I expect there will be writeups about the talk (e.g. on Streetsblog). But meanwhile, here are my slides (warning: 17 MB, because of pictures). These are identical to what was shown at the talk, with two differences: I fixed one small mistake (Fordham Road vs. Pelham Parkway), and I consolidated the pauses, so each slide is a page, rather than a few pages, each page adding a line.
There were light fantasy maps in the talk. Because of size, I’m not embedding them in the post. But there are links:
- Infill stops without new tunnels: low-res/3.5 MB, high-res/20 MB
- The 3-line system (Gateway and realigned Empire Connection): low-res/4 MB, high-res/44 MB
- The 5-line system (the Lower Manhattan lines): low-res/4 MB, high-res/44 MB
Yellow highlights around a line indicate it’s new; Gateway is highlighted in one direction since it’s an existing two-track line to be four-tracked. On the infill map, solid circles are existing stations, gray circles are planned stations, white circles are my suggestions for additional infill.
In New York, there are two dedicated subway tracks on the Manhattan Bridge offering a bypass of Lower Manhattan. Between DeKalb Avenue in Brooklyn and Canal Street in Chinatown in Manhattan, Q trains run nonstop for 3.5 km, while the R train goes the long way, taking 5.5 km and making 2 intermediate stops in Downtown Brooklyn and 4 in Lower Manhattan. The N skips DeKalb Avenue, with a 4.5 km nonstop segment between Canal Street and the Atlantic/Pacific/Barclays station complex.
The Q and N should be immense time savers. Instead, the Q does the trip in 8 minutes and the N in 10, both of which average 26-27 km/h. The subway’s overall average speed, weighed down by local trains stopping every 700 meters, is 29 km/h. The Q and N are still time savers, though, because the R does the 5.5 km in 18 minutes, an average speed of 16 km/h – far less than the systemwide average, and even less than the slowest Paris Metro line, Line 4 with its 500-meter interstations and 20 km/h average speed. Between DeKalb and Pacific, about 800 meters, the R takes 3 minutes. Unfortunately, New York City Transit is not taking any measures that would fix this, and when I asked about one possibility, I got excuses.
There are two reasons why this part of the subway is so slow. The first is something called signal timers. Timers are devices installed at frequent intervals on long interstations, such as the bridges and tunnels connecting Manhattan with Brooklyn and Queens, limiting train speed. These timers have always been around, but after fatal accidents in the 1990s, New York City Transit tightened them, reducing speed further; for some more background, see my Vox piece from last summer. The timers are more safety theater than safety. The biggest conclusion I reached from looking at the accident postmortem on the NTSB and some NYCT information was “make sure your trains’ brakes work as intended”; NYCT derated the trains’ service and emergency braking rates later in the 90s, which marginally reduces maintenance costs but is bad for safety and brutal for train speed.
The second reason is the switches at DeKalb Avenue. DeKalb is a six-track station, with four tracks feeding the Manhattan Bridge and two feeding the tunnel through Lower Manhattan. The two tunnel tracks then continue to the south as local tracks on the Fourth Avenue Line, carrying the R; this is the least used of all subway trunk lines into Manhattan, because the detour and low speed make it useless for most Midtown-bound passengers. The four bridge tracks include two express tracks at DeKalb going to the Brighton Line, and two super-express tracks skipping DeKalb continuing to the south as express Fourth Avenue tracks. Today, there is a splitting and recombining of branches. The B and D run together from Sixth Avenue to the Manhattan Bridge, and the N and Q run together from Broadway, but just north of DeKalb they recombine as B and Q running to Brighton, and D and N running super-express down Fourth Avenue.
This recombination at DeKalb slows down trains considerably, in two ways. First, the interlocking is complex. You can see it on this map on NYCSubway.org; in addition to splitting and recombining the B, D, N, and Q, it also has a non-revenue connection allowing R trains to serve the Brighton Line. Trains on diverging turnouts go at glacial speeds. And second, trains from four lines influence one another’s schedules, and delays propagate. Supervising train movements is thus difficult, and control center has to have a camera watching the trains enter the interlocking to ensure they adhere to schedule; timetables have to take the resulting delays into account.
When I first complained about reverse-branching in New York, I talked about capacity limits imposed by having more trunk lines than branches, a situation that is still to some extent true going north and east of Midtown. At DeKalb, there are six tracks going in and six going out, but the recombination makes things slower, and should be removed. NYCT should make a decision between having B and D trains run on the Brighton Line and the N and Q on Fourth Avenue, or the reverse. The interlocking permits either option, with entirely grade-separated junctions, allowing the trains on the two lines to no longer interfere with each other’s operations.
I in fact asked NYCT about it by proxy. NYCT dismissed the idea, on the grounds that transfer volumes between the B/D and N/Q would be too big. At Atlantic/Pacific, the Pacific side has a cross-platform transfer between the local R and express D/N, but going between the Pacific side and the Atlantic side (the B/Q, and separately the 2/3/4/5) involves a lot of walking. NYCT believes that passengers would flood the corridors looking for a train to their preferred destination, and the transfer volumes would require trains to have long dwell times. NYCT said nothing about whether the overall speed would actually fall, but I believe that based on the large transfer volumes NYCT predicts, passenger trip times (including transfer times) would rise. The only problem: I don’t believe NYCT’s prediction is true at all.
The B and D trains go express up Sixth Avenue, making stops at Grand Street in Chinatown, Broadway-Lafayette on Houston Street, West Fourth Street in the Village, and Herald Square. The N and Q trains go express up Broadway, serving Canal Street in Chinatown, Union Square, and Herald Square. North of Herald Square the two lines are never more than one long block apart until they leave Midtown. Passengers going toward Midtown are unlikely to have strong opinions about which of the two lines they would prefer.
Passengers going to destinations between Manhattan Bridge and Midtown might register stronger preferences. Union Square is the fourth busiest subway station in New York, and is quite far from the B and D. The closest alternative using the B and D is to change cross-platform to the M or F at West Fourth, and get off at 14th Street and Sixth Avenue, two long blocks from Union Square. Three more stations are potential concerns: Canal Street ranks 18th, West Fourth ranks 21st, and Broadway-Lafayette ranks 25th. Getting to Broadway-Lafayette from the N or Q is easy: the station and Canal Street are both on the 6, and passengers can transfer to the 6 at Canal.
West Fourth and Canal remain concerns, but they are not huge ones; they are secondary destinations. Canal is only a major destination for Chinese-New Yorkers, and in Brooklyn they cluster in Sunset Park along Fourth Avenue, suggesting that the Fourth Avenue express tracks should carry the N and Q and the Brighton tracks should carry the B and D. The urban geography of Chinese-New Yorkers is changing due to the combination of fast immigration and fast integration and migration to the suburbs, but this is a service decision, not an infrastructure investment; it can be reversed if demographics change.
Moreover, as a destination, West Fourth is predominantly used for NYU. The Village is a dense residential neighborhood, and West Fourth allows its residents to easily reach Lower Manhattan, Downtown Brooklyn, and two different four-track trunk lines through Midtown. But it has few jobs, outside NYU, which lies mostly between Sixth Avenue and Broadway. Union Square can adequately serve people going toward NYU, and stations on the R and 6 to the south can serve people going to NYU even better. The one problem is that the transfer between the R and the N/Q at Canal Street is not cross-platform; the cross-platform transfers start at Union Square. But with coverage of multiple stations walkable to NYU, the loss of the one-seat ride to West Fourth is not fatal. Even the transfer to the A, C, and E trains at West Fourth has alternative options: passengers from the N or Q going to the E can transfer to the F or M at Herald Square and reach the same stations, and passengers going to the A or C can transfer to the 1 at Times Square and to the A or C at Columbus Circle, both of which transfers are not much harder than climbing two flights of stairs at West Fourth.
With so many options, not many riders would be connecting at Atlantic/Pacific, and trains could keep dwell times short. If anything, dwell times might be shorter, because missing a train would be less fatal: the next train on the same track would serve the same destinations in Midtown, so riders would only need to wait about 3 minutes at rush hour, and 5 minutes off-peak. The gain in speed would be substantial, with the interlocking imposing fewer operational constraints.
NYCT might need to slightly rework the switches, to make sure the chosen matching of the lines in Manhattan and Brooklyn takes the straight and not the diverging direction at the turnouts; typically, the straight direction imposes no speed limit (up to full line speed on high-speed rail lines), but the diverging direction is slow. A matching in which the B and D go on Brighton and the N and Q on Fourth Avenue express to my understanding already involves only one diverging move, if I am reading the track map linked on NYCSubway.org correctly. At the same time, NYCT could fix the switches leading to the R: there was through-service from the Brighton Line to the tunnel tracks the R uses today, but there no longer is, so this out-of-service connection should get diverging and not straight moves. But even with the R, the capital investment involved is minimal.
I do not know the potential travel time gains between DeKalb and Canal Street (or Grand Street) with no timers or reverse-branching. With straight tracks across Manhattan Bridge, and wide curves toward Grand Street, 3.5-minute trips are aspirational, 4-minute trips are still possible, and 5-minute trips should be easy. From Pacific Street, add one more minute, corresponding to cruising at 50 km/h, a speed limit the subway routinely attains even on local tracks. This saves passengers from DeKalb about 4 minutes, and passengers from Pacific about 5. The average trip across the system is about 21 minutes, and the average delay (“excess journey time“) is 3 minutes. The saving would be immense, and contribute to both more casual ridership between Brooklyn and Manhattan, and lower operating costs coming from faster trips.
NYCT should not make excuses for this. The timers may have been originally justified as a safety improvement, but reducing train braking rates had the opposite effect. And, uniquely among the various reverse-branch points in New York, DeKalb feeds two Manhattan trunks that are very close to each other, especially in Midtown, to the point that one-seat rides to every stop have limited value. It should make a decision about whether to run the B/D together on Fourth Avenue and the N/Q on Brighton (switching the Q and D) or the reverse (switching the B and N), based on origin-and-destination data. Some passengers might bemoan the loss of one-seat rides, but most would cheer seeing their trips sped up by 4-5 minutes.
Chatelet-Les Halles has a problem with passenger circulation. It has exceedingly wide platforms – the main platforms, used by the RER A and B, are 17 meters wide – but getting between the platform level and the rest of the station runs into a bottleneck. There are not enough stairs and escalators between the platform and the mezzanine, and as a result, queues develop after every train arrival at rush hour. Similar queues are observed at the Gare du Nord RER platforms. The situation at Les Halles is especially frustrating, since it’s not a constrained station. The platforms are so wide they could very easily have four or even six escalators per access point flanking a wide staircase; instead, there are only two escalators, an acceptable situation at most stations but not at a station as important as Les Halles.
This is generally an underrated concern in the largest cities. In smaller cities, the minimum number of access points required for coverage (e.g. one per short subway platform, two per long platform) is enough even at rush hour. But once daily ridership at a station goes into the high five figures or the six figures, a crunch is unavoidable.
There are two degrees of crunch. The first, and worse, is when the capacity of the escalators and stairs is not enough to clear all passengers until the next train arrives. In practice, this forces trains to come less often, or to spread across more platforms than otherwise necessary; Penn Station’s New Jersey Transit platforms are that bad. The situation at Les Halles and Gare du Nord is a second, less bad degree of crunch: passengers clear the platform well before the next train arrives, but there’s nonetheless a significant queue at the bottom of the escalator pits. This adds 30-60 seconds to passenger trip times, a nontrivial proportion of total trip time (it’s a few percent for passengers within the city and inner suburbs). Avoiding even the less bad crunch thus has noticeable benefits to passengers.
The capacity of a horizontal walkway is 81 passengers per minute per meter of width (link, p. 7-10). This is for bidirectional travel. Unidirectional capacity is a little higher, multidirectional capacity a little lower. Subway platforms and passages are typically around 5 meters wide, so they can move 400 passengers per minute – maybe a little more since the big crunch is passengers heading out, so it’s unidirectional with a few salmons (passengers arrive at the station uniformly but leave in clumps when the train arrives). Busier stations often have exits at opposite ends of the platform, so it’s really 400*2 = 800. Queues are unlikely to form, since trains at best arrive 2 minutes apart, and it’s uncommon for a train to both be full and unload all passengers at one station.
An escalator step can be 60 cm, 80 cm, or 1 meter wide, with another 60 cm of handrail and gear space on both sides. On public transit, only the widest option is used, giving 1.6 meters of width. The theoretical capacity is 9,000 passengers per hour, but the practical capacity is 6,000-7,000 (link, p. 13), or 100-120 per minute. This is more than pedestrian walking capacity per unit of step width, but less per unit of escalator pit width. So a pedestrian walkway ending in a battery of escalators will have a queue, unless the width of the escalator bank is more than that of the walkway leading to it.
Moreover, escalators aren’t just at the end of the station. The busiest train stations have multiple access points per platform, to spread the alighting passengers across different sections of the platform. But mid-platform access points have inherently lower capacity, since they compete for scarce platform width with horizontal circulation. It appears that leaving around 2 meters on each side, and dedicating the rest to vertical circulation, is enough to guarantee convenient passenger access to the entire platform; in a crunch, most passengers take the first access point up, especially if there’s a mezzanine (which there is at Les Halles).
Should New York invest in better commuter rail operations, it will face a bigger risk of queues than Paris has. This is for two reasons. First, New York has much higher job density in Midtown than Paris has anywhere, about 200,000/km^2 vs. perhaps 100,000 around La Defense and the Opera (my figures for both areas in Paris have huge fudge factors; my figure for New York comes from OnTheMap and is exact). And second, Manhattan’s north-south orientation makes it difficult to spread demand across multiple CBD stations on many commuter rail lines. One of the underrated features of a Penn Station-Grand Central connection is that through-trains would have passengers spread across two CBD stops, but other through-running regional rail lines would not have even that – at best they’d serve multiple CBDs, with one Midtown stop (e.g. my line 4 here).
When I computed the needs for vertical circulation at a Fulton Street regional rail station in this post, I was just trying to avoid the worse kind of crunch, coming up with a way to include 16 platform-end escalators (12 up, 4 down in the morning peak) and 16 mid-platform escalators (8 up, 8 down) on a 300-meter long two-level station. It’s likely that the escalator requirement should be higher, to avoid delaying passengers by 1-1.5 minutes at a time. With four tracks (two on a Grand Central-Staten Island line, two on a Pavonia-Brooklyn line) and 12-car trains arriving every 2 minutes, in theory the station could see 240,000 incoming passengers per hour, or 4,000 per minute. In reality, splitting passengers between Grand Central and the Financial District on what I call line 4 means that a sizable majority of riders wouldn’t be getting off in Lower Manhattan. When I tried to compute capacity needs I used a limit passenger volume of 120,000 per hour, and given Midtown’s prominence over Lower Manhattan, even 90,000 is defensible.
90,000 per hour is still 1,500 per minute, or 3,000-4,000 if we are to avoid minute-long queues. A single up escalator is limited to about 100-120 people per minute, which means that twenty up escalators is too little; thirty or even forty are needed. This requires a wider platform, not for horizontal passenger circulation or for safety, but purely for escalator space, the limiting factor. I proposed an 8-meter platform, with space for four escalators per end (two ends per platform, two platforms on two different levels), but this suggests the tube diameter should be bigger, to allow 10-meter platforms and six escalators per end, giving four up escalators per end. This is 16 up escalators. Another 16-20 up escalators can be provided mid-platform: the plan for eight up escalators involved eight access points interspersed along the platform, and 10-meter platforms are wide enough width to include three escalators (two up, one down) per bank and on the border of allowing four (three up, one down).
The situation at the Midtown stations in New York is less constrained. Expected volumes are higher, but Grand Central and Penn Station both spread passengers among multiple platforms. In the near term, Penn Station needs to add more vertical circulation at the New Jersey Transit platforms. The LIRR remodeled its section of the station to add more access points in the 1990s (e.g. West End Concourse), but New Jersey Transit is only doing so now, as part of phase 1 of Moynihan Station, and it’s still not adding as many, since its platforms are shorter and don’t extend as far to the west.
Nonetheless, given the number of proposals out there for improving Penn Station, including ReThinkNYC and Penn Design’s plan, it’s important to think of longer-term plans for better vertical circulation. When I proposed eliminating Penn Station’s above-ground infrastructure, I came up with a design for six approach tracks (including a new Hudson tunnel connecting to Grand Central), each splitting into two platform tracks facing the same platform; the six platforms would each be 15 meters wide, but unlike Les Halles, each of six access points would have six escalators, four up and two down in the morning peak, or alternatively four escalators and a wide staircase (the climb is 13 meters, equivalent to a five-floor walkup). There would be ample capacity for anything; emptying a full 12-car train would take forty seconds, and it’s unlikely an entire 12-car train would empty.
Suspended railways are not a common mode of transportation. In Europe, the best-known example is the Wuppertal Suspension Railway, opened in 1901. Two examples exist in Japan, which is more willing to experiment with nonstandard rail technology. With essentially just these three examples in normal urban rail usage, it is hard to make generalizations. But I believe that the technology is underrated, and more cities should be considering using it in lieu of more conventional elevated or underground trains.
The reason why suspended trains are better than conventional ones is simple: centrifugal force. Train cars are not perfectly rigid – they have a suspension system, which tolerates some angle between the bogies and the carbody. Under the influence of centrifugal force, the body leans a few degrees to the outside of each curve:
If the train is moving away from you, and is turning left, then the outside of the curve is to your right; this is where the body leans in the image on the right. This is because centrifugal force pushes everything to the right, including in particular the carbody. This increases the centrifugal force felt by the passengers – the opposite of what a tilt system does. A train is said to have soft suspension if this degree of lean is large, and rigid suspension if it is small. The depicted image is rotated 3 degrees, which turns 1 m/s^2 acceleration in the plane of the tracks into 1.5 m/s^2 felt by the passengers; this is the FRA’s current limit, and is close to the maximum value of emergency deceleration. There are no trains with perfectly rigid suspension, but the most recent Shinkansen trains have active suspension, which provides the equivalent of 1-2 degrees of tilt.
On a straddling train, this works in reverse. A straddling train moving away from you turning left will also suspend to the right:
It’s almost identical, except that now the floor of the train leans toward the inside of the curve, rather than to the outside. So the suspension system reduces the lateral acceleration felt by the passengers, rather than increasing it. By softening the suspension system, it’s possible to provide an arbitrarily large degree of tilt, limited only by the maximum track safety value of lateral acceleration, which is not the limiting factor in urban rail.
This is especially useful in urban rail. Longer-distance railroads can superelevate the tracks, especially high-speed tracks, where trains have to be reliable enough for other reasons that they never have to stop in the middle of a superelevated curve. Some urban rail lines have superelevation as well, but not all do. Urban rail lines with high crowding levels routinely stop the trains in the middle of the track to maintain sufficient spacing to the train ahead; this is familiar to my New York readers as “we are being delayed because of train traffic ahead of us,” but the same routinely happens in Paris on the RER. This makes high superelevation dicey: a stopped train leans to the inside of the curve, which is especially uncomfortable for passengers. High superelevation on urban rail is also limited by the twist, i.e. the rate at which the superelevation increases per linear meter (in contrast, on intercity rail, the limiting factor is jerk, expressed in superelevation per second).
Another reason why reducing curve radius is especially useful in urban rail is right-of-way constraints. It’s harder to build a curve of radius 200 meters in a dense city (permitting 60 km/h with light superelevation) than a curve of radius 3 km outside built-up areas (permitting 250 km/h with TGV superelevation and cant deficiency). Urban rail systems make compromises about right-of-way geometry, and even postwar systems have sharp curves by mainline rail standards; in 1969, the Journal of the London Underground Railway Society listed various European limits, including Stockholm at 200 meters. The oldest lines go well below that – Paris has a single 40-meter curve, and New York has several. Anything that permits urban rail to thread between buildings (if above ground), building foundations (if underground), and other lines without sacrificing speed is good; avoiding curves that impose 30 km/h speed limits is important for rapid transit in the long run.
Suspended railways are monorails, so they run elevated. This is not inherent to the technology. Monorails and other unconventional rail technologies can go underground. The reason they don’t is that a major selling point for monorails is that their sleek structures are less visually obtrusive when elevated. But underground they can still use the same technology – if anything, the difficulty of doing emergency evacuation on an elevated suspended monorail is mitigated on an underground line, where passengers can hop to the floor of the tunnel and walk.
I’d normally say something about construction costs. Unfortunately, the technology I am plugging has three lines in regular urban operation, opened in 1901, 1970, and 1988. The 1988 line, the Chiba Monorail, seems to have cost somewhat more per km than other contemporary elevated lines in Japan, but I don’t want to generalize from a single line. Underground there should not be a cost difference. And ultimately, cost may well be lower, since, at the same design speed, suspended monorails can round tighter curves than both conventional railroads and straddle monorails.
Despite its rarity, the technology holds promise in the most constrained urban environments. When they built their next new metro lines, disconnected from the older network, cities like New York, London, Paris, and Tokyo should consider using suspended railroads instead of conventional subways.
I said something on my Patreon page about fare integration between buses and trains, in the context of an article I wrote for the DC Policy Center about improving bus service, and got pushback of the most annoying kind, that is, the kind that requires me to revise my assumptions and think more carefully about the subject. The controversy is over whether fare integration is the correct policy. I still think it is, but there’s a serious drawback, which the positive features have to counterbalance.
First, some background: fare integration means that all modes of public transit charge the same fare within the same zone, or between the same pair of stations. Moreover, it means transfers are free, even between modes. Fare integration between city buses and urban rail seems nearly universal; big exceptions include Washington (the original case study) and London, and to a lesser extent Chicago. Fare integration between urban rail and regional rail is ubiquitous in Europe – London doesn’t quite have it, but it’s actually closer than fare integration between buses and the Underground – but does not exist in North America. In Singapore there is fare integration. In Tokyo, there are about twelve different rail operators, with discounted-but-not-free transfers between two (Tokyo Metro and Toei) and full-fare transfers between any other pair.
The reason North American commuter rail has no fare integration with other forms of transit is pure tradition: railroaders think of themselves as special, standing apart from mere urban transit. We can dispense with the idea that it is a seriously thought-out fare system. However, lack of integration between buses and trains in general does have some thought behind it. In London, the stated reason is that the Underground is at capacity, so its fares are jacked up to avoid overcrowding, while the buses remain cheap. In Washington, it’s that Metro is a better product than the buses, so it should cost more, in the same way first-class seats cost more than second-class seats on trains. Cap’n Transit made a similar point about this in the context of express buses.
There are really three different questions about fare integration: demand, supply, and network effects. The first one, as noted by Patreon supporters, favors disintegrated fares. The other two favor fare integration, for different reasons.
Demand just means charging more for a product that has higher demand. This is about revenue maximization, assuming fixed service provision: people will pay more for the higher speed of rapid transit, so it’s better to charge each mode of transportation the maximum it can bear before people stop taking trips altogether, or choose to drive instead. It’s related to yield management, which maximizes revenue by using a fare bucket system, using time of booking as a form of price discrimination; SNCF uses it on the TGV, and in its writeups for American high-speed rail from 2009, it said it boosted revenue by 4%. In either case, you extract from each passenger the maximum they can pay by making features like “don’t get stuck in traffic” cost extra.
Supply means giving riders incentives to ride the mode of transportation that’s cheaper to provide. In other words, here we don’t assume fixed (or relatively fixed) service provision, but variable service provision and relatively fixed ridership. Trains nearly universally have lower marginal operating costs than buses per passenger-km; in Washington the buses cost 40% more per vehicle-km, and perhaps 2.5 times as much per unit of capacity (Washington Metro cars are long). Using the fare system to incentivize passengers to take the train rather than the bus allows the transit agency to shift resources away from expensive buses, or perhaps to redeploy these resources to serve more areas. If anything, the bus should cost more. There are shades of this line in incentives some transit agencies give for passengers to switch from older fare media to smart cards: the smart card is more convenient and thus in higher demand, but it also involves lower transaction costs, and thus the agency incentivizes its use by charging less.
The network effect means avoiding segmenting the market in any way, to let passengers use all available options. The fastest way to get between two points may be a bus in some cases and a train in others, or a combined trip. This fastest way is often also the most direct, which both minimizes provision cost to the agency and maximizes passenger utility. This point argues in favor of free transfers especially, more so than fare integration. Tokyo fares are integrated in the sense that the different railroads charge approximately the same for the same distance; but transfers are not free, and monthly passes are station to station, with no flexibility for passengers who live between two parallel (usually competing) lines.
The dominant reason to offer integrated fares is network effects, more so than supply. Evidently, I am not aware of transit agencies that charge more for buses than for trains, only in the other direction. That fare integration allows transit agencies to reduce operating costs mitigates the loss of revenue coming from ending price discrimination; it is not the primary reason to integrate fares.
The issue at hand is partly frequency, and partly granularity. A typical transit corridor, supporting a reasonably frequent bus or a medium-size subway station, doesn’t really have the travel demand for multiple competing lines, even if it’s a parallel bus and a rail line. Fare disintegration ends up reducing the frequency on each option, sometimes beyond the point where it starts hurting ridership.
In Washington it’s especially bad, because of reverse-branching. The street network makes it hard for the same bus to serve multiple downtown destinations (or offer transfers to other buses for downtown service). Normally, riders would be able to just take a bus to the subway station and get to their destination, but Washington plans buses and trains separately, so two of the trunk routes, running on 14th an 16th Streets, reverse-branch. The hit to frequency (16-18 minutes per destination off-peak) is so great that even without fare integration it’s worthwhile to prune the branches. But such situations are not unique to Washington, and can occur anywhere.
The required ingredients are a city center that is large enough, or oriented around a long axis, with a street network that isn’t a strict grid and isn’t oriented around the axis of city center. New York is such a city: if it didn’t have fare integration, buses would need to reverse-branch from the north to serve the East Side and West Side, and from anywhere to serve Midtown and Lower Manhattan.
The granularity issue is that there isn’t actually a large menu of options for riders with different abilities to pay. This is especially a problem in American suburbs, with nothing between commuter rail (expensive, infrequent off- and reverse-peak, assumes car ownership) and the bus (in the suburbs, a last-ditch option for people below the poverty line). I wrote about this for Streetsblog in the context of Long Island; there’s also a supply angle – different classes of riders travel in opposite directions, so it’s more efficient to put them on one vehicle going back and forth – but this is fundamentally a problem of excessive market segmentation.
This also explains how Tokyo manages without fare integration between different rail operators. Its commuter rail lines are not the typical transit corridor. With more than a million riders per day (not weekday) on many lines, there is enough demand for very high frequency even with disintegrated fares. A passenger between two competing lines can only get a monthly pass on one, but it’s fine because the one line is frequent and the trains run on time.
The rest of the world is not Tokyo. Branches in Outer London and the Paris suburbs aren’t terribly frequent, and only hit one of the city centers, necessitating free transfers to distribute passengers throughout the city. They also need to collect all possible traffic, without breaking demand between different modes. If RER fares were higher than Metro fares, some areas would need to have a Metro line (or bus line) paralleling the RER, just to collect low-income riders, and the frequency on either line would be weaker.
The demand issue is still real. Fare integration is a service, and it costs money, in terms of lost revenue. But it’s a service with real value for passengers, independently of the fact that it also reduces operating costs. The 99.5% of the world that does not live in Tokyo needs this for flexible, frequent transit choices.
Generally, the best guide to where a city should build rail lines is where the busiest buses are. However, there are exceptions. I have written two posts about this giving examples of exceptions, and am going to give a third exception; I also intend to write a separate post soon giving a fourth exception.
The first post, from four years ago, deals with cases where the bus alignment has to stay on a major street, but some major destinations are just away from the street; a subway can deviate to serve those destinations. Examples include Old Jaffa in Tel Aviv near the north-south spine of bus lines 1 and 25, and Century City near the Wilshire corridor. Here, buses are a good guide to corridor demand, but the rail line should serve microdestinations just outside the corridor.
The second post, from last year, is more properly about corridors. It describes street networks that are hostile to surface transit, by featuring narrow, meandering streets. The main example is Boston, especially the Green Line Extension, in a rail right-of-way in a city infamous for its labyrinthine streets. Another example is the Evergreen extension in Vancouver, serving Coquitlam; the bus the extension replaced, the 97-B, meandered through Coquitlam since the streets were so poorly configured, while the extension uses a short tunnel and runs parallel to a railroad.
In this post I’d like to expand on a point I made, obliquely, in the Voice of San Diego. In San Diego, there’s an under-construction light rail extension, in a rail right-of-way, into an area with not-great bus ridership. Consult the following map:
Preexisting light rail (“Trolley”) is in black, the extension (of the Blue Line) in blue, the parallel north-south arterial in purple, and two buses in green and red. The bus ridership on Ingraham is very low: the bus route running on it, 9, has 1,500 riders per weekday (source). The top bus in San Diego, the 7 (going north of downtown, then east), has 11,000. So on the surface, this suggests there isn’t much demand for north-south transit in that area of the city, called Pacific Beach.
But that’s wrong, because in an auto-oriented city like any US city except New York, the major streets are determined by car access. The relentless grids of so many North American cities – Chicago, Los Angeles, Toronto, Vancouver – are not just where the buses go, but also where the cars go. Even in Manhattan, if you have the misfortune to find yourself going east-west in a car, you will probably use one of the major two-way streets, like 14th or 42nd, which are less clogged than the one-way streets in between. Non-gridded street networks for the most part obey this rule too – the commercial streets tend to be the wider ones used by car through-traffic.
Freeways throw a wrench into this system. They offer a convenient route for cars, but are abominable for commerce. Locations 5 minutes by car from the freeway are good; locations right along the freeway are not, unlike ones right along an arterial road. The main car route from Pacific Beach to the CBD is taking an east-west arterial to the I-5, not going south on Ingraham. This means that the demand for north-south traffic actually shows as strong commerce on east-west streets, hosting bus routes 27 and 30, and not on Ingraham. The 27 has weak ridership, and the 30 has strong ridership but not right along the I-5. But in a sense it doesn’t really matter, because, like the car- and bus-hostile narrow streets of old city centers, the freeway-centric road network in that part of San Diego suppresses bus ridership relative to future rail ridership.
In the presence of rail, the strong routes are the ones orthogonal to the rail line. Here, the 27 and 30 already preexist; there is a planned Trolley stop at the intersection with the 27, and presumably the 30 will be rerouted to serve that intersection rather than to duplicate the trains along the freeway. (I tried talking to the transit agency about this, but didn’t get any useful answers.) So the decent east-west bus ridership in Pacific Beach is actually an argument in favor of a north-south rail extension.
Like every exception to a general rule, this is not a common scenario. So where else are there cases where this special case holds? The necessary elements are,
- The city must be auto-oriented enough that car access is crucial to nearly all commercial drags. In Paris, it doesn’t matter how you reach the Peripherique by car, because car ownership is so low.
- The city should not have a strong mainline rail network, which leads to a hierarchical transit network (buses feeding train stations), in which both buses and cars use the same major streets to reach train stations. This means that Sydney and Melbourne are out, as are German cities short of Berlin and Munich’s transit mode shares.
- The city must have a strong network of urban freeways, disrupting the street network to the point of siphoning traffic away from the surface streets that would otherwise be the main routes.
As it happens, all three elements are present in Tel Aviv. North-south travel within the region uses Ayalon Freeway, inconveniently east of the traditional city center; the city has been building a CBD closer to the freeway, but it’s still not quite there. This suggests that traffic is suppressed on the north-south arterials to the west – Ibn Gabirol (hosting the planned second line of the subway) and Dizengoff (possibly hosting the third) – is suppressed, and those streets require subways. This is in part why, before the Red Line began construction, I argued in favor of putting a north-south subway under Ibn Gabirol, and not under freeway-adjacent Namir Road, where the Red Line goes.
In the future, this pattern suggests that Tel Aviv should make sure to build north-south subways under Ibn Gabirol and Dizengoff, and extend them north. The significance of the northern direction is that the effect I’m describing in this post only works when car ownership is high; Israel is poor enough that car ownership is not universal, and in the poorer southern suburbs it is low enough that the buses do give a good guide to corridor demand, whereas in the northern suburbs everyone owns a car. There is likely to be suppressed transit demand in Herzliya, Ramat HaSharon, and northeastern Tel Aviv (including Ramat HaHayal, an edge city with many tech jobs). Thus ridership on a subway line going elevated over Sokolov in Ramat HaSharon and Herzliya, or on Raoul Wallenberg to Ramat HaHayal, is likely to be higher than present-day bus ridership suggests.
An American example is Washington’s suburbs. The Metro extensions are planned with little regard for bus ridership. While the Silver Line is bad for multiple reasons – high construction costs, service to too far exurbs, too much branching on an overloaded trunk – the extension to Tysons Corner is its one good aspect. There is no point in discussing bus ridership at an edge city like Tysons – conventional buses wouldn’t be following the same route that the cars follow, and freeway express buses almost universally have trivial ridership.
Finally, Vancouver. While Vancouver itself is gridded, its suburbs are much less so. In the suburbs served by the Trans-Canada Highway, especially Surrey, it’s likely that car traffic mostly follows roads feeding the highway. People drive to their jobs in Downtown, Central Broadway, Metrotown, or any of Surrey’s internal centers; there aren’t a lot of park-and-rides at SkyTrain stations, which instead emphasize transit-oriented development, and in Surrey there are actually more park-and-ride spaces at the freeways, with express bus access, than at the one SkyTrain stop with parking, Scott Road. This suggests that there is suppressed bus ridership in Surrey and Langley parallel to the Trans-Canada, along Fraser Highway. Extending SkyTrain in that direction is on a distant priority list for the region, and this theory suggests that it should be moved up, to be just behind the Broadway subway to UBC.
I wrote about infill stops on commuter rail two weeks ago, and said I cannot think of any example of anti-infill on that mode. But looking at Muni Metro reminded me that there is need for anti-infill on surface transit. This is called stop consolidation normally, and I only use the term anti-infill to contrast with the strategy of adding more stops on commuter trains.
The root of the problem is that in North America, transit agencies have standardized on 200-250 meters as the typical spacing between bus stops. In Europe, Australasia, and East Asia, the standard is instead 400-500 meters. Even without off-board fare collection, the difference in speed is noticeable. In Vancouver, the difference between the local 4 and the express 84 is substantial: on the shared segment between Burrard and Tolmie, a distance of 4.8 km, the 84 makes 5 stops and takes 10 minutes, the 4 makes 18 stops and takes 16 minutes. A bus with the normal first-world stop spacing would make 10-12 stops and take, linearly, 12-13 minutes. 23 km/h versus 18 km/h.
With off-board fare collection, the impact of stop spacing on speed grows. The reason is that a bus’s stop penalty consists of the time taken to stop and open its doors, plus the time it takes each passenger to board. The former time is independent of the fare collection method but depends on stop spacing. The latter time is the exact opposite: if the stop spacing widens, then there are more passengers per bus stop, and unless the change in stop spacing triggers changes in ridership, overall passenger boarding and alighting time remains the same. Another way to think about it is that judging by Vancouver data, there appears to be a 30-second stop penalty, independent of ridership. Off-board fare collection increases bus speed, so the 30-second stop penalty becomes more important relative to overall travel time; the same is true of other treatments that increase bus speed, such as dedicated lanes and signal priority.
In New York, there aren’t a lot of places with local and limited-stop buses side by side in which the limited-stop bus has on-board fare collection. One such example is the M4, meandering from Washington Heights down the 5th/Madison one-way-pair, over 15.3 km. At rush hour, the local takes 1:45, the limited-stop takes 1:30: 9 vs. 10 km/h. But the limited-stop bus runs local for 6 km, and over the other 9.3 km it skips 26 local stops if I’ve counted right. The B41 has a limited-stop version over 8.3 km (the rest is local), skipping about 17 stops; the time difference is 10 minutes.
One possible explanation for why the stop penalty in New York seems a little higher than in Vancouver is that the M4 and B41 routes are busier than the 4/84 in Vancouver, so every stop has at least one passenger, whereas the 4 in Vancouver often skips a few stops if there are no passengers waiting. Conversely, the higher passenger traffic on buses in New York comes from higher density and more traffic in general, which slows down the buses independently of stopping distance.
On subways, there’s reason to have more densely-spaced stops in denser areas, chief of which is the CBD. On surface transit, it’s less relevant. The reason is that absolute density doesn’t matter for stop spacing, except when expected ridership at once station is so high it would stress the egress points. What really matters is relative density. Putting more stops in an area means slowing down everyone riding through it in order to offer shorter station access times to people within it. On surface transit, relative density gradients aren’t likely to lead to variations in stop spacing, for the following reasons:
- Historically, surface transit stop spacing was always shorter than rapid transit stop spacing because of its lower top speed and the faster braking capabilities of horses vs. steam trains; often people could get off at any street corner they chose. So it induced linear development, of roughly constant density along the corridor, rather than clusters of high density near stations.
- If there is considerable variation in density along a surface transit line, then either density is medium with a few pockets of high density, which would probably make the line a good candidate for a subway, or density is low with a few pockets of higher density, and the bus would probably skip a lot of the low-density stops anyway.
Most importantly, the 400-meter standard is almost Pareto-faster than the 200-meter standard. In the worst case, it adds about 4 minutes of combined walking time at both the start and the end of the trip, for an able-bodied, healthy person not carrying obscene amounts of luggage. The breakeven time on 4 minutes is 8 skipped stops, so 3.2 km compared with the 200-meter standard. Bus trips tend to be longer than this, except in a few edge cases. In New York the average unlinked bus trip is 3.4 km (compare boardings and passenger-km on the NTD), but many trips involve a transfer to another bus or the subway, probably half judging by fare revenue, and transfer stations would never be deleted. If the destination is a subway station, guaranteed to have a stop, then the breakeven distance is 1.6 km.
This also suggests that different routes may have different stop spacing. Very short routes should have shorter stop spacing, for example the 5 and 6 buses in Vancouver. Those routes compete with walking anyway. This may create a spurious relationship with density: the 5 and 6 buses serve the very dense West End, but the real reason to keep stop spacing on them short is that they are short routes, about 2 km each. Of course, West End density over a longer stretch would justify a subway, so in a way there’s a reason short optimal stop spacing correlates with high bus stop density.
The situation on subways is murkier. The stop penalty is slightly higher, maybe 45 seconds away from CBD stations with long dwell times. But the range of stop distances is such that more people lose out from having fewer stops. Paris has a Metro stop every 600 meters, give or take. Some of the busiest systems in countries that were never communist, such as Tokyo, Mexico City, and London, average 1.2 km; in former communist bloc countries, including Russia and China, the average is higher, 1.7 km in Moscow. The difference between 600 meters and 1.2 km is, in the worst case, another 1.2 km of walking, about 12 minutes; breakeven is 16 deleted stops, or 20 km, on the long side for subway commutes.
One mitigating factor is that subway-oriented development clusters more, so the worst case is less likely to be realized, especially since stops are usually closer together in the CBD. But on the other hand, at 1.2 km between stations it’s easy for transfers to be awkward or for lines to cross without a transfer. London and Tokyo both have many locations where this happens, if not so many as New York; Mexico City doesn’t (it’s the biggest subway network in which every pair of intersecting lines has a transfer), but it has a less dense network in its center. Paris only has three such intersections, two of them involving the express Metro Line 14. Even when transfers do exist, they may be awkward in ways they wouldn’t have been if stop spacing had been closer (then again, Paris is notorious for long transfers at Chatelet and Montparnasse).
In all discussions of subway stop spacing, New York is sui generis since the lines have four tracks. On paper its subway lines stop every 600-700 meters when not crossing water, but many trains run express and stop every 2 km or even more. Average speed is almost the same as in Tokyo and London, which have very little express service, and it used to be on a par until recent subway slowdowns. This distinction, between longer stop spacing and shorter stop spacing with express runs, also ports to buses. Buses outside the US and Canada stop every 400-500 meters and have no need for limited-stop runs – they really split the difference between local and limited buses in North America.
On a subway, the main advantage of the international system over the New York system is obvious: only two tracks are required rather than four, reducing construction costs. On a bus line, the advantages are really the same, provided the city gives the buses enough space. A physically separated bus lane cannot easily accommodate buses of different speeds. In New York, this is the excuse I’ve heard in comments for why the bus lanes are only painted, not physically separated as in Paris. Mixing buses of different speeds also makes it hard to give buses signal priority: it is easy for buses to conflict, since the same intersection might see two buses spaced a minute apart.
Buses also benefit from having a single speed class because of the importance of frequency. In Vancouver, the off-peak weekday frequency on 4th Avenue is an 84 rapid bus every 12 minutes, a 44 rapid bus every 20 minutes, and a local 4 every 15 minutes. The 84 keeps going on 4th Avenue whereas the 4 and 44 divert to Downtown, but the 4 and 44 could still be consolidated into a bus coming every 10 minutes. If there were enough savings to boost the 84 to 10 minutes the three routes could vaguely be scheduled to come every 5 minutes on the common section, but without dedicated lanes it’s probably impossible to run a scheduled service at that frequency (pure headway management and branching don’t mix).
The example of 4th Avenue gets back to my original impetus for this post, Muni Metro. Only diesel buses can really run in regular surface mode mixing different speed classes. Trolleys can’t. Vancouver runs trolleys on the local routes and diesels on the limited routes. At UBC, it has different bus loops for diesels and trolleys, so people leaving campus have to choose which type of bus to take – they can’t stand at one stop and take whatever comes first.
On rail, this is of course completely impossible. As a result, American subway-surface trolleys – the Boston Green Line, SEPTA’s Subway-Surface Lines, and Muni Metro – all run at glacial speed on the surface, even when they have dedicated lanes as in Boston. In Boston there has been some effort toward stop consolidation on the Green Line’s busiest branch, the B, serving Boston University. This is bundled with accessibility – it costs money to make a trolley stop wheelchair-accessible and it’s cheaper to have fewer stops. Muni Metro instead makes one stop every 3-5 accessible (on paper), but keeps stopping at all the other stops. It would be better to just prune the surface stops down to one every 400-500 meters, which should be accessible.
If you view rail as inherently better than bus, which I do, then it fits into the general framework: anti-infill on surface transit has the highest impact on the routes with the best service quality. Higher speed makes the speed gain of stop consolidation more important relative to travel time; trolleywire makes it impossible to compensate for the low speed of routes with 200-meter interstations by running limited-stop service. Even on local buses, there is never a reason for such short stop spacing, and it’s important for North American cities to adopt best industry practice on this issue. But it’s the most important on the highest-end routes, where the gains are especially large.
Modernizing commuter rail to run it like rapid transit means a lot of things. It means high all-day frequency, fare integration, good transfers to local transit (which requires fare integration), and ideally through-running in order to hit multiple business districts. In North America, these are absent, resulting in low ridership. So let’s posit that these problems are already being solved: commuter rail is being run frequently, the lines are electrified, the fares are the same as on local transit with free transfers. What should the stop pattern look like? This is not a purely hypothetical discussion, because in Toronto the under-construction RER system includes high off-peak frequency, electrification, and through-running, with fare integration under consideration.
So let’s imagine a city with modernized regional rail, maybe in the early 2020s. Trains run every 15 minutes off-peak, charge the same fare as local transit, and run fast EMUs, on track that’s good for 130 km/h outside station throats. The stop pattern, expressed in kilometers out of city center, is any of the following:
|Line||Stop #1||Stop #2||Stop #3||Stop #4||Stop #5||Stop #6|
|Chicago Metra Electric||1.3||2.3||3.6||4.3||5.2||9.5|
|GO Transit Lakeshore West||3.2||10.8||15.4||20.6||26.9||34.4|
|GO Transit Lakeshore East||8.4||13.8||17.1||20.3||26.6||33.6|
|SEPTA Paoli/Thorndale Line (PRR Main Line)||8.7||9.7||10.9||11.9||13.7||13.7|
|Metrolink Antelope Valley Line||9.3||17.5||24.9||35.4||48.3||55.1|
|Metrolink Orange County Line||14.4||27.1||35.2||41.3||50.2||53.5|
(Metrolink data comes from measuring on Google Earth, the rest comes from Wikipedia.)
Two patterns emerge:
- Metrolink and GO Transit have very wide stop spacing. GO has straight track it owns outright on the Lakeshore lines, and Metrolink is straight with few curves on the Orange County Line (but freight owns much of the route) and straight until stop #4 on the Antelope Valley Line, Sylmar. EMUs could average 90 km/h on these lines, counting schedule padding.
- SEPTA and Metra have wide stop spacing in the core but very narrow spacing in the suburbs; I discussed this issue for Metra in an old post comparing its stop distribution through stop #12 with that of the RER. Metra Electric has a few inner stops, e.g. for the convention center (stop #4), but then there’s a 4-km gap from stop #5 to stop #6 (Kenwood).
Both patterns are compatible with modernized rail operations. But both have problems with dealing with passenger demand. Consider what happens to a transit user in Burbank (stop #2 on Antelope Valley), or Ravenswood on the North Side of Chicago (stop #2 on UP-North), or Danforth and Main (stop #1 on Lakeshore East). Such a passenger would get an incredibly fast train to the CBD, running either nonstop or with one stop. Demand would boom. But is this really the most efficient way of running transit? Not really. This is for the following reasons:
- The downstream locations would still be very attractive infill train stations, with potential for high ridership. After all, they’d get fast service, too.
- If you get a 1-stop, 10-minute train ride to work, then adding four more stops to turn it into a 15-minute ride sounds like an imposition (it raises trip times by 50%), but it really isn’t, because most likely your actual commute is 30 minutes, with 10 minutes of walking at each end.
- With fare integration, buses should really be feeding the major train stations. Making every bus feed a small number of stations with fast commuter rail service compromises the rest of the network, whereas siting train stations at the intersections with the major grid buses in cities like Toronto, Chicago, and Los Angeles facilitates transfers better.
As a result of these principles, my proposal for Electrolink in Los Angeles, built out of the Antelope Valley and Orange County Line, involves extensive infill stops. See map below:
Of note, there is much more infill on the Antelope Valley and Ventura County Lines than on the Orange County Line. This is because there is less residential density near the Orange County Line, and much more industrial land use. Los Angeles has a strong manufacturing sector, using the railroad for freight access, so residential upzoning near potential infill station locations is speculative. In the San Fernando Valley the land use near the railroads is not great, but there is a decent amount of residential density beyond the near-railroad industrial uses, there are strong bus corridors intersecting the railroads (and potential for light rail); the corridors also have less freight, so it’s easier to kick out industrial uses from station sites and do residential and commercial upzoning.
In New York and Boston, there are other caveats, explaining why my various regional rail proposals for these cities call only for mild infill. The biggest caveat is that there exist parallel subway lines. Boston’s Old Colony trunk passes through relatively dense areas in Dorchester and Quincy, with just three stops: JFK/UMass at km-point 3.7, Quincy at km-point 12.7, Braintree at km-point 17.6. But the Red Line runs parallel to the trunk making more stops, enabling commuter rail to work as an express overlay. Thus, only the busiest locations deserve a commuter rail stop, and those are precisely the three existing stops. In Somerville, the Green Line Extension plays a similar role, providing local service that commuter rail would otherwise have to provide under any modernization scheme. As a result, my proposal for how to run the Old Colony Lines and the Lowell Line through Somerville is more intercity rail than local commuter rail.
In contrast, the Worcester Line has no parallel subway except on the innermost few km, so it’s already getting two infill stops (Boston Landing and West Station) and I’ve called for several more. The same is true of the Fairmount Line, which is expanding from five stops including the endpoints to nine.
There is an analog for this in Paris, on the RER. Within the city and La Defense, the RER A mostly runs as an express overlay for Metro Line 1, stopping at the major stations, omitting just Bastille, which is too close to Gare de Lyon. But the RER B really has two separate stop regimes. North of Chatelet, parallel to Line 4, it expresses to Gare du Nord, and then doesn’t stop again until reaching the suburbs. But south of Chatelet, all trains make 5 stops in 5 km to Cite-Universitaire, even ones that run express in the suburbs; this is the older part of the line, much of which predates the Metro, so Line 4 was built along a different alignment via Gare Montparnasse.
In New York, the commuter trunks going north and east are closely parallel to the subway, which is often four-tracked. Since the subway already provides relatively fast service, with stops every 2-3 km, commuter rail should be even faster, with sparser stops. The principles here are,
- Infill stops are more justified at intersections with major orthogonal bus or rail corridors or in dense, transit-deprived areas. Areas with little residential development are to be categorized based on redevelopment potential.
- Infill stops are less justified parallel to a subway line. The faster the subway is, the faster commuter rail should be.
- Infill stops are more justified on a shorter line than on a longer line. Here, “shorter” and “longer” do not mean the length of the line to the endpoint, but the length to the endpoint of local service, if the infill stops would not be served by express trains.
Metro-North is parallel to the four-track Lexington Avenue Line, which has four tracks. Between 125th Street and Grand Central, the 4 and 5 trains make just two intermediate stops, at 86th and 59th Streets. Metro-North runs nonstop between 125th and Grand Central, and because the 4 and 5 exist, it has no reason to make more stops, even at 59th for additional service to Midtown
In the Bronx the trunk line isn’t so close to the subway, but already makes multiple stops. There may be plausible infill in Morrisania between Melrose and Tremont. But even that is marginal – for one, Melrose, Tremont, and Fordham are all located at the intersections with high-ridership east-west bus routes, whereas nothing in between is. This distinction between an inner and outer part of the same line also holds for the Atlantic Branch: west of Broadway Junction it parallels the four-track A/C, so no infill is needed, but farther east it parallels the slower J/Z and isn’t even that close to the subway, so infill is useful.
Going east, the LIRR Main Line is parallel to the Queens Boulevard Line, like the Lex a four-track line. There is no real point in infill, except at Sunnyside Junction, where the line meets the Northeast Corridor. The Port Washington Branch is a shorter line than anything on the Main Line, even the Hempstead Branch (but not by much), and isn’t as close to the 7 as the Main Line is to the Queens Boulevard Line; the 7 is also slower. This means an infill stop or two may be justified – my map has three (at Queens Boulevard, Broadway, and Junction), but that may be too much.
It’s the west direction that is the most speculative, toward New Jersey. I have called for new Hudson tunnels to feature a station at Bergenline (building a station in the existing tunnels would disrupt current service and slow down express trains) based on the above principles. Additional infill is possible, but only subject to transit-oriented development plans alongside the line. The land use from just west of Bergen Hill, including Bergenline, to just east of Newark, is a combination of industrial warehouses and wetland preserves. The warehouses should be redeveloped, but until there is rezoning, it is pointless to add more stops. Moreover, Secaucus Junction is already in the middle of the warehouse area, so rezoning should start from there, and if the newly-built residential neighborhood grows big enough so as to justify a second station, a second station can be added later.
The upshot is that even though New York has very wide stop spacing on commuter rail near the core, it does not need as much infill as Los Angeles or Chicago. What about Toronto, the original impetus for the post? There, Metrolinx is already considering minor infill. But if the principles emerging from how I think about infill in the US and on the RER are correct, Toronto needs far more infill. The Lakeshore lines are not really close to the subway: they run east-west, as does the Bloor-Danforth Line, but Lakeshore West is about 2 km from Bloor, and Lakeshore East is mostly 1 km from Danforth, with just a short segment within walking distance. The inner areas of Lakeshore West are very dense, with some blocks at 30,000 people per km^2, and only served by buses and slow mixed-traffic streetcars; even some areas along Lakeshore East are fairly dense, more than 10,000/km^2. Toronto’s bus and streetcar grid hits or can be extended to hit several potential station locations, offering better connections than riding to the Bloor-Danforth Line and then changing to Yonge to reach the CBD.
The one drawback in Toronto: the commuter lines are very long. Not all trains have to make all stops, but if there’s one stopping pattern making 3 stops in 15 km and another making 12, then it isn’t really possible to mix them on the same tracks at high frequency. The core lines have four tracks, but Lakeshore needs to eventually mix four classes of trains: local commuter rail, longer-range commuter rail, intercity rail, freight. There are ways around four-way mixture (for example, there is little freight on Lakeshore in Toronto proper, where the local trains would run), and intercity trains can probably share tracks with long-range commuter trains. It’s solvable, just like the three-way track-sharing between local, express, and eventual high-speed trains around New York, but it isn’t trivial.
In general, North American commuter trains make too few stops in the urban core. Tellingly, while I can come up with many examples of lines that require infill, I can’t name five good examples of anti-infill, where a station served by commuter trains full-time should be closed. But not all commuter lines are equally good candidates for infill stops, and there are large networks, such as Metro-North, where the current stop spacing is fine, just as there are ones, such as GO Transit and some Metra lines, where some inner segments could plausible see the number of stops quadruple.
This post is inspired by two separate things. The first is my work on a fantasy subway map for Lagos; here is the current live version. There are twelve radial lines, all serving the western half of Lagos Island, converging on nine transfer stations. Under the principle that whenever two lines intersect there should be a transfer station, this greatly constrains the paths the lines can take. Result: the path between two CBD stations, Eko Bridge and Leventis, carries ten tracks underneath it. This is under a wide street, and it might be possible with a double-deck four-track-wide tunnel and two more tracks deep-boring around it, but it’s not easy to construct.
The second inspiration is a post by Brian Stokle about subway line spacing. Brian looks not at spacing between successive stops on one line, but at spacing between parallel lines, averaging a few North American examples. The average is a little higher than half a kilometer, narrower than the typical stop spacing. On Twitter, Joshua Mello notes Boston’s spacing was narrower; in comments, I add examples from New York and Paris, which are a bit narrower than Brian’s examples but wider than Boston (New York is one block in Midtown, so 280 meters, and Paris is 300 between Metro 3 and Metro 8 and 9).
These two examples together illustrate the tradeoff in subway construction. Most subways have a stop every 1-1.5 km; newer systems are at the high end of this range, mostly because of the demographic weight of China. It’s normal for stop spacing to tighten in the core, but not to a large extent. In Tokyo, the average stop spacing is 1.2 km, and in Central Tokyo it’s perhaps 800 meters. In London, the Tube lines seem to tighten from an average of 1.2-1.5 km to 600-800 meters in Central London.
At the same time, subway line spacing is necessarily short. The reason is that modern CBDs are geographically small. Midtown is maybe 4 km^2, from 30th to 60th Streets and from between 2nd and 3rd Avenues to between 8th and 9th. The Paris CBD, from just west of Les Halles to just east of Etoile, is also about 4 km^2 (see job density on PDF-p. 6 here). The Tokyo CBD, defined around Otemachi, Nihonbashi, Hibiya, Shimbashi, and increasingly Roppongi, is maybe 5-6 km^2, in a metro area of 38 million people.
Subway networks in such CBDs are necessarily crowded. The CBD is where people want to go. A subway line can get away with skirting it – Paris M4 does, and is in a near-tie with M1 for highest ridership per km. But avoiding it entirely is a ridership killer, except specifically for circumferential lines concentrating off-CBD travel: in Paris this suppresses M10 ridership, and in New York, it suppresses ridership on the J/Z (even though they serve Lower Manhattan) and the L (even though it serves Union Square). This means that the CBD of a large city will have many subway lines converging on a relatively small area. New York has its five north-south lines through Midtown.
Ensuring that every pair of intersecting subway lines has a transfer in this environment is difficult. Line spacing is usually narrower than station spacing, requiring kludges like the block-long walkways in New York, such as between Times Square at 42nd/7th and Port Authority at 42nd/8th. Paris managed to have an almost perfect network – before M14 was built, it only had one missed connection, between M9 and M12 (built by a competing private company) – but only by having very short station spacing, unusual even by the standards of the early 1900s, ruling out significant suburban extensions of the kind that are routine in London and Tokyo.
The situation in smaller cities is actually easier. The CBD is very small, often smaller than a square kilometer, but there are fewer lines, so it’s easier to make sure lines intersect properly. It’s also much easier to get line spacing right outside the CBD, where there’s less intense demand, allowing line spacing compatible with stop spacing on any intersecting or circumferential line.
The fundamental issue here is really about planning for the future. It’s not hard to gets lines 1, 2, and 3 to intersect nicely, or even lines 1-6. But beyond that, a city will often find itself in a situation where the best street alignment for line 7 happens to be right between two stations on line 1, spaced too far apart for a transfer. This is what happened to Tokyo. In New York, the three constituent systems (IRT, BMT, and IND) were each internally planned cohesively, so when two lines within the same system intersect, there’s a transfer, and, with difficulty, the IRT/BMT intersections have transfers as well. But the IND connects poorly to the other two systems, sometimes deliberately, and the IND’s layout made future extensions and service changes break transfers. My proposal to reduce reverse-branching in New York runs into the problem of breaking the transfers designed by the IND around a specific service plan.
When lines are designed together, it’s easier to avoid this problem. Paris M8 and M9 share a route through the center, as they were built simultaneously as the street is wide enough for four tracks. In contrast, building a line under or next to an existing line is much more difficult; New York did it anyway, under Sixth Avenue, but this led to cost overruns that doomed the IND’s early plans for further expansion. It is also difficult to build a new station under an existing transfer station, as it usually requires underpinning; in Paris, this problem means that transfer stations tend not to have closely-aligned platforms, requiring long walks between lines. When I’m proposing running multiple lines in the same tunnel in Lagos, this is from the point of view of assuming coordinated planning, with sequencing that allows entire streets to be dug up at once.
However, in reality, even coordinated design has its limitations. Subway networks take multiple decades to build, and in the interim, the city changes. Planners can attempt to use zoning to shape city development in a way that facilitates further expansion, but some tendencies are too uncontrollable. For instance, high-income neighborhoods tend to commercialize; I mentioned Roppongi as a growing part of the Tokyo CBD earlier, which is an example of this trend. The hottest new part of New York commercial development, the Meatpacking District, is really not a subway hub. This means that even if a city plans out lines 1-12 to share tunnels appropriately, it may not be able to control where there will arise the most demand for line 13. Coordinated long-term planning makes things easier, but it will not solve the basic problem of optimal subway spacing and CBD size.
Last month, I committed to producing a subway fantasy map for Lagos via a Twitter poll. I’m working on this, but before I go into Lagos itself, I want to talk about the third world in general. Good transportation in poor countries is of independent interest, but it also has some applications to thinking about solutions for rich countries, such as the countries my readers live in. The reason is that every principle of good transportation planning has edge cases, exceptions, and assumptions, and it is critical to evaluate these in the largest variety of situations. Understanding transportation in the United States can yield insights about Europe and vice versa; likewise, understanding the first world can yield insights about the third and vice versa.
The epistemological principle I use is that if I believe that a high concentration of factor A makes solution B work better, then a low concentration of factor A should make solution B work worse. I used that in a post about high-speed rail in Sweden, arguing against it due to the absence of factors that make it work better, namely, linear population distribution. Many good design principles formulated in rich countries depend on those countries’ high incomes, and are less relevant to countries that are only about as wealthy as the US and northwestern Europe were in 1900.
Everything is terrible
On nearly every indicator of technology or living standards, every poor country is worse than every rich country. There are some exceptions involving middle-income countries (for example, Russia and China have very good rail freight), but not in low-income countries. I wrote a piece in YIMBY recently describing the state of New York and Vienna in the early 20th century, which had very high crowding levels; much of the same story can describe many third-world cities today, especially in India, where tight zoning limits housing supply to the point of overcrowding. In Mumbai, the average residential floor space per person is 9 square meters, compared with 55 in Manhattan.
Pollution levels are very high as well, because of the combination of high population density and heavy industry (especially in North India), as well as the proliferation of cars. The amount of pollution caused by 50 or 100 cars per 1,000 people in a dense city where the cars don’t have catalytic converters can be many times worse than that caused by the 200 mostly diesel-powered cars per 1,000 people of Paris, or the 250 cars per 1,000 people of New York. The low motorization levels of lower-middle-income cities like Cairo, Lagos, or Mumbai aren’t a barrier to traffic, either: those cities routinely have traffic jams, just as the United States started having jams in the 1920s. These cities have centralized employment in the CBD, not a lot of road capacity coming in, and a culture in which the middle class drives (or is driven by chauffeurs).
This creates an urgency for improving public transportation in low-income countries that does not exist in the developed world. Third-world countries that build subways spend a much higher share of their GDPs on them than Europe and Japan do, and some, such as India and Bangladesh, spend more than the United States. If Paris hadn’t built the RER, Franciliens would drive or take the slower Metro; if Shanghai hadn’t built the Metro, Shanghainese would still be living in tiny apartments and riding buses in crawling traffic; if Lagos doesn’t build a metro, Lagosians will keep facing multi-hour commutes. The same situation also creates an urgency for improving other areas the government can invest in; good government, capable of making these investments at reasonable cost, without too much corruption, is crucial for economic and social development.
Concrete before electronics
The cost of advanced signaling systems, such as driverless technology, is approximately the same everywhere in the world, in exchange rate terms. The cost of civil infrastructure construction is approximately the same in PPP terms, and if anything may be a little lower in poor countries. The cost of labor that advanced technology avoids is proportional to wages. This means that the electronics-before-concrete principle is less valid in poorer countries, and is sometimes not valid at all. There are practically no driverless metros in developing countries; the only examples I can find of lines in operation include two lines in Sao Paulo and one in Manila, with a small handful more under construction. Brazil is middle-income, and the Philippines are lower-middle-income rather than poor.
This principle also extends to countries with existing rail lines that they could expand. Investments in concrete – additional tracks, grade separation, relief lines – are more valuable than in developed countries, while investments in electronics are less valuable. A city with a desperate transportation situation can expect that every rapid transit line it builds will fill quickly. Tunnels are in a way more future-proof than precise schedules and resignaling.
Regulate cars, not buses
A recurrent feature of transportation in poor cities without rapid transit or BRT is the minibus. It goes by various names; the most famous to the first-world reader is probably the Nairobi matatu, but it also exists in Lagos as the danfo, in the Philippines as the jeepney, and in Jakarta as the angkot. These vehicles are not popular with the segment of the population that the government listens to: they are typically noisy and dirty and the drivers are aggressive. The governor of Lagos State recently announced a plan to ban the danfos, saying they don’t meet the international standards of a great city and should be replaced with air-conditioned buses. This is while the city is still working on its first metro line.
In Delhi, attempts to give buses road priority met an intense backlash from high-income drivers. There was a failed lawsuit openly stating that car drivers’ time was more important. Eventually Delhi scrapped the system entirely.
In contrast, the most successful public transit in cities that were recently poor or low-income, such as Singapore or Seoul, is in an environment where state policy restrained cars and not buses. Singapore has had congestion pricing since in the 1970s, the first city in the world to implement this scheme, and levies high taxes on cars, as does Hong Kong. Seoul restrained domestic consumption, including of cars, in its period of early industrialization from the 1960s to the 1980s.
Nigeria has 60 cars per 1,000 people. Lagos has maybe 150. To a large majority of the city’s population, cars are traffic, not transportation. Numbers in other third-world megacities vary but are not too different: Cairo has about 200 as of 2011, Delhi about 170, Jakarta about 300. (Some car and population numbers are a few years out of date; caveat emptor.) Traffic restraint is the correct policy given massive traffic jams and growing pollution levels, and the sooner the city starts, the better it will look in a generation.
Plan for growth
Developing-world cities are going to be much larger and richer in 30 years than they are today. National population growth rates range from moderate in India and Bangladesh to explosive in Nigeria, Kenya, and Tanzania. Moreover, all of these countries have low urbanization rates today and fast migration from the villages to the cities, setting up fast urban population growth even where national population growth isn’t so high. Economic growth projections are dicier, but at least one estimate through 2024 is quite optimistic about India and East Africa.
The high-density context of most cities in question rules out any auto-based development pattern. The population density of the eastern half of the Indo-Gangetic Plain, from Delhi downriver to Bangladesh, is about 1,000 people per km^2, comprising nearly 600 million people. Nowhere in the developed world is this density seen outside city regions. Lombardy has 400 people per km^2, and is as hemmed by mountains as North India, producing large-scale thermal inversions; with high levels of car traffic and heavy industry, it is one of the most polluted regions of Europe. Southern Nigeria is not so dense, but with fast population growth, it eventually will be. Egypt’s population density along the Nile is well into the four figures.
This also has implications for how rapid transit should be built. A metro line that passes through lightly-populated areas will soon sprout dense development around it, just as the early lines did in late-19th century London and early-20th century New York. Most New York railfans are familiar with the photo of farmland next to the 7 train in the 1910s; between 1900 and 1930, New York’s population doubled, while Queens’ population grew by a factor of 7. Such growth rates are realistic for some developing-world cities. For the same reason, it is worthwhile investing in grade-separated rights-of-way now, when they are cheap.
Another implication concerns capacity. Even Nairobi, which is not a megacity, can expect to become one soon, and requires many different rapid transit lines entering its center. Some of these can be accommodated on existing roads, as els or relatively easy subways under wide streets, but not all can. When the roads are wide enough, cities should consider four-track structures, since the relative construction cost of four-tracking is low for an el or a cut-and-cover subway.
Four-tracking has one additional benefit: local and express service, which is of critical importance in the very largest cities. In forums like Skyscraper City, Tokyo railfans often express concerns over China’s subways, which have no express tracks and little to no commuter rail, since they offer no path through the center faster than about 35 km/h (Tokyo’s express commuter lines, like Tokaido and Yokosuka, approach 60 km/h).
The final implication is that it’s fine to build a central business district from scratch. Shanghai is doing this in Lujiazui, but that is the wrong location, on the wrong side of a riverbend, with only one Metro line serving it, the overcrowded east-west Line 2; a north-south rail line would have to cross the river twice. A better location would have been People’s Square, served by Lines 1, 2, and 8. This is of especial relevance to cities whose traditional center is in a difficult location, especially Lagos but also Dar es Salaam.
BRT is not rapid transit
The failure of Delhi’s BRT line is in some sense atypical. The line was compromised from the start, and global pro-BRT thinktank ITDP expressed criticism from the start. However, other BRT projects draw cause for concern as well. Dar es Salaam’s BRT is instructive: the first phase cost about $8.5 million per km in exchange rate terms, or about $27 million per km in PPP terms, comparable to an average European light rail line or to an American BRT boondoggle. A hefty chunk of this cost comes from importing Chinese-made buses, which are priced in exchange-rate terms and not in PPP terms.
All else being equal, higher incomes strengthen the case for rail vs. BRT and lower incomes weaken it, since one of the major advantages of rail is fewer drivers per unit of passenger capacity. However, there is a countervailing force: the bulk of the cost of rail construction is local construction, priced in PPP terms, and not imported capital, priced in exchange rate terms. Trains still cost more than buses per unit capacity, but the bulk of the cost premium of rail over BRT is not the vehicles, and a weak currency reduces this premium.
And for all of the global marketing, by ITDP and by Jaime Lerner, the Curitiba mayor who invented modern BRT, BRT is not rapid transit. It is surface transit, which can achieve comparable speed to a tramway, but in a dense city with heavy traffic, this is not high speed. The busiest Parisian tramways, T1 and T3, average about 18 km/h. Modern rapid transit starts at 30 km/h and goes up as construction standards improve and stop spacing widens. BRT is still a useful solution for smaller cities, but in the larger ones, which need more speed, grade-separated rapid transit is irreplaceable.
Don’t neglect mainline rail
How are people going to travel between Jakarta and Surabaya, or between Lagos and Kano, or between Nairobi and Mombasa? They’re not going to fly; the capacity of air traffic is not high. They’re not going to take a vactrain. The only real solution is a high-speed rail network; Indonesia is already building HSR from Jakarta to Bandung, using Chinese technology, with plans for a further extension to Surabaya.
The most difficult part of building a new intercity rail network from scratch is serving the big cities. This is the big advantage of conventional rail over maglev or vactrains: it can run on legacy tracks for the last few kilometers. (In poorer countries, which import technology from richer ones, another advantage is that conventional rail isn’t vendor-locked.) Between this and the need to also accommodate medium-speed intercity rail to smaller cities, it’s important that developing-world cities ensure they have adequate right-of-way for any future system. Trunks should have a minimum of four tracks, with intensive commuter rail service on the local tracks, in a similar manner to Mumbai.
It is also important to build the metro to be mainline-compatible, in electrification and track gauge. It is wrong for India (and Pakistan) to build a single kilometer of standard-gauge metro; everything should be broad-gauge. Russia, where everything is on Russian gauge, does this better. African mainline rail networks are usually narrow-gauge and weak, and in some places (such as East Africa) are being rebuilt standard-gauge. Southeast Asia runs the gamut, with reasonable service in Jakarta, which is running frequent electric commuter rail using second-hand Japanese trains; this suggests future metro lines in Jakarta should be built narrow- rather than standard-gauge, to allow Tokyo-style through-service to commuter rail.
The biggest developing-world cities have problems with air pollution, traffic, overcrowding, and long commutes – precisely the problems that rapid transit is good at solving. They have equally great problems with infrastructure for electricity, running water, and sewage, and with access to health care, education, and such basic consumer goods as refrigerators. And they have limited tax capacity to pay for it all.
This makes building good transit – cost-effective, future-proof, and convenient enough to get high ridership – all the more critical. The smallest cities today may be able to get away with looking like smog-ridden midcentury Los Angeles, but even medium-size ones need to plan on models starting from New York or London or Tokyo, and the biggest ones, especially Lagos, should plan on looking like something that doesn’t really exist today.
To that effect, third-world governments need to absorb massive amounts of knowledge of good practices developed in Western Europe and high-income East Asia (and to a lesser extent Russia and China). But they cannot implement them blindly, but have to learn how to adapt them to local conditions: chiefly low incomes, but also weak currencies, import-dependence in technology, high expected future growth, and (in many cases) high expected population density. Nothing prevents a poor country from doing transit well: China, still a middle-income country, has more high-speed rail ridership than the rest of the world combined, and subway ridership per capita in Beijing, Shanghai, and Guangzhou is healthy. But India, Pakistan, Nigeria, and other poor countries with big cities have their work cut out for them if they want to solve their transportation problems.