Lucius Kwok wrote:What if SEPTA were to revise its line pairing system based on current ridership levels?
It wouldn't help SEPTA operate more efficiently.
Vuchic's tunnel operating plan (in which the line pairings and R numbers were introduced) assumed growth in ridership that didn't happen (remember that at the time this was written, gas rationing was a recent memory and another oil embargo was a significant threat--in that era, DVARP also was proposing plans based on a assumption there would be a significant shift from cars to rail). And the more I think about it, the more I question the basic premise of using ridership to determine the pairings.
If you assume that the system will run regular headways in the off-peak, then there's little reason not to pair. You want to run as many through trains between PRR and Reading points as you can, so as to minimize redundant operations through the tunnel, and if everything's running hourly, the trains are going to keep going to the same places each time.
But the premise of Vuchic's pairing was to reduce car miles--it assumed that Paoli and Lansdale trains would be longer than Norristown and Trenton trains. This is generally not the case during off-peak periods--SEPTA runs two-car trains almost everywhere. Furthermore (and this may be something Vuchic didn't anticipate) SEPTA crews have no compunction about closing off cars of a longer train, and many trains are changing crews enroute, so there's no crew savings from matching up consist lengths--only power and per-mile maintenance costs, which are relatively small compared to crew costs. Therefore, there's little or no savings from pairing by ridership.
As Mack alluded to, we did look at this issue in the DVARP newsletter (maybe someone could look through back issues and tell me when), and DVARP has long recommended that SEPTA revisit the whole issue and either re-pair the lines or drop the clockface headways for increased efficiency.
Line length in miles isn't the crucial factor as much as cycle time--the minimum time it would take to run out from Suburban to the terminal station, lay over, and come back. Since cycle times for paired lines have to be an even multiple of hours (excepting Paoli and Airport which can be even half-hours), if your minimum cycle time is one minute more than an even hour, you have to add 59 more minutes of layover time at one end or the other to maintain the clockface headway. That wastes time and money.
We called some of the most egregious examples to SEPTA's attention, and they responded by tightening up the Warminster schedule and cutting out a half hour of layover time there. But there's still a long way to go, and we maintain that the entire operating plan needs to be revisited, both to increase efficiency and to reduce the confusion stemming from the R numbers (e.g. passengers getting on the Fox Chase R8 when they want to go to Chestnut Hill West). SEPTA runs a much higher ratio of off-peak to peak trains than any other commuter railroad, but their operating plan was largely driven by peak operations considerations.
Now in the peak, slavish adherence to the Vuchic line pairings hurts SEPTA's efficiency. I agree that matching trains for consist length is the way to go, but you should do that on a train-by-train basis rather than line-by-line. If you've got two five-car trains coming in from West Trenton, and two three-car trains from Norristown, you're better off running the Norristown trains through to Media and Wilmington and yarding both West Trenton trains, rather than running one of the West Trenton trains to Media and yarding the other, and running one of the Norristown trains to Wilmington and yarding the other. SEPTA has taken some small steps in this direction, but there's still savings to be had.