Why we collect active modes data in New Zealand

Although we’ve been counting cars using technology since 1937, we only started to count people using active modes (walking and cycling) in the 1990s. More recently, technological advances have enabled us to differentiate people using other micro mobility modes (primarily e-scooters) and a wider range of use cases. With a wealth of data at our fingertips, we have a wealth of potential when it comes to improving transportation and public space.

‍

Collecting traffic data on all our road users helps transport professionals understand, predict, and enhance the real-life impacts of our projects. This can impact how we allocate space to each mode (the design of our public space) and help build the case for further investment.

‍

Data collection can also help in pursuing funding applications and supporting sustained development of capital and operational programmes to support pedestrian and cycle growth. Monitoring is also a requirement to receive National Land Transport Programme (NLTP) funding. According to the NZ TransportAgency (2014) Investing in Cycling, “programme information needs to clearly identify …baseline figures on usage and safety and proposed monitoring of change”. Monitoring is also required by the Department of Internal Affairs and for reporting to Waka Kotahi through the Te Ringa Maimoa Transport Excellence Partnership (formerly, the Road Efficiency Group), including on:

‍

• Condition of footpaths
• Expenditure on walking and cycling, road safety promotion
• Fatal and serious injuries for people walking and cycling
• Length of the cycleway network by urban and rural split

‍

The data is presented through Waka Kotahi’s Transport Insights Web Portal. Starting in 2023, count data is also to be factored into a national multi-modal network model including flows by mode. The fact that having good data on hand is not just useful, but required by these various bodies, means the quality of the data collected is vital. Beyond requirements, it is clear that the data we collect, and how accessible that data is, can have a determining impact on positive behavioural change, like – as this paper will focus on – encouraging people to consider walking or cycling. Accuracy, therefore, is a crucial factor.

‍

State of practice and sources of information

‍

It’s an exciting time for walking and counting technologies, which have been rapidly advancing. With so many options out there, it’s useful to consider research regarding the relative limitations, applications, and operation of each:

‍

‍

• Cycling Network Guidance includes a CNG section on evaluating and monitoring
• Traffic Monitoring Guide (TMG, US Federal Highways Administration 2016) includes material on active transport such as data cleaning techniques
• Guidebook on Pedestrian and Bicycle Volume Data Collection (TRB 2016) evaluated the performance of various technologies and describes methods for developing monitoring plan. The usefulness of this guide to NZ practitioners is some what limited due to a prohibition on naming suppliers and the rapid advance of counting technologies since 2013

‍

‍

Active mode traffic counts are the focus of this paper, and are used in three general ways:

‍

• Cordon counts around a central city or major trip generator give useful information but typically only represent a limited proportion of cycling activity occurring city-wide
• Screen line counts commonly follow natural or artificial barriers such as a river. If all potential crossing points are counted, they can validate home interview or other transport survey data
• a sample of network locations selected to be generally representative of the range of trip purposes and geographic areas

‍

Data types

‍

For people managing walking and cycling networks, there are over sixty different metrics to help evaluate the impact of engineering, education, encouragement, and enforcement interventions. These range from mode share (as measured by the census question on the journey to work mode) to 'hands-up' counts in classrooms, parked bicycle counts, local interview or destination surveys, to measures of physical activity through the NZ Health Survey. The most direct measure is traffic counts – but the questions of where, when, how, and for what duration we should count requires careful consideration. This paper focuses on active mode (walking, cycling, and scooting) traffic counts, and primarily the trends in automatic count technology, crowdsourcing, 'big data', and why manual surveys are still important.

‍

‍

There are at least 62 indicators categorised by importance, geography or availability:

‍

• Key indicators designed to “headline”performance measured either continuously, annually or at least every two years (listed in Table1)
• Secondary indicators that add nuances to the understanding of walking and cycling measured every 3 to 5 years
• Sub-indicators used in the production of the headline or secondary indicators
• Regional-level indicators available from the Household Travel Survey or NZ Health Survey
• Other measures which would only be collected and reported if available

‍

Indicators should be matched to the benefits in the Transport Outcomes Framework, appraisal summary tables (AST) or other current business case guidance published by Waka Kotahi.

‍

Each set of data can tell us something interesting about activity in a given area. Crash data analysis can reveal important information about where cycle crashes are occurring and the safety risks on the network. However, under-reporting rates are at least 54% for minor and non-injury crashes (Waka Kotahi 2020: Monetised benefits and costs manual), crash causation reporting is often biased (Wood 2008),and crash numbers may not reflect the reality that a given network element maybe perceived as so dangerous that active mode trip making demand is suppressed. So, it is worth thinking critically about the sources of our data and how we intend to use it – and, of course, it is worth pushing for high accuracy levels.

‍

‍

Manual counts

‍

Manual counts conducted by surveyors with a clicker or a mobile app are quick to organise, low-cost, and capture demographics (age, gender) and behaviours (footpath riding, crossing locations) that cannot be easily measured with automated means. However, manual counts are subject to human and sampling errors. For active modes, the variation in counts from day to day mean that a few hours of counting are often insufficient to obtain an average annual daily value.

‍

‍

In New Zealand, many road controlling authorities (RCAs) conduct manual counts one day each March during the 7-9 am morning commute period at selected sites, which may include a central city cordon or screen lines such as bridges. Full-day counts (with gaps for surveyor breaks) can help establish other peak hours during the day. Counts should always be taken on a fine day during the school term, as rain dampens ridership, especially on recreational trails. Manual count data should be collated into a database, such as an Excel workbook, and maintained annually for trend analysis. Although manual counts certainly contribute a great deal to our understanding of active modes in given areas on a given day, the clue’s in the name: there’s also a lot of manual work to be done to make sure we’re getting the most out of the data collected.

‍

‍

‍

For example, manual count totals should be aggregated before scaling up to account for the day of the week and month-of-year seasonal variation. Ideally, scaling factors should be developed for each type of facility where manual counts might be undertaken (cycle lanes, paths, near the central city, along inter city routes) and then applied to similar sites. All persons conducting counts need to use the same classification system (e.g., a person walking a bike is a pedestrian). Manual counts certainly provide a vital source of information, especially to calibrate newly established automated counting technology, but as mentioned above, require specific circumstances in order to be valid, and need to be scaled artificially in order to create a bigger picture.

‍

‍

‍

Big data, crowd-sourcing and app-based counting

‍

Smartphone apps (active counts) and Bluetooth (passive counts)

‍

With smart phones being ubiquitous, apps such as CounterPoint (http://counterpointapp.org/) make it easy for members of the public to contribute "crowd-sourced" counts. Researchers have also investigated the potential for the capture and use of ‘big data’ from smartphones. However, some smartphone users will likely have Bluetooth and Wi-Fi turned off or restricted. Bluetooth travel time counters are ubiquitous now in the industry, capturing roughly 10% of the traffic that passes a sensor.

‍

‍

Strava app

‍

Strava is targeted at recreational and fitness riders and runners. It is an app-based tool that enables users to compile their exercise and other trip data and share it with others on a social media platform. While Strava data probably has few utility-oriented trips, it is a useful quick tool for screening-level analyses. Data is now available free of charge to RCAs.

‍

‍

Map views (Figure1 shows a cropped area of Palmerston North) are free, while the underlying data is available to government agencies for a fee. The thicker and brighter the line, the more activity is being registered by the app.

‍

‍

‍

Ride Report app

‍

Perhaps one of the most promising datasets grew out of the Aotearoa Bike Challenge’s promotion of the Ride Report app. This app runs in the background and uses the smart phone accelerometer to detect whether the user is walking, running, cycling, or driving. Users do not have to do anything at all to capture and report their travel totals. Errors are easily reported and help to refine the algorithm. Once the user enables the app, it records active transport trips in the background and sends the data for aggregation to a server that Waka Kotahi can access for planning purposes. The app is being used globally and, as of late 2017, had over 248,000 participants logging 4.6 million trips. In New Zealand, the 2018 Aotearoa Bike Challenge saw 37,823 people logging 1,015,523 trips, many of which were logged with Ride Report. There is strong potential for this data to be used to determine the usage level of favoured routes, with privacy concerns addressed through data aggregation.

‍

‍

Micro-mobility sharing datasets

‍

In another application of IoT technology, GPS-equipped bike and e-scooter sharing operators can easily obtain large datasets indicating how many trips are made and where, and share this with the RCA. The data that could be mined from such systems has the potential to be used for network management applications, such as corridor planning and the provision of cycle parking. It may also be used for planning studies of how weather affects ridership across an entire city (or service area), user demographics such as gender and neighbourhoods, and even travel time competitiveness with other modes.

‍

‍

Automatic counters

‍

Automated technology enables large samples (days, weeks, months) or the full population (continuous permanent counting). Technology canal so supply real-time or near real-time information for public open datasets –supporting public information campaigns, transparency, and independent research. Assuming the technology is installed effectively, automated counting provides a lot of benefits, allowing us to maintain a full and unbiased picture of active transport.

‍

There is a wide range of counter types. Pneumatic tubes are well known for short-term use, while inductive loops have been a mainstay of permanent counters. Video data collection was historically only conducted for short durations (e.g. one day) and, therefore, insufficient to scale up to an annual value and monitor overtime. Recently, however, computer algorithms and sensors/cameras used in video analytics have become more robust and lower cost, so video is now are realistic choice for permanent installation.

‍

‍

‍

‍

‍

In order to fully understand a city's cycle volumes and trends, it’s best practice to use a network of automatic count sites. The design of an automatic cycle counting programme depends on the technology chosen, as counter-performance varies depending on the facility type (e.g. on-road cycle lanes versus off-carriageway paths). The principal types of counters include infrared (passive thermal contrast or active beam interruption), ultrasonic, radar, video imaging (computer analysis of pixel changes), piezometric pressure sensitive (above-ground pneumatic tubes or in-ground cables), and inductive magnetic field loops (in-pavement). Each technology has different applications and accuracy considerations: for example, Figure3 and Figure4 provide two examples of how cycle counting with tubes on-road can significantly undercount ridership.

‍

‍

A prospective counter technology user is advised to (a) clearly state the goals of the automatic count programme and (b) carefully assess each proposed site and consider new technological developments before selecting a product for each site.

‍

Rotating automatic counters around the network

‍

In addition to permanent continuous counters, having a few portable automatic counters can significantly expand the RCA’s understanding of how the network is being used. Contractors must be trained on how to ‘rotate’ an inductive loop logger between permanently installed count stations every two months. This will maximise the site hours of data collected while still achieving a robust count duration and leveraging the previous equipment investments.

‍

‍

Data analysis

‍

Average annual daily traffic and variability in the data

‍

Although accurate, up to date data allows for the best outcomes in planning future projects, securing funding, and making public spaces more user-friendly, many RCAs do not consistently maintain cycle count data and do not annualise the data. Perhaps this is because it can be difficult to get all the factors right, and perhaps current methods aren’t doing all they can to promote accuracy. Traffic data should be reported in annualised terms to account for seasonal and weather variations. The CNG includes a method for estimating cycle AADT (average annual daily traffic) from counts done for part of a day and a spreadsheet-based scaling tool for two-week short-term counts. Estimations are fine, of course, but there are more accurate approaches.

‍

Context is important. For cars, a one or two-week duration is generally sufficient to reliably compare year-on-year flows and detect relatively small trends because there are thousands of vehicles counted, and there is a small variation from the mean traffic volumes. But with few riders counted at a given site, a longer duration is necessary to enable statistically reliable comparisons. A duration of seven days per year is suggested to achieve a 90% confidence of detecting a 20% change in volumes, assuming a coefficient of variation (CV) of 0.15 for sites with 250 or more daily riders.

‍

Without significant interventions (such as the UK’s Demonstration Towns or the UCP), a 20% increase per year in cycle volumes maybe overly optimistic for most New Zealand urban areas and thus a longer count duration will be required.  On the other hand, higher cycle traffic sites in New Zealand will have volumes in the 301 to 450 range; thus, a lower CV can be used.

‍

To achieve a 90% confidence in detecting a 10% change in volumes, a CV of 0.10 is required. Given sufficient permanent count sites which exhibit a good correlation with short-term sites and careful matching of site characteristics, a shorter duration for temporary sites may still be possible. In the absence of statistical analysis, two-week durations for temporary sites should be the absolute minimum and longer durations are preferable to apply seasonal and weather adjustment factors more accurately.

‍

‍

Recreational use versus commuter sites

‍

Figure 5 presents a comparison of weekend versus weekday cycle traffic at selected sites in Palmerston North, showing that some sites are strongly recreational and others are strongly commuter-focused, with four times higher traffic during the week than on weekends. It is important to know the difference and clearly state it in reports because using a single method will mask the type of site and lower the average counts.

‍

[1] Coefficient of Variation – a measure of the relative variation of data, taken as the standard deviation of the cycle volume divided by the mean cycle volume.  CV is normally lower for sites with higher volumes

‍

‍

Getting the word out

‍

Monitoring data and using it to encourage the switch to active modes of transport, whether by infrastructure changes or messaging is just one way we can use accurate counting to improve public environments and congestion. It’s a path we’ve already started to go down: the first known annually issued monitoring and progress report was Copenhagen’s 2006 Bicycle Account. Since then, numerous cities, including Auckland, Portland, San Francisco, and Minneapolis, have released their own annual reports.

‍

In Christchurch, the approach has evolved beyond reports: the city uses SmartView to present data on everything from the location of public fruit trees to the active mode route and usage data. The SmartView app permits users to pan around a map of the city and interrogate cycle count data (illustrated by the blue-filled red circles in Figure6). From the home page, a viewer clicks through ‘on the go’, then ‘cycle counters’, then can pan around the city and click on count sites for the most recent data. Putting data into the hands of the user like this means giving the people the tools to make conscious decisions about their transport modes. The app, in fact, also has layers  transport modes, such as the bus network and route planning assistance.

‍

‍

‍

‍

‍

Real-time displays of daily and annual cycle traffic at a particular point are now present in at least three locations across NewZealand: Auckland, Napier, and Christchurch (Figure7). Such displays encourage cycling by communicating to cyclists that they are a priority while ‘suggesting’ cycling to those not currently riding.

‍

‍

‍

‍

Conclusions and recommendations

‍

While we are investing in infrastructure and promotion, we want to know that we are making a difference and that we are investing in the right places at the right time. Monitoring is one way of determining what is making a difference in our effort to provide a more multi-modal, safe environment for all road users. By gauging the gender, age, and geographic distribution of people on bikes, we can better assess if we are delivering for all our residents (and visitors). That means accurate data is crucial.

‍

Because actual and perceived safety is a key barrier for people to choose to cycle, it is important to collect and analyse safety data (e.g.crashes reported in the Crash Analysis System and other proxy measures). However, this paper focuses on volume data collection and analysis.

‍

While collecting and analysing good quality data is essential, it is even more important to make it accessible to the general public and to use it when planning new or upgraded facilities. Annual report cards, real-time displays, and live web applications are now being used around New Zealand and should be emulated more widely to communicate that cycling, in particular, is not a fringe activity but happening everywhere.

‍

To manage an effective count programme, it is recommended that practitioners:

‍

• Use a site ID and maintain accurate count location information to help others who either get involved in the programme or need to use the data
• Regularly review the sites where data is collected to ensure that a representative sample of locations is being monitored within the available budget
• Conduct automatic counts for a minimum of 14 days to allow scaling up to annual averages, reduce variability, and enable the identification of trends
• While automatic counting is becoming more affordable and reliable, continue to conduct annual manual counts to collect age, gender, and behavioural information (e.g. footpath and contraflow cycling)
• Publish an annual report card on the local authority’s website, including progress in developing the network as well as volume data and trends

Easily accessible data will drive the most significant change, whether it be infrastructure changes from government or councils, or conscious decision-making from individuals. Making sure that data is as accurate as possible, therefore, should be priority number one.

‍

‍