Category: New York
The Urban Geography of Park-and-Rides
The urban geography of transit cities and of car cities is relatively well-understood. In a transit city, there will be a strong CBD surrounded by residences with spiky secondary centers, all quite small geographically but dense, centered around train stations and junctions; because density is high throughout, minor trips are done on foot. In a car city, all trips are done by car, the core is weak, and most employment is in suburban edge cities and edgeless cities.
What I haven’t seen is an explanation of how urban geography works in mixed metro areas: there are those in which short trips are done on foot and long ones in cars, such as new urbanist developments, and those in which short trips are done by car and long ones on transit, such as park-and-ride-oriented commuter suburbs. It is the latter that I want to address in this post.
The first feature of park-and-rides is that of all combinations of modes of transportation, they are the fastest and enable suburbs to sprout the farthest from the center. This is because the segment of the trip done in a car is uncongested and so driving is faster than transit, while the segment done on a train parallels a congested road, and conversely makes few stops so that average speeds are high.
On top of this, because intra-suburban trips are done by car, the density in the suburbs is very low, comparable to proper car cities (see the lower end of the density profiles of the New York, Chicago, and Boston metro areas), and this forces sprawl to go outward. New York is the world’s most sprawling city measured in total built-up area; the only other city of comparable size that’s not a transit city or a bus/jitney city is Los Angeles, which is forced to have denser suburbs because of the mountains. Of course Houston and Dallas sprawl even more relative to size, but because they lack New York’s transit-oriented core, there’s an inherent limit to their size.
The other feature is that there’s a definite socioeconomic history to the development of the auto-oriented commuter suburbs of transit cities. First, people move to the suburbs and commute into the city, almost always by train due to road congestion (or, as in the earliest New York suburbs, because mass motorization hasn’t arrived yet). The mass exodus into these suburbs comes from cars rather than commuter rail, and so the local services for people living in those suburbs are built at automobile scale, rather than at the walkable town center scale of 1910.
In North America there’s also a definite class element here – the early movers are the rich rather than the poor. Historically this was partly because poor people couldn’t afford regular train fare, and partly because the impetus for suburbanization was idyllic country homes with access to urban jobs rather than cheap housing for the poor. If I’m not mistaken, this wasn’t the case in Australian cities’ suburbanization, leading to a more urban transit-style mode of running mainline rail. The result of this class distinction is that North American commuter rail styles itself as for the rich: agencies make an effort to ensure everyone has a seat and downplay comfortable standing space, and the expectation is that transit is a last-ditch mode of transportation for when cars just don’t have the capacity to get people downtown, and so nobody needs to take the trains in the off-peak or take a bus to the train.
The result is that the park-and-ride city will still have a strong core with high-capacity transportation, and the primary CBD will maintain its supremacy for high-income jobs. Establishing edge cities in the direction of the favored quarter can happen, but there’s still a congested city nearby, and so from many directions it’s impossible to drive, and taking transit is impossible. Thus jobs in White Plains and Stamford are not nearly as high-paying as jobs in Manhattan.
There can even be secondary CBDs, if the inner part of the metro area, where people take transit more regularly than the suburban commuters do, is large enough. But those secondary CBDs are frequently quite auto-oriented. Brooklyn’s mode share for jobs is only 42-39 in favor of transit (for residents, it’s 60-25), and all other counties in the New York region except Manhattan have more workers driving than taking transit, a situation that is not true if one looks at residents. Those secondary CBDs then have mixed characteristics: they are dense and fairly walkable, as can be expected based on their history and location, but also have plentiful parking and a large share of drivers demanding even more. They can accommodate multiple modes of transportation, but driving is more convenient, and from the suburbs the commuter rail system isn’t always geared to serve them.
Commuter Rail Ridership Distribution
As a followup to my claim that the Northeast Corridor in New Jersey had a more outer-suburban ridership than the Morris and Essex lines, I decided to tabulate the ridership distribution of various commuter lines. This tells you what percent of the ridership originates within some distance of the city center. All lines in New York are included, though some are grouped together because of branching.
Explanation: the ridership numbers for New Jersey Transit come from the New York Times, and those for the LIRR and Metro-North come from files published by the MTA. To maintain comparability with the Metro-North and NJT numbers, ridership in city terminal areas is ignored for calculation purposes; thus, X% really means X% of beyond-city ridership. This means stations from Jamaica west, from Newark Penn east, and from Harlem south are not counted. All km points are calculated from Penn Station or Grand Central, even for lines that do not run through to those stations. Finally, some lines are lumped together, when they share stations beyond the excluded city terminal zone.
| Line\km | 20 | 30 | 40 | 50 | 60 | 80 | 100 |
| NJ NEC/NJC (66,997) | 4 | 14.2 | 31.3 | 44.6 | 59.6 | 78.2 | 98.7 |
| R. Val. (10,639) | 0 | 13.8 | 51.7 | 74.3 | 84.4 | 98.5 | 100 |
| M&E/MB (35,183) | 5.6 | 36.3 | 64.7 | 81 | 91.8 | 98.9 | 100 |
| Erie Main/BC (13,249) | 16 | 35.6 | 60.7 | 73.9 | 81.2 | 89 | 92.9 |
| P. Val. (3,674) | 5.5 | 39 | 70.3 | 97.3 | 100 | 100 | 100 |
| Hudson (25,442) | 5.3 | 16.7 | 30.8 | 52.6 | 66.7 | 76.2 | 88.8 |
| Harlem (45,117) | 4.7 | 27 | 68.8 | 73.9 | 82 | 91.3 | 98.1 |
| NH + NC/D/W (61,973) | 0.1 | 13.6 | 31.4 | 42.7 | 60.8 | 77 | 90.2 |
| Port W. (23,404) | 14.1 | 80.8 | 100 | 100 | 100 | 100 | 100 |
| LIRR Main (65,104) | 0 | 8.5 | 29.1 | 51.7 | 69.2 | 83.7 | 99.9 |
| Mont. + Atl. (58,835) | 0 | 10 | 52.9 | 80 | 88 | 96.3 | 99.3 |
Note that the data isn’t completely reliable. The NJT and Metro-North data sets paper this over by counting just one direction, but the LIRR counts both, and there are discrepancies, for example at Huntington. So the numbers above have a fair margin of error around them.
Observe that the ridership of the Northeast Corridor is so skewed outward that despite having twice the ridership of the Morris and Essex lines in New Jersey, the Morris and Essex lines actually beat it on ridership within 40 km of Penn Station. Similarly, the Harlem Line beats the New Haven Line up to 50 km.
Similar data exists in Boston, and, in harder to search form even if you speak the language, Tokyo (better data for Tokyo can be found here, but for most lines the numbers include only inner and middle segments, up to about 50 km outside Tokyo Station). It’s also quite easy in both cities to set a boundary of the excluded city zone, and with Boston this could allow constructing the same table.
The implication of the difference between various lines is that some lines are more local and some are practically intercity. This relates to the service decisions within each line – more stops or fewer stops – but there aren’t a horde of people in Elizabeth and South Newark who are clamoring to ride rush hour trains into Penn Station and would in large numbers if only stop spacing were narrower, or a horde of people in Sussex County who’d ride if only there were fewer stops between Dover and Penn Station.
That said, the more local lines do have potential for local service on trips that American commuter rail doesn’t serve. There’s an untapped market of people commuting from New Jersey to Jamaica and Brooklyn, or from Long Island to Newark and Jersey City, and this market necessarily needs to be served with more local trains, because most people in it live closer to the city.
Regional Rail for New York: What Can Be Done Now
MTA Chairman Joe Lhota recently proposed to through-route commuter rail lines in the New York area, as was proposed in the past by the RPA, the Institute for Rational Urban Mobility, and more recently myself. Lhota proposed other, less flashy ideas for integration, including better track sharing at Penn Station and lengthening platforms to accommodate 10-car trains. Although a network that looks like my proposal should still be the goal for the next 20 years, there are several things that can be done in the very short run. None is do-it-tomorrow immediate, but neither does any require very difficult modification of equipment or organization or significant infrastructure investment. Most should not require extensive studies.
Note that this is not a wishlist of the most important commuter rail reorganization projects in the region. Many of those reorganizations do not have anything to do with interagency integration, and are therefore not included. Only projects that are very cheap and would come from or benefit integration are on this list.
1. Integrated ticket machines at Penn Station. This requires the physical tickets on New Jersey Transit to look like those on the LIRR and Metro-North (and thus some modifications to the fare barriers at Secaucus and Newark Airport), and some reprogramming of ticket machines, but no change otherwise. Ideally a ticket from (say) Hicksville to Newark should cost less than the sum of tickets from Hicksville to New York and New York to Newark, to encourage reverse-peak traffic, but strictly speaking the discount is not needed. Amtrak and commuter rail machines should also be integrated, though the physical tickets can still be different if switching over is too hard.
2. Integrated concourses at Penn Station. This means treating the upper and lower concourses as belonging to all three railroads. This requires Amtrak to give up its single-file queuing and accept that people already can walk around and get to its trains from other railroads’ turfs. Trains should be announced on all concourses, and all access points to a platform should be clearly signed with the next train’s type and schedule.
3. Timed transfers. Although a clean integrated timetable is impossible, because trains interline on some inner segments to increase capacity, a partial version is still possible. What this means is that, with hourly off-peak service on each branch, Morris and Essex trains should arrive at Penn Station just before the hour, as should one of the several hourly trains on the New Jersey side of the Northeast Corridor, and then two or three branches going to the east (say, to New Haven and Port Washington, and on one additional LIRR line for service to Jamaica) should leave just after the hour, with the tightest connection done cross-platform. This would make trips from New Jersey to JFK and from Long Island to Newark easier, and the choice of services to participate in the system should be consistent with even spacing on interlined trunks.
4. Modification of rolling stock. Metro-North’s M8s can run under 60 Hz catenary and third rail, but unfortunately not 25 Hz catenary; as lower frequency requires a larger transformer, modifying the trains to run on the New Jersey side of the Northeast Corridor may be too hard in the very short term (though not in the medium and long terms). However, NJT’s ALP-46 locomotives and Arrow EMUs can run on 12 kV 25 Hz and 25 kV 60 Hz catenary, and thus modifying them to run on Metro-North’s 12 kV 60 Hz catenary is easy, allowing them to run from the NJT network to the New Haven Line. Unfortunately, because locomotives accelerate more slowly than EMUs and the Arrows are quite old, they do not have very good performance for short-stop service, for which through-running is the most useful.
5. Voltage change on the Northeast Corridor’s New Jersey side to 25 kV 60 Hz. This voltage change was done to the Morris and Essex lines and much of the North Jersey Coast Line. It is somewhere on Amtrak’s wishlist of projects, but I do not know how high it is. This allows M8s to run through, ensuring the better rolling stock is available for the service that needs it the most. It may possibly be bundled with Amtrak’s installation of constant tension catenary south of New Brunswick to reduce costs. Since this eliminates the need for 25 Hz transformers in the future, this meas future NJT rolling stock would be lighter.
6. Depending on 4-5, rolling stock sharing along interlined services. In practice this means M8s on local Northeast Corridor services, which would also allow adding and serving infill stations in New Jersey (for example, more regular service to North Elizabeth, and perhaps a station at South Street in Newark), and Arrows and locomotives on express services from Penn Station and New Jersey to New Haven.
7. Platform raising on the North Jersey Coast Line and the Morris and Essex lines, if service using M8s rather than Arrows is desired. Because of the voltage, it’s actually easier for M8s to serve the Morristown Line other than their inability to serve low platforms: it would only require 8-21 km of reelectrification rather than 88-101. The Morris and Essex lines also have a more inner-suburban distribution of ridership than the Northeast Corridor Line, which gets most of its ridership from more distant stations, and this makes them in one sense better-suited for through-service. (In another sense, the Northeast Corridor is better, since it serves downtown Newark, a secondary CBD that draws some commuters from suburbs and boroughs east of Manhattan.)
It is my belief that all of the above, possibly except #5 and #7, are feasible within months or at worst a very small number of years, and would not require additional environmental work. Even #5 and #7, which are more expensive, are still close to two orders of magnitude cheaper than a full through-running plan with new tunnels serving Lower Manhattan.
The medium term is more expensive – perhaps an order of magnitude less than the full program rather than two – and would include further modernization, allowing full through-service on every line and more efficient equipment utilization. It can also assume friendlier regulations, which a snap integration cannot, and this in particular means better rolling stock in the future and higher speeds even with existing rolling stock. Clockface schedules and frequent off-peak service would allow planning infrastructure repairs and upgrades around specific schedules. For example, the current local Stamford-Grand Central schedule is 1:06, but an express train I recently took from New Haven came to Grand Central more than 10 minutes ahead of schedule, suggesting excessive padding; minor upgrades should allow an M8 to do Stamford-New York in an hour minus turnaround time making local stops, and more ambitiously New York-New Brunswick in 45 minutes minus turnaround time.
Lhota can’t do much in the long term, because this requires an enormous investment into concrete, a political decision and a longer-term one than Lhota’s term as MTA chair. However, he can both implement the above seven points within his term, and also set in motion various work rule reforms and small-scale capital project planning and apply for the requisite FRA waivers to permit the medium-term reforms to succeed.
Bus and Rail Mantras
Bus is cheaper than rail. Paint is cheap. Rail only made sense a hundred years ago when construction costs were lower. Trains have no inherent advantage over buses. It doesn’t cost more to operate a bus than to operate a train. All of those are true in specific sets of circumstances, and Curitiba and Bogota deserve a lot of credit for recognizing that in their case they were true and opting for a good BRT system. Unfortunately, the notion that buses are always cheaper than trains has turned into a mantra that’s applied even far from the original circumstance of BRT.
The advantage of buses is that dedicating lanes to them and installing signal priority are financially cheap, if politically difficult in the face of opposition from drivers. Even physically separating those lanes is essentially cost-free. This advantage disappears completely when it comes to installing new lanes, or paving an existing right-of-way. Hartford is paving over an abandoned railroad at a cost of $37 million per km.
Not to be outdone, New York’s own MTA just proposed to pave about 8.5 km of the Staten Island Railway’s North Shore Branch for $371 million. A light rail alternative was jettisoned because the MTA insisted on continuing the line to the West Shore Plaza, along what is possibly the least developed road in the city.
Another, related mantra is that light rail is cheaper than heavy rail. This contributed to the MTA’s decision not to pursue a Staten Island Railway-compatible solution, which would allow lower capital costs and cheaper maintenance since trains could be maintained together with the existing fleet without modifying the existing yard. As with all mantras, this one has a kernel of truth: it’s much cheaper to build on-street light rail than elevated rail or a subway. As with the BRT mantra, this is not true when the discussion is about what to do in an existing right-of-way.
Worse, because the MTA believed its own hype, it completely missed the point of surface transit. People who believe these mantras about bus, light rail, and heavy rail can easily miss the advantage of on-street running wherever the streets are more central than the railroad rights-of-way. The North Shore Branch hugs the shore for much of the way, halving station radius. The most developed corridor is Forest Avenue, hosting the S48, the third busiest bus in the borough and the busiest in the same area and orientation as the line in question. (The busiest in the borough, the S53, crosses the bridge to connect the North Shore to the subway in Brooklyn.) Of the three other east-west routes in the North Shore, the one that the North Shore Branch parallels the most closely, the S40, has the lowest ridership. It would be both vastly cheaper and better for bus riders to have dedicated bus lanes on Forest, or possibly Castleton, which hosts the S46.
In cities that did not develop around mainline rail corridors but rather around major streets, the only reason to use mainline rail corridors for urban transit is that reactivating them for rail can be done at much lower cost than building on-street light rail. New York is for historical reasons such a city: Staten Island development follows Forest and Castleton rather than the North Shore Branch, and for similar reasons Park Avenue in Manhattan and the Bronx is a relatively unimportant commercial corridor.
Now, these mainline corridors have great use for regional transit. Queens Boulevard can’t be easily used for train service to Long Island, and Lexington Avenue can’t be easily used for train service to Westchester. Staten Island has great potential for regional transit – but only if it’s electrified rail going through a tunnel to Manhattan. It’s expensive, but it’s what it takes to be time-competitive with the ferry and with buses to the subway. A more competent agency than the MTA would keep planning and designing such high-cost, high-benefit projects, to be built in the future if funding materializes; such plans could also be used to concretely argue for more funding from the state and from Congress.
Instead, the MTA is spending more money than most light rail lines cost, to make such a mainline connection from the North Shore to Manhattan impossible in the future. The best scenario in such a situation is that the busway would have to be railstituted, for a few hundred million dollars – an embarrassing reminder of the busway folly, but still a much smaller sum than the cost of the tunnel. The worst scenario is that like on Los Angeles’s Orange Line, the need to keep buses operating during construction would make it impossible to replace them with trains.
There aren’t a lot of lose-lose (or win-win) situations with transportation, even if we ignore driver convenience, but this is one of them. It’s a fiscal disaster relative to predicted ridership and the operating costs of buses, it makes future transit expansion in the borough more difficult, and it follows a marginal route. All this is so that the MTA can say it’s finally making use of an abandoned right-of-way.
High- and Low-Speed Rail Coordination
The debate about what kind intercity rail to build tends to be either/or. On one side, there’s HSR-only advocacy: this represents the attitude of SNCF, especially in the earlier years of the TGV, and such American HSR proponents as John Mica. In this view, legacy rail is inherently slow and money-losing and the best that can be done is to start fresh; generally, this view also looks down on integration with legacy regional rail. On the other side, there’s a legacy-only advocacy, which represents how Britain upgraded its intercity rail network in and after the 1970s and also the attitude of proponents of Amtrak-plus lines in the US.
The problem with this is that there are a lot of different markets out there, and the service levels they justify and the construction challenges they impose are different. Sometimes such markets are in the same general area, and this means some lines should be HSR and some should be upgraded low-speed rail.
Countries that tried to go to one extreme of this debate are now learning the hard way that they need to do both. Britain radically optimized its intercity main lines, which now have the highest average speed in the world except for HSR – but it needs more, and this requires it to build a new HSR line at immense cost. In the other direction, France’s TGV-only strategy is slowly changing. SNCF still doesn’t care about legacy intercity lines, but the regions are investing in regional rail, and one region even uses the high-speed line for local service. Japan gets away with neglecting most of the intercity lines because its physical and political geography is such that markets that can support HSR dominate, but other countries cannot.
This means that best network design is going to have to deal with both approaches’ political difficulties at the same time. Upgrading legacy rail means upgrading legacy rail operating practices, against opposition of workers and managers who are used to old and inefficient ways of doing things. And building HSR on the thickest markets means giving special treatment to some regions with infrastructure that other regions don’t justify; it’s economically solid, but the optics of this are poor.
But the advantage of doing it this way from the start is that it’s more future-proof, and allows integrated design in terms of schedules, which lines are upgraded, how cities are connected, and so on.
Doing it piecemeal may require redoing a connection along a different alignment. The issue is that HSR compresses travel times along the line only. It’s like urban rapid transit this way, or for that matter like the air network. A legacy rail system (or a national highway system, or urban buses) has fairly consistent average speed. This means that in a combined system, the optimal path between two cities may not be the shortest path, in case one is close to the HSR trunks.
For example, look at Upstate New York. None of its four major metro areas is large enough to justify a high-speed connection to New York by itself, but all four combined do. Although international service to Toronto is overrated, it could be justifiable in light of Buffalo’s relative economic integration with Ontario and also the mostly straight, partially grade-separated right-of-way available in Canada; this would further thicken the market.
If we draw a rudimentary map of other desired connections, none thick enough to warrant more than an upgraded low-speed train, the fastest connections are not always obvious. For example, with average HSR speed of 240 km/h and legacy rail speed of 100 km/h, it’s faster to get from New York to Ithaca via Syracuse than directly via Binghamton. This is why the connection to Ithaca is through a line that points toward Syracuse, even if it’s not the shortest route to Binghamton. It’s one of many small local optimization problems.
More interestingly, we get a mini-hub in Syracuse. Although it’s the smallest of the four main Upstate cities, it lies at the junction of the trunk line and lines to Binghamton and Watertown, and also has secondary cities at the right location for regional rail. (The largest comparable secondary city near Rochester is Geneva, which happens to be close to and have a good rail connection to I-90, a prime candidate for HSR corridor; thus it should get commuter service using the trunk line, which would be far faster than an all-legacy train.) This means that schedules should be set up to coordinate transfers in Syracuse.
This is a normal way to set things up in an all-legacy format, as is done in Switzerland, but it can equally apply to HSR. The construction challenges on the Empire Corridor are nowhere near as complex as those in California, Pennsylvania, and other truly mountainous states, but they’re still nontrivial. But now that we know that Syracuse should be a hub, one answer to the question “How many design compromises to make to reduce costs?” is “Build just enough to allow integrated transfers in both New York and Syracuse.”
(In practice this means HSR arriving in Syracuse on the hour and in New York whenever convenient. The main intercity line into New York is the Northeast Corridor, a very thick market that at HSR speed would have enough traffic to support show-up-and-go frequency. This is not true of lines serving Syracuse; Watertown is not Washington and Binghamton is not Boston.)
The main cost of doing things this way is political. It requires willingness to both prioritize markets and cut construction costs, as necessary to build HSR, and improve legacy rail operating practices and carefully integrate services, as necessary to build a working legacy rail network. The fiscal cost is not outrageous – those legacy lines are cheap relative to everything else (rebuilding the unelectrified New York-Scranton line is $550 million), and HSR on thicker markets will at least partially pay for itself.
Once we discard the notion that present-day Amtrak operating patterns are adequate, the question stops being about whether one trusts Amtrak or not, and purely about how to build a new transportation network. And then the correct answer to “High-speed or legacy?” is “Both, seamlessly integrated with each other.”
Quick Note: How Much Tunnels Really Cost
New York is currently building a 3-kilometer tunnel between Brooklyn and Staten Island, using the same EPB method that Madrid uses to build subway tunnels. The cost of the single-bore tunnel is $250 million, and the project will be completed by 2014.
Of course, this is a water tunnel rather than a train tunnel. The diameter of the tunnel is somewhat smaller than that of a single-track train tunnel. Double-track tunnels, even ones built to high-speed rail standards, are substantially wider, but the amount of concrete lining required is proportional to radius rather than to cross-sectional area. For example, the double-track Seikan Tunnel is 9.7 meters wide, little more than single-track HSR tunnels in Europe, as Japanese construction tries to minimize tunnel clearances to cut costs and instead equip Shinkansen trains with elaborate aerodynamic noses. While 9.7 is more than 2.5 times the diameter of the water tunnel in question, 250 million times 2.5 is still far below the construction cost of any recent tunneling project in New York.
The expensive part of tunneling, then, is not the actual tunnel. It’s everything else, especially the station caverns. Both ARC and East Side Access included multilevel deep caverns in Manhattan with full-length mezzanines; of course they’d be more expensive.
For what it’s worth, an 8-kilometer long, 9.7-meter wide tunnel from Staten Island to Manhattan would cost $1.75 billion at the same per-km, per-meter cost of this water tunnel. Of course stations at St. George and especially Lower Manhattan would add much more, forcing a lot of difficult choices about location, but the basic infrastructure is not all that expensive.
What’s a Subway/El?
The rapid transit built in New York beginning with the first els codified two characteristics that spread to the rest of the US, and are often seen in other countries’ rapid transit networks as well. First, it is separate from surface transit – even when it did still have grade crossings, they were controlled railroad crossings, rather than street-running segments as is common on light rail. And second, it is separate from mainline rail.
Not much later than New York started building els, Berlin built the Stadtbahn, also an urban elevated railroad. However, it was meant to be used for mainline rail from the start, with two local passenger tracks and two long-distance passenger and freight tracks. Part of the impetus was to connect different railroad terminals within the city, which American cities did by building union stations disconnected from local traffic. Shortly later, Tokyo built its own mainline rapid transit system – the Yamanote Line bypass in 1885 and Tokyo Station connecting the Chuo and Tokaido lines in 1914. Both cities ran frequent local commuter service early, Berlin doing so even before electrification.
Of course, nowadays US regulations locked in the separation of rapid transit from commuter rail, but at the time, there was no such separation. New York could have built its subway to mainline specifications and run trains through to the LIRR. It didn’t because of historical accidents – it preferred compatibility with the els and even when the BRT chose a wider loading gauge for its own subway network, it still opted for narrower trains than on mainline track. At the time it seemed like no big deal, although some of the subway lines built were redundant with existing commuter lines (for example, the Flushing Line with the Port Washington Line). Again due to historical practice, commuter rail did not try to operate to rapid transit standards, keeping frequency low, and so nearly all urban stations closed. In both New York and Chicago, it’s often easy to figure out where the city ends or where the subway/L network ends because that’s the point beyond which commuter train stop spacing narrows, providing makeshift local service.
In subsequent decades, the German and Japanese approach proved itself much more capable of providing good transit to growing suburbs. In Tokyo, subways are legally railroads, and most lines are compatible with at least one commuter line in order to permit through-service. German cities have mainline rapid transit (S-Bahn) and also separate subways or subway-light rail combinations (both called U-Bahn). Many other cities and countries had to adopt the same system to increase transit ridership, at much higher cost since the necessary viaducts and tunnels connecting stub-end terminals were done much later. This is what led to the Paris RER, and what’s led to Thameslink and now Crossrail in London. Any other approach would require spending even more money on extending urban lines to the suburbs, exactly what’s done now in the two big suburban-focused US rapid transit systems, the Washington Metro and BART.
The kink is that despite the above problems of subways that are separate from both mainline and street rail, there’s now a different reason to build such lines after all: they can be made driverless. Most first-world cities already have legacy rapid transit or else have so much sprawl rapid transit is inappropriate, and third-world cities aren’t saving much money by eliminating drivers, but in the few cases of new builds (Vancouver, Dubai, Copenhagen, the newer lines in Singapore), driverless trains are common, and this allows trains to run more frequently, or even 24/7 in Copenhagen’s case.
This kink aside, there’s really no reason for a city to build a new New York-style subway, i.e. disconnected from light and commuter rail and running with a driver. Extending a legacy system is fine, but for new systems, there’s no point. This could be especially bad in growing third-world cities, which could find themselves paying too much for a subway they don’t need or unable to connect a subway they do need to the suburbs once they start suburbanizing. Third-world construction costs aren’t much if at all lower than first-world costs, but wages are much lower.
Some of the world’s largest cities have made or are making this mistake. Mumbai is building a new subway, on a different track gauge from the Indian mainline network, preventing through-service to the overburdened commuter trains. Shanghai and Beijing have vast subway networks, without express tracks or any ability for trains to run fast through city center; they have widely spaced stops so that they are faster than most other subway systems, but they have nothing on the rapid commuter trains in Tokyo. (Beijing is also developing a parallel commuter rail network, running diesel trains from the exurbs to the traditional city terminals at low frequency.) It works fine now, but when Shanghai grows and suburbanizes to the degree Tokyo has, it may find itself having to spend many billions on digging new tunnels.
Since a New York-style subway is inappropriate for new builds, some cities need to ask themselves which of the three kinds is the most appropriate. A subway-surface solution is mainly an option when one underground line can naturally split into multiple surface lines, as is the case in Boston, San Francisco, Cologne, and Frankfurt; this is because there’s a big difference between on-street and grade-separated capacity.
Tel Aviv, which is building a subway-surface line without any branching, is doing it wrong. For the other choice, I believe it’s a matter of how well-developed the suburban rail network is, and how much future suburbanization the city can realistically expect. In Tel Aviv specifically there’s also a separate element, which is that for religious reasons public transit does not run on weekend. If driverless technology makes the difference between trains that run 24/7 and trains that run 16/6, then it should be used even at the cost of otherwise worse service to some suburbs and destinations easily reached by legacy rail branches.
Finally, in North America, one of the reasons to engage in strong regulatory reform is to allow the mainline option to work. Some lines, for example the Harbor Subdivision between LAX and Union Station, should ideally host a mixture of local and rapid trains on the same tracks, and also allow intercity trains; if the Harbor Sub becomes an electrified commuter line then high-speed trains could serve the airport, providing a connection from the Central Valley to a major airport in addition to SFO, which would only get a station at Millbrae.
More in general, the only real disadvantage of legacy commuter networks is that they tend to not be very dense in the center of the city, requiring new builds; most of the Tokyo subway is just lines offering the commuter lines more capacity into the CBD, overlaying itself to also provide a tight in-city network. There’s no technical reason not to just build an electrified local mainline network as its transportation backbone, and if more capacity is required then build additional lines in the mold of Tokyo.
One-Way Pairs: the Bad and the Ugly
One of Jane Jacobs’ prescient observations about bus service in The Death and Life is that one-way pairs, as practiced on the avenues in Manhattan, are bad for riders. Her argument was that one-way pairs require people to walk too long to the bus line, and this cancels out any gains in speed. (This is truer today, when signal priority is an option, than it was fifty years ago.) Jarrett Walker has formalized this in two posts using station radius as an argument; the issue is that passengers need to be within a short walking distance of both halves of the line, and this reduces coverage.
However, not all one-way pairs are created equal. An underrated reason to keep bus services on one line is simplicity: it’s easier to remember that a route follows one street than that it follows two, and also service to specific destinations can become easier. Taking a cue from proper rapid transit, ITDP’s magnum opus BRT standard treats it as a given that buses should run in the median of a street and only even lists one-way pairs as an option on very narrow streets, and even then as an inferior one. The argument revolves around service identity.
In particular, one-way pairs that preserve a semblance of service identity and simplicity are not as bad as one-way pairs that do not. For the original walk-distance reason, it’s also better to have the one-way pair closer together. Jarrett specifically praises Portland’s light rail one-way pair, located a short block apart, as an example of a good couplet. Manhattan’s one-way pairs are located a long block apart, so the walking distance is worse.
But even Manhattan’s one-way pairs are at least coherent. The First/Second Avenue bus follows First and Second Avenues for the entire length of the avenues; south of Houston, it follows Allen, the continuation of First. This is the advantage of the grid. In Providence, things are not as nice, though still somewhat coherent, if one remembers, for example, that Angell and Waterman Streets form a one-way pair (they’re treated as such for car travel, too, so anyone in the neighborhood would know, though people from outside would not).
In contrast, this is how Tel Aviv’s one-way pairs work. They’re getting worse amidst the various bus reform. The post is in Hebrew, but look at the map at the bottom of bus #5, the city’s busiest (and most frequently bombed back in the 1990s and early 2000s). The travesty is that none of those streets on which the line runs in one direction only is even one-way. East of Ibn Gabirol, the street hosting lines 25, 26, and 189 on the map, the streets are wide and two-way. The reason for the complication is lack of left turns. In order to make car traffic flow a little more smoothly, Tel Aviv has completely eviscerated its bus service.
In principle, Tel Aviv has infrastructure for consistent one-way pairs when necessary and regular two-way service elsewhere. For example, Dizengoff and Ben Yehuda, the two north-south streets hosting buses to the west, function as such for cars. They both have contraflow lanes for buses, allowing buses to use them as two-way streets; some do (for example, #5 on Dizengoff), while others still go one-way (for examples, #9 and #55). Likewise, Jabotinsky, the east-west street feeding into the big circuit, is one-way and narrow west of Ibn Gabirol, and could be a one-way pair with Arlozorov to its south; but Arlozorov is kept two-way, and so #66 is two-way, and #22 uses the two as a one-way pair. (By the way, those are fan-made maps; the official maps don’t use color to distinguish routes, and are thus completely unusable.)
The results of the mess coming from ending any service coherence are predictable. Israeli car ownership, low by first-world standards, is rising rapidly, and the social justice and affordable housing protesters are now complaining about high fuel prices. None of them is anti-transit on principle, and all who I confront tell me they’d ride transit if it were usable. I live without a car in a city with worse transit than Tel Aviv, but to me car ownership is not aspirational. When the only transit people know in their country is unusable, people this generation will get cars. The next bus reform will then take into account more left turn restrictions coming from the need to accommodate more vehicles. The next generation of people will grow up with the expectation of even worse bus service and not conceive of any alternative to automobility.
Transportation-Development Symbiosis
The RPA’s Regional Assembly has included the following idea submission: expand reverse-commuter rail service. The proposal calls for surveying city residents to look for the main available reverse-commuter markets, and for expanding reverse-peak service on the model of Metro-North. It unfortunately does not talk about doing anything at the work end – it talks about looking at where city residents could go to the suburbs on commuter rail, but not about which suburban job markets could be served from any direction.
I don’t want to repeat myself about what transit agencies have to do to be able to serve suburban jobs adequately (if “suburban” is the correct way to think of Providence and New Haven), and so I’m going to sound much harsher toward the idea than I should be. Suffice is to say that talking about development requires a lot of reforms to operating practices. With that in mind, let’s look at some suburban job centers in the Northeast: Providence, Stamford, Hicksville, New Haven. As can be seen, those stations all look very suburban, and even Providence is surrounded by sterile condos, with the mall located a short, unpleasant walk away. Compare this with the urbanity that one finds around major suburban train stations in Tokyo, such as Kokubunji and Tachikawa.
But really, the kind of development that’s missing around suburban train stations in the US is twofold. First, the local development near the stations is not transit-oriented, in the sense that big job and retail centers may be inconvenient to walk to for the pedestrian. And second, the regional development does not follow the train lines, but rather arterial roads, or, in cities with rapid transit, rapid transit lines – for example, one of Long Island’s two biggest edge cities, East Garden City, is diffuse and far from existing LIRR stations (the other, Mineola, is relatively okay).
In both cases, what’s missing is transportation-development symbiosis. Whoever runs the trains has the most to gain from locating major office and retail development, without excessive parking, near the train stations. And whoever owns the buildings has the most to gain from running trains to them, to prop up property values. This leads to the private railroad conglomerates in Tokyo, and to the Hong Kong MTR.
The same symbiosis can be done with government actors, but isn’t, not in the US, and the RPA’s attempts to change this and promote integrated planning have so far not succeeded. Hickville recently spent $36.4 million on a parking garage adjacent to the station plus some extra sum on expanding road access, but none of the relevant actors has made any effort to upzone the station area for commercial, to allow easier commuting. Providence is renovating the station, with pretty drawings, but doing far short of a redesign that would add development to the area.
The importance of this symbiosis, coming back to the original idea, is that the correct question to ask is not, “Where can city residents go to the suburbs to work?” but rather “Which suburban and secondary-urban destinations can be adequately served by rail?” In all four Northeastern cities under discussion, there is more than one direction from which commuters could come. From the commuter railroad’s perspective, a rider who takes the train in the traditional peak direction but gets off in a suburb short of the CBD is a free fare, just like an off-peak rider or a reverse-peak rider.
The task for regional planners (as opposed to service planners and railroad managers) is then a combination of the following priorities:
1. As noted above, ensuring edge city and secondary CBD development is both close to train stations and easily accessible by pedestrians.
2. Aggressively upzoning near potential station sites, with an eye for junctions, such as Sunnyside, Secaucus, and New Rochelle.
3. Examining where people working in secondary centers are living, and which rail lines could be leveraged to serve them and where new construction would be needed. For example, Providence could use rail to Woonsocket and the East Bay and more local service to Cranston and Warwick, but reviving the tunnel to the East Bay could be expensive and needs to be studied carefully. Note that north of South Attleboro, there are very few people living near the Providence Line working in Providence, and so reverse-peak service is useful mainly in the original sense of people reverse-commuting from Boston, in contrast with service to Massachusetts suburbs of Providence such as Seekonk.
The problem with doing all three is political: current regional rail traffic is dominated by suburbanites using it as an extension of driving into the city. This influences local thinking because the economics of residential development are not the same as those of commercial development. Agglomeration and density are less important. Transfers and long access distances are more acceptable. People traveling within the suburb go toward the station in the AM peak rather than away from it, and so parking availability is more important. Take all of these together and you get a powerful constituency supporting continuing to choke suburban train stations with parking and sterile development for city-bound commuters, no matter how many tens of thousands of jobs are nearby.
This is why some symbiosis is necessary. One way to do it is via market mechanisms: if a well-capitalized company gets ownership of the transit infrastructure and is free to develop with few zoning constraints, it could decide to build office towers in Hicksville on top of the train station, or develop the empty lots near New Haven and Providence. This is possible, but may well be too hard politically, even more so than direct zoning reform, because every trope used by the community to oppose the changes (namely, fear of outsiders) would apply and also there would be explicit loss of control.
The other way is the public way, which is where integrated planning comes in. Even on the level of intransigent railroads, it may work if all done together. In other words, there would be simultaneous effort to add reverse-peak service on the LIRR and the MBTA, upzone surrounding station areas and make them more walkable at the expense of some parking spaces, direct major developments such as malls and office complexes to the resulting TOD, and integrate local transit with the changed commuter service in all directions.
But whatever is done, it’s critical to integrate the two functions, of transportation and development. There’s no need for an overarching bureaucracy to take care of it all, even – just cooperation between regional planners, local planners, and transit managers. Transit needs thick markets, and if all development outside the primary CBD is diffuse and auto-oriented, there will not be any thick markets for it to serve. A transit revival necessarily requires new markets, and this means going after what are now hopelessly auto-oriented suburbs. And what needs to be done is not just figuring out where new service is required or where car-free urbanites commute to, but also what kind of TOD can be done at each secondary job center.
Why Long Island Should Get An HSR Spur
Having looked into why high-speed rail from New York to Boston should go through Providence, I want to explain why it should go through New Haven, rather than through any of the fanciful Long Island routings proposed most prominently by the Penn design group. Like Hartford, Long Island should have high-speed trains use the LIRR Main Line, but at medium speed rather than high speed, and with careful consideration to the much more important needs of commuter rail.
Although the LIRR Main Line shares one characteristic with the New Haven-Springfield line, namely that it is very good for 160-200 km/h but bad for 300, the reasons are subtler and less geometric. The most visible is NIMBYism. Even increasing the traffic of existing LIRR trains raised the ire of some suburbs along the Main Line, which opposed the three-tracking project (since canceled due to budget shortfall) on the grounds that extra train traffic would reduce quality of life and that eminent domain would be required. This is not Caltrain, whose local residents do not know what electric trains sound like; this is Long Island, which has lived with these trains for generations. Introducing HSR is asking for trouble.
Of course, the same could be said about any suburb that HSR needs to pass through. Connecticut is full of NIMBYs, just like Long Island. The reasons usually given for avoiding the existing Shore Line are that it’s too developed and has too much local opposition. But those are present on Long Island, and are worse because of the higher population density. For examples, compare Westport and Cos Cob with Brentwood and Farmingdale. The LIRR offers multiple straight rights-of-way, but all are going to have the same speed limits as heavily upgraded and modified tracks on the Shore Line – 250 km/h in the better parts, and 200 in the worse parts.
The Penn design proposal is not even the best Long Island proposal, for three reasons:
1. It insists on proceeding from Penn Station to Jamaica on the Lower Montauk Line. If a connection from the line to Penn Station opens, it’ll be far more useful for local rail, while intercity rail can use the Main Line. The difference between appropriating a Manhattan-accessible Lower Montauk Line for HSR and replacing the Lexington Avenue Line with a truck tunnel is one of degree, not kind; in both cases, local passenger rail is the most valuable use of the infrastructure.
2. It departs from the Main Line to use the Hempstead Branch (necessarily eviscerating commuter service) as well as abandoned tracks through endless residential suburbs, full of urban grade crossings. The Main Line has grade crossings and would need to be four-tracked, but the local NIMBYs actually supported grade separation, and multi-tracking at least could be sold as the local transit improvement project that it is.
3. Last and worst, it sharply veers north after stopping at Ronkonkoma, along a curve whose radius judging by the alignment map is around 900 meters (=150 km/h if superelevation and cant deficiency are set at normal HSR levels, or 170 km/h at cutting-edge levels). Then it crosses the Long Island Sound at its widest, so that it adds more than 20 kilometers to the New York-New Haven route length over the Shore Line, all at medium speed.
A route similar to the Penn design route but using the more feasible Main Line alignment would be 9 minutes slower than the optimal Shore Line route – 41 versus 32 minutes – with stops at Jamaica and Hicksville, enforced by unfixable track curvature near the stations. But in addition to the extra travel time, fixing the alignment through New Rochelle, Darien, and Bridgeport is far cheaper than a long undersea tunnel. A better Long Island route would follow the Main Line to the end and tunnel near Greenport, trading deeper waters for shorter tunneling and a route length comparable to that of the optimal Shore/I-95 alignment, so it could achieve a comparable trip time. But even that’s unneeded: it’s 15 km of deep tunneling, whereas if one is willing to slightly compromise on trip times, the only Connecticut tunneling required for a Shore Line fix is 3 km in Bridgeport.
The other problem is what to do about commuter service. The Providence Line’s traffic level is low enough and its average interstation is long, allowing a blended plan. Shared tracks between New Rochelle and Penn Station would see more commuter traffic, but intercity trains would go slower anyway, and there is more room for four-tracking. The Ronkonkoma Branch’s 10-minute peak service requires at least one overtake between Hicksville and Ronkonkoma and probably two, in addition to four-tracking the Main Line; this is feasible, but less than optimal, and the overtakes would have to be constructed in more constrained locations than those available on the Providence Line. East of Ronkonkoma commuter service may need to be cut, but this is less of a problem on account of its low traffic. On the other hand, the Main Line west of Hicksville is not a problem with four tracks, and neither is the New Haven Line – express commuter trains could weave in and out.
On the benefits side, offering Long Island service to Boston that doesn’t go through New York is better than not doing so. However, the difference in benefits with New Haven, while positive, is smaller than it seems. The New Haven Line has almost as much ridership as the LIRR Main Line, and Stamford is a bigger edge city than Mineola and Garden City. On top of that, since the optimal LIRR option connects to the Shore Line in the far east of Connecticut, there is no hope for service to Hartford except on legacy track. On balance, the advantage of the LIRR option is just service to Jamaica, a larger draw than those smaller cities and suburbs, but there the time saving is the smallest.
On top of that, does such a small benefit really justify the cost? Having some high-speed trains run through to Jamaica, Mineola, Hicksville, and Ronkonkoma at lower speed requires re-electrifying the LIRR with catenary, which is a fraction of the cost of all those urban grade separations and 1-2 order of magnitude cheaper than an undersea tunnel and land connections. On a similar note, since half an undersea tunnel is of no use, it’s harder to break construction into small chunks if it is necessary, putting it at a disadvantage against a route consisting of cutoffs and modifications of the existing line. The route of 1834 may work now that we can build tunnels, but the cost structure favors that of 1846 and 1852.
Pedestrian Observations 
