Image: Bournemouth Echo, picture by Fiona Wellings

The wall on the corner of Howard Road at 42 Queens Park Avenue still bares the marks of a previous accident. The crossing island at the Howard Road junction has been destroyed twice in the past two years. Neighbours who have lived in the area for many more years recall many more accidents and crash map shows 26 accidents in the past 23 years.

This issue has been raised with the BCP Council (and predecessor Bournemouth Borough Council) Highways Department on several occasions. This has included escalation to Richard Pearson, the then Group Manager of Highway Design and Road Safety (May 2014 – January 2020). Mr Pearson explained the technical methodology used to prioritise the Council’s spending on road safety investment.

“It would…be challenging for the Council to explain why it was…ignoring areas where we are aware of more serious issues.”[1]

Mr Pearson has had extensive international and domestic experience, advised Ministers and politicians, drafted legislation, co-wrote sections of the Institute of Civil Engineering Manual of Highways Design and Management and chaired the same organisation’s Municipal Engineering Panel. However, it is difficult to accept that Mr Pearson has an adequate grasp of the speeding problems in specific areas of the conurbation when he resides in the Isle of Man. Whilst his qualifications are not in doubt, this fact highlights the general approach to the problem is a purely academic exercise.

Mr Pearson described how casualty data is used in the prioritisation of investment. It cannot be argued that casualty reduction is an admirable goal, but it fails to highlight the underlying cause of the issue or the types of casualties.  

Cluster analysis also has several limitations, and there are several things to be aware of when conducting cluster analysis:

1. The different methods of clustering usually give very different results. This occurs because of the different criterion for merging clusters (including cases). It is important to think carefully about which method is best for what you are interested in looking at.

2. Except for simple linkage, the results will be affected by the way in which the variables are ordered.

3. The analysis is not stable when cases are dropped: this occurs because selection of a case (or merger of clusters) depends on similarity of one case to the cluster. Dropping one case can drastically affect the course in which the analysis progresses.

4. The hierarchical nature of the analysis means that early “bad judgements” cannot be rectified.

Reviewing the BCP Council Road Safety Report (“Report”) the obvious flaw is starting with headline collision data before then looking at Personal Injury Collisions (PIC) and then Killed or Seriously Injured (KSI). Collision data is not a proxy for inconsiderate or dangerous driving. 

Furthermore, the cluster analysis used in the Report is flawed. It uses data from 2014 to 2018, but changes in severity reporting systems for many police forces in 2016 mean that serious injury figures, and to a lesser extent slight injury, are not comparable with earlier years.

Additionally, collision data is a blunt instrument as it discounts all other variables, all of which are captured in police data. Time of day, conditions, driver experience, and so on, are ignored. Taking headline collision data could lead to false interpretations such as multiple accidents happening in the same location within a narrow timeframe because of some external factor.

Knowing that fatal and serious road accidents are statistically distributed allows the application of statistical techniques to assess longer-term trends, using a confidence interval on the underlying risk remaining unchanged. These confidence intervals can be used as a rough approximation to determine whether the numbers of fatal or serious road accidents in any two years are statistically significantly different from one another. 

To measure the significance more accurately, it is appropriate to use a statistical test. A Poisson distribution can be applied to the number of road accidents per year large enough to approximate a normal distribution. Therefore, a statistical test can be used to determine if the counts in each year are statistically significantly different from one another at the confidence level i.e. whether there has been a true change in the underlying risk. The point is, using time-bound collision data alone is insufficient and BCP Council have used “where there have been seven or more collisions within this five-year period.”[2]

Queens Park Avenue is also used by many children walking to the local schools to the north of the Avenue on East Way and Mallard Road where Strouden Park is used as a cut through.

BCP Council has a “Safer Routes to School” (SRTS) scheme but this appears to focus narrowly on the immediately vicinity of the school, rather than routes one would expect children, particularly teenagers to take from within the wider catchment areas.

Cyclists also use the Avenue, and many have taken to cycling on the pavements, to avoid the high speeds of vehicles. This creates a dangerous situation in and of itself.

Whilst one can appreciate that a methodology must be used to assess the various issues, the Report does not make use of simple and cheap technology that would enable it to look at the various locations and assess the extent to which speeding is a contributory factor.

Apple, Google and Microsoft predict traffic flows and warn of incidents via their free mapping software for smartphones. Google Maps uses shared location information derived from phones and other handsets running Maps to harvest the speed of traffic (fig. 3). Apple Maps relies mainly on data from sat-nav manufacturer TomTom. Many of Microsoft’s Windows Phones come with Here Drive+ navigation, which collects live location data from users.

It is very simple to access traffic flow data that show the average speed of vehicles travelling in a specific area. This is a relatively inexpensive solution and provides robust quantitative data that can be used to identify areas in need of engineering works. This would allow BCP Council to effectively “crowd source” data on vehicle speeds in specific problem areas without the need to undertake expensive on-site studies and would help to narrow down attention to those problem areas to allow studies to take place.

[1] Email from Richard Pearson on 5th September 2017

[2] Section 3.1: 2014-2018 Analysis Process – 2021 Road Safety Report

This study was setup to highlight an ongoing concern from residents with regards to excessive speeding and only intended to highlight in a public way speeding vehicles. There is no affiliation with Here Technologies or Make.

To obtain Traffic Flow Data, an account was created with Here Technologies and a second account created with Make to run a workflow to output the data.

Using the Flow Bounding Box as the source of traffic flow data a box was drawn around the area of interest and a REST API Key was created. A workflow was created to obtain the data using the coordinates and API and then connected to Google Sheets to collate the speed used in this website to highlight the problem to the public.

On an ongoing basis, there will be continued collection of data via an API to the HERE Traffic Service using a bounded box for the section of Queens Park Avenue east of the fixed speed cameras.

The data endpoint provides AVERAGE speeds for the road segment within the given set of coordinates. Data is gathered from devices travelling within the specified area, including mobile applications and in-dash systems. So, in effect, a moving vehicle becomes a traffic sensor.

This information is known as Floating Car Data (FCD), which is timestamped with geolocation and speed data collected directly from moving vehicles. The information helps determine traffic speed and generate traffic reports on congestion and average travel time. This is type of data is used by Sat Nav providers such as TomTom and Waze and used in mobile applications by Google, Apple and Microsoft.

In contrast to the roadway sensors (Automatic Traffic Counters - ATCs) installed by the fixed speed camera, collecting FCD doesn’t require additional hardware. The data is analysed and stored to allow the API to return granular details. It’s important to note that all of the data on the movement of devices is stripped of any personally identifiable information before it reaches the back-end database. Only relevant information, such as location, speed of travel, device type, and time is collected — amounting to billions of data points from millions of devices worldwide. 

Ongoing statistics are found here

The data collected is only intended to serve as an indicator of the speeding problem. It should be considered the baseline or minimum, not the maximum.

The limitation of this method is that it does not show the full extent of the problem. 

1. As it is only running once every 15 minutes, it is only showing the highest average speed in that period. This means that multiple vehicles travelling over the trigger threshold will not be captured.

2. Where drivers do not have a map service enabled on their device or they have chosen not to share location data under their privacy settings, they may exceed the limit and not be captured within the data.

These are important points to highlight, as the data will understate the extent of the speeding problem on Queens Park Avenue.