefore children, older people and financially disadvantaged people are likely to be under-represented. While the resulting data can indicate relative levels of cycling usage on different routes, some specific spot counts would be needed to scale up the data to true network numbers. Someone going for a recreational or training ride is more likely to record their trip than someone commuting by the same route they use every day, so the data also tends to over-represent recreational routes. That said, ‘commuting’ modes are available and data may be able to be sorted by purpose. It is not appropriate to scale up crowd sourced counts at low volume sites.
Crowdsourced GPS data provides extensive coverage across cycling networks, providing rich information on certain kinds of cycling activity. It supplements existing traffic surveys but does not replace them – especially if accuracy is important. Auckland Transport has purchased Strava data for 2013/14 and successfully used those for network planning and corridor optimisation projects (Norman and Kesha, 2015).
Queensland Transport and Main Roads (Langdon 2015) have also evaluated the usefulness of Strava data and concluded that, where there is sufficient data recorded, crowdsourcing data could be used for:
cyclist route choice analysis
assessing the cycle and road network impacts of new cycleways
overview of cycle network usage, including connections between recreational and commuter or transport routes
identifying peak days/hours and indicative cyclist usage
planning to inform undertaking of a more detailed traffic survey
research using revealed preference (as distinct from stated preference data)
wayfinding and focal point mapping
route discovery and asset inventory
locating some route preferences between on-road and off road facilities
identifying gaps where facilities are unsuitable or not present
reviewing cyclist average speeds over particular links.
Don’t use crowdsourced data for:
analysis at sites with low numbers of people cycling
calculating year to year growth
site specific volume comparisons
small distance analysis (less than 5metres) – such as whether on road or footpath
system wide scale up
intersection turning movement analysis.
Consider also crowdsourcing in the form of data submitted by the public from their desktops, eg identifying routes taken, locations of specific problem areas (See Questioning people about their desire to cycle, below).
Questioning local bicycle users, or members of the public who do not currently cycle, can be useful to gather information on:
the types of people currently cycling
their origins and destinations
the routes where they currently cycle (and from this infer existing cycling volumes)
‘desire lines’, ie the routes where they would like to cycle
barriers to cycling identified on their desire lines (ie in terms of the five main requirements for cycling – see Cycle trip types in People who cycle), which it could be useful to collate a map:
locations perceived to be hazardous, or where people have experienced crashes
physical features which sever route continuity, e.g. waterways, motorways, railways, large industrial areas etc
routes that are not suitably direct
routes that are not suitably attractive
lack of network cohesion, e.g. poor signage or inconsistent treatments.
This type of investigation could be done by:
conducting a study or focus group
individual questionnaires, which could be conducted by:
intercepting people cycling on popular cycle routes and either conducting a roadside interview, or giving them a questionnaire to fill in and return later
on-line survey tools and crowdsourcing
via local cycle advocacy networks
cycle shops, libraries or places that cyclists visit often
placing questionnaires on parked bicycles
in classrooms for school surveys, and at tertiary institutions and workplaces
at cycling-related events, eg Bike to Work breakfasts, ‘Ciclovia’ and ‘Open Streets’ days.
Relevant information may also be available from previous initiatives, for example workplace and school travel plan projects and neighbourhood accessibility plans.
People in the ‘no-way no-how’ group (see People who cycle) may have little or no interest in responding to a survey perceived to be solely about cycling, or they may have a response bias that will result in overly negative responses against cycling initiatives. When seeking information from people who do not currently cycle, it may be better to incorporate this in consultation on a wider range of issues, such as a city or district council’s annual citizens’ satisfaction survey.
Note that many of the techniques listed above rely on ‘stated preference’ surveying, ie where the various cycling environments of interest are described to participants (either by words or images) without them actually having first-hand experience. People’s behaviour may not align with their answers given in a stated preference survey. For example, if a neighbourhood greenway situation is described, some people may think they’d be too scared of traffic to cycle there, but if accompanied to cycle on one, they may find they enjoy it. Similarly, someone might cite not having a bike as the reason they do not currently cycle, but, when given a free bike, still choose not to cycle. Giving people real cycling experiences is the best way to avoid inaccuracies from stated preference surveying.
Study groups and questionnaires enable open-ended questions which give insight into people’s preferences and reasons for choosing certain routes, or choosing not to cycle in a certain place or at all. Compare this with the methods in for determining demand above, which only give information on where people currently cycle.
People who cycle usually have excellent local knowledge of the routes they use and their associated problems. This can also be an excellent way of identifying leisure cycle routes, locations where minor maintenance is required and barriers.
Consultation is one of the few ways of assessing latent demand – ie. where and how much people would like to be able to cycle and the associated provisions required to support this level of cycling.
The route and hazard information obtained from study groups or questionnaires is usually plotted on a map of the study area and can be used to identify improvements along routes or at specific sites.
Unless a very large sample is captured, the resulting data may not be completely representative of all trip patterns throughout the study area of interest. The smaller the sample size, the higher the associated margin of error.
It is necessary to survey a representative cross-section of cyclists (and/or people who do not currently cycle). Individuals, unless they cycle many different routes, can talk accurately only about the number of routes with which they are familiar.
Enthused and confident and strong and fearless cyclists may not be able to represent the needs and desires of people who are less confident at cycling, or who do not currently choose to cycle but would like to if the right provisions were made. Thus this method will not necessarily help achieve the best possible increase in cycling.
People may not be motivated to participate in a survey or respond to a questionnaire. It may be possible to encourage responses, for example by providing prizes
There may be a response bias (ie factors that influence many people to give answers that do not accurately reflect their true opinions) especially if the consultation includes people in the ‘no way, no how’ group. People’s stated preferences do not fully represent how they will actually behave – good intentions don’t always result in behaviour changes (think New Year’s resolutions). Hence care is needed in interpreting responses about hypothetical intentions.
The questionnaire has to be developed, distributed, collected, collated and interpreted. This requires time and effort on the part of those conducting the survey, especially if manual techniques are used.
Consult with people who currently cycle and those who would like to cycle more, to obtain more information on their route preferences and existing barriers in the network. Consider whether it is most useful to do this via study groups or questionnaires.
If there is no bicycle users’ group, consider convening one for the purpose of ongoing liaison during cycle planning and implementation. If the target audience includes the interested but concerned, ensure that the users’ group includes members who can represent the views of interested but concerned people. Bear in mind however that people who don’t currently cycle may not know the best cycling routes; typically they will default to trying the same routes they are used to driving.
If questionnaires have not been used for network planning, they should still be considered for route planning.
Developing and using a good questionnaire that will produce meaningful conclusions are not simple exercises; it may be wise to seek specialist survey design advice to ensure cost-effective and useful results.
As discussed, there are advantages and disadvantages to each of the data collection methods outlined above.
Many of the methods identified will produce large data sets relating to specific attributes, for example user volumes at a particular location, for a relatively low level of effort required to collect and analyse the data. However, these methods will not give a good understanding into the specific routes people take. It is difficult to get a large data set of actual routes cycled without either investing a lot of time and effort in choosing a representative cross-section of cyclists and tracking their routes, or by accepting the biases associated with crowdsourcing options. Finally, methods that provide insight into people’s preferences, such as questionnaires and study groups, require significant effort on the part of both the surveyors and the participants, which makes it difficult to include a large sample size.
Using a variety of methods and combining the various data sets collected can give a more meaningful understanding of existing and potential cycling activity within a network or along a route, whilst reducing the effects of biases associated with individual methods.
For example, combining crowdsourced cycle route data (which has a widespread network coverage, but involves a sample of the general cycling population that is possibly skewed towards certain cyclist types) with specific spot counts (full count data but at a limited number of locations) can give a good overall estimate of complete network activity. This can be used to inform questions for questionnaires or study groups which will then reveal more about people’s current choices with respect to cycling and route choice.
Demand for cycling can increase over time as a simple function of population increase, but there are a lot of other factors that can influence the equation. In particular, the concept of latent demand describes potential new cycle trips that are currently suppressed, but that would be made if cycling conditions were improved. The Geller typology (see People who cycle) illustrates this principle of latent demand by showing that there is a large proportion of the general population who would choose to cycle if they felt safer (ie if provision was improved according to their needs). A fundamental aim of cycle network planning is to improve conditions for cycling and therefore the demand estimation undertaken in the planning process should consider the potential effects of releasing latent demand.
Latent demand can be assessed in relation to specific route improvements or to the whole network, assuming it is fully developed and that complementary cycle promotion activities are undertaken. This is an often difficult but important task, and overseas experience has shown that significant problems can arise from not having foreseen the enormous demand that some cycling infrastructure may attract within a decade or two. For example, the VicRoads guideline on shared pathway dimensions (VicRoads,2013) was commissioned following a coroner’s recommendation in a case where a pathway had become congested due to high use (Lloyd et al, 2008).
Note that consultation with people regarding how much they would like to cycle if improvements were made gives a useful measure of latent demand. However, it may only be practical to undertake such consultation for a small population of people or on a specific route. Demand prediction models such as those listed below generally require some sort of consultation to inform their development.
A wide range of methods have been proposed for forecasting cyclist travel demand, and those of relevance to New Zealand are listed below:
network models specific to cycling:
for example Roberts (2014), which details the development by QTP of the Christchurch Strategic Cycle Model (CSCM), based on the city’s existing traffic model, and taking account of changes in demographics, traffic congestion, fuel prices as well as people’s perceptions of the utility of cycling and attractiveness of various network improvement packages
the demand assessment of the Auckland Central Urban cycleways, by Flow Transportation Specialists (Jongoneel & Ormiston 2015), which has asimilar demand modelling approach
route choice models, such as the Abley Route Choice Metric (ARCM) outlined in Rendall et al (2012), which is based on a model developed for Portland and adapted to the New Zealand context (scaling factors are applied to base travel times to represent the desirability of different characteristics between and at intersection to enables evaluation of different route options)
Martin's (2015) examination of future cycle demand in Christchurch by a GIS analysis, which involved data on current patterns in employment and travel and predicted population growth to estimate demand that would be experienced on the planned Christchurch Major Cycleways Network (this analysis also employed the ARCM (Rendall et al, 2012) as an input to identify areas where improvements to the network were most needed; such an approach could therefore serve to inform route choice and prioritisation.
facility based models:
a tool for estimating demand for a new on-road cycle facility (without physical separation) to the existing road environment, which uses a step function to represent the change when the facility is introduced (the required inputs are existing cycle volumes and census mode share growth rate) (McDonald, et al., 2007)
a tool for estimating demand for a new off-road cycle facility parallel to an existing road. The required inputs are cycle AADT and motor vehicle volume on the parallel road; census cycle mode share; and the ratio of New Zealand average trip length by cycles to motor vehicles (from NZ Household Travel Survey) (McDonald, et al., 2007)
willingness to cycle models:
for example the model developed for Wellington City Council by Pettit and Dodge (2014) to assess willingness to cycle for people in different user type categories (different to the cyclist types defined by Geller (2009))
the Transport Agency’s Economic Evaluation Manual (2013b, worksheet A20.1) provides a simple method of calculating demand for new cycle facilities, based on existing use and population, which is designed to be used when cycle counts are unavailable or unreliable and is based on the assumption that the area surrounding the facility is residential in nature (i.e. demand for cycling is generated within a defined buffer) – where this constraint is not met, the method is unsuitable.
Each of the methods presented have different advantages and disadvantages, depending on the inputs required. However, the following general observations can be made.
An appropriate demand assessment will help in determining the type, location and design details of a new facility for cycling.
Demand models can also help to plan and prioritise the implementation of the entire cycle network.
There are time and cost disadvantages associated with each method, and these are generally proportional to the reliability and coverage of the final result. A network model for cycling will require greater input than an isolated facility evaluation using the SP 11 or Research Report 340 method, but the former should yield more meaningful results.
Further research is still required to get an accurate understanding of the interested but concerned target audience in terms of how much latent demand is actually present and what types of provisions are necessary to encourage these people to take up cycling or cycle more often. Note that the stated preference research by Kingham et al (2011b) and Pettit and Dodge (2014) does help towards this issue and additional research is underway.
While these methods require further research for application in New Zealand, and in terms of the interested but concerned audience in particular, the simpler methods may provide a useful starting point until this research can be done.
A comprehensive monitoring programme of existing use at established locations (see Cycle counts, above) can provide useful inputs to produce meaningful demand estimations for future facilities for cycling.
Geographical information systems (GIS) are well suited to analysing data to develop estimates of demand for cycling. GIS then enables the present of the data and model outputs in an accessible, graphical manner. By presenting collected data as layers on common maps, many aspects can be considered together and a complete picture of cycle demand and obstacles developed. Sufficient work should be done to obtain a clear picture of where people wish to cycle, where they currently cycle and where the key network barriers to more cycling exist. The aim is to have usage information that is useful for project evaluation and prioritising improvements in cycle provision. Martin (2015) gives a good example of how GIS has been used to estimate demand for cycling and assess the appropriateness of a planned cycling network.
Note: Each spot represents a bicycle collision. Thickness of buffered line varies in proportion to the number of bicyclists surveyed.