CRUSE Tool project: Recreational trips

1 Introduction

We estimated the total number of recreational cycle trips using the assumption from PAG Unit 13 that there are a mean of 5 daily recreational trips per 100 people.

We have divided recreational trips into three categories:

1. Sport cycling

• For this we use data from Strava. The emphasis will likely be on cycling for fitness. Routes may be chosen for a mix of high speeds or pleasant conditions. Since the Strava data is private, this layer cannot be added to the public-facing CRUSE Tool, but it can be produced as an additional layer for use by TII.

2. Tourism and cycle touring

• This involves longer distance touring and trips to major attractors. Many of these trips will cross county boundaries. Places of interest could include greenways and popular tourist sites. This layer can be released to the public. For this we use a spatial interaction model representing trips to and from these places of interest.

3. Dispersed short trips

• This is characterised by short family trips to the local park and similar outings. We are not currently modelling these trips.

To ensure the estimated number of trips is accurate, we have also validated the recreational and non-recreational trip data against two independent sources - cycle counter data from a set of Dublin count points, and estimates of total km cycled in the Dublin Metropolitan Area from the Bikelife 2019 report.

2 Sport cycling

We have accessed Strava data for 2019 giving the number of app user trips (rounded up to multiples of 5) on each route segment. Strava classifies user trips as either ‘commute’ or ‘leisure’, based on the trip characteristics such as regularity. For this purpose we use only the trips classified as ‘leisure’. In Figure 2.1, we show this data for County Leitrim.

Figure 2.1: Strava leisure trips in 2019 in County Leitrim

According to PAG Unit 13, the mean recreational cycling trip duration is assumed to be 60 minutes. Segment lengths and mean speed across each segment are included using the Strava data, which would enable us to estimate mean trip distances in km. Using this, we could calculate the number of unique trips that are contained within the Strava data. However, this may well be an overestimate, if trips by Strava users tend to be longer in duration than other recreational cycle trips.

Instead, a better approach is to count the number of km cycled in the Strava data. We can compare this to the number of km cycled in the Dublin Metropolitan Area, from the Bike Life 2019 report.

3 Tourism and cycle touring

We have modelled trips to tourist sites, initially using data from the Geodirectory. Figure 3.1 shows location of campsites, hotels and other tourist accommodation according to Geodirectory data.

Figure 3.1: Campsites, hotels and other holiday accommodation (NACE codes 17, 51 and 380)

It would also be possible to use other sites, such as those identified by Failte Ireland. Greenways and similar cycle trails should also be represented (Figure 3.2).

Figure 3.2: Greenways (in green), Eurovelo routes (in black), Sports Ireland cycle trails (in red) and major tourist destinations

With a spatial interaction model similar to the one we used for utility trips, we modelled trips from Electoral Divisions (ED) to campsites. The destinations are aggregated to a 500m grid. Trips to campsites in County Carlow are shown in Figure 3.3.

Figure 3.3: Route network for trips to campsites in Carlow, max trip distance 20km

An important step is to properly parameterise the gravity model. For utility trips, the estimated number of trips was based on the ED population, but that doesn’t make sense for cycle touring, as it is strongly biased towards cities. The best approach may be to equalise based on the ED area. A spatial interaction model with an area-based parameter is shown for County Leitrim in Figure 3.4.

Figure 3.4: Route network for trips to campsites, hotels and other tourist accommodation in Leitrim, using an area-based weighting for trip numbers

For cycle touring it is important to emphasise longer distance trips. This can be done through modifications to the spatial interaction model parameters, perhaps introducing a minimum trip distance.

4 Dispersed short trips

We have not yet investigated these.

5 Validation using cycle counter data

To validate the number of recreational and non-recreational trips, we have obtained cycle count data from a set of counter locations in Dublin. For the year from 12th October 2021 to 11th October 2022, there are 9 cycle counters providing reliable data that can be cross-referenced with the corresponding segments on the recreational and non-recreational trip route networks. Here, non-recreational trips means both POWSCAR (travel to work / school) and other utility trips. The only recreational layer we have attempted to validate so far is the Strava data.

The count data correlates very strongly with our estimates of the number of POWSCAR and other utility trips on these route segments, with an R squared of 0.89. This is slightly higher than the R squared for POWSCAR trips alone (0.88), suggesting that the spatial interaction model we used to model the other utility trip purposes (social, personal and shopping trips) is providing a realistic representation of these trips. As shown by the dotted line in Figure 5.1, there is almost a 1:1 match between the count data and the route network trip numbers at many of these points.

Figure 5.1: Correlation between cycle count data and POWSCAR/utility baseline cycle route network. Dotted line represents a 1:! correspondence.

When Strava data are added together with the POWSCAR and utility data, the R squared falls slightly to 0.85 (Figure 5.2). At some locations, the combined POWSCAR/utility/Strava flows are higher than the cycle counter data, even though the cycle counter data is from 2021-22, while other data is from 2016-2019. Perhaps there is some duplication, for example with Strava being used for non-recreational trips.

Figure 5.2: Correlation between cycle count data and combined POWSCAR/utility/Strava leisure trips. Dotted line represents a 1:! correspondence.

It also appears that some popular recreational routes are over-represented in the Strava data, in comparison to the Dublin count points. The two locations with the highest recorded Strava leisure trips in Figure 5.3 are both on the Clontarf strand, a very popular coastal leisure route. The R squared of the correlation in this graph is 0.04.

Figure 5.3: Correlation between cycle count data and Strava leisure trips

6 Validation using Bikelife 2019 data

The Sustrans Bikelife 2019 report provides a range of statistics related to cycling in the Dublin Metropolitan Area (DMA). This is a slightly different geographical area from County Dublin - it excludes the northern part of the county but includes parts of Kildare, Meath and Wicklow (Figure 6.1). In particular, the report provides estimates of the number of trips and total km cycled in the DMA, broken down by trip purpose. Trip purposes include commuting, travel to school by children, travel to school or college by adults (including adults accompanying children), other utility journeys, and recreational journeys.

Figure 6.1: The Dublin Metropolitan Area

Bikelife 2019 reports a total of 375.1 million km cycled per year in the DMA, of which 163.3 million km are for recreational trips, with 120.8 million km for travel to work, 27.1 million km for travel to school (by both children and adults), and 63.6 million km for other utility journeys.

The data in Bikelife 2019 is based mainly on a survey of 1,106 residents aged 16 and above, conducted in June to July 2019. The number of trips for each purpose apart from children’s travel to school is taken from the survey responses, with corrections for seasonal variation and trip chaining and to infer recreational trips by children, and validation using Dublin counter data. Children’s travel to school is taken from the 2016 census. Respondents provide an estimate of the distance of each journey, and the median trip distances are multiplied by the total number of trips to get the km cycled for each trip purpose.

6.0.1 Comparison with POWSCAR and utility data

Using CRUSE route networks for the Dublin Metropolitan Area, with estimates of km cycled according to the baseline scenario, we found that the number of km cycled for trips to work and school (POWSCAR data) is 86% of the equivalent work and school journeys in Bikelife 2019. The number of km cycled for non-POWSCAR utility trips is 95% of the equivalent in Bikelife 2019. It makes sense that the proportional coverage of POWSCAR trips is lower, because the Bikelife school category includes trips made by adults accompanying children to school, which are not included in the POWSCAR school data, and would come under other utility trips. In total, the CRUSE estimate of non-recreational km cycled in the Dublin Metropolitan Area is 89% of the equivalent Bikelife 2019 estimate.

The CRUSE estimates of km cycled are based on 2016 Census data, so the fact that these are 11% lower than the estimates from the Bikelife 2019 report suggests that cycling uptake in the DMA has increased between 2016 and 2019.

6.0.2 Comparison with recreational data

The number of km cycled in the DMA according to the Strava data (which is from 2019) is 5% of the recreational km cycled according to Bikelife 2019. This suggests that in 2019, around 5% of cyclists used the Strava app.

Across the DMA, Bikelife 2019 gives a mean estimate of 3.2 daily recreational trips per 100 people, when we divide the number of recreational trips by the total population aged 4 and above. This is similar to the PAG Unit 13 assumption of 5 daily recreational trips per 100 people, suggesting that it is reasonable for us to use this PAG estimate to determine the number of recreational trips by residents of each Electoral Division.

We have less information about how recreational trips are likely to be divided between sport cycling (as represented by Strava), cycle touring/tourism, and dispersed short trips. This is important in terms of the overall geographical distribution of recreational trips. Dispersed short trips will be more likely to originate from close to people’s homes, while tourism will be biased towards more scenic and rural areas, especially gravitating towards trip attractors such as greenways. The sport cycling represented by Strava could be halfway between these two extremes.

7 Next steps

The most important steps are to:

• Define the proportions of recreational trips that we expect to be represented by the different categories of recreational cycling. Counter data from greenways may help with validating this. By assessing the counter data together with the usage of these routes in Strava, we can estimate the number of tourist trips. Alternatively, since the other categories of recreational cycling will not be included in the public-facing tool, we could scale the trip totals in the cycle tourism layer so that it incorporates all recreational trips, using both Strava and greenway data to validate the geographical distribution of these trips.
• Refine the methodology for tourism and cycle touring, creating a consistent set of route networks using a spatial interaction model, that account for key trip attractors.
• Create a tourism layer that can be added to the CRUSE web app.
• Create a sport cycling layer that can be distributed for internal use within TII. Strava data privacy issues mean this cannot be externally distributed without explicit permissions. If the trip totals of the tourism layer have been scaled to include all recreational trips, then users of the sport cycling layer will need to be made aware that this cannot simply be added on top of the other layers without rescaling the trip numbers in the tourism layer.