Friday, November 6, 2009

Interregional High Speed Rail: which corridors work where

A recent study (PDF) from a group called America 2050 has put together one of the most data-heavy (and that's a good thing) approaches to examining high speed rail corridors in the country. There are still some issues, most notably the fact that corridors over 500 miles were ignored (yes, they should be weighted less than 200-400 mile corridors, but, no, with proper speeds attained, they shouldn't be dropped) and their map does not seem to fully mesh with their data. Still, they take in to account such factors as transit accessibility in cities analyzed, economic productivity (higher local GDP is better), traffic and air congestion and whether the city is in a megaregion (this seems to be a rather ancillary data point).

Their subsequent phasing map, while better than most, seems to be, well, not completely in-line with their data. This is mainly because each corridor seems to be analyzed separately, and overlapping corridors, from their report, are not shown well.

First, they did get the two big corridors right (the "no-brainers," if you will): California and the Northeast Corridor. Both of these corridors have multiple city pairs in the top-10 of their analysis; in California the San Francisco-San Jose-Los Angeles-San Diego line and in the northeast the Boston-New York-Philadelphia-Baltimore-Washington corridor. Of course, those are obviously the top high speed rail corridors in the country. However, the rest of their "first phase" corridors are less obvious.

In an effort to, perhaps, not leave out the Midwest (where much of the current political support for high speed rail originates), they include, in phase 1, lines from a Chicago hub to Minneapolis, Saint Louis and Detroit. These are all worthy corridors but, according to their analysis, are not in the same echelon as the coastal corridors. Chicago to Saint Louis clocks in at 14th, trailing Chicago to Columbus by a spot. Chicago to Minneapolis ranks 25th, behind corridors such as Cleveland to Washington and Phoenix to San Diego.

With Chicago to Detroit (11th), however, things get interesting. Let's introduce two maps in to the equation. The first is a map of the top 50 corridors analyzed by America 2050, with the color of a line indicating if they were in the top 50 (red), 40 (orange), 30 (green), 20 (light blue) or 10 (dark blue). Opacity is set rather low, so overlapping lines should show up considerably darker (see the Northeast Corridor, where four top-ten corridors intersect from New York to Philly). From Chicago to Minneapolis and Saint Louis, there are single lines. Despite the presence of some smaller cities (Decatur, Springfield, Urbana-Champaign; Milwaukee, Madison, Rochester) none of these corridors crack the top 50. (Milwaukee-Chicago was not calculated as it is less than 100 miles.) East of Chicago, however, there is a web of lines. From Chicago going east, three cities make the top 16: Detroit, Cleveland and Columbus. And east of there, these cities are all linked eastwards. (Any city with at least two corridors is shown with a point, its size corresponding to the number of corridors.)

So it begs the question: which routes are most applicable to high speed rail if we overlap corridors which could share significant trackage. For instance, Chicago to Detroit, Cleveland and Columbus could all share one high speed link, with short spurs to each of the cities. These three cities could all share a link across Pennsylvania (with Pittsburgh) to Washington, Philadelphia and New York. 11 of the top 50 city pairs are between New York, Philadelphia and Washington in the east and Columbus, Cleveland and Detroit in the west. Since most of the capital costs of constructing a high speed rail line is the initial capital cost, combining several corridors could dramatically reduce the amount of line needed, saving billions.

So, the second map. For this map, lines with little or no overlap were ignored. Other corridors were assigned a (rather arbitrary) point value based on their ranking:

1-10: 6 points
11-20: 4 points
21-30: 3 points
31-40: 2 points
41-50: 1 point

(Why did the top 10 get a slightly higher weight than the rest? Well, the numerical rankings of the top 10 ranged from 100 to 91. The rankings of the next 40 ranged from 91 to 85.)

Here's another scheme: assign a route with a score of 85 one point, and an additional point for each increase in the score. This is, perhaps, a more equitable approach for larger corridors, and it really pops out the Northeast Corridor. A possible network of 2450 miles (1870 in the East and Midwest, 580 in California) could serve Boston, New York, Philly, DC, Pittsburgh, Columbus, Cleveland, Detroit, Chicago, San Diego, LA, San Jose and San Francisco (and several smaller cities, like Toledo, Harrisburg and Hartford). Adding up only the top 50 MSAs served (those with populations over 1m) and 2500 miles would serve 90m people. That's not bad.

So, what's the takeaway here? Well, there are two. The first is that, as much as we want to build a multi-regional high-speed rail network, the Northeast Corridor is still, by far, the largest market for HSR in the country. The second, however, is that even when you exclude the Chicago-to-East Coast routes, the New York-to-Chicago Corridor should still be the third-highest priority to build. And if properly built (with top speeds of 200 mph or a tad more, especially across the flat land west of Canton) such a corridor could begin to compete with airlines, even on >500 mile routes.


  1. It's an interesting concept... but your way of implementing is a bit weird. If you have just one Chicago-to-East corridor, then trains can't serve multiple destinations. They run New York-Chicago, or Chicago-Cleveland, or New York-Cleveland. This is bad: New York and Chicago are too far apart from HSR between them to make sense without considering intermediate points, and the other city pairs involve smaller cities, which means frequency to each city pair will be lower, reducing ridership.

    Remember how on the California HSR environmental impact statement, they explain how excessive splitting, for example Altamont with a San Jose/San Francisco/Oakland split, reduces ridership? It's the same here.

    This method of construction won't even reduce costs. From Chicago to Cleveland there are already multiple straight railroad rights of way, which just need an extra pair of tracks built to HSR specs, plus some grade separations. From Cleveland to New Pittsburgh and Philly you'd need new ROW, possibly on Interstates, but your single East-to-Chicago corridor would require new ROW the entire way.

  2. @Alon,

    I'm not sure I've fully addressed this, but I'd contend that if you can do New York to Chicago in four hours (possible) or even four and a half, it's competitive. This is mainly because the airports in New York and Chicago are pretty much the most congested in the country, and because getting to them from New York and Chicago, the two biggest CBDs in the country, is atrocious.

    The problem with trying to serve every intermediate market is that every time you zig and zag, you hurt a larger market. If you go from Detroit to New York via Pittsburgh, you add twenty minutes or more of run time, which deters the Detroit to New York market. Same thing of going from Pittsburgh to Chicago and serving Cleveland, Toledo, Fort Wayne and South Bend.

    Then there's construction. Yes, in the midwest, especially between about Chicago and Canton, you could get away with building separate corridors along existing ROWs and farmland (~$20m/mile). However, Pennsylvania will cost three to five times that much with tunnels and bridges. Consolidating everything in to one corridor in Pennsylvania makes sense.

    As far as serving multiple markets … one idea is to have a transfer station just west of Harrisburg. With proper scheduling, three trains could arrive at the same time on consecutive platforms like rush hour at Jamaica on the LIRR. Trains from New York, Philly and Washington to Cleveland via Pittsburgh, Toledo and Detroit, and Chicago. Everyone makes the transfer, and the trains depart one after the other. This could probably be done in five minutes, far less than transfers at airports. You do eliminate the one-seat ride, but by switching the destinations, people could get one-seat rides if/when they wanted, but also have the flexibility to dependably get where they wanted to go.

    One more thing: as fuel prices rise, railroads will have a distinct cost advantage over planes on routes under 1000 miles, even if the time is slightly longer (as long as you keep the on-time performance at 95% or higher). If a two hour flight from LaGuardia to O'Hare is $500 with potential weather delays, and a four and a half hour train ride from Penn Station to Union Station is $200, rail takes a significant portion of the market.

    I have a Google Doc which ballparks costs for a Chicago-New York line; I should distill that and post it one of these days.