How do transit network designers go about their task? Surprisingly little has been written about this. You can pick up books that appear to cover the "network planning" process and find examples of good and bad networks but rarely a description of how to do the design thinking itself. EMBARQ's recent manual for network planners in India, for example, provides great detail about how to analyze demand and evaluate results, but show no awareness of the really challenging task of network design, which sits in between those tasks.
I'm thinking my next book will probably be a little e-book talking through the process in more minute detail. For now, let me talk through a quick example, just to capture the outlines of how a network designer might approach a problem.
My example is the core of Halifax, Nova Scotia, a dense peninsula built out mostly before 1945 and therefore highly conducive to transit. It isn't the only transit-conducive place in the metro region, but it's the largest, and its peninsular location helps us isolate it as a design problem, so I can discuss it in a brief example.
The peninsula's developed area is about 6 km (4 mi) long and 3 km (2 mi) wide at its widest point.
First, we take stock of the strategic focal points, especially major destinations and chokepoints.
Major destinations are part of why you'd focus on the Halifax peninsula as the core of any effective regional transit strategy. The strongest transit destinations are:
- intense, in terms of number of people coming and going.
- all-day, so capable of using relatively efficient two-way, all-day service instead of just the more expensive one-way peak commuter express service.
Those are usually places with lots of people coming and going, not just employees. What kinds of places are those? Large scale:
- Government institutions
- Medical facilities
They're all on the peninsula. Provincial government, two universities, the major medical centers serving all of Atlantic Canada, and retail centers both downtown and on the west edge.
The other reason you focus on the peninsular core is that it's already dense and walkable, because most of it was built before 1945. So much of Halifax is here, and is so amenable to transit, that a strategy trying to optimize transit mobility must take full advantage of these opportunities. If you care about mobility to an outer suburban community, you must care about circulation within the peninsula, because many of your citizens are going there and they may need to move about within the peninsula while there. What's more, peninsula residents who choose not to own cars (which for many could be a viable and liberating choice) will sometimes need to travel outward, and a transit-intensive peninsula will generate outward transit demand further supporting the radial services that outer suburbs need.
So what do we have on the peninsula?
All access is via just six chokepoints (red circles), plus ferry lines east across the harbor. These will be ideal locations for transit connection points (yellow rectangle with T) because many lines, offering travel to many origins and destinations, must converge there anyway. So the transit agency has done a good job here. Except for the northernmost bridge and the northernmost exit to the west, all the chokepoints have transit connection facilities nearby.
The whole peninsula is dense and there are tall buildings here and there throughout, but the big concentrations are in the areas I've shaded. Downtown is the blue area on the waterfront, while the parallel blue box further west is the medical zone. The magenta zones are the major universities. The southern one, St. Marys, is a Catholic school with about 7000 students. The northern one, Dalhousie, is a major public university (16,000 students) that spills eastward along the green axis toward downtown, mixing with major medical facilities, office buildings, and dense, often student-oriented housing. This green area is the core of a potential sustainable-transport paradise, because all the elements that make Halifax in general such a strong market (listed above) are mixed especially closely here, while the dominance of the universities, medical facilities, and shopping guarantees all-day two-way flow, the best situation for highly efficient transit.
I hope you've followed all this even if you don't know Halifax, because the same kind of assessment needs to be made of any city. This doesn't mean that you have to have all of Halifax's features, but you need to be looking for those features you do have, with particular focus on density patterns and chokepoints, becuase these are the best starting point in defining a strategic outline of the network, within which you can then develop local ideas in more detail.
Now we look at the network of major transit-operable streets:
The grid pattern means that some kind of high-frequency grid is almost certainly in order, so that people can get anywhere to anywhere on the peninsula with at most one connection. Obviously, since all trips within the peninsula are short, connections are a bigger disincentive, so we'll try also to link major destinations with direct services. However, if we made direct service everywhere to everywhere our primary goal, we'd end up with lots of overlapping lines and would struggle to afford enough frequency on them all.
Notice, however, the way the southwest edge of the peninsula runs on a diagonal compared to the street grid. That's an invitation to draw L-shaped lines, an east-west grid element connected to a north-south grid element. Like this:
Grid lines that "bounce" off of diagonal edges, like this one, tie together more destinations without a connection while remaining complete grid lines (i.e. running all the way across the grid) so as to maximize the number of other connections that can be made.
But Oxford Street sort of peters out at the north end, so where do we go? We "tie off" or "anchor" the line by going to a nearby major destination where many other connections will be available. Fortunately, one is nearby. The "T" just southwest of where Oxford ends is also a major shopping center.
Meanwhile, at the other end, we notice that many connections throughout the east-of-harbour Dartmouth area are available at a "T" just over the bridge. A single frequent line to that point, covering many of the major destinations of the peninsula, can plug into all of those connections, so that the whole catchment area has easy one-transfer access to most of these destinations. So we have this:
And in fact, Halifax's Metro Transit has already had the same thought. Their most frequent transit line, Line 1, looks like this:
It's on Spring Garden instead of South, but that's only about 400m difference. The university and most medical facilities in the area are adequately served from either. But Spring Garden is more of a continuous "mainstreet" through this area, although major institutions front on both.
So what have we missed? Well, the next big corridor is Robie, which (along with nearby parallel streets like South Park) is really the medical axis. It would have been tempting to turn our first line north along Robie instead of Oxford, because while Oxford is perfectly fine lowrise density Robie has the highrise intensity of institutions. At the south end, Robie bounces off the diagonal grid-edge as Inglis, which gets us to the smaller university, St. Mary's. If we bounce again at the east end of Inglis we head north along Barrington into downtown. Once we're past Spring Garden we're duplicating the first line, but we're virtually downtown by this point, so duplicative service on one street, adding up to very high frequencies, makes sense and can be useful for internal downtown demand. (Remember, the shorter the trip, the higher the frequency you need to compete for it.) So that suggests something like this:
This is not what existing service does. Robie now has an overlay of several lines doing slightly different things, several of which end in a very odd one-way loop at the south end, like this one, rather than going downtown.
Perhaps it's just trying to turn around. One-way loops are a good idea at outer low-density ends of lines, but they introduce needless confusion and circuitousness in dense areas, where it should be possible to serve all markets two way.
Meanwhile, back in our ideal grid, our Robie-Inglis route encountes a perplexity at its northwest corner. At Chebucto Road, Robie seems to branch into the continuation of Robie and the larger Windsor Road, which takes on a more car-oriented character as it heads for the chokepoint at the far northwest corner of the peninsula, near Clayton Park.
This situation is tricky, and I wouldn't make a recommendation at the high-level scale of this exercise; much deeper knowledge of existing travel patterns and land uses would be needed. For now, let's assume that the smart thing to do is follow Windsor, because it completes the grid, serves a chokepoint, and thus is useful both as an internal corridor on the peninsula and a gateway path to the larger region.
Now, you can see how some east-west frequent lines would largely complete the grid. To work best they'd have to aim for the chokepoints too, so that while offering intense local circulation within the city they'd also provide access to much of the peninsula from the surrounding suburbs. Metro Transit has already drawn most of these lines, and they seem to work pretty well except that many aren't frequent enough to really function in a high-frequency grid, or even to compete for the very short east-west trips that they would serve within the dense peninsula.
The dark blue line, existing lines 2 and 4 combined, is already every 15 minutes, but the other two key grid lines here, orange Line 6 on Quinpool and and green Line 9 via the far north, are still every 30, not enough to really function as elements of a grid. Notice too how 6 and 9 bounce off edges of the grid, completing linear corridors before turning to find a major destination and connection point at which to terminate. (Lines always want to terminate at big destinations, called anchors in the business, because otherwise they tend to empty out near the end, leaving permanent wasted capacity.)
Yes, there are still gaps, but we've hit all the major denstinations and most of the peninsula's density. As we turn to those gaps, network design starts to get fun and subtle, as we have to dig into more detailed data to find the clues for how we should patch routes together. And here I'll stop, as I've reached the end of what I can do in such a "high level" view.
But here's my point: I'm almost certain that the lines drawn here, or something very, very like them, would be part of a successful network for Halifax, because they are big-picture responses to issues that are obvious at this big-picture scale. Network designers sometimes fail to take this high-level sketch planning step, and instead wade too quickly into the million possibly interesting details. If you focus on too much detailed data too soon, though, you end up wandering around inside the data, unable to get any big picture "structure" that you can hang onto while you consider the subtler questions.
Also, while it's very important for many people to be able to follow this kind of thinking, designing optimized networks, especially in difficult geography, is a bit of an art. Like any inductive thought, it involves deeply understanding the data but then being open to ideas that come in unpredictable bursts of inspiration, much the way scientific theories come about. The "having ideas" part isn't in many manuals because it's not really teachable; it's to some extent an innate talent.
Consultants like me can help with the subtler thinking of network design, and of course many professional transit planners are adept at it. But you don't have to have those skills, or want them, you can be a more effective advocate if you understand the kind of thinking I've been demonstrating here. If you want to influence transit in your city, you have to understand the basics of the network design problem, as it arises from facts of geometry, facts of transit, and the unique geography of each city. That way, even if you're not ready to do network design yourself, you can assess whether the designers of your own network have done a competent job, and make suggestions that they could actually use.