Monday, 2 March 2015

A Review of the Scottish Bike Counters


Introduction
Last year a bike counter was installed on Millburn Road in Inverness.
It is one of nine bike counters sited around Scotland.
As well as the current date, time and temperature each counter displays the number of bikes that have passed the counter that day since midnight along with the total since the start of the year.
There are also a pair of master counters at Victoria Quay, Edinburgh. They don't themselves do any counting but simply display the values from the other nine.


Online Sources
Clickable photos of most of the Scottish counters (and one in Cardiff) can be seen online at
http://www.falco.co.uk/products/live-cycle-counter-data/

The location of 52 bike counters around  northern Europe can be seen on a map at
http://www.bicyclecounter.dk/BicycleCounter/BC_cykelbarometer.jsp
Individual counters can be selected and displayed from the side bar.
Then, clicking 'See statistics' shows two graphs
- hourly data over the previous 24 to 48 hours;
- daily data for the previous fortnight.


My Interest
I was curious to see how the Millburn Road counter could be used to support local cycle campaigning.
Although the operation of the physical counter looked good, I soon noticed problems with the online graphs.
I then noticed problems in the other Scottish counters.
I haven't spent much time looking at counters in other countries but haven't observed any issues.

To give a fair assessment, I monitored the Scottish online graphs from the beginning of February.
I visited Edinburgh on Sat 21 Feb to attend the Campaigner Training Day.
When I got home, I discovered that four of the online counters had stopped.
February may have been an unfortunate month but, at any time, only the previous month is available online.

The graphs shown here were created by Olsen Engineering for the ITSTeknik website.
Click on any graph to enlarge it.


BC1 Edinburgh Middle Meadow Walk

Data for 01 Feb - 20 Feb. No data after 20 Feb.

The daily count for Wed 11 seems rather high compared to the typical count of about 1500.

The consistently lower weekend counts point to people cycling to get to work rather than for leisure.

BC2 Glasgow Anderston Footbridge

Data for 01 Feb - 20 Feb. No data after 20 Feb.

The typical daily counts of around 100 to 150 are dwarfed by the obviously anomalous count of 700 for Tue 11.

The consistently lower weekend counts point to people cycling to get to work rather than for leisure.

BC3 Coatbridge The Fountain

Data for 01 - 28 Feb.

The graph appears continuous for the whole month. But two parts are suspect.

The peak for Sun 08 / Mon 09 is unusually high and the same for both days.

The ramp down from Thu 12 to Tue 24 is too steady to be believed. Interpolated?

BC4 Perth Bells North Inch

Data for 01 - 28 Feb.

The graph appears continuous for the whole month. But is surprisingly similar in profile to BC3.

The peak for Sun 08 / Mon 09 is unusually high and the same for both days.

The ramp down from Thu 12 to Sat 28 is too steady to be believed. Interpolated?

BC5 Edinburgh Union Canal

Data for 01 Feb - 20 Feb.

This counter normally gets little traffic and appears to have stopped recording after Sat 07.

The counts for Thu 05 and Fri 06 being unusually high and the same is suspect.



BC6 Stirling Laurendecroft Road (sic)
(Should be Laurence Croft Road)

Data for 01 Feb - 12 Feb.

The variation between 300 and 1000 seems rather large.
These may be correct but not much can be said with so few values.

BC7 Clydebank Whitecrook Street

Data for 01 Feb - 20 Feb.

The graph shows a ten-fold increase over 15 days followed by a dramatic halving in one day.
It would be interesting to know if this corresponds with reality and what the explanation is.

BC8 Inverness Millburn Road Hitrans

Data for 01 Feb - 20 Feb. No data after 21 Feb.

The count of about 530 on Sun 01 is clearly wrong and unfortunately dwarfs the other values.

The count on Wed 04 was caused by false peak of about 100 at 4pm. The count on Thu 05 may have a similar cause.

There seems to be no

BC9 Benderloch Lochnell Primary School

Data for 25 Feb - 28 Feb.

The counter was not active until Wed 25.

The automatic labeling of the horizontal axis has inserted dates twice.

There are really only four values. Early days.




Summary

Only BC3 and BC4 have data for the complete month but even that seems to be of low quality.
BC1, BC2, BC7 and BC8 show nothing after Fri 20. Of these, only BC1 seems to show reliable data.
The others contain unfeasible peaks and apparently missing data.
BC5 only has data for the first week and even that may be dubious.
BC6 only has the first twelve days. The values cover a wide range but could be correct.
BC9 only shows the final four days. Values are to low and too few for any comment.


Conclusions

Only BC1 (Middle Meadow Walk, Edinburgh) appears to be reliably informative.
In general, it would be better to indicate interpolated sections of a graph with a broken line.
Anomalous high values and missing values can be observed on all of the counters.
This is likely to be caused by a problem with the online system rather than individual physical counters.

The physical sensors are the most sensitive part of the system.
It is difficult for them to distinguish individual bikes in a group and different materials.
But they appear to do this job reasonably well.
Transmitting, processing and publishing the data should be the easiest part.
I'm disappointed as I find it difficult to understand why this system doesn't work flawlessly.
Unfortunately, I'm not sure if it will be fixed as it's not clear (to me) who is responsible for what.

Your comments are welcome.

5 comments:

Sara Dorman said...

if you've not already done so, you should send your info to John Lauder /sustrans scotland. They are very aware of the problem with the Union canal counter (which is also in the wrong place). The Meadows one I have followed since it went in. It is also not always accurate. Perhaps the errors/corrections there are less obvious because of the larger numbers overall? I know these are monitored and compared with 'hidden' counters that are less likely to have kids on skateboards playing on them...

More generally, I am struck that people love to see the counters and watch them. I ride past the meadows counter most days and often there are people standing there waiting to see it change when I go past. So, I think it does do something for public awareness. Also a morale boost when going home at the end of the day and seeing evidence of much travel, even in poor weather.

Daniel said...

I noticed an unusally large count of over 600 (as compared to 200-300) at 08:30 one morning earlier last month on the Middle Meadow Walk counter, which would probably coincide with the spike in the appropriate graph. Such results occurred sporadically last year and are almost certainly a result of drunken antics during the early hours of the morning, but I've also witnessed children playing with the counters as suggested by Sara...

FreeYourInnertube FYI said...

I have been doing some software development that uses the Scottish counter stats, met data, and Glasgow holiday data (public and school) - could easily add in Edinburgh holiday data too.

I've been recording hourly values for the last 2 weeks, though some of the counters stopped working just over a week ago.

The end result is an online pivot chart that allows questions such as, how many cyclists are there on a rainy Tuesday in February?; or how many cyclists stop cycling on a Wednesday if the wind speed is > 20 mph?; or how many cyclists are there on a cold Thursday in January, cf a cold Thursday in February? - or a miriad of other questions. The granularity of the data that I've been collecting is 1 hour.

If you're interested the tool I'm using has a demo http://nicolas.kruchten.com/pivottable/examples/index.html

I've contacted sustrans re reuse of their data, and am waiting for them to get back to me - as the Danish site has not copyright or reuse statement.

If I'm not allowed to publish their stats in a pivot chart, I'll release the software so those interested can run it themselves.

I contacted the Danish counter people today as I believe there's a bug in the graph software they're using. This weekend BC4 shows a spike around midnight of 175 cyclists per hour for 2 consecutive hours. However, the data that my server has been obtaining shows no such spike, and it looks like the graphs sometimes are using DayCount values instead of changes in cyclist numbers each hour ie DeltaYearCount (175 just happened to be the daycount value at that time). So while I cannot confirm a bug, I think there is one.

Zen Biker Maniac - I'm not publishing my site yet, but if you would like to have a look at a demo dm me @fyinnertube

Anonymous said...

While waiting for some members of our group to catch up, we were stopped at one of these counters and one guy went over to the trigger strip, and somehow as he was stopped over the strip, counter jumped up by a dozen or so. He was as surprised and anyone else, but it seems it is possible to fool the counter if you were so inclined.

aedan said...

I've seen the Coatbridge one not counting people.

It never counts me.