Scramble Crossings: Friend or foe for passenger interchange?

London got a new pedestrian crossing earlier this week. Instead of waiting twice when a diagonal crossing is desired, the new 'scramble crossing' clears all roads of vehicles during the walk cycle. This allows pedestrians to cross diagonally on a single 'green man' and in turn increases road space available exclusively for both cars and pedestrians.

London was by no means the first; busy crossings in Japan, Canada, New Zealand and Australia (to name a few places) had them before. Having all cars stopped while the entire intersection fills with pedestrians is a striking sight compared to existing crossings where cars are always moving somewhere and pedestrians are confined to narrow lines. Instead of having to wait for two red men, diagonal crossers only need to wait for one. And they're safer. So the idea has allure and may have applicability in Melbourne CBD.

The BBC report missed mentioning any possible downsides of scramble or 'barn dance' crossings. Their cycles can be longer so you may wait more. The safety gains may be negated by increasing impulsive crossing.

Although the proportion of people who cross straight versus those who need to cross diagonal depends on trip generators around the intersection, my own observations of Perth show that most pedestrians are not making a diagonal crossing. While this comment is subject to what changes are made to cycle times, changing from a conventional crossing to a scramble crossing disadvantages those crossing straight, ie the majority.

A trend in Melbourne in the last ten years has been reduced direct access to city stations. A remodelling of Melbourne Central reduced direct access from Swanston Street while the reconstruction of Spencer Street/Southern Cross Station closed the subway under Spencer Street.

Increased waiting and access time increases end-to-end trip times, as shown on the table below:

The difference between the two lines is that the first trip takes longer as passenges need to cross a road to get to a tram stop. In the second example they do not; the tram stop is directly outside the station. A zebra crossing is near enough to a seamless crossing but a signalised pedestrian crossing, especially one with long cycles, is not.

For the purpose of examining end-to-end transit times, it may be useful to regard a road crossing just like a short but frequent bus or tram service. After all familiar transit planning concepts like as span, frequency, reliability, forced transfers, DDA access and overall service levels all equally apply to pedestrian crossings.

For example, a CBD traffic light crossing may have the following service characteristics:

Service Span: 24 hours
Frequency: 120 seconds*
Max Wait time to walking/transit time ratio: 120 seconds/20 seconds (ie equivalent service level to an hourly bus route for a 10 min trip)
Reliability: 100% (you will always get a green man within the advertised frequency)
Forced transfers: Present for diagonal crossings on non-scramble crossings

(*) This represents the maximum frequency, not the average frequency. 120 seconds is clearly 'turn up and go' so there are no timetables for pedestrian crossings. Given that we are dealing with tranfers to scheduled services, we must always use the maximum wait figure to guarantee a 'connection', even if the average will be less.

What about other types of crossings?

A zebra crossing is the 'highest and best' form of pedestrian access. This is because its frequency is effectively infinite since you have right of way when you step on it (after previous cars have gone). While there is a forced transfer (you can't go diagonally through an intersection) in practice this doesn't matter as the frequency is effectively infinite. The same is true for a quiet unsignalised street as the delay caused by one or two passing bicycles or cars is neligible.

Conversely a busy road near a roundabout will have a constant stream of traffic. There is no minimum 'service frequency' for pedestrians - this is set by traffic speed, traffic volume, and the risks they are willing to take. Some times there may be a lucky break in the traffic, other times one may walk to the nearest traffic light, or simply give up. Hence both service span and reliability are both undefined, with service level being lowest during busy traffic times (up to a point - if traffic slows to a crawl this may aid pedestrian access).

In this extreme example the overall level of service provided is like an unreliable bus or train that does not reliably feed passengers to the real bus that picks up on the other side of the road. In other cases, although direct walking time to a bus stop might be 5 minutes, passengers may need to double or triple this to increase the likelihood (but never an assurance) of finding a gap in the traffic. This of course increases end-to-end travel times and reduces both overall speed and reliability.

Let's extend our service level parallel to scramble crossings. Being a traffic light type crossing their service characteristics are like those mentioned above. But there is a parallel with bus network design that I wish to tease out further.

Consider a signalised crossroads with four corners. From any one corner you may wish to go to any one of the three others.

With a conventional intersection if you want to cross diagonally you need to wait, cross one road, wait again and cross another. The extra waiting is effectively a forced interchange if we continue the transit system analogy. You didn't want to go to the intermediate spot, but the signals and traffic require you to.

In contrast, if it was a scramble crossing, you can go straight to the corner you want. There is no forced transfer via an unwanted corner and the trip is more direct.

But this comes at the expense of frequency or cycle length. Just like with buses. For a given route kilometres budget you can either have multiple infrequent routes that run to each of your preferred destinations, or you have one frequent service from which you will need to change (to other frequent services) for some trips. The infrequent option is inflexible and doesn't suit many people's needs, although those who it does satisfy get a direct trip. On the other hand the frequent model provides an attractive service for most people to most places, but at the cost of transferring.

The analogy is only partial since while scramble crossings may result in reduced frequency through longer cycles, the level is still at 'turn up and go' levels. This is unlike a bus system where the choice might be between 60 minutes for the direct service network or 15 minutes for a transfer hub based network. Still, the 60 minute approach is roughly akin to a scramble stop on a long cycle (but where you can go anywhere on the 'green man'), while the frequent service approach would be like zebra crossings or short-cycle lights where it's one at a time but the wait for each is short.

For best public transport interchange, it is essential that the act of transferring from train to bus not be regarded as a trip in itself (with its own 'forced transfer' disincentive). This requires direct 'infinite frequency' pedestrian access with a platform to stop transfer time of (say) thirty seconds rather than two minutes or more. Such 100% reliability access could be provided by a stop directly outside the station, a zebra crossing or buses that enter a train platform or Perth-style overhead ramp connected by escalator.

How does this answer the original question about the appropriateness of scramble crossings at interchange points? The answer is that, as with good transit networks, frequency is key. Scramble crossings are good if their cycle times are not significantly lengthened. But if offered the choice between shorter cycle times and scramble crossings, the importance of frequency makes the former look more attractive.