The Cardiff bike share scheme was launched in March 2018 an now boasts 94 stations. Operated by Nextbike, the scheme allows bikes to be left both at designated stations or left un-docked once a journey is complele.

As with comparable schemes studied earlier here in Dublin and Glasgow, Cardiff's scheme has been changed significantly by the unfolding pandemic. As with other schemes, this has seen a shift to usage in weekday afternoons, a great increase in usage at the weekends, and a slackening in morning commuting activity. In this post I'd like to examine in some detail how the scheme has changed over the course of 2020.

Data for all charts, except monthly cycling time, was captured up until August 9th 2020. If you're interested in the data presented, feel free to check it out for yourselves using the dedicated Dashboard, Live Map, Metrics Map, and Day Profiles tools. I hope you find it useful.

Contents

1. Increased Aggregate Usage
2. Evolving Day Profiles
3. Increased Stress on Bike Availability
4. Dispersing Commutes
5. 25th of May: A Very Busy Day

Increased Aggregate Usage

While we cannot collect data on rentals directly here at SchemeStats, it is possible to indirectly measure usage. The quantity of bikes in transit is already estimated based off the quantity of bikes in stations at any given point in time, so it is a short hop to extrapolate the amount of time spent cycling from that figure. If we take it that every minute one bike is in circulation is one minute of cycling time, then the amount of cycling time per day is simply the summation of the minute-by-minute quantities of bikes in circulation. This allows us to plot usage month-on-month in terms of minutes cycled, and that's what I've done below to view usage in 2020:

While this doesn't give us a handle on the quantity of journeys taken, we can say for sure that aggregate usage has increased over the course of the year, peaking in May and declining by about 40% by July. Still, the figure for July is approximately double that of January. This is in stark contrast to the sharp decrease in traffic congestion levels, as reported by Wales Online.

Evolving Day Profiles

A ‘Day Profile’ is a plot of the average values of a given metric on a given day of the week – e.g. ‘Available Bikes on Tuesdays, January to March’. I've used the day profile tool available here at SchemeStats to try and understand how typical day profiles have changed, both between 2019 and 2020 in aggregate, and over the course of 2020, week by week.

The 2019 Day Profile

Before March, the typical weekday on Cardiff Bikes was characterised by:

• A weekday pick-up in activity during the commutes, peaking at about 9AM and 5.30PM.
• The evening commute declined more slowly than the mornings, corresponding with what we see elsewhere - people probably doing some post-work errands, finishing at more flexible times than they might stary work, etc.
• Weekend activity was characterised by slightly more aggregate usage than weekdays (in contrast with busier commuting schemes like Dublin), with a peak at around 4PM each weekend.
• Saturdays were a little busier than Sundays.

Plotted on the same scale as 2020's (significantly larger) figures for comparison, the following is the averaged day profile of estimated bikes in circulation over all of 2019:

A Look at 2020

It's instructive to look at 2020's figures in two lenses: aggregated over all of 2020 thus far (i.e. until August 9th) and each week, to demonstrate the bounce in usage after the pandemic set in.

2020 To-Date in Aggregate

Here is the chart for average bikes in circulation over 2020 to-date:

Some observations:

• Weekday peak usage is up about 100%!
• The morning commute has a smaller peak than before.
• Weekdays now look almost the same as weekends, however activity peaks a little later.
• Weekday activity, instead of peaking twice per day, now is composed of one steady rise in activity from around 11AM.

However this average profile for 2020 to-date doesn't tell the whole story - the data has been changing every week, as we'll see below.

2020 in Timelapse

To get an understanding of just how dramatically the usage for 2020 has been changing, I found it instructive to produce some animated GIFs, the first of which I've used to illustrate changing week-on-week day profiles for bikes in circulation. Note that the week being plotted is in the top left. Forgive the bug in my charting tool causing the two dips on one of the Saturday frames, alas there is always something to iron out.

Some observations on this evolution:

• For weekdays, we never saw a fall off in activity in March when the effects of the pandemic set in - instead usage went from typical double daily peaks to one main peak in the afternoons.
• Activity became dramatically busier in May, and has declined significantly since.
• Some of May's weekends say peak usage exceeding 200 bikes in circulation simultaneously, whereas of late the figure is half that.
• The double daily peak of commuting activity during the week has not yet reasserted itself.

Increased Stress on Bike Availability

After a point, increased bikes in circulation will result in some stations emptying out. This is precisely what we see over the time from the start of the year, as demonstrated by the following day profile time lapse, this time for the number of empty stations over the course of the day:

Regarding empty stations:

• The baseline number of empty stations appears to be drifting upward from 15-20 at best in January, to 35-40 in recent weeks. Clearly daily activity is not leaving bikes evenly distributed by evening.
• The peak number of empty stations appears somewhat correlated with earlier data for busy times of day according to bikes in circulation, as one might expect.
• In the context of a scheme with 94 stations, a baseline of 35-40 empty is high, reflecting the strain on the scheme that increased usaged has caused.

To delve a little deeper into the relationship between bikes in circulation and empty stations, here is a scatter plot bikes in circulation data versus empty stations, with a linear approximation of the trend in orange above. The data is fuzzy, but the trend is there.

Dispersing Commutes

Using SchemeStats’ Metrics Map tool, it’s possible to plot the geography of a given station measure over a given time period. That time period can either be a simple from- and to- value, or a slice of a day. We can plot things like ‘The average net change of all stations between the hours of 7AM and 10AM, Monday to Friday’. Using this tool, I've plotted the change in net flow figures for every stations during the morning and evening traditional commuting periods during the week. As above, each frame in the animation is one week of activity. Data is for Monday to Friday only. Red stations are gaining bikes, and blue stations are losing them.

Morning Commute - 7AM to 10AM

Evening Commute: 4PM to 7PM

Commuting observations:

• At the outset of the year, there was a clear pattern of commuting into the centre of the city in the mornings and back in the evenings.
• In mid March, this trend ceased altogether.
• In recent weeks, we can see a mild 'warming' of activity in the same pattern as before, however the trend is not as strong as it once was.
• As noted above, afternoon and evening weekday activity levels has become stronger since before Covid, but this has not been matched by a clear pattern of bikes changing location during the traditional commuting period. Combined with a weak recovery in morning activity levels (i.e. people going to work), I think is is fair to say that the large levels of activity we see on weekday afternoons is not down to traditional commuting.

The 25th of May: A Very Busy Day

I'd like to wrap up this post on Cardiff with a time lapse of SchemeStats' own Live Map for the busiest day of the year on the Cardiff bike share scheme - Monday May 25th, a bank holiday. Each station below is represented by a circle composed of blue, representing bikes docked, and white, representing empty stands. Green dots are bikes parked up outside of stations. Note the docked bikes being almost completely taken out of their stations by 1PM.