Policy Bash: Redesigning a Roundabout

NB This is a DRAFT (and images to follow!)


Redesigning a Roundabout

Although both links (ie. stretches of road) and junctions are both important, designing decent cycle infrastructure at junctions is much harder and less likely to have been done in the UK, where bike lanes notoriously just give up at difficult junctions.

Burnbrae Roundabout, Milngavie

We therefore decided to do a 'worked example' taking a real junction and applying the Crow Manual guidelines to it as best we could.  We chose the Burnbrae roundabout in Milngavie because it was sent to us by an Embassy member as part of his commute - scene of a nasty near miss he had recently. (caution, video contains a lot of swearing - not safe for work).

The roundabout can be seen on Google Maps here

As you can see, it's a five-arm roundabout with a variety of types of roads coming off it. It already has a narrow cycle lane running north-south on one side of the road which bypasses the roundabout entirely although it does give way to a small access road. All the other cycle lanes disappear on approach to the roundabout. The roundabout itself has a single lane  although the southern arm of the A81 has two lanes approaching it. There's plenty of space and the usual sweeping corners that allow vehicles to approach it at speed and take a 'racing line' through it - features that caused Dave Brennan to nearly end up under the wheels of a tanker despite him being a fast and confident cyclist who had taken up the primary position on the roundabout in full view of approaching traffic.

Limitations of the exercise

Please note here, that we've not gone into the fine details of the geometry of the roundabout, although that is important. We were drawing freehand with no access to detailed plans of the roundabout. We had to assume that the finished result would have tighter corners and a smaller radius than the existing roundabout and that this would slow traffic significantly - as well as making room for cycle tracks.

Note also, that we've had to make assumptions about the nature of the roads and particularly the volume and speed of the traffic, based on the video and the surrounding area. 

What Would the Crow Manual Do?

Our first step was to look at the Crow Manual guidelines for similar roundabouts. The nearest match for an urban area was a single-lane roundabout where bikes have right of way (generally the guidelines are for bikes to have right of way in urban areas; cars in rural areas). This is suitable for intersections between distributor roads or distributor and access roads. It could be argued that the A81 is more of a main or through road. This design calls for a track all the way round the outside, with give way markings for the bike track and a separate zebra crossing for pedestrians. There's a traffic island for the pedestrians but not for the bikes and the coloured tarmac and raised profile of the bike track reinforces the need for cars to give way. The bike track crosses 5 metres away from the roundabout exit to allow for a car to come off the roundabout and still give way to a crossing bike. There are other details - for instance, bikes turning off the roundabout track turn off as soon as possible so that drivers on the roundabout can quickly see if they need to stop for them or not. The circulating lane is quite small but the island has an outer layer of paving to allow lorries to mount it if necessary to get round.

How would this work in the UK?

Accordingly we drew a UK equivalent, assuming separated tracks leading up to the roundabout, one-way circulation, and zebra crossings and bike tracks crossing the roads with right of way. Frankly, this frightened us. Knowing UK road conditions as we did, it just seemed implausible that anyone in the UK would implement such a design and, if they did, that it would result in carnage as HGVs swept off the roundabout and across any crossing cyclists. We started to wonder whether in fact the cars should have right of way instead of the bikes, at least on the bigger roads, with a signalised crossing across the A81 and maybe the B8030. Obviously we didn't want to end up with something like this where bikes and pedestrians are corralled behind railings and sent on great detours to cross three or four roads, having to wait for two cycles of the lights each time. But it seemed to us that the gap between ideal and reality was just too wide to bridge safely.

A Staged Approach

Having reached an impasse, we took a step back, and remembered the principles from day one: we needed to ask what these roads were for and how they fit into the wider picture. Looking at the surrounding area, and assuming that the existing bike lanes indicated where the bike traffic was heaviest, we assumed that the southern arm of the A81 and the B8030 were of the most importance for bikes as well as being important for cars. The side road to the left was not very important for either (it's a cul-de-sac) and the side road to the right was important for bikes, while the northern arm of the A81 was important for cars but likely less important for bikes, except for longer distance commuters.

Accordingly, we decided that we would start with some two-directional bike tracks of decent width approaching the roundabout. On the side roads, we put in zebra crossings and gave the bikes right of way. Crossing the main roads to and from the bike tracks we put in signalised crossings. We decided against toucans because they slow down bikes to the speed of pedestrians and tend to require a two-stage crossing. Ideally, they would be timed so that bikes could cross in a single light cycle, with no railings or detours to slow them down. Pedestrians would get a traffic island, which allows them to cross the roads more easily when the lights are against them if there's a gap in the traffic and their own lights. There should also be a divider between the bike track and the entrance to the roundabout, making sweeping turns less likely. 

As bike (and pedestrian) traffic increases, additional tracks could be put in, moving them from being mainly two-way on one side of the road to being one-way on either side of the road (but hopefully keeping the same width). More bikes and pedestrians would serve to slow traffic coming off the roundabout and make crossing safer. Priorities could be changed to give bikes right of way - or light cycles improved to make crossing the roads easier (for instance with detector loops in the pavement to change the light to green for bikes on their approach). The roads approaching the roundabout could be re-engineered to enforce their current 30mph speed limit - or the roads could be changed to 20mph reflecting their primarily urban location. At this point, signalised crossings would no longer be needed and bikes could cross quite safely without them. In fact, we would end up exactly where we would have been had we followed the CROW manual from the start.

Paul James has drawn up a UK version of a CROW style roundabout to give people an idea of how it might look:

He's also blogged about it here


The CROW model is sound and safe, as long as the roads can be modified to ensure that drivers have plenty of time to see bikes and yield to them. Taking a staged approach would actually end up more expensive in the long run, so ideally on important bike routes, the 'full Dutch' approach should be used from the start. On others, a staged approach could provide a series of stepping stones without compromising the quality of the final result. 

What do you think? There's a thread on the forum (login required) to discuss this in more detail.