Sunday, June 21, 2015

What is the actual capacity of BRT?

This is the second in a series about the ITDP bus rapid transit report for Boston, and the ITDP standard in general.

When proponents of Bus Rapid Transit—

You know what? I need to redefine this. I am a proponent of BRT. But I am a proponent of BRT in context. When the ITDP talks about transit, they only mention BRT. Heavy rail, light rail, commuter rail, these are seen as competition, and need to be denigrated whenever possible. BRT is the solution, anything else is not even worth mentioning. 

This is myopic. Bus rapid transit is a tool, but just a a tool box needs more than just a hammer, transit needs a variety of modes working together depending on a city's existing infrastructure, needs and geography. BRT needs to be used where and when it is appropriate, but it is not a one-size-fits-all solution for every transit need. I've already discussed how BRT is not particularly compatible with narrow streets, and how the cities used as analogs to Boston are anything but. 

—So to begin again:

When propagandists of BRT (yup, I went there, ITDP) talk about the benefits of bus rapid transit, they don't tell the whole story. Their argument is that bus rapid transit has the ability to transport as many people as any other mode (45,000 per hour!), at a fraction of the cost. In very isolated cases, this may be true. However, they don't mention that this is an extreme outlier. The infrastructure required for that number takes up enough space that it is compatible only in urban areas with long, wide thoroughfares with space to build. Without this, capacities are an order of magnitude lower, and BRT is much harder to scale than rail.

Here is what the ITDP shows for capacities in people per direction per hour:

This is somewhere in the neighborhood of being true (it's, shall we say, rosy), but it shows absolute maxima, which for BRT are often attained in conditions which, in most cities, are unworkable. (Let's also set aside the fact that 6000 people per hour on a non-BRT bus system equates to 1 minute headways, that a four-track metro like the 6th Avenue Line in New York runs at a capacity of 60,000 per hour and theoretically could run at 100,000 and that light rail is capable of more than 20,000 passengers per hour in, for example, Calgary. So, it's basically not true then; see below.) The BRT number is from Bogotá, and it is an outlier. The way that Bogotá attains that number is by having the BRT system in the center of a highway with wide stations and two lanes for buses on either side, necessitating about 70 feet of street width. This requires four bus lanes at stations, and the street width to accommodate that something many cities just don't have.

Without this width, BRT carries many fewer people. Bus and rail transit scale in two very different ways. Imagine (or look at the chart to the right) a graph where the X axis is the route, and the Y axis is the width of the corridor or the number of lanes/tracks. Rail scales along the X axis, by adding vehicles to the train, so that going from one car to 10 cars gives ten times the capacity. However, adding a second track (increasing the Y axis) only doubles capacity, there are no similar economies of scale. BRT can only lengthen the vehicle so much; most BRT buses top out around 100 feet (carrying about 160 passengers). However, doubling the number of lanes a BRT uses increases capacity by 10 times (or even a bit more; the most frequent route in Bogotá has 350 vehicles an hour—a bus ever 10 seconds!). So while rail can scale by an order of magnitude within a narrow corridor, BRT scales best in another dimension. However, this requires four lanes of width, plus stations, to have the same increase in capacity.

This becomes an issue when capacity is an issue. For a line transporting 1000 or 2000 people an hour, rail is no better than bus: a single-car light rail train every 8 minutes has about the same capacity as a 60-foot bus every 4 minutes. (This is assuming they have similar signal priority, level boarding and fare collection mechanisms to minimize dwell times and unnecessary stops.) Both these frequencies are show-up-and-go frequencies; the average wait time for a three minute headway versus a six minute one is a negligible 120 seconds, a small percent of total trip time.

But if demand increases, a rail line can easily add capacity while a BRT system can not. Increase demand to 3000 people per hour, and a rail line will handle it fine: a two-car light rail train every seven minutes does the trick. However, a BRT system maxes out around 60 trips per hour, and even at this point, even a minor load imbalance (say, from connecting services) or a traffic light cycle missed (say, to allow pedestrians to cross*) will cause bunching. There are diminishing returns at very low headways as being slightly out of sync can cause bunching and crowding issues. There's a reason the BRT line in Los Angeles (the Orange Line) has four minute headways, and not less. Beyond that, bunching, and accompanying diminishing returns, are inevitable.

[Update: Mexico City has more frequent service, it's just that Google Maps transit doesn't show that. Thanks, Google Maps! And I didn't go in to the GTFS file to see what was going on, and it's a somewhat complex file! So, Mexico's BRT system has higher throughput, especially given their longer buses, maxing out around 12,000 per hour. Of course, with vehicles every minute at-grade, bunching is inevitable as crossing phases have to be a certain length on wide streets, so speed declines. It's certainly faster than minibuses in mixed traffic, which the system replaced.]

Beyond 3000 people per hour? A two-lane bus system has problems; crowding will increase dwell times, and capacity or speed may actually go down. A light rail line will reach this point as well, but will be carrying many more passengers when it does so. Boston and San Francisco run 35 to 40 light rail trains per hour underground, with a capacity of 15,000 passengers per hour (Boston, with some three-car trains, actually has a slightly higher capacity). Calgary runs 27 three-car trains (with plans to increase to four) through downtown at rush hour, at-grade! 27 four-car trains will give it a capacity of 22,000 per hour. (Their system carries more than 300,000 riders per day, higher than Boston or San Francisco.) That's on par with pretty much any BRT system (Bogotá's is over capacity, and they are actively looking to build parallel lines to reduce the demand on the main trunk routes.), but the stations and track only take up about 40 feet of street width, enough for a lane of traffic and wide sidewalks in an 80-foot building-to-building downtown corridor, still narrower than any BRT street in Bogota.

In any case, the chart that the BRT report has should actually look something like this, accounting for typical loads and outliers:


Typical loads are lines such as the Broadway IRT for the four-track metro, the Red Line in Boston for the two-track metro, a single branch of the Green Line for the LRT, and the Orange Line in LA for BRT. I took a guess at the typical throughput of a four-lane BRT; I couldn't find any specific schedule or loading data.

Maximum loads are theoretical maxima. For a four-track metro, this is double a two-track metro (the 6th Avenue line is the busiest trunk line in New York, running about 30 trains per hour with a capacity north of 60,000, but could carry more). For two-track metros, several are in the 40000 range: the Victoria and Central lines in London (33 trains per hour, 1150 passengers per train), and the L train in New York (20 trains per hour, 2200 passengers per train). For BRT, four lane, the number is from Bogota. For light rail, the number is from Calgary, assuming they implement four-car trains as scheduled this year. And for BRT, two lane, the number is from a single-lane, one minute headway system with 100-foot buses (which don't exist in the US).

Two notes:
  1. Bogotá's system is an outlier. Most BRT systems carry many fewer passengers, especially the majority of lines which do not have passing lanes at stations to increase their capacity. While light rail can scale dramatically, BRT can not, unless the streets are wide enough. Which, in Boston, they're not.
  2. Four-lane BRT is akin to four-track metros in capacity enhancement (a four-track metro can carry, in theory, more than 100,000 passengers per hour). However, a four-track metro is only necessary in very high demand situations; most two-track metros can be scaled to meet demand. Four-lane BRT, however, is necessary even when demand is well below what a typical metro line, or even light rail line, might carry. 
Here's another way to look at capacity. It shows how different transit modes attain capacity: rail by adding vehicles (and, to get very high frequency, extra tracks) and BRT by adding passing lanes and frequency. It also shows a dotted line at 60 trips per hour—a one minute headway. Most systems operate to the left of the dashed line. In the case of rail systems, this is because more capacity is generally not needed. In the case of BRT, however, it is because the system is operating near its maximum. In reality, the lines should curve flatter beyond 30 trips per hour (except for four-lane BRT) as bunching and load imbalances cause diminishing returns.
In any case, it's another way to show that while BRT is a useful tool in the transit toolbox, it has a very finite capacity unless it can be expanded to four lanes (plus stations). If you are trying to design a system which can scale, you either need to have that corridor space available (as is the case in Bogotá), or build a rail line. Without that, bus rapid transit can carry about 2500 passengers per hour, but it can't scale higher.

[ * A note on pedestrians: surface BRT is constrained by the length of crossing traffic light cycles. Even with full signal preemption, a crossing cycle needs to be long enough to clear crossing traffic, and for pedestrians to cross the street. In most cases, a BRT corridor will be wide enough to require 30 seconds of pedestrian crossing time. At 5 or 6 minute headways, this is not a problem; the BRT only requires 10 or 15 seconds every two to three minutes, or so. At three minute headways, it requires 15 seconds every 90 seconds, and at two minutes, 15 seconds every 60, and at a minute, BRT requires half of the signal time. It is likely that buses at this frequency would, at times, be forced to stop because of the length of the pedestrian phase (and to keep cross traffic flowing at all), which would create bunching and crowding problems downstream. Again, most single-lane BRT networks operate at four minute headways, which constrains capacity. Beyond that, they lose signal priority advantages, which constrains speed. In other words, there's a fair argument that for surface transit, a three minute headway may be better than a one minute headway.]

16 comments:

  1. One interesting observation about capacity is most of the time, the limiting factor is actually station or junction capacity, rather than plain line capacity. In the case of subways, it's almost always stations, because even if the signal system allows trains to run with zero separation, there is always going to be some amount of dwell time for passengers to get on and off, and some amount of time for a train to accelerate out of the station and the next train to decelerate to a stop. For BRT, especially in its four-lane incarnation, this is less of a problem, because a bus in a station doesn't necessarily block the ones behind it. But the station ultimately still has a finite size, and there's a minimum dwell time required to get everyone on and off the bus, so unless you have a station the size of the Port Authority Bus Terminal, stations are going to constrain capacity at some point along the line.

    The other capacity constraint is junctions: for metro trains, it's generally waiting for opposing moves to clear through a crossover at a terminal, because the highest capacity lines have no other junctions at all. For at-grade transit, there's also a need to deal with cross-traffic, whether it's pedestrians, cars, or other buses/trains, and when the headway gets close to the frequency of the light cycle, the traffic lights will start to enforce bunching (or just plain limit throughput).

    ReplyDelete
    Replies
    1. All true. The L, 1, 6 and 7 trains are able to provide three minute headways because they don't intersect with any other lines, so terminal space is the issue. The L has higher throughput because the loading gauge is so much larger for those cars.

      The 1 and 6 trains operate as part of local-express services, but both are fully grade-separated from the express trackage. This is not the case with BMT/IND services, which have more complex trunk-branch systems. Very good points, and basically supports the thesis: buses scale because they are more nimble, but only if they have enough room. In a constrained corridor, rail has much higher capacity.

      As far as how it relates to BRT, yes, terminal space is an issue. In Bogotá, there are huge, huge bus yards at the end of each line. With loops and tail tracks, you can't get good throughput.

      Delete
    2. The problem with this hypothesis is simple. . . BRT does not take care of increased demand in a linear fashion. Hence, as stated in the article, additional lanes are required to simply double capacity while other modes, like heavy and/or light-rail, can accomplish this in a linear fashion.

      Further, does it take much imagination to envision the nightmare of clogged entry/exit points to bus transfer facilities because of the number of additional buses required to keep pace with demand; while similar demand can be accommodated by fewer light or heavy rail trains? What happens to bus dwell time when facility parking becomes a challenge?

      Also, consider the astronomical bus maintenance and operational costs associated with hiring legions of men and women to operate the number of additional buses necessary to meet ambitious headway demands, as opposed to hiring one operator per railcar consist? The math is scary!

      Delete
  2. Good piece. I've also written a bit about BRT. The point is that a bus line's capacity is pretty much limited to about 3 000 pphd or so for articulated buses (5 000 for bi-articulated buses) before bunching and severed overcrowding significantly affects service quality. The way that BRT can offer higher capacity is if many lines overlap on a bus corridor with a passing lane and huge stations with many quays to allow many buses to stop at the same time.

    The result is that the capacity is really the corridor's capacity, not a line capacity like for rail transit. As a result, the practical capacity is going to be lower than theoretical capacity because some lines are probably going to be under-used. For example, if at station A there are 8 000 people per hour going to station B and 1 000 going to station C, but only two lines serve station B and 1 line serves station C, with each line having a capacity of 3 000 pphd, then the two lines serving the A->B trip will be overcapacity and will leave people at the station, but the line serving A->C will be under-capacity. So the corridor, despite having a theoretical capacity of 9 000 pphd will have a practical capacity of just 7 000 pphd.

    Well, to be fair, there is another way, and that is to run buses in convoys, having two or three buses follow each other, stop at the same time and accelerate away from stops at the same time. It can work to increase capacity, but it significantly increases dwell times. An attempt to do so showed that the practical capacity of the line was 12 000 pphd... versus 27 000 pphd theoretical capacity.
    https://goo.gl/zXM0xw

    Anyway, thank you also for seeing through the ITDP's insane BRT and anti-rail obsession. It greatly annoys me that they are portrayed as a reasonable, non-partisan institute when they are the single most dogmatic, fanatical soldier of the "mode war", whereas people who defend rail against their accusations by pointing out some advantages rail has over buses are quickly accused of "railfanning" or of having an anti-bus bias.

    BTW, some additional data for the maximum capacities... Tokyo's Yamanote line, which is essentially a metro line running on two tracks, actually has a capacity of 80 000 pphd while maintaining an average speed of 40 km/h (25 mph). Manila's Line A, an elevated LRT line, has also achieved 40 000 pphd and goes at about 30 km/h (18 mph).

    An additional point that I think should be considered on at-grade lines with intersections is that the shorter the headway between vehicles, the lesser the transit signal priority that can be implemented in traffic lights on the way. For example, in Paris, some of their tramway lines run 2-car trains with a capacity of maybe around 500 people each, with an headway of 5 minutes. As a result, they can implement very strong signal priority for their trams, which have near train-like priority and are almost never stopped at traffic lights. It's okay, because they call their priority only 24 times per hour (12 times in each direction). If they replaced these trams with articulated buses, they would need headways of about 1 minute only between vehicles, leading to 120 vehicles crossing intersections each hour. At this frequency, the tram signal priority would result in the intersections being completely jammed for every other user at these intersections: cars, buses, cyclists and pedestrians. As a result, only a lesser signal priority could be used that would significantly delay and cause unreliability in the bus line.

    ReplyDelete
    Replies
    1. Good point about the service complexity in BRT. There's a video on the topic of "what is wrong with the TransMilenio" (https://www.youtube.com/watch?v=HfrBFM8L64Y) and most of it is making pretty much exactly this point.

      Delete
  3. You say, "a BRT system maxes out around 30 trips per hour." Portland has streets that serve 160 buses per hour. Substitute 160 for 30 in your calculations and see how BRT compares with rail.

    ReplyDelete
    Replies
    1. He is correct. You are comparing apples to oranges. Maybe there are 160 buses per hour going through on the street, but do they all stop and serve the street or are most of them express buses that skip all the stops and end right to a terminal?

      Looking at Portland transit data, it seems that one of the most used, non-terminal, bus stop is stop number 9301, at the corner of 5th and Davis. At its maximum, it has 36 buses that can stop there per hour, very close to the 30 bus limit described here, which is a limit for the best operational service. In reality, the practical limit may be around 40-45 buses per hour, but at that point the service is clearly problematic with buses interfering with one another and a high likelihood of bus queuing at stops.

      The bus stop that is most used is 5th and Pine (ID 7631). Around half the buses that stop there are expresses that either start or end the line there and have no other stop nearby, despite that, it serves a maximum of just 59 buses per hour, and it has a massive 250 ft bus stop to allow buses to queue up, sign that the bus stop is overcharged.

      You clearly don't understand transit. Transit needs to serve a corridor to be useful, if there are 120 buses going straight on a street during an hour but only 8 stop, then for the transit user on that street, the effective transit service is just 8 buses per hour.

      Just like intersections are what limit road capacity, stops/stations are what limit transit capacity, hence why the 30-bus limit exists, though one bus lane can carry hundreds of buses per hour IF THEY DON'T HAVE TO STOP. Though you can use expresses to bypass stops, their usefulness is limited, and an overreliance on expresses creates a confusing, fragmented transit network where local trips end up requiring a transfer downtown, and where transit authorities are faced with the dilemma of either running empty express buses to maintain the network in off-peak periods or not running expresses outside the peak, resulting in the transit network disappearing for many residents during hours of every day. It is better for a transit network to have few lines with high frequency than to multiply lines, each having low ridership.

      I have a good example of this in Montréal, there is a bus lane on the Champlain bridge that carries hundreds of buses per hour to a downtown bus terminal, where buses from suburbs on the South Shore end up. These buses have no stop within Montréal itself, just the terminus, in order not to jam the bus lane. If it weren't for the subway system allowing people to quickly head up to their actual destination, bus use would be significantly less, because not everyone can have the chance of working within walking distance of the bus terminus. Though the bus lane on the Champlain bridge carries nearly as many people as a subway line, it doesn't actually provide service to the surrounding area, whereas the subway carries as many people and serves every single station on its way.

      That is why all transit planners of the Montréal region want the bus lane converted to a high-capacity grade-separated LRT system, with a projected capacity of 20 000 pphd, a project that will also save untold amounts of man-hour for suburban bus companies as they will no longer be forced to run every bus to downtown at every hour of the day, they will be able to simply run them to the nearest LRT station, from which people will be able to go downtown BUT ALSO to go to other suburban stations, where they will be able to take other buses to other parts of the suburbs, trips that now require a transfer downtown.

      Delete
    2. And if you do have 160 buses per hour running down a street, but each one only stops every 4 blocks so as to spread out the demand on the stops, you've just made your system that much more user-hostile, since now a passenger can't just take the next bus that comes, since that might require running two blocks to catch it. Now, that isn't a problem during rush hour, since there are so many buses, and hopefully stop locations are grouped sufficiently well by destination. But what about off-peak or late evening when buses are much less frequent? Then you have the choice of having to wait halfway between stops to see which bus shows up first and sprinting one way or the other to catch it, or else just picking a stop and waiting there as the bus you could've taken drives by while the route whose stop you're waiting at gets delayed and won't show up for another 15 minutes.

      Delete
    3. Both of the stops referenced here by simval have two lanes for the exclusive use of transit. That's the only way they can easily get this kind of throughput. I think that kind of proves the point I made in the post, doesn't it? Also, as a faithful reader of the Antiplanner Comedy Hour®, thanks for dropping by, Randall.

      Delete
    4. Simval84: You clearly don't understand the Portland system. Careful timing has proven that each bus stop can handle 42 buses per hour. Four stops every two blocks can potentially handle 168 buses per hour, but in actual practice were scheduled for 160 buses per hour. You can move a lot more people with 160 buses per hour than with any light-rail system.

      Arcady: Portland's transit mall has buses stop every two blocks. Every block has two transit stops on it. It has worked for nearly 40 years. The only glitch was when they put light rail on the same mall, reducing its overall capacity.

      Ari: The mall has one lane for exclusive use of transit and one parking lane. Besides, what if it does have two lanes? ITDP's Gold Standard requires dedicated lanes, and you claimed it could run only 30 buses an hour.

      Delete
    5. Parking lane?

      For those not familiar with Portland, south of burnside the transit mall is ~30 feet. Wider sidewalk to accommodate waiting passengers, a lane for stopped transit vehicles, and a lane for passing transit vehicles. Then the next block over, ~250 feet away, you have the same going in the other direction.

      If you have a street grid then you can split the needed road space. 70 feet might be too
      much to expect to find. But 30 feet is often doable.

      Randall this post criticized quoted BRT capacities as misleading because the typical implementation has one lane in each direction, and it's argued here that that is not enough to offer anywhere close to the capacity of a metro. You respond by giving an example of an implementation with two lanes. Well done.

      Delete
    6. Dear antiplanner, careful timing for at-grade bus lines with no pre-boarding payment nor all-door boarding is a pipe dream. Dwell times fluctuate wildly and so cannot be predictable, vehicles may be delayed at intersections. There is no way to guarantee that buses will remain on their schedule, especially not on a bus stop with that many buses stopping there. Again, you clearly don't understand transit operations.

      160 40 ft buses means about 10 000 pphd. That is still within the realm of what is achievable with light rail. 3-car light rail trains can carry up to 800 people, so they would need only 13 trains per hour to match that capacity. Elevated LRT can do even more, Manila's Line A LRT achieves a capacity of 40 000 pphd, the equivalent of 670 40-ft buses or 400 articulated 60-ft buses per hour.

      More importantly, when 160 buses per hour travel along a road, but are still limited to 30-35 buses stopping at each bus stop, then the big problem is that, for users, not all buses are relevant to them, because bus lines skip stops, so if a bus line either skips the stop the transit user is at or skips the stop he's going to, that bus line is not actually relevant to him. On a macro level, bus stops will not be uniformly used, so that means that some trips will be more frequent than others, as a result, you are never going to see all bus lines at 100% capacity, you are going to see some bus lines at capacity because the bus stops are full, with no possibility to add service, and then you are going to see some bus lines at 60 or 70% capacity, or even lower, because the route they serve just isn't as much in demand as others. As a result, the practical capacity of such complex bus network is always significantly inferior to their theoretical capacity, meanwhile, rail is frequently observed with practical capacity being over the theoretical capacity due to overcrowding vehicles with more passengers.

      The great thing for light rail is that EVERY vehicle is useful for EVERY user on that corridor, so that simplifies things a lot both for the transit planner and the transit user. You can also get away with simple network maps that allows users to find out their itinerary in a couple of seconds, no matter where they are going.

      Delete
  4. Okay, let me clear some things up:

    Yes, this is an example of using narrower rights of way to move more buses than a single two-way street could. See also the Marq-2 project in Minneapolis where a number of routes—mostly freeway express routes—were consolidated on two parallel avenues downtown. It's a great idea. It works well. It's not BRT.

    What it is is a solution to the terminal problem: if you have a lot of buses coming in to a dense area, you need a lot of space for them to stop. You can build one huge terminal (like the PABT in NYC, or the Champlain Bridge in Montreal) but then you need a distribution network (subway). (Thus extending the 7 train to New Jersey and moving the PABT there; you don't lose any utility and you gain very valuable land in Midtown.) The one-way pair solution works to take buses which come in from the periphery and let them act as distributors along a corridor in the core, spreading this terminal out linearly. It works, but it's not a perfect solution, and requires a good deal of street space. It has never been replicated over long distances, and it likely won't be.

    Why not? Because in a downtown area, it's possible to have long stretches of roadway without any vehicle access to one side of the street. Most people arrive without a car, so you can have half the street inaccessible to vehicles without disruption. Portland and Minneapolis both have a few parking driveways accessible across transit-only lanes; in a BRT system this would be discouraged. But outside of downtown, you need access to both sides of the road for things like driveways and deliveries and even, say, bike access, especially since longer blocks and smaller lots mean that most people won't be able to access their homes and businesses from a cross street. Alleys may help, but not all cities have them. So you'd need two bus lanes, plus a station, plus room on both sides. (Yes, it's conceivable you could have a single lane in the middle and then, in areas with stations, two lanes beside the curb, the inside one for stopping and the outside one for passing, but you'd still be cutting off the sidewalk in those areas. You could let vehicles stand in those stops, but then you're kind of defeating the exclusivity of the lanes and introducing myriad other potential conflicts. Also to my knowledge it hasn't been done anywhere.)

    So for something like this you need at least 50 feet of road width for four lanes and a parking/stopping lane somewhere along the way, although likely a bit more; even in Mexico City there are no BRT corridors less than 55 feet wide. (I'm not counting Line 4, which runs 40 foot buses through the Centro Histórico; it's another area where there is no vehicle access but not portable to North American cities.) But that only gets you one direction. So you need a parallel road that also has this road width, and it should really be very close; as both Jarrett Walker and Alon Levy have pointed out, one-way pairing is bad for transit, causing people to have longer walks to access service. And more to the point, pairs of wide-enough one-way streets are few. Most cities developed with wide streets every half mile or so (in parts of the midwest, every half mile, religiously), far too far apart for this type of service. There are probably a few examples, but it's not widespread. In any case, this is a possibility, but it hasn't been put in to place anywhere.

    Finally, this series, so far, is in reference to a study of BRT in Boston. And Boston is a city with no grid to speak of, no buses that run through downtown, and certainly not a city with pairs of wide streets running parallel hither and yon.

    ReplyDelete
  5. Another fun fact about light rail in Boston: on Beacon St at least, the tracks are built with a 9'6" centerline spacing, which is narrower than you would ever be able to build a busway, at least if you want the buses able to move at any reasonable speed. On our narrow streets where every inch counts, this does make a difference.

    ReplyDelete
  6. Excellent summary. I would like to talk with you about including both the Prime Law of Networks and Personal Rapid Transit network into your analysis. Here is a link to the Prime Law (http://www.jpods.com/primelawofnetworks) and JPods capacity as an illustration of PRT (http://www.jpods.com/capacity).

    ReplyDelete