Traditionally, proposals for new mainline rail tunnels across the Hudson enter Penn Station’s southern tracks, which are used by New Jersey Transit. This includes the ARC incarnations that connected to Penn Station’s preexisting tracks, Amtrak’s Gateway, and the Institute for Rational Urban Mobility’s through-running counter plan to ARC. To my knowledge, ARC-North proposals, entering the northern tracks used by the LIRR, have not been investigated. I am not ready to sign off on ARC-North yet and in fact called it a troll proposal, but I believe it deserves more study and is probably superior to southern proposals except at high levels of investment.
While this in principle only concerns the configuration of the tunnels across the Hudson, it has implications about the configuration of Penn itself and the service plan supported by the infrastructure. The one element of the various Penn Station redesign ideas that isn’t relevant is the look of the station itself; this is also the least important element for passenger throughput. Penn Station’s failure to look like a cathedral is a lesser problem in a city that is full of multi-billion dollar starchitect-designed ugliness.
First, some principles for tunnel design. A good infrastructure proposal should have the following elements:
1. A through-running service plan. It matters which line at the New Jersey (or Hudson Line) end connects to which line at the Long Island or Connecticut end, since this influences the interlockings. It also matters where people are expected to transfer, since this influences platform crowding; ideally, transfers should be handled at Secaucus and Sunnyside.
2. Simple interlockings. Complex interlockings limit train speed, and switches especially do. It should be possible for trains to enter and exit Penn Station at speed. Avoiding slow zones in station throats is an underrated way of improving line haul time at relatively low cost, and the fancier the trains and more upgraded the tracks elsewhere are, the heavier the time penalty of slow throats is.
3. Adequate platforms. Penn Station has less bad platforms than people think – when I timed rush-hour LIRR trains, they emptied in about 90 seconds or a little more – but they’re still not good, especially if we’re assuming large increases in ridership coming from better service. The LIRR has better platforms than New Jersey Transit because it has more access points per platform, but the platforms are still narrow. In the worst case, a plan should consider paving over some tracks to widen the platforms, since Penn’s 21 tracks are more than enough for its traffic.
4. Adequate speed for intercity trains. The current tunnels are limited to 60 miles per hour (97 km/h) because of air resistance generated by non-aerodynamic trains in narrow tunnels. The cross-sectional area of the trains going through the tunnels is about half the cross-sectional area of the tunnels; in new high-speed rail tunnels, the corresponding ratio, called the obstruction ratio, is about 15%. It is possible to squeeze more speed out of the existing tunnels with better aerodynamics and sealing, but 200 km/h is probably impossible, and even slightly lower speeds are a problem if the tunnels are very busy and there is a speed mismatch with unsealed, non-aerodynamic commuter trains. So ideally, the larger-diameter new tunnels should be used by intercity trains. We can plan around one minute of travel time difference; this figure can in reality be anywhere between zero and 2.5 minutes.
5. Separation between intercity trains and Grand Central trains. If there is a Penn-Grand Central tunnel, it should be used exclusively by commuter trains because of the high local travel demand, and because to reduce real estate acquisition cost the curve radius should be low, possibly too low for Shinkansen equipment. This means that if a Hudson tunnel points toward the Penn-Grand Central tunnel, intercity trains should use the other one.
6. Two platform tracks per tunnel track. This improves capacity in two ways. First, rapid transit capacity is a combination of tunnel capacity and station dwell, and splitting each tunnel track between two platform tracks allows slightly higher capacity by deemphasizing the dwell since successive trains use different platform tracks. Paris is limited to 30 tph on the RER A with moving-block signaling, whose central segment has one platform track per tunnel track, but the shared RER B+D tunnel between Gare du Nord and Chatelet-Les Halles gets 32 tph with fixed blocks, where the B and D serve separate platforms at each station. And second, because each train can dwell at the platform for longer, this reduces the need for wider platforms, allowing violations of #3. Ideally, the two platform tracks would face the same platform to improve wayfinding and allow unscheduled track changes in case of train delays.
Now, Penn’s tracks are numbered 1-21, from south to north. There are platforms between tracks 20-21, 18-19, 17-18, 15-16, 13-14, 11-12, 9-10, 7-8, 5-6, 3-4, and 1-2; the 18-19 platform is wider than the rest, as if there was supposed to be another track immediately north of 18. The Hudson tunnels connect to tracks 1-19, the southern pair of the East River tunnels connects to 5-15, the northern pair connects to 15-21. The Hudson tunnels and the southern pair face each other, and trains on tracks 11 and 12 go straight through every switch; the northern pair offers no switch-free option, since the eastbound track faces track 21 and the westbound curves into the interlocking.
Since the Hudson tunnels face the southern pair, a simple proposal for new Hudson tunnels should face the northern pair. This would give 4 tracks between New Jersey and Sunnyside. Each of the two track pairs could point toward either Long Island or Connecticut, because of the Harold Interlocking, flawed as it is. This means intercity trains would use the northern pair and go to Connecticut, other regional trains would use either pair and go to either Connecticut or Long Island, and Harold would be superfluous. New Penn-Grand Central tunnels could be constructed branching from the southern East River tunnel pair. It is possible to also construct the tunnels around the southern pair, with trains from the southern pair either merging heading into the Hudson tunnels or terminating and reversing at Penn Station.
Under this plan, all of the numbered principles could be satisfied except #3, and to satisfy #3 every plan requires track paving or other platform modification. It’s also simple to construct: it’s just a new tunnel pair, and ideally also some work on the preexisting southern East River tunnel pair to construct a connection to Grand Central. The one drawback is that, unlike in ARC Alt G, Gateway, and the IRUM plan, the Penn-Grand Central tunnel shares approach tracks to Penn Station with the southern tunnels, reducing capacity. This becomes a problem if ridership from Grand Central, the LIRR, and points east grows to the point of overwhelming three tunnel pairs (the two heading into Penn, and East Side Access). The alternative with the Penn-Grand Central tunnels going around the southern tunnel pair and only merging heading into the Hudson tunnels has more capacity, but interferes with principles #3 and #6.
Track-paving for any plan is hard, because many of the straightest, longest-platform tracks have to be removed. To widen the platforms and improve throughput, there are two ways to pave over tracks, each of which gets rid of about half the station tracks. One is to pave over every other track, guaranteeing each track access to two medium-width platforms; trains could open both doors then, improving egress. The other is to pave every other pair of adjacent tracks lying between platforms, giving each track access to just one very wide platform. The former option is difficult at Penn because there are support columns between adjacent tracks, and a look at New Jersey Transit rolling stock suggests that each train would have 1-2 doors facing a column. We are left with the latter option, paving over, for example, tracks 20, 17, 16, 13, 12, 9, and 8. Each tunnel track would get two platform tracks facing the same platform, except for the westbound northern tunnel track, whose two tracks (21 and 19) would be split because of the aforementioned columns. The lowest-numbered tracks would not be used; the LIRR’s West Side Yard would not be used regularly but instead trains would run on the tracks more often off-peak. The now enlarged platform between tracks 7 and 10 would be lengthened to allow 16-car trains at track 7, which is currently 13 cars long.
Since principles #3 and #6 are both satisfied, the capacity per track can be quite high. The RER B+D achieves 32 tph, but this is split as 20 tph B and 12 tph D, which is suboptimal since two successive trains could both be B, making the dwell a problem. With even alternation between each platform’s two tracks and moving-block signaling, even higher capacity may be possible, reducing the capacity disadvantage of having just two tunnels coming into Penn from the east rather than three.
The IRUM plan, building on ARC Alt G, is quire different. It has three tunnels in each direction: the northern East River tunnels are paired with the Hudson Line via a new short tunnel linking the Empire Connection with the northern Penn Station tracks, the new Hudson tunnels come into the southern station tracks and then continue to Grand Central via new tunnels, and no tracks are paved. Intercity trains have to keep using the old tunnels, necessitating Harold, but that’s money already spent. Principle #3 is violated, or alternatively #6 is if tracks are paved, but there is more track capacity. The risk there is insufficient demand from the Empire Connection, which would leave the northern tunnels underused, reducing the system’s capacity advantage over a two-tunnel option to just a few tph. This plan has to route intercity trains through the old tunnels. Conversely, the advantage is that it easily shoehorns Empire Connection service, which ARC-North does not.
The main difference between ARC-North and IRUM is investment levels. ARC-North is cheaper; it’s more comparable to the bare-bones proposals for ARC(-South) and Gateway that are hinted at but never formally published. At the low investment levels of a bare-bones proposal, ARC-North is superior because it provides better capacity because of the relatively straight train paths through the station. If there’s no connection to Grand Central, then it is not at a capacity disadvantage since Long Island and Connecticut don’t have the demand, and barely have the capacity, to overwhelm three track pairs (including East Side Access again).
Conversely, ARC-North is harder to retrofit for somewhat higher investment. The best that can be done is tunneling to connect the Empire Connection to the southern tracks, digging new tunnels to Long Island as in Alt S, and running trains to just one line, probably Lower Montauk, which goes through serviceable neighborhoods but lacks a direct connection to the existing tunnels to Penn. This provides much better service coverage because of the Montauk connection, but at higher cost since there’s an additional underwater tunnel. It avoids the expensive components of Gateway and ARC, but so does IRUM, which becomes an intermediate level of investment between ARC-North proper and ARC-North with a Montauk tunnel.