Early in my career, I did a number of network designs for free-standing small cities in the American West. These cities, say populations of 30,000-100,000, tend to have a similar set of problems and opportunities, and could probably benefit from a little more theoretical focus. The same issues arise in most of these cities across North America, Australia, and New Zealand, including:
- Low traffic congestion, which makes it possible to offer a reliable operation built around pulses.
- A tendency for destinations important to public transit to be scattered all over the city, often in response to low land value, thus creating an “everywhere-to-everywhere” demand pattern.
- A predominantly “social service” justification for the network, as opposed to a vehicle-trip-reduction justification, though the latter can arise in cities heavily influenced by universities.
You can see almost all of the issues, and the approach I usually recommend, in Great Falls, Montana. I led a study that refined the current network for Great Falls Transit in the late 1990s. Recently, local commenter Kathy called my attention to their network map. It was interesting to discover that it’s hardly changed since I worked on it. (As always, click graphics to enlarge.)
It’s a network of routes mostly running every 30 or 60 minutes. The central idea of the network is the pulse, a meeting of buses from each route every hour or half hour at a central point, so that you can connect from any bus to any other even though the frequency is poor. Geometrically, there is no other solution to the problem of connections in low frequency everywhere-to-everywhere networks. It’s routine in small cities across North America.
In Great Falls, the pulse occurs at a downtown transit center which is also, very usefully, the station for intercity buses. However, the design also provides for an outer pulse point at the far east end. Most pulse networks create these opportunities, but not all take advantage of them.
In Great Falls, the ideal would be that Routes 1, 2, 3 and 4, which all cross near the east edge of the city, should make a timed connection there as well. After all, if the buses are all at the downtown end at the same time, they will tend to be at their outer end at the same time, too, so you should also be able to do a pulse there as well. If Routes 1, 2, 3, and 4 all had exactly the same running time, this would work perfectly.
It doesn’t quite work in Great Falls, because the lengths of the route are too unequal. We tried to make it work, but ultimately, Route 1 was just too much longer for the others, so to get all the buses to downtown at the same time, Route 1 has to leave the east end before the others arrive. Note that their map has a warning, near the east end of Route 1, that you cannot make a timed connection there. Connections among 2, 3, and 4 are possible.
Another trick I often use is the through-routing at the ends of lines, signified by the little boxes that say things like “5 becomes 6, 6 becomes 5.” This is both a network design strategy and a legibility strategy.
Operationally, Route 5 and 6 buses aren’t going back and forth on individual routes. Instead, they are operating continuous two-way loops: the bus that goes out of downtown as 5 comes back as 6, then back out as 5, then back in as 6, etc. Another bus is doing the opposite.
So why do we number the routes as we do? Why don’t we describe them as they are: a clockwise loop and a counter-clockwise loop? Because on balance, that would be more complex. If most people are travelling two-way into or out of the city, why tell them that they need Route 5 to go out and Route 6 to go in, and why clutter the map with overlapping loops and piles of one-way arrows? It’s just simpler to present them as two way routes, then provide the note, for the benefit of the relatively few people who may find it useful, that you can ride through from one route onto the other.
Because of the nomenclature system, every single route segment in the Great Falls network is presented to the customer as two-way, with the trivial exception of pairs of adjacent one-way streets. This is quite an achievement if you compare this network to many networks in similar-sized towns, which often present dizzying tangles of overlapping one-way loops. Here, for example, is a bit of the network in Wodonga, Victoria, Australia (PDF of full map). Compared to the Great Falls map above, how welcoming does this network look to you?
One-way loops mean that the way you get from A to B is different from the way you get from B to A, and may be much longer or shorter. My insistence on presenting two-way services, as much as possible, is about offering a transit system that’s as legible as the street system, so that people find it welcoming not just for the trip they make every day, but for any new trip they may want to try.
I take pride in having designed the underlying structure of several small-city networks in the American West that are still running a decade or more later, including those of Chico, California (PDF, now part of Butte Regional Transit); Corvallis, Oregon (PDF); and Klamath Falls, Oregon. You’ll see some of the same techniques at work in those network structures. If your local small-city transit system needs unscrambling, send them a link to this post, and encourage them to contact me!
The same thing is done with the huge two-way loops that buses undertake in the far bigger network of Leeds, England; the loop is described in two halves radiating from downtown, that run off each other. This has permitted a really brilliant system of legible coloured lines that wouldn’t be possible had the loops been presented as huge circular lines.
Meanwhile, Helsinki’s tram network (mostly reserved track, local stops) contains two two-way loops, and for maximum confusion, they are described in different ways. You can see them on a small map here:
The yellow 3A and 3B are a clockwise loop and an anticlockwise loop. An outward journey on the 3A returns on the 3B.
The green 7B and 7T, meanwhile, run off each other at a northern point and a southern point, so an outward journey on the 7B returns on the 7B too.
In both cases, it’s firstly really difficult to remember which letter suffix you want, and as such, I’ve found myself going the wrong way more than once, or missing trams at stops while trying to work out where they’re going to take me. Secondly, the numbers bear no relation to other trams that are also going out to the same part of the city, instead they only relate to the tram going to opposite way. Thirdly, the network map, as you can see, is a complete mess.
I worked on a vast simplification of Helsinki’s tram network, and it included grouping corridors into simple and extremely frequent (every 3.5 minutes at peak times) corridors. This retained, renumbered, the 7’s loop and the northern end of the 3’s through running, and I presented it like this:
Note that the 3 and 4 are the same colour, as they both leave Helsinki the same way in this frequent network; the joint section is labelled the 3-4.
So labelling loops like Helsinki presently does helps some very few passengers who actually make use of the ability to travel around the loop, but destroys the possibility of any sort of legibility for the majority of trips.
So why not demand-responsive transit in the town of 30k people, with an “everywhere-to-everywhere” need?
“So why not demand-responsive transit in the town of 30k people, with an “everywhere-to-everywhere” need?”
Is anyone aware of a demand responsive system that even vaguely works for daily commutes? The ones around here don’t even have a system for long term automatic daily pickups, so at best you would be left calling in daily to schedule a pickup that is quite likely to change in both time and/or location on a daily basis.
Quite frankly, demand responsive service has it’s uses, but fixed routes are so much easier for everyone involved that IMO they should be the goal whenever possible. Even where demand responsive is the needed, systems should look to establishing scheduled pickup points at high demand times and locations to simplify the scheduling; essentially layering a quasi scheduled service on top of a basically demand oriented system.
Great Falls (as you point out, thanks in part to you and colleagues at the time) does a far better job than Huntington, WV (population 49,129) at making anywhere-to-anywhere trips easy.
Huntington works off hourly pulses at a downtown transit center. Unlike Great Falls, despite routes overlapping and meeting at various places, transfers outside of downtown are more or less never possible. This is because the routes that go furthest are given the quickest routes out of downtown, and will have already left the potential transfer points before the slightly slower routes.
This means that, for example, to travel between the relatively dense areas of social housing in the eastern suburbs to the large mall on the Eastern Edge of the city, you will get to the intersection of US60/Farmdale Road about ten minutes after the bus to Huntington Mall has already left. Instead, you’ll have to leave 30-40 minutes sooner in order to backtrack to the downtown transit center.
This would be all well and good if downtown accounted for the majority of employment and retail activity in the city – as is common in similar cities, however, it accounts for a small fraction of it.
A second problem with Huntington’s structure is that in all of the three most populous sections of the city bus routes are far closer together than a 400m walk radius would require, and in a few cases follow each other in pairs. This is because it has a single hourly pulse, incorporating buses that spend most of their time entering/leaving downtown in similar ways, but for the final part of their route doing different things.
Obviously, it would be better to have a single half-hourly route in these areas. Actually, that would have made the bus useful to me where I spent a brief period in the city; instead walking two miles downtown was almost invariably quicker.
It would mean, however, that the downtown pulses wouldn’t provide transfers for quite everywhere, though suburban pulses would mitigate this if they existed. Here small cities like Huntington can often find themselves debating frequency vs. connectivity.
It gets even better in places like Joliette, QC where long-distance bus routes are synchronized with the local buses.
Jarrett, a couple of things struck me about Great Falls, which I agree has a delightfully simple network for a small city.
1. North-south travel. One weakness appears to be the lack of options for north-south travel.
Also downtown doesn’t appear to be very strong, given all the trip generators along Route 1 (I’m guessing some sort of edge city or business park?).
I wonder how tough the decision was to send Route 6 to downtown rather than (say) the university, which would create a partial orbital route, at the cost of inferior downtown access?
2. I’m not sure of the spacings between routes, but I’d suspect that there’d be people on either side of Route 2 who would be willing to walk to either 3 or 4 if a bus was more suitably timed.
However around midday Routes 2, 3 & 4 all depart downtown at h:10, making the effective frequency hourly, even for those between two routes.
Was thought given to offsetting the central route (2) by 30 minutes, so its departures were h:40?
If Routes 3 and 4 were left at h:10, this would double effective frequency (to 30 minutes) for those willing to walk from the next route.
Another possible gain is that there could be improved serendipitous connectivity with Route 6 from either 2 or 3, where the routes cross, without going downtown.
Similarly Routes 5 and 6. Both leave downtown at h:10. So if you live near where the routes meet you’ve got 2 buses an hour, but still a 60 minute effective frequency.
Whereas if one of the routes was offset by 30 minutes (eg Route 6 departs downtown at h:40) would that not boost effective frequency to 30 minutes?
Even if you don’t live right where the routes meet, is it possible that people would prefer the option of the higher frequency even if their travel time home in one direction is longer (perceived waiting time being longer than in-vehicle travel time, especially in poor weather).
Of course doing this might stuff up connectivity downtown. No longer would everything neatly leave at :10. The main pulse would be at :10 (Routes 1, 3, 4, 5, 7) along with a secondary pulse at :40 (Routes 1, 2, 6). But a large proportion of the catchment would have a new option of leaving at :40, and because it already has a 30 minute headway, connectivity to the busy Route 1 would be maintained.
Taxi-buses in Quebec are a great response to the transportation problems of small North American cities. There are specified pick-up/drop points in the city and a specified schedule. You call an hour in advance to reserve a spot(but can always schedule earlier multiple departures) on a taxi. The operator later tries to fill up your taxi with people going to the same part of the city at the said time. You have a schedule transportation with flexible trajectory to accomodate everyone. But it’s only good for realy small cities. Over 30k-40k, a schedule bus often becomes appropriate.
For an exemple, see Rimouski below. At 45k inhabitants, this regional center plans to introduce buses though. The limit is there.
Great Falls has a sort of demand response with their para transit service. They have wheelchair accessible vans which a disabled person can call and they will transport them. I don’t know how many they have or how often they are used but as the population ages I imagine there will be more demand and more service.
I don’t know how common this type of service is either. I think some towns have this done by private companies and some have it as a public service.
I think it is pretty cool that I live in a town where Jarrett helped design the bus routes. One of a select few in the world.
I don’t even use the bus that often (I ride my bike far more often or drive. It only takes 15 minutes to drive from one end of town to the other but the bus takes an hour and a transfer to do the same trip) but I appreciate the bus service. By the way all the buses have bike racks. There are a lot of bicyclists here so that is appreciated.
And, a bus geek moment here, the other day, my daughter and I caught a bus downtown and it was brand new. The driver had never driven it before as they had just gotten it so we were on his maiden voyage. It was low floor and very nice. I had not been on a low floor bus before. The low floor has a ramp that comes out so wheelchairs and strollers can easily get on. Then the front part they like you to leave for the disabled and there are stairs going up to the back part of the bus for those who are healthy or who do not have little children with them.
At first we were the only ones on it and we sat in front so we could visit with the driver but then some disabled and elderly people got on so we moved up the stairs to the back which was also very comfortable.
Thanks. This article will be useful for a potential re-design of Lorain County (OH) Transit. I am advocating a re-deploy to accommodate a commuter rail.
I did want to point out that Key West, FL does call their routes clockwise and couter-clockwise. I used it as a model for Vermilion, OH. And researching Key West is not at all a bad thing, either.
I thought about what you said about the 6 running north-south in detail
We could send the 6 along 4th Avenue and 23rd Street straight down to the College and University, which are obviously the biggest trip generators in the city. All the inbound trips would have to run 2 minutes earlier and all the outbound trips 2 minutes later.
People from the north who want to go downtown to connect or for other reasons would have to transfer to the 3:
6S -> 3W (5 minutes)
3E -> 6N (5 minutes)
A number of other transfers become possible, enabling easy north-south travel in the eastern section of the city, a very basic grid:
6S -> 2E (18 minutes)
2W -> 6N (18 minutes)
6S -> 4E (17 minutes)
4W -> 6N (17 minutes)
6S -> 1E (20 minutes)
1W -> 6N (20 minutes)
6N -> 4E (9 minutes)
4W -> 6S (9 minutes)
6N -> 2E (4 minutes)
2W -> 6S (4 minutes)
These transfer opportunities greatly facilitate travel from the college area to the eastern neighborhoods.
I also saw another “completing the network” opportunity in the west. The 7 could be routed north to the Twilite Center during rush hours only. This would enable transfers between the 6 and 7. Also, since the 7 does not offer transfers at the other end of the line, there is no reason why every inbound departure couldn’t be delayed 5 minutes to tighten transfers downtown. These transfers could be done at Twilite Center during rush hour:
6W -> 7S (12 minutes)
7N -> 6E (8 minutes)
Please let me know what you all think.
A number of other transfers become possible along the 6 corridor at rush hour:
6S -> 1W (12 minutes)
1E -> 6N (12 minutes)
2E -> 6S (12 minutes)
6N -> 2W (12 minutes)
4E -> 6S (13 minutes)
6N -> 4W (13 minutes)
Thanks for this post. I spent my career at small transit agencies and it’s good to see their issues discussed here along with those of larger transit systems.
Looking at the Great Falls schedules, they seem to have 30-40% layover time. Certainly increased frequencies could be squeezed out of such a “loose as a goose” system, even with much lower, but still very generous 15%-20% layover times, e.g., 9-11 buses should be able to do the work of the current 13 during the peak hours…
I WAS wondering why the system’s productivity was so low despite a relatively high level of service for the size city–0.5 annual hours per capita and about 12 passengers/hour, respectively–now I know why!
@ant6n, one big reason why cities aren’t as willing to implement demand-response transit, even in small cities, is because the costs would be prohibitively high for such a service if people start using it.
It’s one of those weird conditions where it’s cheaper to run an empty bus than a full demand-response van.
“It’s one of those weird conditions where it’s cheaper to run an empty bus than a full demand-response van.”
How can that happen?