Why You Need Through Running

Through-running usually comes up in the context of New York (ARC vs Alt-G; NJ Transit, LIRR, & potentially MNR all turning back at PSNY; East Side Access not connecting to existing Grand Central) and sometimes Boston (no connection between South Station & North Station).

Well guess what you East Coast pikers, there is a city about to rectify its dead-end downtown terminal – Los Angeles. As is often the case in LA, the plan involves a lot of concrete and aesthetics that would make other cities blanche.

Let’s leave aside the question of whether current Metrolink and Amtrak volumes warrant the run-through tracks at LAUS, and review that particular plan in a future post. For now, let’s explore why through-running is important. This is something that people who know railroad operations would take as self-evident. But most people, whose frame of reference is auto traffic or bus operations, might not understand what dead-end operations mean for a railroad.

Capacity Curves

First, let’s borrow a basic graphic from freeway engineering. This plots traffic speed (x-axis) against traffic volume (y-axis).

On the right end, we have high speeds with little traffic. As you add more cars, increasing traffic flow,  the speeds get lower, which is what you’d expect. Then at some point you reach a maximum possible throughput. Add more cars to the freeway, and traffic volume actually starts to go down. This is a little counterintuitive for some people – when the freeway is totally full of cars, the traffic volume is very low. It makes sense if you think it through, though – if the freeway is totally gridlocked and no one is moving, the volume of cars passing a fixed point is zero because, well, no one is moving.

That curve is the reason every freeway on-ramp in Los Angeles has ramp metering. By regulating the flow of cars onto the freeway, you hope to stay on the right hand side of the curve, maximizing throughput. Even if average volume is below maximum throughput, a small burst of cars entering all at the same time can push you onto the left side of the curve. Once you go over the top onto the left side of the curve, you’re doomed to have congestion and unstable traffic flow, which won’t clear until after rush hour when the volume of cars trying to enter the freeway drops way off. (If your state DOT doesn’t have ramp meters and they want to widen freeways, well. . .)

Traffic

Ok, back to trains. Some basic about rail operations. Trains move proceed along the tracks as directed by the signal system. Since we don’t ever want to have opposing trains heading towards each other on the same track, the signal system is set up to enforce the direction of traffic between interlockings. Interlockings are locations on the track where switches allow trains to move from one track to another.

So, in the graphic above, there is a train moving east between B and C. Traffic on this section is to the east. There’s a train waiting to go westbound. The dispatcher can request a route for the westbound train, but that route won’t clear (be given a green signal) until the eastbound train clears the interlocking at C and the signal system checks that there are no other trains between B and C, and no conflicting routes requested.

Even in this simple case, the dispatcher has to make a decision. When the route is cleared for a westbound train, the following eastbound train (between A and B) will have to wait at B. If there are a lot of eastbound trains, it’s possible that they’ll start to get backed up at B. And just like that, you’re on the wrong side of the capacity curve, and all your trains are delayed. If your station throats are operating near maximum throughput – like, say, the Hudson River Tunnels entering Penn Station – it doesn’t take much of a delay to make that happen.

Smooth Operator

That’s a simple case with an obvious capacity constraint in the single-track section. Now, let’s consider a more complex case. Suppose that we have a terminal station like Penn Station New York, where the tracks run through, and that the service is planned with through-running.

This is a really easy system to dispatch. Eastbound and westbound trains never conflict with each other. The dispatcher never has to change the direction of traffic. We could even set up the signal system entering the station to automatically throw the switch to whichever platform track is clear. All the dispatcher has to do is set routes for trains leaving the station, in the proper scheduled order.

Rough Sailing

Now take the same station and have service from each side operated by different agencies, with no through-running; each agency treats the station like a stub-ended terminal.

Consider the red trains entering eastbound. Should the dispatcher send the westbound train out in front of the first eastbound train, hoping it doesn’t cause delays to tumble back? Or should the dispatcher wait for the first eastbound train to enter, and then send the westbound train? What if the westbound train is running a minute or two late? The platform needs to be made available for following trains to enter the station. Is it better to delay that one westbound train a minute or two than to risk delays to eastbound trains? Even in this simple terminal, the dispatcher faces tough decisions, and if train volumes are high, they must be made every few minutes.

Consider a more complex stub-ended terminal like LA Union Station.

The dispatcher faces many difficult choices. In addition to deciding which trains to prioritize, the dispatcher must decide the track on which to berth each train. Poor decisions can come back to haunt you. For example, suppose a train is entering from the Ventura Line and is berthed on Track 13. If that train’s next trip is to the Antelope Valley Line, it will have to cross the entire terminal interlocking on its way out. During that time, no other trains will be able to enter or leave the station. However, if that train is heading out to the Orange County Line, putting it on Track 2 when it comes in might be a good decision, because it won’t interfere with any other moves on the way out!

The situation gets worse if, in addition to the two stub-ended operations, we have some through trains, like Amtrak service at Penn Station. Now, the choice to delay the westbound red train could cause delays to a westbound through train, which could tumble back to the operator of the blue trains to and from the east.

If the services are all being operated by different agencies, it’s inevitable that one service is going to come out on the top of the heap – probably, whichever service is provided by the operator that dispatches the terminal. That employee has an incentive to try to keep their trains on time, meaning they might dispatch their trains first even if the overall result is more delay on the network. Consider, for example, South Station in Boston, where Amtrak has dispatch rights. How much is Amtrak going to worry about the on-time performance of the Worcester Line if its trains are late?

Of course, all of this can be ameliorated by good scheduling, which includes a plan for assigning trains to the right track. But even in that case, service disruptions are harder to deal with and have a greater chance of getting out of control.

Don’t Dwell On It

This analysis hasn’t even touched on another issue, which is the platform dwell time required to switch train directions. In the US, this is generally a minimum of 10 minutes, enough time for passengers to alight, the engineer to switch ends and do the FRA-required brake test, and new passengers to board.

I’ve heard that in other countries, they turn trains faster. Even so, it’s obviously more efficient if the engineer doesn’t have to switch ends of the train. You can avoid that if you have a new crew waiting to board the train at the station, and in fact, this is how LACMTA runs things at 7^th/Flower on the Blue and Expo Lines. The downside is increased labor costs, because you always have at least one crew just waiting around the terminal.

Through-running eliminates these problems and reduces dwell time, which increases the capacity of the terminal. This is one of the reasons why a place like Shinjuku Station can handle 750,000 passengers per day on 16 platform tracks while places like Boston South Station are planning expansions because they’re having trouble handling 45,000 passengers (6% of the ridership) on 13 platform tracks. (After glancing at track maps of Shinjuku, I also suspect that they keep things running smoothly by blocking off the terminal into operating segments, rather than sending trains all the way across a massive terminal interlocking like US operators are wont to do, but more on that another time.)

Run Through If You Can

So that’s a simple summary of why you need through-running. US operations don’t need more tracks; they need better operations. As Alon Levy says, organization before electronics before concrete.