Dublin Bikes During the Pandemic
May 24, 2020 by Philip Lowney

The unfolding Coronavirus event has altered the transport dynamics in our urban centres. While the effect on public and private transport usage has been widely publicised, here I will examine the changes in usage and dynamics of the Just Eat Dublin Bikes scheme.

Data is examined over the period from Monday February 3rd to Thursday May 21st. This is inclusive of the period during which the Government went through multiple stages of controls – from initial closure of public facilities on March 12th, to the Stay at Home order issued on March 29th, and the first few days following the loosening of restrictions from Monday May 18th.

Before delving into the data, a quick reminder: Cycling affords us the ability to get around while more easily meeting the necessary social distancing requirements designed to keep us from spreading this dangerous disease. However, please be sure to sanitise your hands after using the facility. For further information and advice, please see the Dublin Bikes News page and the Government’s Coronavirus information page.

Pre-Pandemic Review

As outlined previously, the Dublin Bikes scheme is very well-utilised and sees significant daily activity, both during the week and at weekends. While explained in much detail in earlier posts, the dynamics of the scheme might be best summed up in two figures: day profiles and the average net flow during the morning weekday commute from 7 AM to 10 AM. Here are both figures rendered for data from January 1st to February 29th:

To remind ourselves:

  • The morning commute sees a huge drainage of bikes from the northwest quadrant of the scheme each morning, with bikes flowing into the south-east.
  • The activity of the morning commute is mirrored in reverse each evening.
  • Both the morning and evening commute see many stations emptied and filled up rapidly.
  • Weekdays see a reliable triplepeak of activity, with large morning and evening spikes and a smaller pick-up in activity during lunchtimes.
  • Weekends see far less usage, with both Saturday and Sunday exhibiting a gradual rise in activity until around 2pm before slowly declining.

For more information please see the earlier posts.

General Trends

Bikes in Transit

It is instructive to look at the utilisation of the scheme across a span of time covering the progression of the virus mitigation efforts. Below is a graph of the average utilisation over time from Monday February 3rd until Thursday May 21st. To give the graph some context:

  • On Thursday March 12th, the Government announced that school closures, along with colleges and public facilities would take effect from the following day. This is highlighted in green.
  • On March 29th, Taoiseach Leo Varadkar announced that Ireland would be moving into more intensive phase of lockdown, with the Government requiring everyone to stay at home except in a limited set of circumstances. This is highlighted in purple.
  • On Monday May 18, the restrictions were relaxed somewhat, with designated businesses such as garden centres allowed to reopen. This is highlighted in orange.

Some observations:

  • In the first six weeks of the graph, we see the normal weekday/weekend rhythm repeat, with five tightly packed triple peaks followed by two more gradual ones.
  • Activity drops off precipitously after the government announcement of the closure of schools and other public facilities on the afternoon of Thursday March 12th, to take effect the following day. This decline suggests commuters were ahead of the government in deciding to stay at home, even before the formal order to do so was introduced two weeks later.
  • Activity appears to rise for a few days after the second announcement on March 29th, however this is an artefact of the bikes in transit estimation algorithm being put off by bikes being withdrawn from the scheme (see below). Given this, activity did not drop significantly more after March 29th on this view.
  • Activity appears to have risen somewhat since May 1st. More on this below.

Fleet Reduction

As noted above, from March 29th there is an anomaly whereby the values for estimated bikes in transit are artificially elevated. This is because the algorithm which determines this statistic takes the current number of bikes in stands and subtracts it from the maximum number of bikes in stands in the preceding 24 hour period. If a significant number of bikes are taken out of circulation however, the estimate will be inflated for the following 24-hour period. This is borne out by looking at the total number of bikes in stations over the same period:

During the three-day period above, starting March 30th, in which bike circulation was artificially inflated, we can see that there was a simultaneous daily removal of bikes from the scheme, with a total of approximately 150 bikes removed from circulation. Given the dramatic reduction in usage of the scheme however, this did not result in a negative impact on the availability of bikes or slots after their withdrawal.

Day Profiles

Overall Period of Restrictions

Given the massive drop in usage outlined above, it is not surprising to see a change in the trend for day profiles either. Let us start by examining the average daily trends for estimation of bikes in transit each day, for the period during which we see declined usage above – Friday March 13th to the end of the period, May 21st:

Some observations:

  • Weekdays have gone from a peak of about 400 bikes in circulation to one of around 90 – a reduction of over 75%
  • Weekday peak usage has shifted from the mornings to the evenings.
  • Weekday usage has lost its former characteristic of three peaks for morning, lunch, and evening, and instead we see a gradual rise from 7AM onwards before dropping off rapidly after 7PM.
  • What remains of the morning commute levels off at around 5060 bikes.
  • Usage on Sundays remains at similar levels and is similarly shaped to before the period of restrictions commenced.
  • Usage on Saturdays has risen by approximately 40% on average, although there is wide variance in the figures, see below.

Variance of the data

It is useful to know to what extent this data varies day to day, and to get a sense of that I’ve plotted the same chart as above except with a +/- one standard deviation (SD) band around each line, with the value of SD re-calculated per line and per time of day. 68% of the sample points should fell within the bounds of +/- 1 SD. In other words, about two-thirds of all the source data occurred within these bands:

Some observations:

  • Morning data seems reliable – this suggests remaining commuting activity.
  • Mostly, weekday activity seems reliable, with the exception of Wednesday which looks to occasionally get busy in the evenings.
  • At the weekends, and Saturday in particular, there appears to be a marked variation in numbers in circulation. This could be explained by people choosing to take advantage of the good weather when it occurs.

Day Profile Evolution over Period

While the chart above indicates average usage, it does not tell us if there is a trend towards greater usage over the period. To do this, I have split the data up per week, since Monday March 16th. I then overlay the values for each week in three charts: 1) The average of all weekdays in that week, 2) Saturdays, and 3) Sundays. The weeks in question are labelled as follows:

Week Number

Starting Monday

Note

12

March 16

 

13

March 23

 

14

March 30 [1]

Weekday data for week 14 is excluded because of the artificial inflation issue mentioned above

15

April 6

 

16

April 13

 

17

April 20

 

18

April 27

 

19

May 4

 

20

May 11

 

21

May 18

Partial data, running up to Thursday 21 May inclusive

 

Here are the average weekday numbers for bikes in circulation for the weekdays (averaged), and the two individual weekend days:

Observations:

  • For weekdays, the last three weeks of the period are the busiest, but in the evenings only (note data is partial for w21).
  • Each week has similar numbers of people using it during the (former) morning commute
  • The two most recent Saturdays are by far the busiest, with the fine weather on May 9th getting far more people on their bikes than even in normal times.
  • Three of the last four Sundays were nearly twice as busy at peak times than the other Sundays examined. However Sunday the 10th of May was very quiet.
  • While Saturday May 9th was exceptionally busy, the following Sunday was very quiet. This may be explained by Saturday afternoon having temperatures of nearly 20 C whereas Sunday was around 10 C.

Daily Flows

Let’s take a look at how bikes are flowing in the scheme. We’ll begin by looking at the period which was formerly characterised by the morning commute, then we’ll look at the newly-emerged dynamic of a gradual evening peak each weekday. All the maps herein were created using SchemeStats.bike’s Metrics Map tool. Except where otherwise stated, the maps below are based on data over the period of March 16th to May 21st.

Morning Weekday Commute

Here is a map of how stations are gaining/losing bikes each weekday morning on average, from 7AM to 10AM:

Observations:

  • The usual areas to which bikes flow each morning are now much quieter than before. In particular the docklands, which now is losing bikes in many places, and staying neutral across north wall.
  • The important arrival point for many commuters, Heuston Station, is now virtually net neutral.
  • Stations near two hospitals – St. James’ and the Mater – are now gaining bikes where previously they were losing them.
  • The whole picture is now much less clearcut than before: movement is more dispersed.

Weekday Afternoons

It looks as though, in particular in the past three weeks, that weekday activity reaches a maximum by about 6:30PM each day. Because the last three weeks appear to show much greater activity in the evenings than the preceding weeks, I’ve broken this map in two: the one on the left is based on weeks 12 to 18 inclusive, and the one on the right is based on the more recent weeks 19-21 inclusive. In both cases, I’m plotting the afternoon period of 1PM to 7PM as this is where the curve seems to be picking up. The map on the right is the more recent one:

Observations:

  • To an extent these two charts represent the traditional behaviour of the scheme, with stations losing bikes that were gaining them earlier in the day, largely.
  • In the three most recent weeks, we have seen a slight increase in the numbers of people arriving at Heuston Station in the afternoons, counterbalanced by an uptick in the numbers of bikes leaving the Docklands. This suggests increasing commuting activity recent weeks.

Bike/Slot Availability

Finally, let’s take a look at the availability of bikes and slots over the period versus beforehand. I’ve split this into two comparative charts, one comparison for empty stations and the other for full. In each case, the chart on the left is based on January and February’s data, and the chart on the right is based on data from weeks 12-21, the period during which various restrictions have been in place. Time was against me as I put these together, so you’ll have to forgive the different axes per graph – but that’s an important detail.

First let’s look at a comparison of the day profiles for empty station quantities:

The same comparison for full station day profiles:

Observations:

  • The number of empty stations has gone from a peak of 30-35 during the week, to a steady figure of 3-6, an improvement of 80% or more.
  • The number of full stations during the week has gone from a peak of around 810, to one or none.
  • The daily strain on the number of empty & full stations seen during weekday mornings has disappeared.

Summary

  • Bike usage dropped immediately after the government announced that schools were to close, but two weeks before the Stay at Home order was announced.
  • Bike usage has remained far lower than the period before restrictions, with peak usage during weekdays down by between 50% and 75%.
  • Weekday morning usage is down to a peak of 40-80 bikes, down from a reliable peak of 350 in circulation before restrictions.
  • Weekday evening usage has been picking up in recent weeks, with average usage rising from around 80 bikes at 6PM up to May 3rd, to increasing to a peak of 180 in the most recent week evaluated.
  • Approximately 150 bikes have been withdrawn from circulation, but this has had no negative impact on their availability.
  • In recent weeks, weekend usage has been increasing in the evenings, but this is more variable than weekday usage. Fine weather appears to result in much greater usage. Sunday May 10th saw usage far higher than typical Sundays before the crisis.
  • The morning commute is far weaker in terms of bike movement than before. The pattern is also more disperse than previously, with the northwestsoutheast commuting axis less apparent.
  • Two hospital locations, the Mater and St. James’, have transitioned from experiencing a net loss each morning to a net gain.
  • There is no longer a lunchtime pickup in usage.
  • Afternoon usage during the week now climbs steadily from 1PM to about 6.30PM.
  • In the three more recent weeks, we have seen greater numbers of people arriving a Heuston Station over the afternoon, and likewise we have seen greater numbers leaving the docklands area, suggesting marginally rising commuter activity.
  • The greatly reduced load on the scheme has meant far lower numbers of empty or full stations. Empty stations around down in excess of 80%, and there are almost no no full stations any more.

Finally – stay safe, keep your distance, and don’t forget to wash your hands!