Saturday, 19 December 2020

We Need To Talk About Roundabouts

We have a scourge stalking the Kingdom. It creates a maelstrom of danger in the search for capacity and yet designers simply cannot see what is wrong.

I'm talking about the UK's approach to roundabouts. But before I delve in, it's worth taking a step back and ask what roundabouts are for. Whenever roads and streets meet you have a junction. A place of both interaction and conflict depending on the situation. Where things are at the interaction end, then we don't need to be too worried about interfering with how people deal with it whereas at the conflict end, we will want to manage things to keep people safe and moving.

Now maybe those comments are more around how we manage motor traffic, but we can throw walking and cycling into the mix. The difference now is we're dealing with modes moving at different speeds which will immediately pull interactions to one end of the scale. In other words, non-managed interactions are the stuff of low (motor) traffic situations.

When we are managing conflicts, as traffic speeds and volumes increase we need to increase the amount of control and this is where roundabouts come in along with traffic signals and grade separation. Roundabouts are a UK favourite method of control because they can run at very high capacities and are scalable with good space efficiency and operational safety (in terms of collisions). What's not to love?

The problem with the UK approach, however, is how roundabouts are set out to enable drivers to easily find a gap in traffic to enter them and to move round them quickly - the idea is people can see a gap in the traffic on the roundabout and simply join the flow. Of course if the person in front doesn't fancy doing the same and you aren't paying attention, then there's going to be a shunt.

The photograph above is of a large roundabout in Stevenage where you can see how the approach to the give way points puts drivers onto a trajectory where they don't have to reduce speed too much if they see a gap on the roundabout. You can also see an exit from the roundabout on the right which enables people to exit at high speed. The grade separated (different level) walking and cycling space which doesn't place people moving more slowly anywhere near the motor traffic was a conscious decision for the town's original planners. Where traffic flows are light, you'll see some drivers pass through the roundabout taking as straight a line as possible, switching between lanes as they do so. It is very easy to drive through the typical UK roundabout.

Have a look through CD116 - The Geometric Design of Roundabouts, a document from the Design Manual for Roads & Bridges (DMRB). Now the DMRB is for trunk roads and motorways which usually means a high speed situation. However, many local highway authorities will use it for road schemes which are also high speed designs and so we see the DMRB permeating both rural and urban situations (with the latter often being completely inappropriate). The drawing below is an extract from CD116 showing a "normal" roundabout which is the UK standard approach. You can see how easy it is to drive around. On the approaches, the single lane becomes two lanes as the nearside "flares" out which increases capacity. With a two lane approach, you'll often see a flare to three lanes and so on.

The actual intricacies of the design of normal roundabouts are complex and not really my area of expertise, but the high level issues of the geometry and the way that high speeds and capacity are prioritised (and enabled) are all I'm really interested in at this point. The issue I have with their use is where we then mix this type of design with people walking and cycling (and horse riding in some situations) who need to cross one or more arms of a roundabouts. CD116 has a whole section devoted to designing for pedestrians, cyclists and equestrians at roundabouts. 

The table above from CD116 gives advice on the type of crossings to be considered for situations under 40mph. Zebras should not be used where the 85th percentile speed is over 35mph which rules out many situations and is unlikely to be suitable for a rural situation. Traffic signals are for higher flows on dual carriageways. The fact that under 16,000 vehicles per day (on the busiest) arm (or up to 16,000 vehicles per day) on a dual carriageway still allows an uncontrolled crossing to be considered is pretty scary in my mind and zebra crossings on multi-lane approaches are something which makes me very nervous.

My main concern with multi-lane approaches for zebra controlled situations comes from the position that in order to cross, one must place a foot on the crossing. As one looks for approaching traffic, it's entirely possible that the viewing angle could have a vehicle in the far lane blocked from view and so the person crossing cannot see or be seen. If the driver in the second lane isn't paying enough attention, then there is a risk that the person is getting hit.

In my crude little sketch above, the lorry driver in the nearside lane is stopping to allow a person to cross and while the driver of the car in the offside lane should be reacting to the lorry driver stopping, the pedestrian crossing from the right cannot see anything in the visual shadow of the lorry. This arrangement feels unsafe. Some people would feel nervous crossing there and if they had a bad experience, they may be put off from even going there again. Signalised crossings will make people feel more confident because drivers in the visual shadow will see the signals, although being an administrative control, a red signal does not necessarily make it safe. For uncontrolled crossings with multiple lanes, you're being left to find a gap in multiple traffic streams which can be very difficult.

For uncontrolled crossings, the guidance suggests that these should be provided on kerbed islands within 20m of the give way points (otherwise people may not use them which makes sense). However, the problem here is that this places them on the "splitter" islands associated with providing the flare and high speed entry/ exit and therefore situations which are potentially very difficult to cross, especially with multi-lane situations. For speeds over 40mph, uncontrolled crossings of single carriageways of less than 8,000 vehicles per day on each approach are deemed acceptable, although we will often still have flares to two lanes at the crossing point and a wide exit lane from the roundabout. 

The image about is the Weeley Roundabout near Clacton-on-Sea in Essex. I've added the shared-use footway/ cycle tracks in the area and you can see that they cross two lanes on the approaches and wide exit lanes. To the northwest, cyclists are expected to rejoin the carriageway! The aerial view shows a newer layout on the southern arm than you can't yet see in Streetview. It used to be a single lane flaring to two at the last minute (still where people are expected to cross), but as you can see there's a long 2-lane approach. Because the roundabout is often congested (even after this "improvement", it's hard to find a gap in both lanes at once.

Give or take, this roundabout layout "complies" with the DMRB standards. Between 2005 and 2019, there were 17 collisions in the vicinity of the roundabout. 16 involved a slight injury and 1 involved a serious injury (with a motorcycle rider being hit). In the whole of this period, there was 1 slight injury collision involving a cyclist (2006) and none involving pedestrians. In the conventional wisdom we have around 0.8 injury collisions a year with people walking and cycling barely registering and therefore this must be a safe layout.

The problem is, you only have to look at the layout with the difficult crossing points and the walking/ cycling next to traffic on a narrow shared path to understand that this is subjectively unsafe and really only for the fit and the brave. I imaging walking and cycling flows are insignificant compared with motor traffic flow and therefore there simply isn't an impetus to do better. Designers will carry on delivering stuff like this because it is in a National design standard document, because obviously, it has been put together by clever people who know their stuff.

CD116 actually manages to be even scarier. Let's have a look at the segregated left turn lane (SLTL) layouts. These are arrangements where separate left turns are provided to allow that traffic movement to bypass the need to slow down for the roundabout completely - a mini-bypass. There are two main types of SLTL arrangements; where drivers move into the lane and then merge after the roundabout and where the nearside lane peels away from the approach and rejoins on the exit (lane drop/ lane gain). For people wishing to cross, these can be signalised or uncontrolled.

The image above shows an SLTL layout with a lane drop/ gain and uncontrolled crossings. Let's zoom in to the SLTL on the bottom right;

Imagine trying to cross this from bottom right to top left. Three approach lanes, a wide exit lane. Then a wait with traffic passing both sides of you before you cross the slip road. Then a similar arrangement over the next arm of the roundabout.

It took me a while to find, but there is a real live SLTL roundabout just near Crawley, West Sussex - the Cheals Roundabout and it's nuts;

OK, it could have been worse with another entry lane to cross in that southwestern arm, but look at what the designers are expecting people to use;

Imagine being in that island to the right with slip road traffic on one side and traffic entering the roundabout on the other side. This island is around 2.4m in width and so not wide enough to fit many non-standard cycles and so small, you're going to struggle to fit a family on there. As an aside and if you find yourself on the roundabout but fancy getting off it, there's a curious gap in the SLTL island (which is not in CD116);

The odd cycle gap to one side, this SLTL arrangement is a legitimate layout. Even though the Cheals Roundabout is a local authority road, the DMRB has been used to guide the design in a situation one might find on a trunk road. The designers will no doubt argue that it's a reasonable design to use and because "it's in the book" it is safe. But again, it is subjectively unsafe with a design that shouldn't have soft bodied people anywhere near.

The problem with all of this is, that if we start from a sustainable safety point of view rather than a traffic flow point of view then we would either provide fully signalised layouts for people cross (notwithstanding the safety issues which remain) or providing grade separated arrangements. Ideally, we would have a completely separate walking and cycling network. However, sometimes this does coincide with the motoring network and as Stevenage taught us many years ago, grade separation is a good solution. 

It's of course no surprise that I'm going to mention the Dutch approach and that they would not provide layouts that I have mentioned in this post. The Dutch do use high capacity roundabouts (called "turbo roundabouts") which is a whole post in it's own right, but where people are not mixed other than in certain urban situations (although these are risky and should really be avoided). If a junction is very busy, then the Dutch will grade separate it to provide a safer environment for drivers with the walking/ cycling network being kept separate.

In fact, it's not just the Dutch, there's lots of rural grade separation on Danish roads in situations where the UK would go for a roundabout. Although the Danish don't have a well developed rural walking/ cycling network to the same extent as the Dutch.

Where people are crossing Dutch roundabouts, these will be one lane in/ one lane out affairs (compact roundabouts) which are developed as a safer type of junction. Generally they are moving to walking/ cycling priority in urban areas with traffic priority retained in rural areas. I wrote a whole post about this back in 2017. 

Anyway, the UK is baking in roundabouts which are subjectively hostile to people because of a design standard which comes at the issue as a traffic capacity question, rather than a making people feel safe question. When I was looking for an SLTL example, I came across lots of design reports for locations where they are being considered, so expect to see more of these as we keep trying to add capacity.

We really do need to talk about UK roundabout design especially as the compact or continental style roundabout appears in CD116 but is often rejected by designers on capacity grounds;

This is the last general post of 2020. Next week will be my annual roundup and it certain to highlight a very strange year. It just remains for me to wish you all a Merry Christmas and thank you for reading. Stay safe and let's hope 2021 is a better one.


  1. "Where traffic flows are light, you'll see some drivers pass through the roundabout taking as straight a line as possible, switching between lanes as they do so. It is very easy to drive through the typical UK roundabout."

    This is virtually a design feature. The governing factor is the 'entry path radius'. This is the flattest path a vehicle will take if the driver tries to straight-line through the roundabout. For a normal roundabout this value should be no larger than 100m, which, when all's said and done, is still a very large, relatively flat path, which is going to have limited effect on drivers' speeds.

    The other significant problem is that to get a roundabout design accepted in the first place it must be shown to have sufficient capacity, worked out using ARCADY. Anyone who has been involved in transport planning in the slightest way knows which factors will notionally increase a roundabouts capacity - the entry width and the flare length - even if the changes are a few 100mms. This leads to overly wide entries which many designers then seem to create by reducing the size of the splitter island so that it can barely accommodate a child, never mind someone with a pushchair or on a bike. Plus the unnecessary lane width makes crossing to the island in the first place a bit of a lottery.

    Unfortunately, somewhere along the line UK highway engineering has lost its link to ingenuity and independent thinking and become an exercise in picking the desirable minimum value out of the right table in the DMRB.

    Stepping back, thinking about whether a design satisfies the original brief and is safe for all its users seems to be forgotten in the rush to meet the all important design programme.


  2. Re: SLTL - there's one on the A35 Dorchester bypass at the junction with A354, which you have to cross on the Weymouth and Dorchester cycleway.

    Some pictures of people struggling to cross it, including an electric wheelchair user, from our Infrastructure Safari in 2014:

    The council officer who joined the safari was adamant that people prefer surface crossings to subways.

    (Yes I am very behind in reading my RSS feeds, Merry Christmas etc.)

    1. And a Happy New Year! Well people like surface crossings when UK underpasses are awful!