Green Line Reconfiguration

Re: Park Street loop and system capacity

I take everything Alon Levy writes with a grain of salt. He's not wrong but he thinks only in pure numbers and not how people actually want to ride transit. Everytime a rider needs to transfer their desire to use the system drops. The psychology of a one seat ride will always win out over a faster trip with transfers. Besides, both NYC and the Green Line were designed for interlining and it would be more expensive to redesign the systems to eliminate it than just dealing with the consequences.

Broadly speaking, I think we're now discussing two different arguments that Alon has made:

First, the argument that where interlining ends up happening, using the same headway on all services is likely optimal; this is where I think I agree with him.

Then there's the discussion of disentangling the New York City interlining of 13 services on most of the lettered lines. I was under the impression that Alon had come to the conclusion that the construction costs of actually separating things out would probably not be worthwhile because so many of the stations that would need to become major transfer points if the interlining were reduced have small platforms because it was assumed that interlining would mean people wouldn't transfer; the disagreements around whether rider reaction to transferring matters don't seem like they're likely to make a major difference for any actual real world service pattern decisions.

However, I also thought I'd seen something somewhere claiming that a lot of New York City subway riders gravitate towards trains with ``express'' in their name, even when the number of local stops that will be skipped is so small that transferring or waiting extra for the express won't really get the passenger to the destination any faster.
 
Per demand, I'm willing to bet that Park St will still be the #1 transfer spot so that any train looping will mean that passengers will still need to transfer to get to Park. This obviously doesn't hold true for games at the Garden but normal service will probably by 100% through.

But for the end of an event at the Garden, you probably want every southbound slot at North Station allocated to a train that will run through to Park St and no pseudo deadheads or actual deadheads wasting that capacity; although the challenge is that if you have to park all of the northbound trains in the yard at Reservoir until you know when the event will end it's difficult to get the northbound trains to North Station quickly at that point.
 
Re: Park Street loop and system capacity

Broadly speaking, I think we're now discussing two different arguments that Alon has made:

First, the argument that where interlining ends up happening, using the same headway on all services is likely optimal; this is where I think I agree with him.

The problem is that's not how things work out in the real world. I'd have to look at the ridership numbers but I'm willing to bet that each branch of the GL is not equal. To have crowded B trains and empty D trains is a waste (just using those as examples). However if ridership is similar enough then obviously matching headways is best.

Then there's the discussion of disentangling the New York City interlining of 13 services on most of the lettered lines. I was under the impression that Alon had come to the conclusion that the construction costs of actually separating things out would probably not be worthwhile because so many of the stations that would need to become major transfer points if the interlining were reduced have small platforms because it was assumed that interlining would mean people wouldn't transfer; the disagreements around whether rider reaction to transferring matters don't seem like they're likely to make a major difference for any actual real world service pattern decisions.

He's right about the costs. Interestingly, the GL actually has the infrastructure, not totally, for this. If there was a second trunk entering at Pleasant St Portal then you could loop C trains at Kenmore, B trains at Park St, and D/E would still interline but have their own terminals. This is only something that would make sense if B ridership was considerably higher than C ridership.

However, I also thought I'd seen something somewhere claiming that a lot of New York City subway riders gravitate towards trains with ``express'' in their name, even when the number of local stops that will be skipped is so small that transferring or waiting extra for the express won't really get the passenger to the destination any faster.

This has been proven time and time again. It's human psychology, truthiness, losing out to the facts. This is the heart of my argument about a one seat ride. Even if the transfer takes 1 min the psychological impact is enough to make the rider second guess their trip.
 
Re: Park Street loop and system capacity

The problem is that's not how things work out in the real world. I'd have to look at the ridership numbers but I'm willing to bet that each branch of the GL is not equal. To have crowded B trains and empty D trains is a waste (just using those as examples). However if ridership is similar enough then obviously matching headways is best.

I think at one point in the past the T claimed that one of the Green Line branches ran 5 minute headways at peak, another ran 7 minute headways at peak, and the other two ran 6 minute headways at peak, and more recently each line runs 6 minute headways at peak.

Maybe you could elaborate on exactly how you get consistent 5 minute headways mixed with 6 minute headways mixed with 7 minute headways if you typically need 1.5 minute spacing between one train and the next? It seems like you're doomed to not actually follow the headways because meeting the headways would require having trains on top of each other at less than the 1.5 minute spacing. Can you list which branch you think should show up at Copley's westbound platform at the top of the hour, which branch should show up 1.5 minutes after that, which branch should show up 1.5 minutes after that, etc, for a whole hour to best get the mix of 5 minute headways for one branch, 6 minute headways for two other branches, and 7 minute headways for the fourth branch?

Alon's blog post points out that if you have a major imbalance, running different headways by a factor of two can work.

And different train lengths might also work in some cases, although the T has phased out the four car trains that used to run on the Red Line, and the Red Line is now all 6 car trains always AFAIK.

This has been proven time and time again. It's human psychology, truthiness, losing out to the facts. This is the heart of my argument about a one seat ride. Even if the transfer takes 1 min the psychological impact is enough to make the rider second guess their trip.

The specific thing I was thinking of, I think, is that most express stops in the New York City subway have one island platform for each direction, so that you have a cross platform transfer between the same direction local and express (sort of like how Kenmore and the Green Line level of Park are laid out, except that the Green Line has different branches instead of local vs express). However, NYC has at least one express station which has an island platform serving the express tracks for both directions with side platforms for the local tracks, and IIRC that station may not have been built to prevent same direction cross platform transfers initially.
 
Re: Park Street loop and system capacity

If the center loop is taken out and a double switch installed just south of Park St then the turning capacity is around 20-25tph. This might be more than enough. I don't know enough about the ridership to say which is best but the basic number we need to know is how many riders from the branches (past Copley and Kenmore) are going to and past Park St. Riders along the trunk don't matter in this case because a rider can just wait for the right train to take. But if we know what the branches bring in then we can better adjust the terminal points.

The more I think about the operations of the moves the more I realize that the loop is probably the best option. Park St is basically the length of two full trains (3 cars each?) and allows for different trains to berth at specific locations. This would need to change if the loop was removed meaning that the center tracks would work best with only one branch. Otherwise with two branches the dance of trains terminating and leaving becomes more complicated. Even in NYC each branch of each line has their own track space for terminating (Forest Hills-71 Av being the most complicated but there are tail tracks and yard leads past the station to help). There is no room past Park St to build terminal tracks which would help with multiple branches so ultimately keeping the status quo seems to be the best option.

There are several stages of system expansion:

While the Lechmere Viaduct is closed to build the Green Line Extension, we have the four branches, and trains have to turn at Park or Government Center or the North Station yard. We probably want one branch turning on the Park Street loop at this point.

Once the Green Line extension opens, with there initially being the same familiar four south side branches, two of the south side branches need to continue past Lechmere. The other two could in principle turn at any of Park, Government Center, the North Station yard, or the Inner Belt yard just north of Lechmere. In this configuration, if the tight radius of the Park St loop makes it difficult to keep in service, getting rid of it is no great loss (although it does limit options for short turning a late train a bit); and then the question is whether continuing trains further north for the flexibility of progressively larger yard space is worth the costs of the extra mileage. However, there's also the option of normally turning two branches at Government Center, and extending a few trains to North Station or even the inner belt yard if the northbound trains are getting too bunched and some need to be delayed heading back south.

If we find ourselves with the initial two north side branches from the Green Line Extension plus the familiar four south side branches plus service to Dudley, that's where the question of turning trains at Park becomes most potentially challenging. But if we were to turn B trains at Park, 10 TPH, reversing on the fence track(s) probably works, along with perhaps sending E trains to Union Sq, D trains to College Ave in Medford, and maybe looping C and F at Government Center, and each branch can probably have 10 TPH.

There's the question of whether the Pleasant St Incline is ever going to get more than 10 TPH. I'm thinking the answer might be no, but obviously there's the discussion of the possibility of the D/E connector or the Ink Block to South Station transitway. But if the Pleasant St Incline only ever gets 10 TPH then we probably don't need the Park St loop if there are significant benefits to getting trains that can't use that loop.

There's also then the possibility of 275'+ trains at 8 TPH per branch, which with 40 TPH would mean that five lines could run through, which may not be worth doing unless we end up with both five north side lines and five south side lines. On the south side, just adding Dudley / Franklin Park service and rerouting D to Needham and then running Riverside to Kenmore service would get us to five branches. Maybe we'd end up with something like:

  • B Boston College to Casino / Chelsea / Amazon HQ2
  • C Cleveland Circle to Casino / Revere Beach Parkway / US Highway 1
  • D Needham to McGrath Highway / Mystic Ave / Medford Sq
  • E Hyde Sq to College Ave in Medford
  • F Franklin Park (or maybe Mattapan or even Milton on the Mattapan Line) to Union Sq in Somerville (and maybe Porter and Arlington and Lexington)

(That's attempting to go west to east on the south side and east to west on the north side in choosing the pairings. Some of these routes would probably not be worth riding from end to end if we end up with more direct service via the Grand Junction Railroad.)

In this five branch scheme, if we wanted the two easternmost Chelsea busway stations to get Green Line service, and if we wanted the Watertown branch to get service, it might make sense to just run Watertown to Eastern Ave in Chelsea trains that wouldn't ever run through Lechmere Station, and let the Eastern Ave Chelsea folks transfer at Sullivan or Casino if they want to get to Lechmere, and let the Watertown folks transfer at Porter or Union if they want to get to Lechmere.
 
Re: Park Street loop and system capacity

I think at one point in the past the T claimed that one of the Green Line branches ran 5 minute headways at peak, another ran 7 minute headways at peak, and the other two ran 6 minute headways at peak, and more recently each line runs 6 minute headways at peak.

Maybe you could elaborate on exactly how you get consistent 5 minute headways mixed with 6 minute headways mixed with 7 minute headways if you typically need 1.5 minute spacing between one train and the next? It seems like you're doomed to not actually follow the headways because meeting the headways would require having trains on top of each other at less than the 1.5 minute spacing. Can you list which branch you think should show up at Copley's westbound platform at the top of the hour, which branch should show up 1.5 minutes after that, which branch should show up 1.5 minutes after that, etc, for a whole hour to best get the mix of 5 minute headways for one branch, 6 minute headways for two other branches, and 7 minute headways for the fourth branch?

Without knowing the demand I can't say for sure but what I would argue is that on the trunk it might not really matter. Lets say both B/C trains run to Park St and D/E past. The it doesn't really matter if you get a B or a C train and what order they run in; same for D/E. What causes the problems is terminating different trains at different spots along the GL so that it DOES matter how often a C or D comes. This is why I want to see the GL trunk split up so that B/C will run down Boylston and D/E down Huntington via Pleasant. Obviously that is the most expensive option but it also is the only fool proof way to deal with interlining. Throw traffic lights into the mix and you can basically throw all balance out the window.

And different train lengths might also work in some cases, although the T has phased out the four car trains that used to run on the Red Line, and the Red Line is now all 6 car trains always AFAIK.

This is a great point. The GL has plenty of space for more cars so maybe they try a balanced headway with more cars on heavier lines.

The specific thing I was thinking of, I think, is that most express stops in the New York City subway have one island platform for each direction, so that you have a cross platform transfer between the same direction local and express (sort of like how Kenmore and the Green Line level of Park are laid out, except that the Green Line has different branches instead of local vs express). However, NYC has at least one express station which has an island platform serving the express tracks for both directions with side platforms for the local tracks, and IIRC that station may not have been built to prevent same direction cross platform transfers initially.

There are two stations like this, both at Penn Station, one on 8th and one on 7th Ave. The basic idea is that if you need to transfer you can do so at 42nd St while having a separate express platform allows express trains to service 34th St as well.

The famous other example is trains on Queens Blvd which have rather long express stop spacing. Going into Manhattan the crush point is Roosevelt Ave where many people cram into express trains which only have 2 or 3 more stops into the city rather than stay on the local trains which have 6 or 7. The actual running time difference is 3-5 min. Now one could argue that that does add up but if you consider the time the express train then has to wait as the crowd squeezes on then the time saved shortens. Still riders do this everyday.

What the city did when it designed it's IND system in the 1920s and 30s was to just build more express stops in areas where more people were going. This lessens the crush in lower Manhattan and midtown. The problem with Queens Blvd was that it hadn't been developed yet so additional express stops weren't necessary. Today an in fill express station would help but realistically would only spread out the crush.
 
Re: Park Street loop and system capacity

And different train lengths might also work in some cases, although the T has phased out the four car trains that used to run on the Red Line, and the Red Line is now all 6 car trains always AFAIK.

This is a great point. The GL has plenty of space for more cars so maybe they try a balanced headway with more cars on heavier lines.

If I remember correctly the GL can't run more cars per train because it would draw to much power.
 
Re: Park Street loop and system capacity

If I remember correctly the GL can't run more cars per train because it would draw to much power.

I'm guessing with the GLX they would be installing more transformers to pump in more power. However this might just be enough for the new sections and not enough for new trains on the existing network.
 
I was in town briefly and rode the GL for a bit. One thing that stood out was just how loud those turns are. Expanding the turning radius would go a long way in speeding up trains, reducing wear, and reducing sound pollution.
 
Re: Park Street loop and system capacity

I'd have to look at the ridership numbers but I'm willing to bet that each branch of the GL is not equal. To have crowded B trains and empty D trains is a waste (just using those as examples). However if ridership is similar enough then obviously matching headways is best.

Page 29 of https://d3044s2alrsxog.cloudfront.net/uploadedfiles/documents/2014 BLUEBOOK 14th Edition(1).pdf shows that B and D ridership are quite close, but I was surprised to discover that C ridership is apparently a bit less than half of B ridership. And comparing the E branch data on page 29 to the other branches is a bit potentially misleading in that the E branch data includes Prudential and Symphony, but that page excludes Hynes and Kenmore. When you subtract Prudential and Symphony from the E branch count, and account for Hynes + Kenmore having more than 3x the Prudential + Symphony ridership (see page 28), C and E have ridership pretty close to each other.

The ridership data seems to suggest that if the Central Subway can handle a train every 1.5 minutes, it might make sense to run 4.5 minute headways on B and D, and 9 minute headways on C and E, with the goal that the westbound platform at Copley would see something like B, C, D, B, E, D and then repeating.

Also, instead of running a Riverside to Kenmore shuttle if Needham takes over the D branch slots in the Central Subway, it might be worth building the less than 300' of track it would take to have a connector along Swathmore St to the east end of the yard at Reservoir if that would allow C branch trains to be redirected to serve the Reservoir platfoms instead of the Cleveland Circle platforms and extend the C branch out to Riverside, although it appears that the C branch is 6 minutes longer than the D branch from Reservior / Cleveland Circle to Kenmore. Then again, if the ridership numbers mean that C and E can tolerate half the trains per hour that B and D get, maybe 9 minute headways to Riverside along its current route and 9 minute headways to Needham is good enough.
 
Looking at the E ridership drop off as dramatically as it does is more for the case of connecting the D/E via a new subway under Huntington Ave and leaving the 39 bus to pick up the South Huntington traffic (which wouldn't be so bad with a transfer at Brigham Circle).

There is an argument to be made that you could terminate the C at Kenmore and boost B service so that it runs every 2 min or so. Given how close the C runs by the B and D at times riders could chose to take a different train or transfer at Kenmore.

Looking at the drop off on the D after Newtown Highlands is certainly food for thought about what a Needham Branch would look like.

The thing about the GR is that because it is more flexible than the heavy rail lines maybe you don't need to eliminate interlining. Obviously there are real bottlenecks that need reworking (those tight curves don't help either) but it's really only the westbound traffic that needs straightening out so that it's more balanced. Obviously I would love to see an expanded Huntington Ave Subway for balance but it's the train schedules that need tweaking and that's a whole lot cheaper.
 
Looking at the E ridership drop off as dramatically as it does is more for the case of connecting the D/E via a new subway under Huntington Ave and leaving the 39 bus to pick up the South Huntington traffic (which wouldn't be so bad with a transfer at Brigham Circle).

I believe that several years ago when there were service cuts because the T was underfunded, they decided to truncate the E branch to Brigham Circle, at least on weekends. They subsequently decided to reduce headways slightly so that they could continue running E trains to Heath St. I suspect that the lesson here is that truncating the E branch to Brigham Circle to get rid of the street running is probably politically impossible.

There is an argument to be made that you could terminate the C at Kenmore and boost B service so that it runs every 2 min or so.

So how many trains per hour do you expect Copley to see on all the branches combined if you try to do that? (Or is this assuming the construction of a new very expensive tunnel that we don't actually need anytime soon?)

Given how close the C runs by the B and D at times riders could chose to take a different train or transfer at Kenmore.

If we were to move to 4.5 minute headways on B and D and 9 minute headways on C, I think folks halfway in between would likely avoid C for inbound trips.
 
Here's a random idea that just came to me. Would it make sense to install a third express track on the C line from Cleveland Circle to St Mary's? The highest ridership stations are Cleveland Cir, Washington St, Coolidge Corner, and St Mary's so they could be the express stations. And to get the most out of it why not use the short connection to the B and extend service to Boston College?

Looking at the ROW there is plenty of space. Hell, you can still keep the parking in many places!

Now obviously I doubt ridership on the line warrants such an investment but it would greatly speed up service from the suburbs to the city.
 
Cleveland circle is just a short walk from Riverside. Riverside provides what basically is express service.
 
Cleveland circle is just a short walk from Riverside. Riverside provides what basically is express service.

But Washington St and Coolidge Corner aren't. But given that it's the end of the line I suppose express service isn't really needed.
 
really? short walk?

I am sure he meant to say Reservoir on the D Riverside line.

And actually the B, C and D lines are all close together at Cleveland Circle (Chestnut Street - B; Reservoir - D). Perhaps free surface transfers at that point would encourage people on the B line to use the D line as an express?
 
I am sure he meant to say Reservoir on the D Riverside line.

And actually the B, C and D lines are all close together at Cleveland Circle (Chestnut Street - B; Reservoir - D). Perhaps free surface transfers at that point would encourage people on the B line to use the D line as an express?

No one is going to walk that far to shave off a few min, that's not how people think.
 
Would it make sense to run occasional trains to Boston College down the D to Reservoir, up chestnut hill ave, then on the B the rest of the way? Might reduce load on the rest of the B
 
BC + South Street + Chestnut Hill Ave (which would require new platforms) account for less than 2,000 daily riders - about 13% of B Branch ridership.
 

Back
Top