FANDOM


Welcome to the Bike Sharing Analysis WikiEdit

An analysis of bike sharing schemes with a focus on Brisbane's CityCycle scheme. A source for usage statistics and comparison of the scheme to other schemes worldwide.

MotivationEdit

Brisbane CityCycle is a great scheme on a very large scale, with huge potential for changing people's transport habits in Brisbane. The good news is mainly in the potential rather than the reality, and it's important to understand why the scheme has a relatively low usage rate compared to other schemes of its size (ie 150+ stations) worldwide, in order to help the scheme realize its potential.

Suggested StudyEdit

In an article on The Conversation Tim Churches suggested a study in a comment:

I think it is important to investigate why the Melbourne bike hire scheme is unpopular (and the Brisbane scheme, when a bit more data from it is available).
However, ALL of the likely factors (including pricing) need to be considered, quantitatively, in order to estimate the influence of mandatory helmets on bike share utilisation. Such investigation involves tedious collection of data on as many bike share schemes around the world as possible - data which includes:
a) the size of the scheme (number of bikes, number of docking stations and geographical extent);
b) the residential and business/educational population densities of the area which each scheme covers;
c) the pricing of each scheme (relative to local income levels);
d) levels of household bike ownership;
e) measures of the extent and quality of bike-specific infrastructure in the bike hire area;
f) other indicators of cycling safety (such as cycling injury and fatality rates);
g) the duration that the scheme has been in operation; and of course
h) whether helmets are required.
All of these explanatory variables should then be regressed simultaneously in a suitable statistical model against one or more metrics of the utilisation of each scheme. From such a statistical modelling exercise, it should be possible to estimate the influence of mandatory helmets on bike share scheme utilisation. The results should then be submitted to peer-review through publication in a scientific journal. At that stage we would be in a much better position to decide, based on a thorough scientific investigation as opposed to mere supposition, as to whether helmet exemptions should be granted to bike hire schemes. Carrying out such a study is not quick nor easy, but that is the degree of care and effort we ought to expect of public health academics before they propose changes to public policy.

I would say that (e) can include speed limits and actual speed of travel - which could be measured with GPS data where available or actual speeds, or openstreetmap data. Also, (h) can vary in practice, depending on enforcement - the enforcement within a scheme as large as Brisbane's differs by area - for example, West End with Brisbane's highest proportion of commuters ancedotally has less observance of helmet laws; and observance can vary by time of day. So (h) can include actual helmet wearing rates as well as the level of the fine vs income ($176 in Melbourne and $110 in Brisbane). A good proxy for factors (e) and (f) is the commuting modal share at SA1 level in the census; or the percentage of females cycling at SA1 level. Further investigation might show a correlation with the percentage of Strava users who are female on a particular segment. (e), (f), and (m) are the quantitative ways of measuring "culture" and "driver attitude" mentioned in such articles as John Nightingale's article on CityCycle.

To these I would add:

i) the topography of the city and the scheme in terms of hills - since GPS co-ordinates for the stations are readily available in each scheme and ASTER GDEM data is accurate enough to allow inter-city comparisons
j) a study of the weather - heat and humidity - excellent monthly averages for most cities are available for free from weatherbase.com - eg Rio de Janeiro.
k) operating hours - there aren't many comparable schemes which cease operation at 10 pm, as CityCycle does. Ahillen has noted that 7.6% of Washington DC's bike share trips are taken between 10 pm and 5am. CityCycle has cited "concerns about public safety"  as one of the reasons for this limit, although Melbourne's hiring is available 24 hours and it is the larger schemes which tend to have the 24 hour access. There is a parallel with the helmet law here - public safety is cited as the reason for imposing the wearing of helmets by adults.
l) ease of use of the scheme - for example, credit card for immediate access (Melbourne), versus signing up with ID (Tehran) or a phone link (Rio de Janeiro) - I suspect this is much more relevant than (c). Customer service issues have been raised but are difficult to quantify.
m) related to (d) - OBIS discovered that there is a strong negative correlation between existing modal share of bicycles and use. For example, cities where the existing modal share was < 2.5% have much higher usage rates.
n) the competence of the redistribution efforts. Users are discouraged if the redistribution is not competent - The Standard story about Londonmy own experience with Brisbane CityCycle, and Washington Post story about Washington DC.

Hypothesis and remarks on the suggested studyEdit

I agree entirely with Tim Churches that quantitative analysis is lacking in this area and would be useful. However, I believe that, even without conducting an extensive study, the evidence is now clear that the key factors are (e) and (h) combined, with (b) affecting usage rates to a lesser extent; and it is very difficult to separate out (e) and (h) as they are inextricably linked - we can't find a scheme with great cycling infrastructure and a mandatory helmet law, because the law incorrectly places the onus for safety on the cyclist; so cities with mandatory helmet laws tend not to have the best cycling infrastructure or high modal shares.

This has become clear with the recent Fishman et al study showing that CityCycle subscribers tend to own more bikes than non-subscribers (which excludes (d) as a factor); the fact that CityCycle is one of the larger schemes in the world (which excludes (a) as a factor); and most crucially, recent research by Forrest showing that (h) is the most important deterrent from usage. Usage does increase gradually over time but I don't believe (g) is a factor - Dublin's scheme was an immediate success.

If and when the Seattle and Vancouver schemes begin, there will be more data points for the quantitative study, as both of these cities have mandatory helmet laws.

Melbourne versus BrisbaneEdit

In 2013, the Melbourne scheme has been recording about 0.93 trips per bike per day versus 0.35 trips per bike per day for Brisbane. Again, my hypothesis would be that the main explanatory factors for the higher usage in Melbourne are (l) - ease of use for Brisbane and (f) - measured by the proxy of commuting modal share for cycling safety. In fact, anecdotally, many Brisbane residents are not aware that riding on the footpath is legal, so promoting footpath riding, which is the norm in cities like Tokyo, would raise the usage levels.

Other "success" measuresEdit

Another important measure of "success" which is not often mentioned is the percentage of subscribers and users who are female. Jan Garrard has pointed out that this correlates with usage rates at every level of geography.  In the CaBi system, in Washington DC, this has been reported for casual users as 51.33%  (with a helmet wearing rate of 7.4%); I am waiting for Ahillen's paper to be published to find the figure for Brisbane. Ahillen's paper may also have information on the most popular routes taken - ie start and end points for trips, which I don't have access to. I have looked at the most popular stations - consistently among the top few are 119 (south end of Goodwill Bridge next to east end of Southbank), 98 (south end of Kurilpa Bridge near west end of Southbank), and 25 (Brisbane Square, near east end of Bicentennial Bikeway and Victoria Bridge).

Weather effectsEdit

Oliver Nash has also examined the effects of rainfall and other weather on usage for particular days - this is well recorded by the BOM in Australia and worth looking into.

Washington DC Bike Share DataEdit

Are the most popular trips on flat ground, according to the ASTER GDEM data, and vice versa - the least popular trips involve going up hills?

Is there a correlation between the steepness of a scheme and the most/least popular stations in the scheme? How does this ratio between the most and least popular stations differ over the schemes? What sort of distribution can we see in the station popularity - does it follow Zipf's law or something similar?

In fact, this data is extremely useful - can we build a model predicting the popularity of any pair of stations based on just a few data points - distance from "centre" of the scheme where population is densest, the gradient between the two stations, and the distance between the two stations? Travel density models such as those of Dudek et al can then be used to predict the popularity of stations in a general bike share model.

The open source nature of this data is excellent - the openness should be encouraged in the Brisbane scheme. The operators have a twitter account which responds to all questions and in a marked difference from the Brisbane CityCycle scheme, the survey results are made prominently available on the front page.

DataEdit

Data on the CityCycle usage rates from April 2012 to date is at a Google Docs spreadsheet.

Strava data for Brisbane shows the percentage of riders on some segments around Brisbane who are female.

Filtering for more than 500 riders to avoid spurious results, I think it works pretty well - there is some correlation between the number of riders and the percentage of female riders, but what I was really looking for was a correlation between separated bike paths and percentage of female riders, and that's very strong.

The average value is 10.73%, and then I highlighted the segments which are >12% or <9%.

The top ones are Boondall Wetlands (14.5%), Kedron Brook bikeway around Nudgee Rd (13.6%), another Kedron Brook bikeway segment (13.5%), Meiers Rd (12.8%), Riverside Drive (12.4% and 12.2%), and Macquarie Street and Sandford Street in St Lucia (12.2% and 12.1%).

The bottom ones are Kingsford Smith Drive, Waterworks Road / Musgrave Road, Victoria Bridge, and Moggill Road around Kenmore, Brookfield (6.5-9%).

Sylvan Road is a fair bit below average - various segments on it are 10.05%, 10.01% and 9.93%.

Latest activityEdit


Ad blocker interference detected!


Wikia is a free-to-use site that makes money from advertising. We have a modified experience for viewers using ad blockers

Wikia is not accessible if you’ve made further modifications. Remove the custom ad blocker rule(s) and the page will load as expected.