Intercity trains usually only call at the main cities, and only at their main stations. However, in some cases, they make more than one stop in the same city. Examples include the Northeast Corridor with its two Boston stops, the Shinkansen stopping at Shinagawa and Ueno each a few kilometers out of Tokyo Station, Israel Railways with every train stopping multiple times in Tel Aviv, and ICEs stopping not just at Berlin Hauptbahnhof but also a station in each of the cardinal directions out of the city. This practice is a useful tool in the kit of an intercity rail planner, but is not always appropriate – indeed, anti-examples exist throughout France where this never happens and in most German cities that are not Berlin. The goal of this post is to discuss when it is or isn’t appropriate.

Opposite-side stops

If a city is at the end of a line, then the train may continue shortly past it and serve a secondary center. The main examples are in Berlin: it is at the northeastern end of the country, so trains running north-south either run through northwest to Hamburg or terminate just north of Hauptbahnhof at Gesundbrunnen, and trains running east-west entering Hauptbahnhof on the Stadtbahn run through to Ostbahnhof. This is not done in the other examples I have in mind: Israeli trains run through Tel Aviv, Shinkansen terminate at Tokyo Station rather than continuing to the opposite side, and the Northeast Corridor terminates at Washington Union Station rather than continuing one stop south to Alexandria.

In all cases, the question of whether to do it hinges on infrastructure more than operations. Unlike near-side stops, nobody is being delayed by the extra stop – Ostbahnhof isn’t between Hauptbahnhof and anything except Poland, which is barely even peripheral to the German rail network. The main cost is the extra few minutes of train operating time, on which ridership is necessarily going to be lower than the average because most people will have gotten off at the primary station. Whether it’s worth it depends on where it’s easier to turn, based on specific infrastructure considerations.

In Berlin, the reason for Ostbahnhof, a station with no connections to any public transport except the Stadtbahn, which also serves Hauptbahnhof, is infrastructural. The station has one more track than Hauptbahnhof (five vs. four on the mainline Stadtbahn tracks); more importantly, the railyard is to the east of the station, so any ICE using the Stadtbahn has to travel via Ostbahnhof anyway, so it might as well stop there. The railyard is also why some east-west trains even use the Stadtbahn with its 60 km/h speed limits to begin with, when those trains entering from the west can divert to the faster North-South Main Line.

In contrast, the Northeast Corridor’s electrification ends at Washington Union Station. In theory, it can be wired to Virginia and some trains can continue to Arlington while still running under wire. In practice, other works are involved – Washington Union Station is a mix of through- and terminating tracks and so far the through-tracks are low-platform, and the only bridge on the Potomac is primarily used by freight and a separate bridge to allow for passenger-dedicated infrastructure is under construction but won’t open until 2030.

Distance from the primary station

The farther away a node is from the primary station, the more appropriate it is as a prospect for a near-side station, because it is less likely to share a travel market with the primary station, and the cost to travelers of backtracking is higher. In a number of cases, it’s debatable whether the station should even properly count as a secondary station in the city, as opposed to a separate city’s stop. In New York, for example, the nearest station to Penn Station is Newark, 16 km away, in a city center that evolved separately and only became part of the New York metro area in the 20th century, long after its location as a station was established.

I emphasize that this logic is only for near-side stations, because for far-side stations, the logic goes in the exact opposite direction: the train is necessarily emptier going past the last primary city on the route, so the longer it has to run relatively empty, the weaker the idea of serving the far-side stop is. Not for nothing, Ostbahnhof and Gesundbrunnen are both very close to Hauptbahnhof, both well within the city center urban fabric.

To return to the case of New York, this is what separates Newark from Long Island City. There is currently no serviceable station in Queens on the Northeast Corridor, but there are occasional plans to build on, either at Queens Boulevard or a bit farther east at what I (and people following me) call Sunnyside Junction to allow for a transfer to East Side Access trains to Grand Central (the map below calls it Harold after the interlocking). MTA plans have called for the former, and occasionally one sees advocates call for the latter (ETA calls it Queens Junction), with official plans for decking Sunnyside Yards that may be connected.

Newark Penn Station is a smaller job center than Long Island City. Within 1 km of Newark Penn are 48,000 jobs, vs. 70,000 within 1 km of the intersection of the NEC/LIRR Main Line with Queens Boulevard (and 35,000 of the plausible Sunnyside Junction location). But any project that builds commuter rail stations in Long Island City automatically makes the area easy to get to from Penn Station, whereas Newark Penn is a longer distance from New York, forcing a longer detour for passengers going to points south.

Track speeds

The main cost of a station stop is the extra time it forces on through-passengers. This means that it’s more justifiable to include more stops in a slow zone than in a fast zone.

In practice, this principle works in the opposite direction from that of distance to the primary station, since slow zones are more commonly encountered in major cities than outside of them. The best location based on synthesizing this and the distance principle is that stations are best placed near slow zones that are nonetheless far from the primary city and yet very difficult to fix. Newark Penn Station is at such a slow zone, the approach from the Dock Bridge to the station having some surprisingly tight curves, and the Harrison curve farther north even tighter. New Rochelle is another such example, next to a tight S-curve with flat junctions; this curve can be eased with a necessary grade separation, but the current maximum track speed there is so low, oscillating between 30 and 45 mph (48-72 km/h) based on track maintenance, that it’s understandable that Amtrak Regionals stop there. Back Bay, in an unfixable slow zone, is likewise a natural candidate for a station despite its proximity to South Station.

On urban through-tracks, the maximum speed is usually low, because the costs of engineering the right-of-way through an already crowded city center underground are high and economies are required. The curves on the Ayalon Railway look good for maybe 100 km/h, Stuttgart 21 is designed for 80-90 km/h and so was the North-South Main Line in Berlin, and the Shinkansen doesn’t go faster than 100-110 km/h through Central Tokyo or faster than 130 km/h to Omiya. Such environments encourage secondary stops, whereas the ability of LGVs to get out of city centers fast helps explain why it was never considered.

Of course, TGVs have another reason not to make secondary stops: they’re not set up for them. The TGVs are the worst high-speed trains in the world from the perspective of either acceleration, since they are hauled by locomotives whereas every other piece of equipment save some Talgos is a multiple unit, or interior car design, since they are single-door pair double-deckers with narrow passageways, not at all designed for fast egress since the concept has always been nonstop trips from a Parisian terminal to a secondary city. Regardless of the speed, a TGV making a second Paris stop would have a minimum 5-minute dwell time. At the other end, the Shinkansen and the Israeli trains have level boarding (as do Chinese high-speed trains but they tend to avoid slow city centers).

Relative strength of destinations

Stop spacing on every public transport vehicle is a matter of relative intensities of usage. Absolute density or intensity matters little; the real question is the intensity of the stop considered for addition or deletion compared with the intensity of the stop pairs through the stop that would be delayed if it were added. On intercity rail, the intensity of demand is a combination of all of the following:

• Commercial density near the station
• Density of other destinations near the station, such as tourist attractions, which can be approximated by looking at hotel capacity
• Connecting urban rail lines and the residential density thereon

Back Bay has, within 1 km of the station, 95,000 jobs, compared with 212,000 within 1 km of South Station. This ratio, 0.45, is atypically high for any American secondary central business district, to the point that people perceive Back Bay as a Boston central business district and not as secondary at times. It also is on the Orange Line, which doesn’t serve South Station, and close to the Green Line, which doesn’t get to South Station either, both with solid ridership (the Green Line is comparable to South Station’s Red Line, the Orange somewhat lower than either).

In New York, in contrast, Manhattan overwhelms everything. The 1 km radius from Penn Station if anything undersells it – it has 565,000 jobs, but Midtown continues beyond it and the 2 km radius has 1,415,000, whereas a 2 km radius from Newak Penn has 83,000 and a 2 km radius from Sunnyside Junction or LIRR/Queens Boulevard has 117,000-122,000. Secondary stations must be justified as origins rather than destinations: Newark Penn keeps up by collecting some commuter rail lines, but suffers since the Morris and Essex Lines only go to New York, not Newark Penn, and holds up mostly because of the slow nearby speeds.

No data as granular as OnTheMap exists in Germany, but one can still model connecting lines and their traffic intensity. The Ring is a high-traffic local corridor, especially its eastern half, including both Gesundbrunnen and Südkreuz; at both stations, its passenger traffic density is barely less than that of the Stadtbahn at Hauptbahnhof. Both also have the North-South Tunnel, which intersects the Stadtbahn at Friedrichstraße rather than Hauptbahnhof, and Gesundbrunnen also has the U8 connection. From quite a lot of the city, even relatively central parts, one of the two north-south secondaries is better than the primary. The east-west situation in Berlin is dicier, since Spandau is much more peripheral, more comparable to Newark than a secondary station within the primary city, and Ostbahnhof really exists as a byproduct of where the railyard is.

One should expect that European cities, with their lower-kurtosis centers, should be better places for secondary stations than American ones, which have very high kurtosis, even outside New York. Other than Back Bay, there really aren’t any strong secondary city centers on American intercity rail lines. California High-Speed Rail’s secondaries, like Burbank, are there for residential catchment, often with easier access by car than the central business district. And yet, one element goes in the opposite direction: larger metro areas tend to have more secondary destinations. Munich is small enough that it can monocentric within walking distance or a short S-Bahn ride from Hauptbahnhof. The sort of American cities that need to think about intercity rail are not; this is how Boston has such a strong secondary, and even New York has decent secondaries immediately outside city limits. Tokyo, with a very strong center but lower kurtosis than New York – high job density is not just in areas near Tokyo Station but also includes Shinagawa, Shinjuku, Shibuya, and Ikebukuro – is a natural place for secondaries, and if there were a Shinkansen line heading west rather than north or south, it goes without saying that it would stop at Shinjuku.