Integrated Timed Transfer Schedules for Buses
I’ve written a bunch about integrated timed transfer (ITT) scheduling based on Swiss and Dutch principles, developed for intercity and regional trains. Here, for example, is how this schema would work for trains connecting Boston and Worcester. But I’ve also seen interest in how buses can connect to one another, so I feel it’s useful to try to adapt the ITT to this different mode. Two particular places where I’ve seen this interest are a statewide plan for intercity buses in West Virginia, and regional integration around Springfield and the Five Colleges; I’m not going to make specific recommendation for either place, since I don’t know them nearly well enough, but I hope what I write will be helpful there and elsewhere.
The ITT principles for trains
ITT for trains relies on total coordination of all aspects of planning. The centerpiece of this is the triangle of infrastructure, rolling stock, and timetable, all of which must be planned together. Decisions on infrastructure spending should be based on what’s required to run the desired schedule, based on tight turnarounds, maximal utilization rates of equipment, and timed connections.
The even broader principle is to trade state complexity for money. It’s harder to plan everything together – different departments need to talk to one another, planning has to be lean or else the back-and-forth will take too long, regulations may have to be adjusted, government at all levels has to push in the same direction. The reason to do this chore is that it’s far cheaper than the alternative. Organization is cheaper than electronics and concrete at all levels; American households spend around 20% of their income on transportation, mostly cars, whereas households in transit cities like Paris or Berlin or Tokyo spend a fraction of that, even taking into account residual car ownership and operating subsidies to public transit.
On buses, there’s no such thing as electronics…
The Swiss maxim, electronics before concrete, concerns trains exclusively. On buses, no such thing exists. It’s not really possible to get higher-performance buses to make a more aggressive schedule. Acceleration rates depend on passenger safety and comfort and not on the motors (in fact, they’re higher on buses than on trains – rubber tires grip the road better than steel wheels grip rails). The closest analog is that electric buses are lower-maintenance, since the diesel engine is the most failure-prone part on buses as well as trains, but what this leads to, IMC, is not really a strategy for improving timetabling – IMC’s main benefits are less pollution and lower maintenance costs.
…but there is a surplus of organization to be done
All the little things that on trains go in the electronics bucket go in the organization bucket on buses. These include the following operating treatments to improve local bus speeds:
- Off-board fare collection
- Stop consolidation to one every 400-500 meters
- Dedicated lanes in congested areas
- Signal priority at busy intersections
In addition, bus shelter does not increase actual speed but does increase perceived speed, and should be included in every bus redesign in an area that lacks it.
These are all present in Eric’s and my Brooklyn bus redesign proposal, but that doesn’t make that proposal an ITT plan – for one, it’s based on 6-minute frequencies and untimed transfers, whereas ITT is based on half-hour frequencies (for the most part) and timed transfers. Of note, in a 6-minute context signal priority should be conditional to prevent bunching, but if buses run on a 30-minute or even 15-minute timetable then bunching is less likely, especially if buses have prepayment and some dedicated lanes.
That said, it’s important to talk about all of the above in this context, because a bus ITT belongs in areas where public transport ridership is so low that people view a bus every 15 minutes as an aspirational schedule. In such areas, the politics of giving buses more priority over cars are harder than in a city with low car usage like Paris or New York or Barcelona. There are some positive examples, like Rhode Island’s eventual passage of a bill giving six key bus corridors signal priority, but in Tampa I was told that DOT wouldn’t even let the bus agency bump up frequency unless it found money for repaving the street with concrete lanes.
What about intercity buses?
Prepayment, stop consolidation, and dedicated lanes are important for speeding up local buses. But intercity buses already stop sporadically, and often run on highways. There, speedup opportunities are more limited.
But there may still be some room for signal priority. If the bus only runs every hour or every half hour, then driver resistance may be reduced, since the vast majority of stoplight cycles at an intersection will not interact with a bus, and therefore the effect of the change on car speed will be small.
This is especially important if buses are to run on arterial roads and not on freeways. The significance is that highways are noisy, especially freeways, and do not have the concept of a station – freeways have exits but one takes an exit in a car, not on foot. Therefore, development does not cluster near a freeway, but rather wants to be a few minutes away from it, to avoid the noise and pollution. Arterials are better at this, though even then, it’s common for American big box stores and malls to be somewhat set away from those, requiring bus passengers to walk through parking lots and access roads.
Arterial roads, moreover, often do have stoplights, with punishing cycles optimized for auto throughput and not pedestrian-friendliness. In such cases, it’s crucial to give buses the highest priority: if these are intercity buses rather than coverage service to a suburb where nobody uses transit, they’re especially likely to be full of passengers, and then a bus with 40 passengers must receive 40 times the priority at intersections of a car with just a driver. Moreover, if it is at all possible to design stoplights so that passengers getting off the wrong side of the street, say on the east side for a northbound bus if the main development is west of the arterial, can cross the street safely.
Designing for reliability
The principles Eric and I used for the Brooklyn redesign, as I mentioned, are not ITT, because they assume frequency is so high nobody should ever look at a timetable. But the ITT concept goes in the exact opposite direction: it runs service every 15, 20, 30, or even 60 minutes, on a consistent clockface schedule (“takt”) all day, with arrival times at stations given to 1-minute precision.
Doing this on a bus network is not impossible, but is difficult. In Vancouver, the bus I would take to UBC, the 84, came on a 12-minute takt off-peak, and ranged between on time and 2 minutes late each cycle; I knew exactly when to show up at the station to make the bus. When I asked Jarrett Walker in 2017 why his American bus redesigns assume buses would run roughly every 15 minutes but not on such a precise schedule, he explained how American street networks, broken by freeways, have more variable traffic than Vancouver’s intact grid of many parallel east-west arterials.
So what can be done?
Dedicated lanes in congested areas are actually very useful here – if buses get their own lanes in town centers where traffic is the most variable, then they can make a consistent timetable, on top of just generally running faster. Signal priority has the same effect, especially on arterials as noted in the section above. Moreover, if the point is to make sure the noon timetable also works at 8:30 in the morning and 5:30 in the afternoon, then driver resistance is especially likely to be low. At 8:30 in the morning, drivers see a bus packed with passengers, and their ability to argue that nobody uses those bus lanes is more constrained.
Pedestrian Observations 
When you talk about dedicated lanes do you mean interlined segments where you have buses far more frequent that every 30 minutes like in the center? Otherwise it feels like a waste of road space to reserve a lane for a vehicle the comes twice an hour.
Yeah, plus some short queue jump lanes at complex intersections.
This is an almost wilful misreading.
Identifying the needs for, let alone implementing, “off-board fare collection” (why? how?), “stop consolidation” (where? qui bono?), “dedicated lanes” in “congested areas” (where?) and “signal priority” (how?) at “busy intersections” (where?) are all “Elektronik”.
Vehicle tracking, holding connecting services based on real-time arrival information? Elektronik.
Passenger flow monitoring, identification of service-significant slowdown locations and times, identification of sources of schedule variance, reconciliation of planned versus achievable schedules? Elektronik.
Monitoring of vehicle health, communication with maintenance force, identification of failure modes? Elektronik.
The single most basic useful thing: an LED sign at a bus stop and a URL on a web site predicting, with reliability that actual humans can accept, that the next bus on line XXX will arrive in about YYY minutes? Elektronik.
I know all the actually advanced Asian megalopolises and all the cool kids are all about their endless tunnels and non-interlined subway lines and all, but in the shitty real world, even with shitty real “first” world, doing something, anything, lots of things, to help the buses carry actual people where they actually want to go, today, is a huge fucking deal. And yes, “elektronik” cyber-brain computers can help you mere humans with this, despite there being “no such thing as electronics [on buses]” .
How many padding should buses with Integrated Timed Transfer have?
“Padding” has different meanings in different cities. In Edmonton’s great era of timed-transfer-focal-point service design, we scheduled the meets for four minutes. On a 30-minute cycle feeder route that also met the contractual time for a restroom visit or a quick smoke. In Denver I sometimes had to use seven minutes. This seemed to be due to less supervision (and was before the Init GmbH dispatch network and Automatic Passenger Counters were installed.
I do not know.
I’m guessing a hub and spokes design (with multiple hubs if you want everywhere everywhere service) is better for timed transfers here? Timed transfers on a grid system just seems like it would be a nightmare for 2 reasons: First. buses would have to wait for people to make the transfer at every intersection, and that time would add up. Second, there would be hardly any room between two transfers for enough padding to recover from a delay. A hub or distributed-hub system would not have these problems.
Yeah, I should make it clear, this is intended for a mesh of nodes – the example I was thinking of for a local or regional system was Pioneer Valley, with nodes in Springfield, Northampton, Amherst, etc. You can’t do any of this on a grid, same reason why you can’t do this on a circumferential regional railway. (PVTA does have one crosstown route in Springfield and it sucks, for the same reason.)
It works really well for hub-and-spoke, but the US is actually pretty decent at that, with a lot of timed transfers at downtown hubs. Usually the problems for these kinds of systems are that a) the buses meander a lot so the speed relative to straight-line distance sucks, b) the frequency is often hourly or irregular rather than half-hourly, and c) where trains exist, there is no integration with them. The principles I’m proposing are for going a bit beyond this, i.e. having multiple hubs, or timing a transfer every 15 minutes, which requires a good deal more reliability than a lot of small-city US bus networks have because there’s so much traffic variability.
I should add that the Edmonton TTFP network was based on Dutch principles conveyed by U of Alberta Prof. John (Jaap) Bakker. As city engineer in Medicine Hat, Alberta he pioneered the concept (while nobody was looking) and then in 1963 (±) talked the Superintendent of Edmonton Transit, D.L. MacDonald into using the concept to solve a problem created by annexing a large suburb which City Council had promised good bus service with little cost. MacDonald in turn was recruited by Portland to introduce the concept there. Later, I was hired by Denver for the same reason. All three cities have struggled with the concept because whenever an economic downturn came along political and bureaucratic forces would align to disrupt the network.
Norfolk also did a network redesign based on work by Bakker and then the system was merged with another and headed off in a different direction. A common thread is that it requires some extra work that is specialized and when people are hired without experience they sometimes don’t even check to learn where the transfers are. The standard North American software for scheduling actually encouraged sloppiness.
We need some rules of thumb for how many takts a city should have, and how far a bus should detour from an otherwise straight route to get to one.
My city has a takt every half hour. (the frequent route is every 15 minutes and and so only every other actually meets). However there are many around the city, so every 15 minutes the buses all meet at a takt. Many buses leave the midtown takt and go to the downtown one by a different route. (the outbound buses also meet there and go to an edge of town takt synchronized with the downtown takt. As a result my trip downtown which a car can do in 10 minutes takes 20 minutes (when I’m the only one on the early bus, but it is slightly longer in the afternoon going home because of those getting off). There is one bus that detours from an otherwise straight route for 15 minutes to get to the middle takt.
This seems wrong, but I don’t know if I’m just annoyed because I’m the person who is negatively affected by good practice, or if the system is bad.
You have right not to remain silent. My own rule of thumb is that the round-trip added time for the deviation should not be more than one-quarter of the standard headway. And in designing a network the focal point ideally will be on a main line service. Edmonton’s goal until 1984 or so was for that main line to be either LRT or a trolley coach line so that it would be difficult to add deviations later on. Denver’s service standards had a board-adopted anti-deviation formula that was mathematical but didn’t work well with major traffic generators or transit centers, but I found that rule of thumb workable in practice.
A lot of factors turn up in real life low density systems. In Klamath Falls, Oregon I was asked to set up a two-bus system. No matter how hard I tried I could not get each route down to a 60 minute cycle with a 60-minute headway without iffy connections or crazy driving. I ended up with a 70-minute headway and it worked perfectly. Two added factors entered with actual operation: the transfers were at the regional shopping mall next to the coffee and doughnut shop and the second success factor was that a retired NYCTA operator turned up thousands of miles from New York City and helped teach the other operators. In a small set-up I learned that it pays to look for local enhancements.
Is off-board fare collection significantly faster than payment onboard but with all door boarding, as is practiced on some smartcard systems? It’s a lot cheaper to fit every door with readers than it is to equip even a majority of bus stops.
At least in the US I believe there is significant mileage to get out of improving bus dwells by improving the bus itself to be more like a streetcar/train, particularly if all door boarding is implemented; doors should be much wider and towards the center, and the driver should be in a separate compartment raised above the floor level with the wheels and other equipment underneath. The current American low-floor bus layout takes up too much space with wheel wells protruding into the passenger area.
Also one major issue with takt-like schedules is that buses are so low-capacity that they are awful at dealing with sudden, large overflows of riders. The big example of this in New York where the relationship between the Department of Education and the MTA is apathetic at best and acrimonious at worst, where all schools let out nearly simultaneously; but even without this, a single school releasing students would quickly break down a takt-like schedule. More distributed but similar problems can be found at college campuses, where generally classes start and end at roughly the same times.
You can just add vehicles to the takt to do this. The Netherlands has more buses (including many entire routes) that only operate at the peaks to cater for situations like the one you mention.
As for trains, the Netherlands (and Switzerland?) are lucky to have no primate city and many dispersed job centres, meaning train commuting is more many-to-many than many-to-one like it is in the UK and especially France. In the latter case, you can just add trains to the takt. So the train operator where I went to ‘high school’ just added an extra local stopper timed to deal with the sudden outpouring of hundreds of students at 16:30 each day, on top of the standard takt.
It’s not a maximally efficient use of vehicles (and in the case I mentioned, it stretched the infrastructure to its limits) but sometimes you’ve got to do this.
There are several different ways of operating school trippers to supplement regular service. As Henry indicated there are cities with poor relations between transit and school systems and then it doesn’t matter which technical solution is chosen. And some of these issues go way back. A Denver public school superintendent launched into the RTD’s General Manager about uncooperative bus system management and it puzzled him. He had only been in Denver about four years and it turned out that she was talking about the private provider that was in the process of collapsing when the unrelated court ordered racial busing began. For two decades she had carried this anger. In the meantime, two or three articulateds had already been operating on school trippers for the high school in question, interlined from a neighborhood feeder route so that the kids did not have to transfer. The base service bus line was on a 30-minute headway.
One other typically American problem is that FTA regulations intended to force the use of yellow school buses is involved, but the school district only sees the transit system telling them that they can’t do something. Something that is true in many places is the amount of field work that is necessary at the start of semesters. Automatic Passenger Counters in the past decade have proved helpful in tipping off transit schedulers to unexpected overloads but someone may still have to go out and investigate. That brings me back to the issue of TTFP service design. In some cases the solution is double or triple-heading on the base clock schedule times. In others it’s better to split the base headway with more frequent trips.
BVG split the 20-minute headway on Linie 68 with a high-floor Tatra car after they had otherwise disappeared. That made two ten-minute intervals (or “spaces” as Denver tradition calls it). By Murphy’s Law I showed up with my trans-Atlantic suitcase expecting a low-floor car, but the kids helped me board and stow it. Without the tripper there would have been pass-ups hurting the regular customers.
If you pay on board with all-door boarding, you still need proof of payment enforcement. In practice I haven’t seen how this works – Singapore has front door-only boarding for some reason (it had POP once, with EZ-Link readers at bus stops). It can be pretty fast with a fast-read smartcard, but if you have the infrastructure for all-door boarding, might as well put readers at bus stops so that people don’t have to pay on board.
Oh yeah, you still need POP.
SF Muni moved to all-door boarding with validators at every door. I think Paris also has this implemented on its buses.
At least in San Francisco this reduced total dwells by 38%: https://www.bloomberg.com/news/articles/2015-03-13/san-francisco-s-all-door-bus-boarding-program-has-been-a-huge-success. And of course I can’t find this now when I need it, but I do remember seeing a source saying that the cost to install fare readers on all doors at all Muni buses was just shy of $1M.
It’s a lot easier to install readers on a bus than it is at a bus stop, since generally speaking there are simply fewer buses than bus stops, the interior of a bus is protected from the elements, and wiring on a bus is usually readily available whereas the same is not necessarily true where bus stops are currently located on the sidewalk.
That might be due to the distance fare system that requires validation on exit as well, so not only do entering passengers have to queue for the validators, they have to wait until all exiting passengers are done alighting (and processing their cards). Not that I don’t think it isn’t worth a try, but all-door boarding could well need a rework of the fare system and maybe even route design itself to make flat fares more palatable for both consumer and operator, so they only have to validate on entry.
Though I think this may be more prevalent in East Asia – from what I vaguely remember of HK buses they also did something similar, again with distance fares. And even in Kyoto, where there is nominally flat fares, the farebox (and IC validator) is at the front and you have to show your day pass to the bus driver, so you board from the rear door and pay on exit: https://www2.city.kyoto.lg.jp/kotsu/webguide/en/bus/howtoride_bus.html
From what I remember of Hong Kong, they have distance-based bus fares but with no tap-off. So riders boarding at the origin station, regardless of destination, will pay the highest fare, and the meter ticks down as the bus continues on its ride. With a bus route ABCD where each segment is $2, A to D is $8, A to B is also $8, B to D is $6, B to C is also $6, and C to D is $2.
People don’t take buses very short distances in Hong Kong as a result, and another result is that rather than clearly labeled trunk routes as is the fashion today with bus redesigns, the Hong Kong bus network is bewilderingly complex with all manner of permutations to get riders on buses that don’t cost too much. Although unlike other Anglosphere areas with labyrinthine bus networks, Hong Kong is dense enough to get all those routes with 15-minute double decker service throughout the day.
Argh, I messed up the fare cost. Math isn’t part of my day job.
Ah yes, reminds me of why the Octopus validators have a fare readout. But because the government collects fares in Singapore, doing that exactly would have political heads rolling pretty quickly.
Consequently over here, while the concept is similar at entry and there are POP inspectors to make sure you’ve validated on entry, we need exit validators to return $4 at B and $2 at C as per your example. In limited cases at terminals where there’s only boarding or only alighting they do use all doors, though, but that’s depending on the staffing on the day.
I find it hard to get such a system to work without going to flat fares or having validators at bus stops – not so difficult with decent 4/5G coverage islandwide and powering validators from the same sources as the lights.
Unfortunately, HK also doesn’t use POP. I’ve been in situations where it easily takes 5-10 minutes at an origin to load up a double decker until it’s full.