Green Line Reconfiguration

[Riverside]

@Teban54, thank you (as ever) for the detailed and thoughtful reply!

I think I agree with you about the theoretical dimensions of limitations here. Obviously the particular practical effects of each of those dimensions can be discussed and examined in much greater depth, but I think I agree about the framework and which factors influence.

Like I said, I think my personal preference is for a setup that ends up rendering most of the concerns discussed here moot (admittedly in part by settling for something less than the full 80 tph capacity -- 16 tph on Huntington-GC, 16 tph on Nubian-GC, ~32 tph on Kenmore-Seaport, and x tph using Park Street Inner for extra service as needed).

Two relatively small notes:

Crossovers (switches): yeah, I noticed the same southbound crossover situation you did; to be honest, I assumed (perhaps naively) that it would be trivial to add a new diamond along there. If we're talking about building a new subway along Marginal Road (to say nothing of the more elaborate suggestions in this thread), my hope is that it would be easier to add a new crossover in an existing tunnel. But it's totally a fair point that I don't know any of the potential complexities around that.

Crossovers (trains):

Assume that you have 32 TPH from Huntington and 16 TPH from Nubian. Under your proposal, all trains need to switch north of Boylston (Huntington trains from Boylston Inner to Park Outer, Nubian trains from Boylston Outer to Park Inner). All trains cross each other's paths, so you'd then need to slot every Nubian train in-between adjacent Huntington trains -- but if Huntington is operating at full capacity, there won't be any such gaps.

Strictly speaking, I don't think this is entirely the case. 32 TPH from Huntington means a departure every 1m52s. I took a quick look at a couple of crossovers out there: at Union Square, the crossover is 180'; in Santa Monica, their E has a 210' crossover. At 10 mph, 210' could be cleared in 15 seconds. Even at worst case, at 5 mph it could be cleared in 30 seconds. So even if you had a Huntington and Nubian train departing Boylston at the exact same time, the worst case is that the Nubian train has to hold for 30 seconds -- but hopefully more like 15 seconds, which seems manageable (particularly if Nubian trains are "only" running every 3m45s on average).

All of which is to say, "full capacity" does not necessarily entail the crossover being blocked at all times.

(I am mindful that TheRatmeister made a similar argument to me a couple of months ago and I voiced skepticism there. So perhaps I need to go back and reexamine my assumptions there -- or perhaps my argument here isn't particularly strong.)

 

[Teban54]

Strictly speaking, I don't think this is entirely the case. 32 TPH from Huntington means a departure every 1m52s. I took a quick look at a couple of crossovers out there: at Union Square, the crossover is 180'; in Santa Monica, their E has a 210' crossover. At 10 mph, 210' could be cleared in 15 seconds. Even at worst case, at 5 mph it could be cleared in 30 seconds. So even if you had a Huntington and Nubian train departing Boylston at the exact same time, the worst case is that the Nubian train has to hold for 30 seconds -- but hopefully more like 15 seconds, which seems manageable (particularly if Nubian trains are "only" running every 3m45s on average).

All of which is to say, "full capacity" does not necessarily entail the crossover being blocked at all times.

The interesting question is, if the only concern is "how long it takes to clear a switch"... Then why can't the capacity limit of a track be higher than 32 TPH in the first place?

In my amateur view, a capacity limit comes from:

1. Turnback capacity at each individual terminal
   • At each terminal, the train not only needs to go through the switch, but also reverse direction, load and unload, etc. The bottleneck is often largely due to how long the train occupies the terminal track (beyond the last switches), not the switch itself.
2. Dwell time within each station; signals between adjacent trains; as well as other typical variations in runtime, even when just running along a straight track
   • Headway = Time between (the front of the first train entering the station) and (the front of the second train entering the station)
   • Between these two time stamps, a lot of things need to happen:
     • a) The front of the train needs to reach the end of the platform and stop
     • b) Loading and unloading passengers
     • c) The entire train then needs to leave the platform
     • d) The second train then needs to traverse through the gap between the two trains -- as directed by signals -- that's meant to prevent collisions
   • Between Feb and Apr 2025, the median dwell time at Park St is 74s southbound and 73s northbound; Government Center 56s southbound; Copley 53s northbound
     • I'm not even sure how many of (a,b,c,d) these account for, so the actual minimum gaps between two trains may be even longer
   • Based on my understanding, this is why people generally assume that most heavy rail lines have a minimum headway of 2 minutes. A few systems can achieve 90 seconds with CBTC, but they're exceedingly rare.

The thing is, I believe #2 is a more crucial factor than #1. If #1 were the bottleneck, you would be able to unlock higher headways by building multiple terminals or operating a ring route -- which is exactly what the ultra-frequent heavy rail lines do.

Back to the specific question of X-merges. In terms of crossover conflicts, you can avoid (b), but (a,c,d) still apply, just that "platform" is replaced with "crossover". In other words, after the preceding Huntington train passes through the crossover, and then after (d), you need a Nubian train to complete (a,c) at the crossover (aka go through the switch) while the next Huntington train performs (a,b) at the platform (aka enter Boylston and load passengers).

The only way this can work out is... If (b) is slow enough but (a,c) are quick enough. Or in essence, dwell time is primarily limited by passenger flow rather than travel. But due to the following bullet point, I doubt that's the case:

• A factor that you might not have factored in: Train length also counts towards the "crossover length". A crossover is occupied from the moment the front of the train enters it to the moment the rear of the train leaves it. A two-car Type 10 would be about 227 ft long, which should be added to the effective length.

I'm willing to be convinced otherwise, but so far I'm not: I believe that any capabilities that would allow Nubian trains to "nicely" slot between Huntington trains could have also been used to add more capacity on any single track in the first place, thus violating the 32 TPH assumption.

 

[Riverside]

I'm willing to be convinced otherwise, but so far I'm not: I believe that any capabilities that would allow Nubian trains to "nicely" slot between Huntington trains could have also been used to add more capacity on any single track in the first place, thus violating the 32 TPH assumption.

Yeah this is fair, I agree.

 

[TheRatmeister]

Strictly speaking, I don't think this is entirely the case. 32 TPH from Huntington means a departure every 1m52s. I took a quick look at a couple of crossovers out there: at Union Square, the crossover is 180'; in Santa Monica, their E has a 210' crossover. At 10 mph, 210' could be cleared in 15 seconds. Even at worst case, at 5 mph it could be cleared in 30 seconds. So even if you had a Huntington and Nubian train departing Boylston at the exact same time, the worst case is that the Nubian train has to hold for 30 seconds -- but hopefully more like 15 seconds, which seems manageable (particularly if Nubian trains are "only" running every 3m45s on average).

All of which is to say, "full capacity" does not necessarily entail the crossover being blocked at all times.

(I am mindful that TheRatmeister made a similar argument to me a couple of months ago and I voiced skepticism there. So perhaps I need to go back and reexamine my assumptions there -- or perhaps my argument here isn't particularly strong.)

I believe I was talking about merging street service from Hyde Square into the Huntington Ave subway there? Correct me if I'm wrong.

There's a pretty big gap in the service levels between the two cases, and also a big difference in the actual switching going on. If an underground Mission Park station was built like Kenmore, Hyde Square trains could use the middle (or side, doesn't matter) platforms and then merge into the Huntington Ave subway. No crossing paths, just wait for a gap and then go. If they need to hold for 30-60s to keep service running smooth, so be it, that's just what happens to surface light rail sometimes. But the frequency assumption there is different, the 4-8 Hyde Square trains per hour need to slot into the gap left by the 18-24 Huntington trains per hour. At worst that's a gap of 2.5 minutes for a train to go, very doable.

Another possible problem with a crossover service like this is that it has a major single point of failure. If one of those switches, a moving part which will inevitably break at some point, gets stuck one way or the other (or worse, in the middle), now service is completely borked on at least the Huntington branch until someone gets out to fix it. It's a similar 'risk' to that introduced by switches at terminals to turn trains. If one gets stuck there you lose a bunch of capacity from single-tracking, if one gets stuck here you lose a bunch of capacity from needing to send 32 Huntington TPH through the ~20TPH Park St Loop. If it's not necessary I see no reason to introduce that risk, and quite simply I don't think it's necessary here. I think 32 TPH for Nubian+Huntington is probably enough, and if we want more then the Charles St Subway is there as an option to complement 'SL Phase III Revised.' That also leads into why I like the Essex St route. Despite it's many complexities it balances distance, transfers, and the ability for a future Nubian/Huntington Split better than any alternative I've really seen so far.

Spoiler: Service Breakdown for 32TPH Combined

Branch	TPH
Nubian	10/12
Riverside	7
Needham Jct.	7
Hyde Sq	4/6 (To be suplemented with additional trains to Kenmore)

 

[Teban54]

Question about Silver Line Phase III (for those who know about engineering and historical studies much more than I do):

Why did they do away with a combined Boylston-Chinatown station and opt for two separate stations?

The more I think about it, the more I feel that the two-station option has many issues:

• Cost of building two stations (with station boxes, platforms, egress etc) instead of one
• Having to put Boylston station immediately next to (and below) GL tracks in parallel
  • Not directly under, because GL runs under the northern edge of Boylston St
• Short stop spacing that incurs travel time penalties
• Concerns with Central Burying Ground
  • Yes, that's primarily due to the turnback look directly under it, but I'd think that construction impacts from the Boylston station box also doesn't help
• Due to platforms extending further towards the ends, it reduces the length available for descents, thus worsening required grades

According to Wikipedia:

A proposal to consolidate the Chinatown and Boylston stations was rejected due to steep grades required and lack of cost savings, and a proposal to eliminate the loop was rejected because of the need to short turn most buses at Boylston.[49]: 2.15–22

I find it very hard to believe that a single station would be even more expensive that the two stations as planned, nor that it would require steeper grades.

F-Line mentioned earlier that a single station would also have high ridership, which would lengthen dwell times and create crowding concerns. IIRC he said it would even eliminate the time savings from making one fewer stop.

Note that the section of Boylston St between the two crossing lines is roughly as narrow as Essex St further east: 44-45 ft building-to-building. Obviously, the eventual plans did manage to place station boxes east of Washington St, but still interacted with building foundations and required mitigation. (That's also true at Boylston station, which just scraped by between the GL tunnel and buildings to the south.) The Boylston station box is 45' 4'' wide; I didn't find the figure for Chinatown after a quick look. I wouldn't expect this to be the distinguishing factor in station choices.

So which of these reasons resulted in them going with the intuitively more expensive and more tedious proposal? Or was it all of the above?

 

[F-Line to Dudley]

Question about Silver Line Phase III (for those who know about engineering and historical studies much more than I do):

Why did they do away with a combined Boylston-Chinatown station and opt for two separate stations?

The more I think about it, the more I feel that the two-station option has many issues:

• Cost of building two stations (with station boxes, platforms, egress etc) instead of one
• Having to put Boylston station immediately next to (and below) GL tracks in parallel
  • Not directly under, because GL runs under the northern edge of Boylston St
• Short stop spacing that incurs travel time penalties
• Concerns with Central Burying Ground
  • Yes, that's primarily due to the turnback look directly under it, but I'd think that construction impacts from the Boylston station box also doesn't help
• Due to platforms extending further towards the ends, it reduces the length available for descents, thus worsening required grades

According to Wikipedia:

I find it very hard to believe that a single station would be even more expensive that the two stations as planned, nor that it would require steeper grades.

F-Line mentioned earlier that a single station would also have high ridership, which would lengthen dwell times and create crowding concerns. IIRC he said it would even eliminate the time savings from making one fewer stop.

Note that the section of Boylston St between the two crossing lines is roughly as narrow as Essex St further east: 44-45 ft building-to-building. Obviously, the eventual plans did manage to place station boxes east of Washington St, but still interacted with building foundations and required mitigation. (That's also true at Boylston station, which just scraped by between the GL tunnel and buildings to the south.) The Boylston station box is 45' 4'' wide; I didn't find the figure for Chinatown after a quick look. I wouldn't expect this to be the distinguishing factor in station choices.

So which of these reasons resulted in them going with the intuitively more expensive and more tedious proposal? Or was it all of the above?

Page 2-15 of the Phase III Draft Environmental Impact Statement details at length why combined station at the mid-block was rejected after being evaluated.

• No overall cost savings over the 2-station option, owing to the extremely complicated mezzanine level connecting walkways to GL-Boylston and OL-Chinatown stations.
• Higher building mitigation costs than separate stations because of the narrowness of Boylston St. at the mid-block and depth of building foundations (meaning, mid-block siting has way more potential to bust its budget from unforeseen circumstances than the 2 separate stations and may in-practice exceed the cost of separate stations, which were studied out enough to more or less put a upper-bound on cost inflation)
• Reduced ridership over the 2-station option, owing to the much longer transfer penalty for using the labyrinthine connecting walkways. Cost per rider was worse because of the lower topline, and that didn't wash well when the total costs projected so close to begin with.
• Problems with headway adherence, because the station dwells at a combined stop would much outstrip the target headways and the platform dimensions would be more constrained at the mid-block (3 bus berths vs. 6 at the separate stations).
• Incompatible grades with the then-preferred Tremont St. portal: 6.4%, in violation of the 6% maximum spec.

The last two may not be as much a factor with a GLR repurposing, but you're still not going to find any cost savings with mid-block given the building mitigation problems (especially with how much a degree that ended up inflating costs on all else for the core Phase III alignment) and the overly complex pedestrian connections to Green and Orange.

 

[Teban54]

Page 2-15 of the Phase III Draft Environmental Impact Statement details at length why combined station at the mid-block was rejected after being evaluated.

• No overall cost savings over the 2-station option, owing to the extremely complicated mezzanine level connecting walkways to GL-Boylston and OL-Chinatown stations.
• Higher building mitigation costs than separate stations because of the narrowness of Boylston St. at the mid-block and depth of building foundations (meaning, mid-block siting has way more potential to bust its budget from unforeseen circumstances than the 2 separate stations and may in-practice exceed the cost of separate stations, which were studied out enough to more or less put a upper-bound on cost inflation)
• Reduced ridership over the 2-station option, owing to the much longer transfer penalty for using the labyrinthine connecting walkways. Cost per rider was worse because of the lower topline, and that didn't wash well when the total costs projected so close to begin with.
• Problems with headway adherence, because the station dwells at a combined stop would much outstrip the target headways and the platform dimensions would be more constrained at the mid-block (3 bus berths vs. 6 at the separate stations).
• Incompatible grades with the then-preferred Tremont St. portal: 6.4%, in violation of the 6% maximum spec.

The last two may not be as much a factor with a GLR repurposing, but you're still not going to find any cost savings with mid-block given the building mitigation problems (especially with how much a degree that ended up inflating costs on all else for the core Phase III alignment) and the overly complex pedestrian connections to Green and Orange.

Thanks! A few interesting bits from a read of the section (2.3.4.2 and 2.3.4.3):

Planned service levels: Unsurprisingly, they planned for much more Boylston-Seaport trips than full Nubian-Seaport trips. But what surprised me was just how many Seaport trips were planned: 74 vehicles per hour!

Trips	60' Buses Per Hour	TPH with two Type 8 cars (one Type 10 car), with equivalent capacity*	TPH with two Type 10 cars, with equivalent capacity**
Seaport total	74	17.8 (crush capacity) 23.8 (policy capacity)	8.9 (crush capacity assumed)
Nubian total (incl. Nubian-DTX surface route)	22	5.3 (crush capacity) 7.8 (policy capacity)	2.6 (crush capacity assumed)

* LRT conversions were made using the 2014 Blue Book's figures for dual mode buses and Type 8. Two sets of estimates were made, one using crush capacities for both, and one using policy/planning capacities. IMO, crush capacity sounds more realistic for reasons below.
** The only source for Type 10's capacity that I could find is 400 per car (likely an estimate), which is stated as equivalent to two current GL cars. The only way I can reconcile this is by using Type 7/8's crush capacities (269 and 199 respectively). The Type 10 column assumes 800 passengers per two-car train.

In some sense, this highlights more about how inefficient buses are, not just the demand for Seaport.

FYI, the SL1/2/3/W buses currently run 21 BPH during 7-8am (average headway 2.9 min). This is not just a far cry from the 74 BPH under the Build Alternatives, but barely half of even the 2025 No-Build Alternative, which called for 42 BPH. The Silver Line is seriously underperforming today.

• Likewise, the No-Build Alternative assumed 18 BPH on the surface SL5, but today, SL4/5 run about 14 BPH combined.

Also interesting to note that they proposed a route from Back Bay (surface) to the Seaport Transitway that runs 6 BPH. So the idea of a Back Bay-Seaport connection is nothing new.


Where the "mid-block" station actually is: While I know that the 2005 DEIR didn't focus on accurate portrayal of the 2003 New Starts Alternatives (which include the two mid-block alternatives that were rejected), the figures seem to indicate that the "mid-block" stations were actually much closer to OL Chinatown than the name implies. Figure 2.3-9 shows this particularly well -- in fact, this is virtually where the two-station alternatives placed Chinatown:

If we go by this location, which obviously has terrible transfers to the Green Line, it seems that the loss of Green Line transfer riders should played a big part in the lower ridership being modeled.


A Green Line branch from Park St had been considered before: Interestingly, the other mid-block alternative apparently showed connections from GL Boylston's (presumably outer) tracks, allowing a Green Line branch from Park St to Seaport:

Obviously, this is very different from the way we'd do a "Green Line" conversion today, and GL under this proposal wouldn't have capacity for 5 branches. But it's interesting that this was even a real-world proposal.


Considerations for LRT conversion and curve radius: The report mentioned LRT operations in the conclusions section -- with no commitment for future conversion, but designed in a way that would not preclude it:

One interesting implication is curve radius. We have been throwing around proposals with tight curve radii on this board, largely because of several tight curves already existing on the Green Line system today. But this document apparently considers curves with <100 ft radii undesirable, even if they're technically feasible?

(Given that the tightest revenue curve on GL today is 60' at GC, hopefully this won't be a big issue if we stay above that. But still...)

 

[TheRatmeister]

If we go by this location, which obviously has terrible transfers to the Green Line, it seems that the loss of Green Line transfer riders should played a big part in the lower ridership being modeled.

More thoughts later, but I think it's worth pointing out that this "terrible" transfer is really not that bad. It's around 400ft walking, or around 100 seconds. It's pretty comparable to the long transfer at State, which we know people use. For a transfer that would (mostly) just be used for people going between Huntington and Seaport, (Many Longwood Pax would just use the truncated D branch for a 1SR, while passengers on the other branches would just get on one of the Park St trains and transfer there.) I'm honestly tempted to call that an acceptable sacrifice if it makes a meaningful difference to project cost/feasibility.

 

[TheRatmeister]

Also, if you exclude Boylston, are there any other underground stops which don't allow for triple, or maybe even quadruple Type 10s?

 

[F-Line to Dudley]

One interesting implication is curve radius. We have been throwing around proposals with tight curve radii on this board, largely because of several tight curves already existing on the Green Line system today. But this document apparently considers curves with <100 ft radii undesirable, even if they're technically feasible?

(Given that the tightest revenue curve on GL today is 60' at GC, hopefully this won't be a big issue if we stay above that. But still...)

Guardrail and cycled maintenance are not major considerations for a ruling curve radius. Guardrail is not expensive or hard to handle on maintenance, and any curve that gets enough daily reps will wear out much faster than tangent track so deferred maintenance is never an option. Pretty much all LRV's and maintenance equipment on the planet can fit within an 80 ft. curve radius, so that ends up being the recommended generic minimum. 100 ft. and "no guardrail" is overly, overly cautious.

 

[Teban54]

More thoughts later, but I think it's worth pointing out that this "terrible" transfer is really not that bad. It's around 400ft walking, or around 100 seconds. It's pretty comparable to the long transfer at State, which we know people use. For a transfer that would (mostly) just be used for people going between Huntington and Seaport, (Many Longwood Pax would just use the truncated D branch for a 1SR, while passengers on the other branches would just get on one of the Park St trains and transfer there.) I'm honestly tempted to call that an acceptable sacrifice if it makes a meaningful difference to project cost/feasibility.

To clarify:

• My comment was about the historical, real-world SL Phase 3 proposal's "terrible" Silver-Green transfer for the mid-block option (as pictured) and its impacts on ridership, not about any hypothetical GLR. (I agree that most reasonable GLR plans have reduced needs for a Boylston transfer, but if anything, the Nubian branch has its transfers affected the most.)
• If you only have a single station at OL Chinatown (as pictured), I measure 750' between the edge of the southbound Boylston platform and Liberty Plaza, and 480' northbound. This average to 615'.
  • In comparison, State (The EGE didn't draw its layout diagram, but some supporting materials here and on OSM) seems to be around 100' to OL northbound, and 600' to OL southbound; average 350'
  • Note that many other metro systems worldwide have even longer transfer walkways, especially those with legacy stations
• Also note that stacked platforms that are 80' deep will add to vertical travel times, partially reducing the advantage of shorter walkways
  • (Speaking of which, do we know why each SL tunnel planned seemed so much taller and deeper than even the Green and Orange tunnels?)

As for whether having one station makes the engineering easier, I found this table:

You can see that the Core Tunnel has a crazy number of underpinnings, while the branches south of Boylston have virtually none. Among the list of core tunnel underpinnings:

• 695 Atlantic Ave is at the Essex/Atlantic hook-in (unavoidable)
• The next 7 or so buildings, until 39 Boylston St, are in the Chinatown vicinity
  • No underpinnings on the mined tunnel segment in-between, until the Chinatown station box
• The last 9 buildings are for Boylston station
  • From the Boylston cross-section, looks like you can avoid those and even put single-level, bidirectional tracks under Boylston St's road width, if you have GL-dimension tunnel (although you need stacked platforms at Chinatown anyway)

That said, plenty of other buildings on Essex St require mitigation, even if not underpinned directly:

(I cannot for the sake of my life figure out why the Charles St alternative required much more mitigations than Charles-Stuart-Tremont, when the southern segments of Charles St and Tremont St seem similar in width. Any ideas?)

My best guess is that a single station at Chinatown, as opposed to either the two-station alt or a single station literally between Tremont and Washington Sts, will bring some cost savings -- but not enough to make the Essex St alignment economical.
For this reason, I myself still wouldn't advocate for it. But if you really need to do it, I think you'll need the following package: (assuming Kenmore-Seaport trains stay entirely on Boylston-Essex Sts, and Kenmore-Tremont trains, if any, use the existing route)

• A single station at Chinatown, nothing at Boylston (for possible cost savings)
• 2x2 setup probably needed: Huntington-Seaport trains likely necessary
  
  • Having Huntington-Seaport trains can basically eliminate any within-GLR transfer, or at least move them outside of Boylston, except Nubian-Kenmore
• If you send all Kenmore trains (32 TPH) to Seaport, you end up crucially relying on a Boylston transfer: (or an even worse Copley transfer)
  • Having no Boylston station means lack of transfers for Huntington-Seaport
  • Even with a more convenient transfer, too much Huntington-Seaport transfers will kill crowd control, same as SL Phase 3 (although slightly better because Kenmore riders have other options)
  • Nubian-Kenmore and Nubian-Seaport transfers are a lost course; compensate for them via other means (e.g. buses)

Spoiler: FYI: Transfer lengths to OL for other alignments

If Seaport trains have a Chinatown-TMC station on Stuart-Kneeland Sts:

• Connection to Chinatown: Chinatown's platforms (The EGE's map) are 150' from Stuart/Kneeland to OL southbound, and 540' to OL northbound -- average 345'
• Connection to TMC: TMC doesn't have an accurate station map, but based on exit locations (at both ends of the platform, if not a little bit beyond on the Washington St side), it's also about 540' to Stuart/Kneeland
• A Chinatown transfer is definitely better, essentially a State replica; the only question is feasibility

If Seaport trains have a Bay Village-TMC station on Marginal Rd:

• Even the Tremont St exit itself is 440' from Marginal @ Shawmut, so any transfer is at least this long
• May be even longer depending on where the Bay Village platforms are on Marginal St

Also, if you exclude Boylston, are there any other underground stops which don't allow for triple, or maybe even quadruple Type 10s?

This may have been brought up at some of the HRT conversion discussions before. IIRC, existing platforms at Courthouse and WTC are not long enough for that. I'm not sure how easily they can be extended, however.

Government Center would probably be the other biggest concern of mine, and maybe Symphony.

 

[The EGE]

TL;DR: you're never going to see anything longer than two Type 10s on the Green Line.

Kenmore: 310
Hynes: 340
Copley: 320
Arlington: 270
Boylston: 230 active, 260 including closed portions
Park Street: 320
GC: 280 NB, 300 SB, 340 loop
Haymarket: 230
North Station: 390
Science Park: 220 (250 including the space currently used by mini-highs)
Lechmere: 350
GLX stations: 225 (provisioned for expansion to 300)
Symphony: 290
Prudential: 300
South Station, Courthouse, WTC: 220

Everything in the subway and GLX is long enough for two Type 10s, as are the Silver Line stations*, but it's highly unlikely you can go beyond that. You'd have to extend platforms everywhere except Hynes, North Station, and Lechmere. Technically it's probably feasible, but it would be incredibly difficult. Same number of underground stations as the 1980s Red Line modification (which was a bigger jump in capacity), even worse complications. Kenmore, Boylston, Park, GC, and Haymarket would require realigning tracks / moving crossovers to get the platforms long enough. Science Park would require modifications to the NRHP-listed viaduct. There would also be a lot of complications on the surface - almost everything on the B, C, and E would need lengthened, and Reservoir and Riverside would need substantial modifications.

Conversion to all Type 10 doubles is vastly, vastly simpler - and even that is a $3.8 billion project when you include the yard expansions, platform retrofits, traction power, etc. It has no major lengthenings except some surface stops that need rebuilt for accessibility anyway, though platforms will need to be raised systemwide once the last Type 7s and 8s are gone.

That project will result in twice the capacity currently available, and increased reliability from TSP, dedicated lanes on the E, better dispatching, and all-door level boarding** will increase that even further by reducing wastes of capacity. The system that struggled with 160,000 daily boardings in 2019 will be capable of handling 300,000+ by the mid-to-late 2030s. Given that, it's hard to imagine that kind of mode shift happening without there being the political willpower for one or more of the projects that would relieve the Green Line - be that Gold Line or related Green Line reconfiguration, the NSRL taking some of the Back Bay demand, a Blue Line extension westward, Urban Ring, etc.

*Two Type 10s will be about 228 feet long. You can get away with a slightly shorter platform because the doors aren't at the very ends of the car, but it's best to have a little breathing room. The Transitway platforms (particularly Courthouse) do seem to have been designed with some futureproofing in mind, but likely not to 330+ feet.

** The all-door level boarding will have two kinds of reliability and travel time benefits. One is that level boarding results in faster and more consistent dwell times compared to stepping up/down. The other is that the current method of using a ramp takes several minutes, meaning that every single time it's needed puts a train substantially off headway (and delays following trains in the subway). With true level boarding, that gets reduced to next to nothing.

 

[Teban54]

TL;DR: you're never going to see anything longer than two Type 10s on the Green Line.

Kenmore: 310
Hynes: 340
Copley: 320
Arlington: 270
Boylston: 230 active, 260 including closed portions
Park Street: 320
GC: 280 NB, 300 SB, 340 loop
Haymarket: 230
North Station: 390
Science Park: 220 (250 including the space currently used by mini-highs)
Lechmere: 350
GLX stations: 225 (provisioned for expansion to 300)
Symphony: 290
Prudential: 300
South Station, Courthouse, WTC: 220

Everything in the subway and GLX is long enough for two Type 10s, as are the Silver Line stations*, but it's highly unlikely you can go beyond that. You'd have to extend platforms everywhere except Hynes, North Station, and Lechmere. Technically it's probably feasible, but it would be incredibly difficult. Same number of underground stations as the 1980s Red Line modification (which was a bigger jump in capacity), even worse complications. Kenmore, Boylston, Park, GC, and Haymarket would require realigning tracks / moving crossovers to get the platforms long enough. Science Park would require modifications to the NRHP-listed viaduct. There would also be a lot of complications on the surface - almost everything on the B, C, and E would need lengthened, and Reservoir and Riverside would need substantial modifications.

Conversion to all Type 10 doubles is vastly, vastly simpler - and even that is a $3.8 billion project when you include the yard expansions, platform retrofits, traction power, etc. It has no major lengthenings except some surface stops that need rebuilt for accessibility anyway, though platforms will need to be raised systemwide once the last Type 7s and 8s are gone.

That project will result in twice the capacity currently available, and increased reliability from TSP, dedicated lanes on the E, better dispatching, and all-door level boarding** will increase that even further by reducing wastes of capacity. The system that struggled with 160,000 daily boardings in 2019 will be capable of handling 300,000+ by the mid-to-late 2030s. Given that, it's hard to imagine that kind of mode shift happening without there being the political willpower for one or more of the projects that would relieve the Green Line - be that Gold Line or related Green Line reconfiguration, the NSRL taking some of the Back Bay demand, a Blue Line extension westward, Urban Ring, etc.

*Two Type 10s will be about 228 feet long. You can get away with a slightly shorter platform because the doors aren't at the very ends of the car, but it's best to have a little breathing room. The Transitway platforms (particularly Courthouse) do seem to have been designed with some futureproofing in mind, but likely not to 330+ feet.

** The all-door level boarding will have two kinds of reliability and travel time benefits. One is that level boarding results in faster and more consistent dwell times compared to stepping up/down. The other is that the current method of using a ramp takes several minutes, meaning that every single time it's needed puts a train substantially off headway (and delays following trains in the subway). With true level boarding, that gets reduced to next to nothing.

Excellent analysis as usual. Do you know why the GLX platforms were built or provisioned for 300 ft, when it seems unnecessary given that platform extensions to this length elsewhere seem infeasible?

 

[F-Line to Dudley]

Excellent analysis as usual. Do you know why the GLX platforms were built or provisioned for 300 ft, when it seems unnecessary given that platform extensions to this length elsewhere seem infeasible?

4-car Type 7/8/9's. In the earliest planning that was still a far-future provisioning target to keep as a rainy-day hold, with Lechmere needing it built up-front because of that station's structural properties. 3-car Type 10's overshoot that figure enough to make it infeasible, but as noted there are a lot of other advantages to dwells (and thus frequency and capacity) to get level-boarding long vehicles established.

 

[TheRatmeister]

If you only have a single station at OL Chinatown (as pictured), I measure 750' between the edge of the southbound Boylston platform and Liberty Plaza, and 480' northbound. This average to 615'.

I think the distance between Inbound/Outbound should only be 230ft, the length of the platform, but yeah still an average of 595ft, more than 2 minutes walking. If you shifted the mid-block station over further, that would obviously help a lot and possibly get the average transfer distance under 600ft to Boylston Outbound, but that's still far from ideal. If you could somehow get a direct pedestrian tunnel over to Boylston, that would be more like 425ft which is much more reasonable. Obviously that would re-add cost, but presumably less than a second station.

But, is there a better option? As you pointed out, anything on Marginal Rd or Stuart St will have a similar or longer transfer distance to the Orange Line, and there is plenty of possibility for construction problems with those. A Stuart-Chinatown transfer for example requires a new foot tunnel dug in parallel to the Orange Line to avoid clogging the Forest Hills platform. Or you could just suck up the added 5-6 minutes from the ride to Back Bay for southbound riders, but even that won't probably be an especially fast transfer, probably around 300ft walking. There is also the problem of Seaport-Kenmore transfers, but with 2x2 using the "long pike" subway that's manageable at least. At that point there's also the question of which 2x2 solution is cheaper, a 3200ft tunnel under or besides the Pike to Hynes or a 2200ft tunnel under Columbus Ave. The Columbus Ave alt was eliminated from consideration for SL Phase 3 but without the need to surface as close to Washington St as possible, it should be possible to eliminate and grade exceedences, and as the report shows it's relatively light on required building mitigations.

too much Huntington-Seaport transfers will kill crowd control,

With 2x2, this is almost definitely not a concern. Without 2x2, it could be but Longwood Service would still run from Seaport to Brookline Villages, Hynes/Copley still cover the Back Bay job centers reasonably okay-ish, and Northeastern is still covered by Ruggles/Mass Ave. I think enough riders could find a suitable alternative in that list to mostly cover transfers and ensure crowds don't get out of control.

Essex St, but with the Columbus Ave Subway, seems like the most promising alternative to me at this point. I don't currently see any other possible route that doesn't suffer from at least one critically important yet terrible quality transfer.

 

[Teban54]

[...]

Several good points here that I agree with:

• Having any kind of 2x2 virtually eliminates concerns about within-GLR transfer, leaving Orange Line transferas the main problem
  • "2x2" here doesn't distinguish Park Inner and GC
  • Nubian-West transfers may need to use Boylston and/or Park
• Stuart-Chinatown will likely require a new pedestrian tunnel
• If the Kenmore trunk retains a through-running branch to Brookline Village (and further south/west), it greatly alleviates the Huntington-Seaport transfer
  • Note that in some earlier proposals, I was more in favor of keeping Kenmore-Fenway-Brookline Village as a shuttle, primarily to let Kenmore accommodate for more unique branches: Harvard, Watertown, B, C already take up the 4 slots
  • But through-running downtown-Kenmore-Brookline Village (to either Needham or Hyde Square) can make sense if you're willing to give up one of the 2+ "A branch" options

Things that I don't agree with:

• Marginal Rd and Stuart St alignments: Their main advantage is with the "Core Alignment", where Essex St's building mitigation is unavoidable and proven to be expensive
  • Essex St has widespread 45' between buildings
  • Stuart-Kneeland is consistently 70'+ east of Charles St; starting at Arlington St only adds one single 60-65' pinch point
  • Marginal Rd eliminates building mitigation on the east-west segment (can use the Pike when needed), and also has known utility mapping (and is friendlier to C&C)
  • Lincoln St (north-south segment) is 60' between buildings
  • => The cost of building Essex's Tremont-SS section alone is almost guaranteed to be far higher than Stuart-Kneeland and Pike Hugger West (possibly despite length advantages).I don't think you ever truly realized how narrow Essex St is.
    • (Building mitigation table for the Core Alignment alone also justifies this)
• Kenmore - Pike Hugger West: I don't think it needs the full 3200' under the Pike -- an alternative is to use the Exeter St tunnel in the opposite direction

Re: Boylston distances, the issue was that existing NB and SB platforms don't align at their ends: there's a small gap between them with tracks but no platforms. That's why the two distances I gave differ more than the platform length.


Alternative OL transfer locations
Besides Chinatown/TMC, we should identify which half-line connections are adequately served at other stations with better transfers, some of which may even be better for the network.

• Back Bay, for whichever line gets Pike Hugger West; likely convenient
  • I'm actually a lot more optimistic on this than you are: if the GLR station is under or north of the Pike, it's within 150-200' from Back Bay OL
  • Even a GLR station under Stuart St is 300' from OL, if needed
  • It's also an even better transfer location for Kenmore-Forest Hills transfers than Chinatown/TMC is
• North Station, for any GC service (except South End Local); very convenient
  • Assuming that North Station will accommodate at least 3 branches, with one terminating there (or branching out away from Science Park). The one branch terminating at GC will most likely be South End Local (Washington St streetcar).
• Park St-DTX, for any PI/GC service; inconvenient
• Symphony-Mass Ave (OL), for any Huntington service; inconvenient
  • This is a last resort if Huntington gets absolutely nothing, but the distance (610') is the same as Park-DTX
  • But an out-of-system walking transfer can be interesting for some trips (see below)

 

[Teban54]

Demand and priority for transfers
Another aspect is: Which connections actually need a transfer between OL and GLR? Some of them can be handled by alternative routes.

(Reminder: Not all of them are replaceable by Urban Ring, because "Kenmore" also has Hynes, "Huntington" also has Prudential/MFA, "Nubian" has South End, "Oak Grove" has North Station, "Seaport" has South Station, etc.)

• Seaport - Oak Grove: High priority; Preference for Chinatown or TMC, especially the former
  • Another line (bus/streetcar/subway) that runs NS-SS-Seaport can be (even more?) competitive
• Seaport - Forest Hills: High priority; Back Bay also works well
• Kenmore - Oak Grove: Moderate priority; Back Bay (and North Station) also work well
  • Likely for Malden-BU trips
  • (North Station doesn't contribute to such demand under 2x2; but if all Kenmore trains go to Seaport, this becomes moderate-high)
• Kenmore - Forest Hills: Moderate-low priority; Strong preference for Back Bay
  • Likely for Forest Hills-BU trips
  • Several alternative crosstown buses, though they don't cover all stops
• Huntington - Oak Grove: High priority; Back Bay and North Station also work well
  • Likely for Malden-LMA trips
  • (and possibly North Station-LMA trips if all Huntington trains go to Seaport)
• Huntington - Forest Hills: Low priority; Preference for Back Bay
  • These two corridors are parallel for too long, and only start diverging near MFA/LMA; most realistic rides are Forest Hills - LMA, Brookline, Newton etc.
  • Symphony-Mass Ave OOS transfer seems like a better solution
  • Also have crosstown buses and the 39
• Nubian - Oak Grove: Moderate-low priority; North Station also works well
  • Likely for Nubian/South End - Malden trips, traveling across downtown Boston to the north; and for Malden-BMC trips
  • This region has many well-established buses that connect to OL: Several for Nubian, 1 for Mass Ave, 8 (BNRD) for South End
• Nubian - Forest Hills: Low priority
  • OL-connecting buses are even more effective here
• Nubian - Seaport: Moderate-high priority; Preference for Bay Village
  • Supplemented by the 12 bus (BNRD), but neither it nor Urban Ring can help most of South End

Without 2x2, you'll add the following:

• Kenmore - Seaport: High priority; Back Bay also works well
• Huntington - Seaport: High priority; Back Bay also works well
  • Includes South Station-LMA trips, especially for RL South and Old Colony
• Nubian - Kenmore: Moderate priority; Preference for Bay Village
  • Likely for Nubian/South End - BU trips
  • Buses serve as good substitutes for South End, less so for Nubian
• Nubian - Huntington: Low priority; Preference for Bay Village
  • Similar reasoning as Nubian-Forest Hills (despite strong demand)

A few emerging themes. With 2x2:

• Seaport-Oak Grove is the highest priority, which argues for Essex and/or Stuart-Kneeland alignments.
  • However, the real solution for Seaport-Oak Grove may be improvements on Congress St (7 bus improvements and/or new subway?). If materialized, this will do better than even a Chinatown transfer. While this is outside GLR's scope, it does suggest that the discussion may be overshadowing the real problem: Seaport connectivity from the north.
• Second highest non-trivial connection is Kenmore-Oak Grove (as Kenmore-North Station OSRs are unlikely), followed by Kenmore-Forest Hills.
  • The former has plenty of flexibility, but Kenmore-FH strongly prefers Back Bay, and Kenmore has the strongest argument for a Forest Hills connections in ways that Huntington and Nubian don't.
  • This leads to a very intriguing conclusion: It seems to argue for Kenmore-Seaport route to go through Back Bay.

Without 2x2, you suddenly add a lot more problems into the mix (beyond transfer loads).

• If you want to stick with no 2x2, the Kenmore-Seaport-only plan (Alternative 1, the original Criss-Cross) is clearly the better of the two, from a route design perspective.
• However, the Huntington-Seaport-only plan (Alternative 3) has much easier engineering via the full Pike Hugger.
• Alternative 1's biggest problem is the transfer between the two trunks.

So, a couple of very interesting observations come out of this... Which I'll try to organize at a later date.

 

[Teban54]

One last note: In reasoning about all the components (including contents and thoughts planned for subsequent blog post series that I haven't expressed publicly yet), I keep drawing back to this idea that I posted in February 2024:

This is essentially Alternative 2 (Tripod + Kenmore-Seaport) in my subsequent blog post, but with concrete engineering -- Stuart-Kneeland, but using the Charles St subway (studied for SL Phase 3) as Seaport connectors from different directions.

While I'm not ready to say this is my ultimate most preferred design yet, I think it resolves many issues that even I myself didn't fully realize when drawing this:

• Coincidentally achieved Alternative 2, without having even reasoned about TPH back then
• Stuart-Kneeland as a middle ground between cost-effectiveness (favoring Marginal) and network design (Kenmore-Seaport detour and OL transfer, both favoring Essex)
• Likely the most realistic Stuart-Kneeland Alternative 2 design, and only requires minor additions to Stuart-Kneeland Alternatives 1 & 3 (if you impose Stuart-Kneeland as a fixed constraint)

If you really want to do Essex... You can shift the entire Stuart-Kneeland portion (as pictured) north under Essex:

• Kenmore-Seaport trains (Brown) still use the Public Garden portal to descend
• Huntington-Seaport trains (Magenta) still continue north under Charles St, but only turn at Boylston St -- identical to the real-world SL Phase 3
  • Advantages of Charles St over Columbus Ave: Shorter tunneling distance (given that Pike Hugger West is still needed for Huntington-GC); retaining the Nubian transfer at Bay Village; serve nearby neighborhoods

Given this interchangeability, there's likely no practical difference between them if they both stay as Alternative 2 (2x2 with Kenmore-Park Inner). Any real-world study would probably consider both the Essex and Stuart-Kneeland alignments as alternatives, where it will come down to cost per rider. The Pike Hugger East alignment, however, will likely pose enough differences that it's best considere separately.

 

[TheRatmeister]

I'm actually a lot more optimistic on this than you are: if the GLR station is under or north of the Pike, it's within 150-200' from Back Bay OL

I'd revise my estimate down to 225ft. Not nearly as bad as I thought but it doesn't really make the long detour for the Marginal Rd worth it. If it was a cross-platform transfer I think it would be arguable, but even then it's still a 5 minute detour for Oak Grove passengers.

Stuart-Kneeland as a middle ground between cost-effectiveness (favoring Marginal) and network design (Kenmore-Seaport detour and OL transfer, both favoring Essex)

Is it though? I don't think the OL transfer is substantially better. My understanding of the layout at TMC Station is that the north entrance (by the SL stop, at the main hospital entrance) is also roughly at the north side of the platforms. That's around 450ft from Stuart St, again I think you'd need a new tunnel underneath OL tracks to connect the two stations. It's not huge but it's all these little things that erode the cost savings of avoiding Essex.
(It's actually pretty much the same distance that Marginal Rd is from the southern end of the TMC platforms, so in some ways Stuart St is the worst of both worlds. You get a higher cost compared to Marginal Rd without getting a better transfer. However the station is much closer to Downtown at least, which is a substantial benefit and why it's not really the worst of both worlds.)

Marginal Rd eliminates building mitigation on the east-west segment (can use the Pike when needed), and also has known utility mapping (and is friendlier to C&C)

But its North-South segment, which has at-present unknown mitigations is longer than the Essex St alignment.

where Essex St's building mitigation is unavoidable and proven to be expensive

Which is why, in my opinion, it's (at least partially) a "The devil you know" situation. We do know roughly what it would take to build the Essex St tunnel, which is far from insignificant. I'm aware that their is some sunk-cost fallacy risk here but starting over (mostly) from scratch is another erosion at any cost-savings for an alternate route. It's also not really like any of the North-South alternatives you'd need for an alternate route is that much better than Essex in terms of feasibility. Hudson/Kingston is even narrower than Essex. Lincoln St is more promising in that aspect but that turn to hook into Essex is rough. Then there's Surface Road and the CAT underneath. Maybe I'm significantly overestimating the difficulty here, but mining out directly under the Big Dig tunnels does not sound like a good time.

=> The cost of building Essex's Tremont-SS section alone is almost guaranteed to be far higher than Stuart-Kneeland and Pike Hugger West (possibly despite length advantages).I don't think you ever truly realized how narrow Essex St is.

• (Building mitigation table for the Core Alignment alone also justifies this)

But again, those mitigations aren't for the tunnel. They're for the station(s). I'm not sure you can build a station on Stuart St without underpinning buildings either. No, you probably wouldn't need to do as much underpinning, but there go the cost savings being eroded again. Stuart also seems to be either roughly equal to or actually worse than Essex in terms of dealing with listed buildings. The Wilbur and Wang theaters are both listed for both interior and exterior, as is the Jacob Wirth restaurant, and the whole block bounded by Knapp, Beach, Washington, and Kneeland steets is listed for its exterior.

Again, I don't doubt that Essex is the most expensive option of the three. But can the sacrifices to transfers justify the price gap? I've listed a bunch of factors that each eat away at the price difference, so if it doesn't end up being that large, which seems like the more likely outcome to me, then I'd lean towards no it's not worth it to sacrifice the OL transfer.

• Note that in some earlier proposals, I was more in favor of keeping Kenmore-Fenway-Brookline Village as a shuttle, primarily to let Kenmore accommodate for more unique branches: Harvard, Watertown, B, C already take up the 4 slots
• But through-running downtown-Kenmore-Brookline Village (to either Needham or Hyde Square) can make sense if you're willing to give up one of the 2+ "A branch" options

I'd probably break it down like this. Yes Cleveland Circle and Packards Corner-BC get small service cuts. I'm willing to accept that sacrifice for more service to BU and the new Harvard branch.

Branch	TPH
A (Harvard-Seaport)	7
B (Oak Sq-Park St)	6
C (BC-Park St)	6
D (Cleveland Circle-Park St)	6
E (Brookline Village/Hyde Sq-Park St)	7

 

[Teban54]

I'll save most details on engineering and cost for future blog posts, but one point that I want to address now:

Which is why, in my opinion, it's (at least partially) a "The devil you know" situation. We do know roughly what it would take to build the Essex St tunnel, which is far from insignificant. I'm aware that their is some sunk-cost fallacy risk here but starting over (mostly) from scratch is another erosion at any cost-savings for an alternate route.

Strong disagree.

Yes, we do know how much it takes to build Essex -- expensive enough to kill a real-world project with political momentum.

We don't know how much it takes to build Stuart-Kneeland or Pike Hugger East -- but they're possibly (if not likely) cheap enough to actually be built.

Assuming an Essex redux in 2025 will get killed for high costs again just like in 2010, there are really two possible scenarios here:

• World A: Either Stuart-Kneeland or Pike Hugger East is cheaper than Essex. The former will have a real chance to get built; the latter will not.
• World B: Neither Stuart-Kneeland nor Pike Hugger East is cheaper than Essex. Nothing will get built -- not even Essex.

(Of course, a cheap proposal still needs enough ridership for a decent cost-per-mile. But even if you forgo Oak Grove-Seaport rides entirely, ridership from Back Bay-Seaport, Back Bay-South Station, and all westside "Green Line" is likely enough to justify it. This real-world study from PMT 2003 -- which runs entirely on Stuart-Kneeland -- had insanely high ridership despite an even worse OL transfer.)

Repeating the exact same struggles of SL Phase 3 and expecting a different result shouldn't be the hill to die on. It's precisely because we know why it failed (regardless of the southern alignment) that we should look for alternatives that avoid Essex's problems.

But again, those mitigations aren't for the tunnel. They're for the station(s).

False. Section 6.2.3 (Page 6-8), particularly Table 6.2-2, lists other forms of building mitigation that go far beyond underpinning alone.

This clearly involves tunnels between stations, not just the station boxes, for at least two reasons:

• If building mitigations were only at stations (as you claimed), all entries in Table 6.2-2 other than the Core Tunnel Segment would be (close to) zero. None of the four alternatives -- which are all south/west of Boylston station -- involve stations.
• The explicit mention of the word "tunnel" in the explanations above do not seem to describe only the cut-and-cover station sites.

Table 6.2-3, the one where I commented about most buildings being at station sites, is specific to Foundation Underpinning. Notice that the 17 buildings requiring Foundation Underpinning is fewer than the 36 buildings requiring Compensation Grouting in the Core Tunnel Segment (Table 6.2-2 above).

My guess is that Foundation Underpinning is the worst (and thus likely most expensive) form of building mitigation, but Compensation Grouting, Soil Stabilization etc. are still costly, just a bit better. Essex does poorly on all of them.

(The "Stuart St Alternative" refers to a Boylston-Charles-Stuart-Tremont route, utilizing a short segment of Stuart St between Charles St and Tremont St. As seen in Table 6.2-2, it involves the fewest building mitigations of any alignment, despite passing by both Wilbur and Wang theaters.)

I'm not sure you can build a station on Stuart St without underpinning buildings either.

Essex St is 45 ft building-to-building just east of Washington St, where part of the station box was supposed to be.

Stuart St is 90 ft building-to-building west of Washington St. Kneeland St is 75 ft building-to-building east of Washington St.

It should be common sense that if buildings are further apart, your station box has less need to interact with building foundations. For reference, Malden Center's OL tracks and platform take up less than 50 ft of width. Even if 75' is somehow not enough for a single-level station, you can still do stacked platforms.

Likewise, notice that in Table 6.2-2, the 45-ft-wide Core Tunnel Segment has higher density of building mitigations (on a per-mile basis) than all southern alignments (all of which have at least 55' of width building-to-building, and all except "Tremont North" have 70-75').