Over the past week, I’ve been iterating on modified versions of the T’s official subway map to illustrate the closures and shuttle services that begin tonight and will continue for 30 days. This map will likely continue to evolve, and I will continue to post the latest revision here. As always, please note that this is not an official map — always refer to the MBTA’s website and to the City of Boston’s website for up-to-date information.
Notes for transit and design nerds
This exercise started relatively simple: show the Orange Line and northern Green Line in some alternate manner to indicate the bustituted segments. This was relatively straightforward: I borrowed design language from the Arborway bustitution in the late ’80s, with a colored outline, white fill, and colored circles for the stops.
On the further advice of someone with better aesthetic sense than I, I shifted the white fill to a lightly colored fill, to better differentiate the lines, and avoid the perception of a total absence of service. The light fill seemed to strike a good balance between maintaining the line’s identity, showing the continued existence of service, and also indicating a significant difference in service.
But, as happens with many projects, I kept on thinking of, “Oh, just one more thing I can add!”
Which brings us to the current design, which pushes the original map’s information design to the limits. I wanted to show:
The bustituted segments
The un-bustituted segments
Text notes on significantly relocated shuttle stops
The one-way service at Haymarket
The early-morning/late-night shuttle to Chinatown and Tufts Medical
The bus routes the T suggests as alternatives to the Orange Line (39, 43, 92, 93, CT2)
The suggested walking transfers between Orange Line and Green Line stations
That is a lot of information to cram onto a diagram that was originally designed to be rather sparse. The current official subway map is an evolution of a design from the early 2000s that primarily showed the rapid transit routes, with commuter rail and ferries being shown secondarily, and limited-access highways being shown tertiarily. In the late 2000s, the key bus routes were added, and a subsequent redesign shifted some parts of the map around while maintaining the same visual language overall.
Evaluating my attempts
Was I successful? Ehn.
I was pleasantly surprised when an earlier version of this map gained a small amount of traction of Twitter, so it’s nice to know that at least some people found it useful. But at a certain point, I fear the level of detail hinders rather than helps. Part of the brilliance of Cambridge Seven Associates’ original “spider map” design was in its simplicity; even if you didn’t memorize the whole thing, the visual concept was highly memorable: four lines, crossing each other in a square and radiating out. That basic schema was easy to recognize and recall, and created a foundation to understand the rest of the system, even if it wasn’t put into one single map.
The eventual addition of commuter rail lines, key bus routes, and now all of the additional information I’ve added here is all very reasonable, especially when done incrementally. But I find myself questioning the ultimate usefulness of the diagram I’ve created. Is it really useful enough for journey-planning? Or is it too confusing to parse?
Simple maps and specific signage
Ultimately, I’ve come to believe that clear and specific wayfinding signage in and around stations is much more important than a detailed system diagram, both under ordinary and extraordinary circumstances such as the Orange Line Closure. (This despite my own love for detailed system diagrams.) In that way, perhaps my earlier, simpler diagrams were more effective.
Shuttle routes only
In this simplest version, the shuttle routes are shown and nothing else:
The advantage of this design is how minimally it alters the original, and (hopefully) how starkly clear the changes are: the most important thing is that the Orange Line and northern Green Line are different and need to be planned around. The question all of this hinges on: can the diagram provide enough information to adequately re-plan the journey? And that’s the part I don’t know.
The second-simplest iteration added the walking transfers:
Including the walking transfers worked better than I expected. Quite frankly, I’d like to see these added to the official map (though hopefully a little more elegantly than I’ve done here). There are a lot of walking transfers that ought to be indicated on the system diagram, such as the ones I’ve included here, but also additionally:
State – Downtown Crossing
Government Center – Park
Riverway – Brookline Village
Reservoir – Cleveland Circle – Chestnut Hill Ave
Kenmore – Lansdowne
These transfers would not be suitable for everyone — and it should be noted that they are not free transfers under the current model — but if you are able-bodied and have a monthly pass that doesn’t charge per ride, these transfers are useful, speedy, and potentially can relieve congestion on key sections of the network.
Adding these transfers to the map is a good idea in general, but does it help in the case of the Orange Line & Green Line Closures? Again, I’m not quite sure. In most of these cases, I would guess that regular commuters are pretty familiar with the areas in question, and likely are well-aware that, for example, State and Gov’t Center are practically a stone’s throw apart. And if you aren’t a regular commuter… well, the pretty clear (and dire) direction from both the City and the T has been, “Please, stay away.”
Working on this diagram has been fun. It also has been nice to see positive response from numerous folks on Twitter. (Shout out to Jeremy Siegel at WGBH for sharing it with his followers!) And at least some of those positive responses have made comments to the effect of, “This is easier to understand than the materials the T has put out.” A few comments on Twitter aren’t necessarily a representative sample; however, the negative reaction to the T’s materials have been widespread and resounding — the Boston Globe going so far as to publish a parody of the official closure diagram.
That negative reaction suggests that there is room for improvement in how the T communicates these closures. I’d argue that the positive reaction to my diagram has been driven by its recognizable similarity to the “normal” map, combined with the clear-and-obvious differences that are blatant and draw attention to themselves.
With rumors swirling of partial shutdowns of the Green and Red Lines later this year, perhaps the T might consider adopting a similar strategy to what I’ve presented here.
“Wait”, you say, “that’s not right. The Green Line turned 100 in 1997, with the centennial of the Tremont Street Subway’s opening.”
True enough. But the Tremont Street Subway, for its first quarter-century of operation, looked quite different from the modern Green Line. It was only in the early 1920s that it began to resemble the system we know today, and it was only in 1922 – not 1897 – that the modern Green Line was born.
An Underground Street
BERy (the Boston Elevated Railway Company, the MBTA’s primary private predecessor) saw the tunnel more like a substitute for the crowded street above than as a proper rapid transit subway.
Streetcars funneled in from all over the city and distant suburbs, squeezing into the subway and crawling along at the speed of a modern bicycle. Contemporary accounts describe crowds surging down the platform at Park Street when the boarding location of the next trolley to such-and-such suburb was announced. The destination board at Park Street in 1899 resembles a departure board at a mainline station like South Station or Grand Central much more than that of a rapid transit station:
The streetcar subways draw a clear contrast with the services which BERy did consider rapid transit (the predecessors to today’s Red and Orange Lines). Beyond the difference in rolling stock (high-platform third rail vs low-platform wired), the rapid transit lines were also distinguished by their use of transfer stations.
A Tale of Two Transit Trips
Compare these two rider experiences:
A commuter from Somerville boards a trolley at the intersection Highland Avenue and Willow Avenue.
The car trundles down Highland, stopping every couple of blocks to pick up passengers.
After 2.7 miles, the car reaches Lechmere Square…
…where it departs from the street and enters the streetcar-only Lechmere Viaduct…
…which snakes over the Charles and around the West End before…
…diving into the subway just north of Haymarket.
After another 1.5 miles, the commuter disembarks at Scollay Square
A commuter from Dorchester boards a trolley at the intersection of Blue Hill Avenue and Seaver Street (equidistant from downtown to the Highland/Willow intersection in Somerville).
They actually have a choice of trolleys – bound for Egleston, or bound for Dudley (now Nubian).
They travel by streetcar for 1 to 1.7 miles, stopping every couple of blocks to pick up passengers,
before arriving at their rapid transit station where they get a (free) transfer to the Elevated.
The Elevated speeds into downtown, with stops roughly every three-quarters of a mile.
The commuter disembarks at State Street.
The Somerville commuter takes a long slow ride from the suburb to the core, while the Dorchester commuter takes a short slow ride followed by a short fast ride, enabled by a transfer station.
A Tunnel Filled With Buses
For BERy, in those first 25 years, the Tremont Street and Boylston Street Subways were just ways of getting the huge volumes of streetcars off of the streets in downtown. In today’s terms, those tunnels were essentially filled with buses. This was useful transit service, to be sure, but it wasn’t rapid transit.
The same was true of the East Boston Tunnel. Extended from its Court Street terminus in 1916 to a portal on Cambridge Street (with a turnback loop at Bowdoin, still used today), the tunnel’s primary use was to bring East Boston trolleys under the Harbor into downtown. In fact, early in the planning of the East Boston Tunnel, one option that was considered was to have trolleys exit the tunnel from Maverick through a portal in the North End, and continue on street-level into downtown – similar to today’s Sumner & Callahan Tunnels. The primary use of the tunnel was to get trolleys (or “buses”, if you will) under the harbor – not to create rapid transit service.
In the early 1920s, that all began to change.
Conversion to Rapid Transit
Birth of the Blue Line
The 1924 conversion of the East Boston Tunnel to rapid transit is well-known and offers a clear example of how a streetcar tunnel can be transformed into a rapid transit subway. The vestigial surface route on Cambridge St was converted to bus, and the half-dozen streetcar routes to the east were cut back to terminate at a new transfer station at Maverick. Thereafter the local streetcar services fed into a dedicated rapid transit service, mirroring similar designs at Dudley, Harvard, Sullivan, and others.
This, I would argue, marked the birth of the modern Blue Line; while it is true that today’s State (f.k.a. Devonshire) and Aquarium (Atlantic) stations opened some 20 years prior, it was only with the 1924 conversion that the service became anything like its modern form.
(There’s also an argument to be made that the true birth of the modern Blue Line actually occurred a bit further to the east, during the early and highly successful years of the Boston, Revere Beach & Lynn Railroad, which was essentially running Indigo Line-style service 100 years before its time.)
As a matter of comparison: mainline rail service with today’s rapid transit stop spacing had been running along both the Highland Branch and what is now the Southwest Corridor for 70 years and 100 years respectively before their conversion to rapid transit; but I don’t think we would say that either the modern Green Line or Orange Line were born circa 1888. We would consider those to be predecessor services, and I would argue that we should view the streetcars in the East Boston Tunnel as a similar predecessor service (albeit of a different character).
Birth of the Green Line
Less well-noticed – but I would argue equally important – was the 1922 construction of the transfer station at Lechmere. Like at Maverick, local streetcars were now short-turned at a rapid transit station where passengers transferred to service that was dedicated to bringing riders downtown at high speed.
In fact, in those early years, BERy ran a dedicated service to Lechmere for this purpose; a “shuttle” service ran from Lechmere to the Pleasant Street Portal for the first six months, which was rerouted to Kenmore in early 1923. Note the similarity to the East Boston Tunnel’s rapid transit service – “shuttle” services whose sole purpose is to run between downtown and a transfer station. More information in a contemporary newspaper account here.
Over the following ten years, BERy experimented with extending Lechmere service on to the Commonwealth and Beacon branches, which eventually both through-ran to Lechmere until the early 1960s. With their dedicated medians, the Commonwealth and Beacon branches were indeed the most “rapid-transit”-like of the various services feeding into the Central Subway at the time. They still intermingled with “local bus”-like services to Watertown, Huntington, Egleston, Dudley, City Point, and Sullivan, but the clear intent was to create a “rapid transit” service, as much as possible.
The efforts to replicate the success of the “rapid transit transfer station” model were originally envisioned to go even further. As discussed previously in my Blue Line series, plans were made for another transfer station in Allston, which is why Kenmore station – not constructed until 1933 – was built with a loop for the Beacon line; the hope had been to convert Kenmore into a transfer station for a Beacon streetcar and a Commonwealth rapid transit line, Maverick-style. You can see a 1926 proposal for such a network here:
The construction of the Lechmere transfer station marked the turning point from BERy treating the Tremont Street Subway as a collection of independent streetcar routes into BERy treating the Subway like a trunk line with multiple feeder branches – in short, the modern Green Line.
Death of the Streetcar Network
Following the cutback of the Lechmere services, the once-expansive network of streetcar routes running into the subways (a subject for a later post) started a rapid winnowing in which the same story played out again and again: a transfer station was constructed and the streetcar route was cut back to the transfer; once a route no longer needed to travel into the subway, bustitution almost always quickly followed.
This is one place I want to draw our modern attention to. I’d always thought of bustitution as a phenomenon of the latter 20th century, driven by the post-war embrace of the automobile. In fact, the vast majority of bustitutions happened between 1922 and 1941. If anything, the pace rapidly slowed following the war, and it is in fact remarkable that Arborway survived all the way to 1985.
1922: Lechmere Network cut back once transfer station opens
1924: East Boston Network cut back once Maverick opens
1925: the Ipswich Street Lines (the predecessors to today’s 55, 60, and 65) were truncated at Massachusetts (now Hynes) station
1932: the routes along what is now Route 9 to Chestnut Hill were bustituted and redirected to Kenmore’s surface station
1935: last year that foreign streetcars from the North Shore ran into the subway, halted by the loss of the bridge over the Mystic River
1938: the local streetcar running below the El along Washington St between Dudley and the subway was substituted with a bus (described here)
1941: the Huntington Avenue Subway opened, and streetcars stopped using the Public Garden Incline.
At this point, the winnowing slowed significantly, and the system had largely transformed into the modern Green Line we know today, with a few extra branches (to Egleston, City Point, and Charlestown) hanging around.
Late ‘40s: the two Charlestown branches are eliminated, just barely outliving BERy itself
1953: the City Point branch is eliminated by the MTA
1956: the Egleston branch is cut back to Lenox Street
1959: the Riverside Line is converted to light rail, expanding the MTA’s reach to Route 128
1961: the Lenox Street branch (formerly running to Egleston) is cut back to the Pleasant Street Portal, and briefly runs as a shuttle service between Boylston and Pleasant Street
1962: the shuttle to Pleasant Street is canceled, and the Pleasant Street Portal – once an anchor of the streetcar subway – falls into disuse
1969: the “A Line” – having survived long enough to actually be called “the Green Line” – is eliminated
And in this context, we now see that Arborway’s survival all the way to 1985 almost seems improbable by comparison – the last in a 60-year effort to remove all streetcars from the subway.
The cruel irony is that this effort – originally intended to speed up service and improve reliability in the subway – contributed to an overall degradation of transit access in the region over the following century. Once streetcar routes were taken out of the subway, they were very easy to replace with buses, and once they were replaced with buses, it was very easy to quietly degrade or eliminate service.
Conclusion: 100 years of the Green Line
In writing this, I read a lot of old reports written about Boston transit in the early twentieth century. One thing that struck me is that the reports written in the 1920s have much more in common – in terms of priorities and perspectives – with today’s approaches than they do with the reports written by their predecessors a mere 30 years beforehand.
By the 1920s, recognition had set in that the streetcar subways could and should be converted into rapid transit, using the same transfer hub model that had been successfully deployed at Sullivan, Dudley and Harvard.
(Interestingly enough, it had also become clear by this point that those elevated railways – built barely twenty years earlier – were awful and needed to be replaced as soon as possible; both the Southwest Corridor alignment and the current Haymarket North alignment to Sullivan were explicitly described in the 1926 report.)
The 1922 opening of the new Lechmere transfer station marked this pivot point, after which every single capital exercise carried out on the streetcar network was done with the aim of turning the subway service into a rapid transit service – or getting it as close as possible.
This continues to this day! The Green Line Extension project is undeniably a rapid transit project, and not just a resurrection of the former Lechmere streetcar network, and the same will be true if the Green Line is ever extended to Needham. Moreover, it seems all but certain that a Green Line extension to Nubian Square would need to find ways to make itself as “rapid transit”-like as possible – or else face a century of institutional inertia to overcome.
The notion of a “Green Line” with branches feeding into a trunk – as opposed to an urban streetcar network whose density called for a tunnel in key locations – arose in the early 1920s. Today’s Green Line is much more closely related to the LRT subway service of the 1920s than of the 1900s, in structure, operation, and public branding. It arises out of 1920s ideas about hub-and-feeder networks, which had previously been applied on other routes, and were then brought to the streetcar networks following their success.The creation of the rapid transit line that would become the Green Lineoccurred not in 1897, but in 1922.
And so, that is why, this summer I’m celebrating the Green Line’s (true) 100th birthday.
I am indebted to a host of transit historians, who have produced reams of carefully researched accounts detailing the stories of Boston’s transit system over the decades; most of them did so as volunteers, producing labors of love that exemplify the root of the term “amateur”. I want to specifically mention the names of Ron Newman, Bradley Clarke, O.R. Cummings, Frank Cheney, and Anthony Sammarco, as well as the volunteers who maintain the Wikipedia pages on Boston’s transit network.
In a previous post, I described the mathematical inevitability of decreased frequencies on branchlines. But I also alluded to a particular feature that certain systems have that allows them to bypass this particular pitfall. And that brings us to today’s post.
Let’s take a simple trunk-and-branch system, which sees 10 tph six-min headways on the core and 5 tph twelve-min headways on the branches:
If those branches are out in the suburbs, those lower frequencies may be justifiable. But what if your system needs to branch within the core?
Introducing the Aldgate Junction:
(“Aldgate Junction” is a term I’ve coined to describe this kind of infrastructure and service pattern; there are other related terms, but I think the way I’ll be using this term is a useful addition to our vocabulary.)
An Aldgate Junction enables a second layer of service to travel in one branch and out the other, without touching the core and (usually) without a reverse move. Sometimes this can be created via a relatively minor infrastructure change, such as adding a third leg to create a wye junction, but sometimes a more significant capital project is required.
The original Aldgate Junction
The original “Aldgate Junction” is actually a trio of junctions and is located at the interchange of the District, Metropolitan, Circle, and Hammersmith & City lines:
The Aldgate Junction concept is deployed here to create a triangle of one-seat rides:
Liverpool St to Aldgate East via the H&C
Aldgate East to Tower Hill via the District Line
Tower Hill to Liverpool St via the Circle Line
Modifying my original diagram above to match the colors allows us to see how the Tube’s topology maps on to the generic model I’m discussing here:
The downside of the original Aldgate Junction in London is that it’s a level crossing, meaning trains block each other as they pass through, which creates a bottleneck. Some Aldgate Junctions utilize flying interchanges to avoid conflicts and increase capacity.
Real-life examples of Aldgate Junctions
There are lots of ways to deploy an Aldgate Junction. It can provide distributed service to decongest central transfer points, or can be used to serve pluricentric regions with multiple “downtowns”. It can also be used to increase frequencies within the corridors of the branchlines, which is valuable for local trips as well as cases where the branch line is unusually long.
In my previous post, I mentioned that BART – a rather exceptional rapid transit system due to its enormous reach and mid-low frequencies on each branchline – uses an Aldgate Junction to increase frequencies on its East Bay services.
BART’s Orange Line (also still sometimes called the verbose “Richmond-Berryessa/North San Jose Line”) plays the role of the “branch-to-branch bypass service”. For example: without the Orange Line, stations like Fremont would only see 15-min headways; Fremont is about 30 miles away from San Francisco, so 15-min headways might be appropriate for that service. But Fremont is only 15 miles away from San Jose, and there is a steady level of density running all the way up the East Bay – there are plenty of local journeys that sit squarely in “Rapid Transit Land”, and 15-min headways underserve that.
Providing the Orange Line overlay raises headways to less than 10 minutes throughout the entire West Bay, without impacting capacity in the San Francisco core. Key to this is the Oakland Wye, a three-way flying junction to enable crossovers without interference. This is an example where an Aldgate Junction required significant infrastructure.
There are numerous other systems that use variants of the Aldgate Junction.
First, we see that the Orange County, Inland Empire-Orange County, and 91/Perris Valley lines form an Aldgate Junction in Anaheim. This creates a three-way link between Orange County, LA County, and Riverside County.
But there is a second, more abstract Aldgate Junction, formed by the Orange County, Inland Empire-Orange County, and San Bernardino Line, between LA County, San Bernardino County, and Orange County.
In both cases, these exemplify the use of Aldgate Junctions to serve distributed areas of density. Los Angeles is the core of the system, but Riverside and San Bernardino are major urban centers in their own right, and the use of the Aldgate Junction model creates two sub-networks based in Riverside and in San Bernardino.
These examples, it should be noted, are cases where an Aldgate Junction is not functioning to increase frequencies; Metrolink’s frequencies are pretty poor, usually only about 1 tph at peak, and little-midday service. In this case the focus is on providing one-seat-rides across a decentralized network, enabling transit journeys to non-Los Angeles destinations that otherwise simply wouldn’t be possible.
New York & New Jersey
Finally, the reigning North American monarch of the Aldgate Junction is New York City.
The New York City Subway has several physical Aldgate Junctions, a notably consistent design feature of the old IND network, highlighted below.
The PATH system is built around a pair of Aldgate Junctions (forming an impressive set of underground flying junctions):
And the Hudson-Bergen Light Rail offers a classic example of a simple Aldgate Junction near Hoboken Terminal:
PATH and the HBLR offer an additional lesson about the operational realities of Aldgate Junctions; the maps above represent the daytime versions of those networks. On weekends (and in PATH’s case, at night), the Aldgate Junctions functionally disappear.
All journeys remain possible, but some require transfers.
PATH offloads its HOB-WTC journeys into a two-seat journey with transfer at Grove St; meanwhile, its two routes out of 33rd St are combined into one, with a reverse-move at Hoboken as part of the revenue service pattern. (Functionally, Hoboken is serving as a transfer point in the same way as Grove St; PATH just does the transfer for the passengers by throwing the switches rather than swapping out trains.)
HBLR maintains weekend service to Hoboken, but riders from the north must transfer at Pavonia/Newport and backtrack.
Aldgate Junctions provide a way to distribute service across a larger area and offer a way to increase service levels on branches without being constrained by trunk line capacity. Their physical form and implementation vary widely, from level wye junctions to complicated flying interchanges to dispersed routes spread across miles. These forms defer in capacity, reliability, and cost.
Despite these differences, Aldgate Junctions are united by consistent contributions to network topologies, and are a valuable tool when designing (and future-proofing) complex systems.
Some of the earliest public feedback on the MBTA’s Bus Network Redesign (mapped in a previous post) came from residents of Somerville (and, to a lesser extent, Cambridge) who were – almost unanimously – unhappy with the proposals.
Based on my anecdotal observation on social media platforms, it appears that initial reaction from other communities has been more muted; this may change in the coming weeks with the further feedback sessions the T has planned. But still, I thought it was interesting that there was such an immediate and resounding response by comparison from Somerville.
I don’t envy the Redesigners their task with the Northwest Quadrant um, Sextant (?). Uniquely, they had to design for a system that doesn’t exist yet: GLX will be a seismic shift in transit access for Somerville, and while some of its effects are predictable, some are not. That’s a pretty big wildcard to toss into the mix.
The early signs suggest the Redesigners missed the mark, at least from the perspective of the community. I wanted to dig in to this and – of course – wanted to make a map to add to the conversation. I believe that I have been able to piece together a visualization that offers context and some explanation for this initial pushback; based on these, I have also generated some modest suggestions for revisions to the Redesign.
Further details are below, but the core of my suggestion is “swapping” the proposed T39 and proposed 90; this provides an increase in service to most riders, but is a more conservative change that does not disrupt existing travel patterns. This extension can be “paid for” by a dramatic shortening of the proposed 87 to its core service area, a reroute of the 90 to a shorter well-established corridor, and the use of a lower-freq crosstown route and a shuttle service to address connectivity gaps to Sullivan and Assembly.
I had two goals with this map: visualize frequency, and visualize potential destinations. I chose these goals because they reflected themes I saw in the initial feedback: frustration at multiple transfers (meaning, need for direct access to a larger number of destinations), and loss of service, particularly for lower-income residents (which I believe in some cases was a result of the Redesign’s shift away from “mid-low frequency” routes – think 45-min peak headways, that kind of thing).
Line color denotes reachable destinations. This system isn’t perfect, but I think captured some key dynamics.
Dark red lines reach Red Line stations
Dark orange lines reach Sullivan Square specifically, and light orange lines serve other Orange Line stations
Dark green lines reach Lechmere
Dark blue lines serve Longwood Medical Area
Some lines see multiple colors, meaning a rider might board a bus for multiple destinations; a good example of this is the 87 & 88 north of Davis, where a rider might board destined for Davis or for Lechmere.
Line width indicates frequency.
The thinnest lines have peak headways more than 30 minutes
such as the 85 or 90
The next size up, forming a large swath of the network, denotes headways between 15 and 30 minutes at peak, including
the 95 along Mystic Ave,
the 87 along Somerville Ave,
the 83 along Somerville Ave and Beacon St
Thick lines indicate peak headways of 15 minutes or better – matching the target for the Redesign’s frequent network, and including major corridors along
The thickest lines – matching the width of the rapid transit lines – see peak headways of less than 10 minutes, including
the cumulative 87 & 88 between Davis and Clarendon Hill,
the 101 & 89 on Broadway running into Sullivan
Here’s a version of the current system only showing routes that see 15-min peak headways or better.
Here we see six distinct corridors emerge:
Medford Square (and Malden) – Main St – Broadway – Sullivan
Powder House Square – Broadway – Sullivan
College Ave – Powder House Square – Davis
Clarendon Hill – Davis – Highland Ave – Lechmere
Arlington Center – Porter – Harvard
Sullivan – Union – Harvard – Allston/Brighton and Reservoir
I then created a map using the same design language to visualize the Redesign proposal.
(I recommend opening the “before” and “after” images in separate tabs, and then switching between them to hone in on the differences. I tried creating a GIF, but was unsatisfied with the results.)
Impact to current high-freq corridors
Let’s first review the impact to the current high-freq corridors:
Medford Square (and Malden) – Main St – Broadway – Sullivan
Goes near but does not provide transfer to GLX at Ball Sq
Extended beyond Sullivan to provide a (long) one-seat ride to Lechmere and Kendall
Powder House Square – Broadway – Sullivan
Moreover, this route is actually composed of a pair of branching routes:
a mid-freq route to Davis
a low-freq route direct to Clarendon Hill
Davis and Clarendon Hill lose direct service to Ball Sq GLX and Broadway
Davis and Clarendon Hill maintain roundabout access to Sullivan via Mystic Ave, an increase of 0.5 miles
College Ave – Powder House Square – Davis
Intact and strengthened
Newly anchored at north by Medford/Tufts GLX and extension to Medford Sq
Worth noting that the twice-hourly one-seat ride beyond Davis to Porter and Harvard is eliminated
Clarendon Hill – Davis – Highland Ave – Lechmere
Significantly reduced, and partially eliminated
Clarendon Hill – Davis gets a pair of the Redesign’s “30-min-or-better” routes, which could approximate 15-min headways, but still does not meet the current service’s average sub-10-min frequencies
Highland Ave drops into the Redesign’s “30-min-or-better” routes, but does not connect to Lechmere, or to Union GLX, or even to Gilman GLX (though it comes close), but instead is rerouted over toward Sullivan… which it doesn’t actually reach, instead ultimately diverting to Assembly
Arlington Center – Porter – Harvard
Sullivan – Union – Harvard – Allston/Brighton and Reservoir
Sullivan – Harvard is maintained, and combined with through-service to Everett
However, ridership from one side of Harvard to the other is actually surprisingly high, so this elimination is not trivial
Of the six current high-frequency corridors, three are maintained and three are significantly impacted. This does seem counterintuitive – as far as I can tell, all other existing high-freq corridors across the system were maintained in the Redesign, and most were significantly strengthened. I suspect the difference in Somerville was the future presence of the Green Line Extension.
Green Line integration (or lack thereof)
One of my biggest surprises on seeing the Redesign was how little integration there was to the GLX stations. Ball Sq and Gilman Sq will both see bus routes pass less than a quarter mile from the station without offering door-to-door service.
In addition, despite the arrival of the Green Line, Union Square now sees less connectivity than it did before, including the loss of one-seat rides from Clarendon Hill, Davis, and Kendall. As Councilor Pineda Neufeld pointed out, this also reduces access to the Market Basket in Union Square, as well as the Star Markets on Beacon St and Elm St.
I would argue that the Redesign treats the Green Line Extension more like one of its high-freq bus routes than like a rapid transit line – as if the GLX has said, “Don’t worry buses, I’ve got this stretch”, and thus bus routes are routed away from it. That is not done with any other rapid transit lines, and with good reason: rapid transit access is concentrated in discrete areas around stations; a key role of surface transit is to provide access to a station from beyond its walkshed. Unlike rapid transit, the close stop spacing of bus routes provides continuous access along the entire route, which GLX lacks.
To be clear, for over a century now, Somerville’s surface transit routes have had to play double-duty, providing both feeder service into rapid transit stations, as well as longer-distance service to support journeys that would in other areas be covered by rapid transit. GLX relieves them of the second burden, but not the first.
To wit: the entire Red Line from Davis to Kendall is doubled by surface transit routes in the Redesign. It is true that GLX’s stop spacing is closer than the Red Line’s, but it’s still well-above typical surface transit stop spacing. The Green Line is not a wholesale replacement of local buses.
Highland Ave vs Elm St & Somerville Ave
In the current system, Highland Ave sees mostly 15-min-or-better service at peak on the 88, from Davis to Lechmere. Pre-pandemic, the average AM Peak frequency was 10 minutes. (The 90 layers on about one trip per hour to Sullivan.)
(A short stretch of half-a-mile between Elm and Park Sts on the northern half of Somerville Ave sees additional 20 minute headways on the 83, running to Central Square via Inman. North of Elm, the 87 and 83 split, though remain only a couple of blocks apart for some distance. Pre-pandemic, the 83’s average AM Peak frequency was also 22 minutes.)
In the Redesign, Highland Ave drops into the 30-min-or-better category, and loses direct connections both to the Green Line and to Sullivan. (I should again note that 30-min-or-better routes still may see high-frequency peak service, which perhaps is the vision in this case.)
By contrast, Somerville Ave sees a significant increase: the 83 holds its place, while the 87 between Union and Porter is supercharged as the T39, receiving 15-min-or-better service. Somerville Ave also maintains a direct connection to the Green Line – only one of two radial routes in Somerville that do so, the other being the T96.
In essence, the Redesign swaps the frequency tiers of Highland Ave and Elm St & Somerville Ave. As mentioned above, Highland Ave is one of the current high-freq corridors, so its exclusion is puzzling.
I am guessing that the Redesigners saw the extent to which the 88 parallels GLX, and believed the Green Line would be an adequate replacement. As discussed above, however, surface transit and rapid transit are different beasts; the 88 is being relieved of its rapid transit duties by GLX, but not its surface transit duties.
Moreover, while it’s true that the 88 parallels GLX for a lengthy distance, they diverge significantly after Magoun Sq GLX – GLX aims for Tufts and beyond, while the 88 aims for Davis and beyond. And as numerous people noted in the reactions I linked earlier, the 88’s corridor north of Davis is not meaningfully accessible from GLX.
Increasing frequency on Somerville Ave is not necessarily a bad thing – the problem is how the increase is achieved. The current Somerville Ave corridor runs uninterrupted from Clarendon Hill to Davis to Porter to Union to Lechmere. The Redesign shears off the southern half of this corridor, and joins it to an extended T39, running Porter to Union to Central and beyond to Longwood. That is a bold proposal, but it also dismantles a one-seat ride that has been in place for over a century.
This reflects a larger challenge introduced by treating the Clarendon Hill buses as feeders to Davis Sq, rather than acknowledging their actual roles as parallel spines threading the city together, radiating not from the Red Line at Davis, but from the Green Line at Lechmere.
Lack of through-running from Clarendon Hill
It is worth highlighting that the current system offers one-seat rides at least half-hourly from Clarendon Hill to:
Powder House Square
The Redesign reduces that list significantly:
Powder House Square
Sullivan station (bus journey lengthened by half mile)
Southernmost Broadway, near Sullivan
Ball Square (via a 6 minute walk)
Most of Broadway
For over 100 years, Somerville has been tied together by crosstown routes on Highland and Somerville. The Redesign makes the unfortunate decision to reduce the Highland corridor, and breaks what is currently a single seat journey across the city from Clarendon Hill to Union Square via Somerville Ave into an almost-certainly three-seat journey.
On paper, it may look like Clarendon Hill is all set with feeding into a hub at Davis; in reality, the Redesign takes numerous journeys that have been single-seat rides for – again I emphasize – over 100 years and makes them essentially impossible.
I want to emphasize that I don’t mean to besmirch the Redesigners or their efforts: this was a herculean task that is impossible to get perfect. Tradeoffs had to be made, and balances struck. I can understand how they arrived at their current proposal, and I also can understand why community members are deeply unhappy.
In this post, I have attempted to detail the nuances of the current system, to illustrate some of the features that I believe underlie many of the criticisms that have been levied against the Redesign. I want to conclude by offering some modest suggestions for revising the Redesign to address the concerns that have been raised.
Methodology and fudge functors
While I am not sure that it has been stated explicitly, I believe the Redesign was undertaken with a “net zero” assumption – meaning the Redesigners assumed they could only work with the existing buses the system has today, and not increase the number of buses overall. As such, I am going to frame my suggested modifications through a reallocation lens: if I propose lengthening one route, I will attempt to balance it out by shortening another.
There are also fudge factors at play here that may provide some wiggle room on these proposals. First, infrastructure upgrades such as bus lanes and transit priority signaling may enable faster service and thereby reduce the number of buses needed to maintain frequencies. Second, it may be possible to spread minor frequency decreases across multiple routes to free up enough buses to add a new route – enabling the creation of a new high-freq route without eliminating routes elsewhere.
These fudge factors cut both ways, of course – in some cases, they may make things harder, not easier. So while I’ve done my best to consider these suggestions as carefully as I can, I am of course limited to the data available to me; these suggestions should thus be considered preliminary.
Shorten the proposed 87 along Mystic Ave to terminate at Harvard St
Shorten and reroute the proposed 90 to Union Sq GLX via Somerville Ave (essentially recreating the current 87’s route)
Institute an MBTA-run frequent shuttle service between Sullivan and the Assembly Row development, potentially leveraging public-private partnerships
Cumulatively, these modifications generate a surplus of 6.18 “route-miles” at 30-min-or-better frequencies.
Utilizing the surplus
Reroute and extend the proposed T39, running Union Sq – Highland Ave – Davis – Clarendon Hill
Revise the proposed network to include a 60-min-or-better route running Davis – Highland Ave – Sullivan, along the corridor of the current 90 (it may be possible to run this service at 30-min-or-better)
Suggested revisions in review
The core of my suggestion is “swapping” the proposed T39 and proposed 90; this provides an increase in service to most riders, but is a more conservative change that does not disrupt existing travel patterns. This extension can be “paid for” by a dramatic shortening of the proposed 87 to its core service area, a reroute of the 90 to a shorter well-established corridor, and the use of a lower-freq crosstown route and a shuttle service to address connectivity gaps to Sullivan and Assembly.
As detailed in my appendix, I believe these changes
are achievable with a net-zero impact on the overall Redesign
maintain better continuity with established travel patterns and community corridors
still provide specific and general enhancements of service to most riders, per the Redesign’s objectives and philosophy
I commend the Redesign team on their thorough and innovative work. I hope they will carefully review these suggestions and the further feedback provided by the community, and consider adopting these changes into a revised version of the Bus Network Redesign.
Thank you, as always, for taking the time to read!
In short, the MBTA is undertaking its first bottom-up redesign of its surface transport network in 100 years. The vast majority of the T’s bus routes once were streetcar routes; some routes have seen minor-to-moderate modifications in the ensuing decades, but the need to avoid disrupting the live system — which governs the day-to-day realities of thousands of people — has always capped how much could be done. As part of the long-running Better Bus Project, the T has conducted a deep dive review of its existing routes, including their ridership and reliability (see the Better Bus Profiles) and user research speaking directly to riders and community members.
Years in the making, the T this month released a proposed redesign of its bus network — details at the first link. This is a massive undertaking, and I give the redesigners credit for very clearly trying to avoid the “We’ve Always Done It Like This” Syndrome that plagues so much of American transit planning (especially in Boston).
Like any proposal, it has its imperfections and flaws. From my perspective, there are some things I like, and some things I don’t. There already has been some initial community feedback, and the T has a docket of public meetings (in-person and virtual) through mid-summer to collect feedback.
For the most part, I don’t intend to use my platform here to evaluate the merit of these proposals. The most important voices here are those of community members — their opinions should be listened to first, and given paramount consideration. Instead, my hope is to add to the discourse by providing additional ways to view and conceptualize the redesigned network — mainly through maps.
(There is one area of the proposal which received swift and strong public criticism. I have a post, and a pair of maps, in the works on that, where I will attempt to illustrate the flaws that have been pointed out by the community, and hopefully offer some modest suggestions to improve the proposal to address those problems. Stay tuned.)
In this post, I will share a map I have created to illustrate the Redesign’s “15-minute network”: a series of bus routes that are proposed to have 15-min-or-better headways all day from 5am to 1am, seven days a week. I’ll use the map to highlight some system-level features of the Redesign, and hopefully provide a framework for deeper discussion.
Very few circumferential or crosstown corridors — almost everything was radial
Morning peak frequencies were often higher than afternoon
The bus network “breathes”
An entire subnetwork of high frequency services turns on and off during the peak, providing much more comprehensive service during rush hour, but a signficantly sparser network during middays, evenings, and weekends
The entire northern quadrant of the network — everything between the Red Line and the Blue Line — was bereft of (intentional) high-frequency all-day routes, with the sole exceptions of the 111 in Chelsea and the 116/117 in Chelsea/Revere
Some communities, like Everett, don’t show up on the “Gold Network” (high freq all day) because they are instead served by a more diffuse number of routes spread across the city, operating at lower frequencies
The absence of the North Shore network — meaning the absence of consistent high frequency service — was conspicuous
The Dorchester network is one-of-a-kind, with features that don’t exist elsewhere
The network itself is actually three networks superimposed: a “10-minute all-day network”, a “15-min peak, low off-peak network”, and a “low frequency network” — most routes fall very cleanly into one of these three buckets
(Interestingly, that last point about the Dorchester network[s] seemed to be on the mind of the Redesigners as well — they’ve also adopted three primary tiers: a “≤15-minute all-day network,” a mid-frequency network that I’m guessing will be “15-min peak, 30-min off-peak”, and a low frequency network that, like Dorchester’s, would mostly see hourly services. I won’t really go into much detail here, but in my previous analysis, I did note consistent characteristics about each of the Dorchester subnetworks, and I see many of those ideas applied systemwide in the Redesign.)
I mention all of these points because I believe the Redesign recognized these features as well, and explicitly designed their proposal to address them.
The Proposed 15-Minute Network
The Redesign calls for a series of 26 high-frequency routes that would see 15-min-or-better headways all day everyday from 5am to 1am. This proposal goes much farther than the system I described above — in particular with its commitment to late night and weekend service. Vanishingly few corridors see any level of service approaching this currently. Routes on this network would be indicated by a “T” prepended to their route number: the T39, the T111, and so on.
I’ve spent a while digging through the weeds of the Redesign, and have concluded that the 15-Minute Network is composed of three kinds of routes.
These are straightforward: routes that radiate out from the core, and which feed into major transfer stations such as Harvard or Forest Hills. On my map below, I have used a dark blue for these routes. Most of these routes are unsurprising, and many of them are identified on my Gold Network map above.
These routes offer crosstown service that goes around the core rather than pointing toward it. Some of these routes behave like radial routes as they approach their terminals; for example, the southern half of the T96 essentially radiates out from Porter and Davis. So these routes still will be used by commuters going to downtown — but they also will enable journeys between multiple subway lines that can avoid going all the way to downtown. The T1 is a classic example, connecting Roxbury and Cambridge without requiring riders to change at Downtown Crossing.
There are two proposed routes that do not fit cleanly into the categories above or below: the T109 and T101, both running through Sullivan. North of Sullivan, they behave clearly like radial routes, to Medford/Malden, and to Everett/Malden. South of Sullivan, they do something that bus routes historically have not done: continue on from one transfer station to another.
Traditionally, these would be considered circumferential routes. However, I have mapped them as radial routes — I believe the Redesigners are trying to reconceptualize these corridors as radiating instead from Harvard and from Kendall, passing through Sullivan somewhat incidentally. I of course have no insight into their actual thought process, but I think it reflects general trends they’ve shown in favor of longer routes that pass through multiple quadrants of the system.
Longwood Medical Area Routes
This is by far the most seismic shift in the redesigned network. For the first time, Boston is proposing a transit network that acknowledges that Longwood is a major employment center, a third “downtown” equivalent to Back Bay and the Financial District. See for yourself:
The Redesign proposes extending most major routes from Ruggles beyond to Longwood; it extends Cambridge crosstown service beyond Central to Inman, Union, and Porter; it adds a new crosstown route to the Seaport; and it increases frequencies on most existing routes.
When I did my analysis in 2020, the Kenmore-Longwood-Ruggles corridor saw 15 buses per hour during the morning peak, 4 per hour midday, and 6 per hour during the evening peak. The Redesign notes that exact routings through LMA are tentative pending further study, but by my count, the number of buses per hour between Longwood and Ruggles, Nubian, or Roxbury Crossing is proposed to increase to at least 20 buses per hour, all day.
I’ve colored the Longwood routes in dark red on the map. Some also act as radial routes to other hubs, and some do double duty as circumferential services in the larger network. But Longwood is undeniably the center of gravity, and makes for a distinct subnetwork, worthy of its own identification.
A few additional notes here:
This map is my creation, based on materials published by the MBTA; it is not an official map and any errors are mine. I recommend using my map as a jumping off point before reviewing the official materials.
Each route is marked separately and indicates a minimum of 15-minute headways (4 buses per hour) all day every day — I am sure that many if not most routes will see significantly higher headways during peak
Certain major bus stops are indicated, largely for the purpose of indicating transfer points to rapid transit; stop placement is not exact
Some potential “transfers” would require some walking, indicated by a dotted black line
The proposed Blue-Red Connector and potential Silver Line Extensions are indicated with dotted lines
Some corridors see high-frequency service provided by layered mid-frequency services; these are mostly indicated by the line splitting and ending with arrows
This map is not precise and is meant to illustrate the overall network, not detail individual routes
Some routes have been simplified to reduce clutter. For example, the T39, T9, T70, and most of the routes in Chelsea, all have significant stretches where the route splits on to parallel one-way streets; most of these, I have simplified by drawing the route through the block in between the streets
Here’s a detail view on Back Bay and Longwood, probably the most visually cluttered part of the map:
I do want to emphasize again that I am not trying to evaluate the quality or suitability of the Better Bus Project’s Bus Network Redesign proposals. There are elements of the proposal that I believe are transformative in that they are shifting the conversation in ways that are vitally necessary: centering Longwood, insisting on consistent high frequencies all day everyday, and creating wholesale new corridors that do not descend from the old streetcar network.
But the devil is always in the details, and there are many details to sort through in this proposal. My hope is that my overview and map can make it easier for you to wrap your head around this sprawling project, and from there, dive into the details, well-armed with a larger context.
In this diagram, the trunkline sees high-frequency headways of 4 minutes (which is better than many subway lines in North America). With such a high frequency, it’s easy to think that there’d be enough trains to serve a bunch of branches.
But as you can see, four-minute headways equals 15 trains per hour. If you have three branches, that means each branch gets 5 trains per hour – which yields 12 minute headways. If these branches are out in suburbia, 12 minute headways might be appropriate, but you’ve nearly reached the limit. If you were to add a fourth branch, each branch would see less than 4 trains per hour, at 16 minute headways, at which point you really no longer have a claim to “frequent service”.
In my last post, I recommended, as a rule of thumb, no more than two branches per line. However, I didn’t explain why. It comes down to a combination of typical throughput capacities, and mathematical inevitabilities.
Divvying up trains-per-hour among multiple branches
Consider this chart:
On the left side, we have a list of potential trunkline capacities, measured in trains per hour. These indicate how many trains in each direction you can squeeze through your trunkline in an hour. (Don’t worry about converting these numbers to headways – I’ll get to that below.)
For some perspective (most numbers pre-covid):
BART ran 16 tph through its core section from Daly City to West Oakland
CTA ran between 12 and 20 tph on its Red and Blue Lines during peak
WMATA ran 15 tph on its Red Line during peak
MBTA ran 15 tph on its Red Line during peak
London Underground’s throughputs vary widely from line to line, with some lines seeing over 30 trains per hour, following major infrastructure and modernization improvements
Beijing’s subway runs between 30 and 35 tph on several of its routes
Shanghai’s subway runs between 15 and 32 tph on most of its inner routes
All of which is to say, the top few rows represent trunkline capacities that require major investment in transportation infrastructure.
To the right of those trunkline capacities are the number of trains available to each branch, depending on how many branches you have. So, for example, in the first row, a 40 tph trunk will provide 20 tph to two branches, 13.3 tph to three branches, 10 tph to four branches, and so on.
As mentioned before: you’ll notice that, as you move from left to right across the chart, the numbers in each row drop dramatically. In fact, the decrease is literally exponential; you can describe the chart above using
y = n-x
where n is the capacity of your trunkline in tph, x is the number of branches, and y is the resulting tph per branch.
Decreasing frequencies due to decreasing tph
It’s helpful to start this discussion using tph as a measure, because it’s easier to recognize the patterns in the numbers’ decrease. However, once we convert those tph into headways, it’ll become that much clearer why branching quickly leads to decreased frequencies.
I’ve added some (opinionated) color-coding, meant to suggest the various “levels” of service these different frequencies provide.
The bright green (every 5 minutes or better) represent the highest tier of frequent service an agency might provide. “Turn up and go.”
The pale green (every 5-10 minutes) are still comfortably in the realm of rapid transit, but are probably better suited to off-peak periods and lower-ridership networks. “Turn up and wait a couple minutes and go.”
The yellow (every 10-15 minutes) are the lowest tier of what could be considered “turn-up-and-go,” describing services where riders don’t need to check a schedule when planning journeys. Call this tier “turn up and wait.”
This tier should be approached cautiously, with careful attention paid to the specific corridor where these frequencies would be deployed.
(Sadly, consideration should also be given to reliability; if a bus is scheduled to show up every 13 minutes, but one run gets dropped, suddenly you have folks waiting nearly a half-hour for their bus.)
The subsequent tiers are rarely going to be considered “frequent service”; some could be “salvaged” by adhering to a strict clockfacing schedule: a route that reliably comes exactly :13 minutes and :43 minutes past the hour can be a useful service that isn’t “turn-up-and-go”, but doesn’t require consulting a schedule either. These tiers break down something like this:
Orange = “plan when to leave, but journey whenever”
Light grey = “plan when to journey”
Dark grey = “schedule around the schedule”
As you can see, our frequencies drop through the tiers I’ve described above quite quickly. Once you hit the yellow tier, you’re teetering on the edge of frequent service, and once you hit the orange and beyond, you’re definitely over the edge.
Different regions will have different definitions of “high frequency”. For example, it’s pretty rare to wait more than 4 minutes for a tube train in central London; a six-minute headway would be considered sub-par. In Boston, we are sadly accustomed to six-minute peak headways on the Orange Line, while Baltimore’s subway sees peak service every 8 minutes.
It’s worth highlighting that reaching 90-second headways at 40 tph on heavy rail is exceptionally difficult. If you have multiple tracks in the same direction, it becomes more manageable, but a standard two-track subway is nearly impossible to operate at 40 tph per direction, outside of the world’s most advanced subway systems. And notice – even the top examples I listed above, such as Beijing or London, you’re still looking at that second row as your baseline, at around 30 tph. Even in those systems, having more than two branches knocks each branch down into a lower “tier”, meaning it’s not suitable to do within the urban core.
If your crayon map requires shoveling 40 trains per hour through a trunk line, then probably it’s worth trying to plan a second trunk line!
Gaming out examples, and dealing with the unideal real world
Many of the North American systems I mentioned above see 4 minute headways on their core. If we go to that row, we can see two branches gives us 8 min (good so far), three branches gives us 12 min (borderline), and four branches gives us 16 min (no good).
Some North American systems see base headways of 7 or 8 minutes. In that case, our drop-off happens even faster: 2 branches becomes borderline, and 3 branches sinks us with headways longer than 20 minutes.
On the London and Chinese systems mentioned above, trunklines see headways around 2 minutes or better. In theory, those trunklines could accommodate five or six branches; but if you drop the core headways by just one minute, suddenly you can only accommodate three or four branches at similar frequencies.
And that “in theory” caveat is what’s really going to get us. Even if you can squeeze 30 tph through your trunkline to get 2-min headways, if you are feeding that trunk from five branches, that’s five times as many opportunities for delays and disruptions. We just saw above that the difference between a 2-min headway and a 3-min headway is worth two whole branches of throughput. That’s a very thin margin for error – meaning that you need reliability to be extremely high, or else the whole system will unravel, with cascading delays across your network.
The exceptions that prove the rule: North American legacy subway-streetcar networks
There are only two networks that I’m aware of which see sub-10 minute headways on four or more branches, feeding into a trunkline handling 40 tph: the MBTA’s Green Line and SEPTA’s Subway-Surface Lines.
This level of throughput is achieved (see caveat below) mostly because they are light rail lines rather than heavy rail. This means shorter trains which can start and stop faster, and therefore can be run closer together (albeit at lower speeds). The MBTA allows multiple Green Line trainsets to enter certain stations simultaneously, and SEPTA actually treats a couple of its subway stations as “request stops”, with trolleys rolling through non-stop if no one signals to board or alight.
(I haven’t done the math on this, but it would be worth someone calculating the actual capacity of those two systems compared to heavy rail equivalents. It is true that the TPH levels are higher, but since the trains are shorter, I don’t know that you actually end up carrying more passengers.)
The reality, sadly, is that both of these networks are infamous for their reliability issues. (Full disclosure: I’m much more familiar with the MBTA than SEPTA, but I believe most of the T’s problems also exist in Philadelphia.) Delays are common, both on the branches and then resultantly in the core, potentially resulting in less than forty actual trains per hour through the core.
Having four or five branches is somewhat workable on these two systems due to their unusual characteristics. Replicating that success elsewhere would likely require replicating those characteristics as well.
The other exceptions that prove the rule: hybrid rapid transit/commuter rail systems
BART’s core trunkline from Daly City to West Oakland feeds out into four eastern branches (Richmond, Antioch, Dublin/Pleasanton, Berryessa/North San Jose). Likewise, the London Undergroud’s Metropolitan Line runs to four termini in the northwestern suburbs (Uxbridge, Amersham, Chesham, Watford).
Both networks accomplish this by running service to distant suburbs that is more like high-frequency commuter rail than traditional rapid transit service. Each branch on the BART runs every fifteen minutes, which isn’t super unreasonable given that most of those branches travel over thirty miles from downtown San Francisco. (BART is also a pluricentric network, which makes the ridership patterns a bit different than, for example, Chicago’s.)
The Met in London is a bit more complicated, but the concept is similar. Similar to New York, and unlike BART, the Metropolitan is quad-tracked for certain stretches, and therefore runs both local services and express services. During peak hours:
Uxbridge sees 10 tph (6 min) spread across local and express services, with 4 of those trains short-turning at Baker Street
Amersham sees 4 tph (15 min), half of which are express
Chesham sees 2 tph (30 min)
Watford sees 8 tph (7.5 min), again with some short-turns and some expresses
A couple of further notes for context:
Almost all of the Uxbridge branch is also served by Piccadilly trains – 12 tph at peak – providing a robust 22 tph, which creates headways under 3 minutes
The Amersham and Chesham services run together until Chalfont & Latimer, before splitting off and going one stop each to their terminus, meaning most of the branch in fact sees 6 tph; these services are also supplemented by at least 2 tph that run on Chiltern Railways to Amersham and beyond.
Chesham is also the most distant Underground station – 25 miles from Charing Cross in central London, and technically outside of Greater London – well into territory where 30-minute headways might be reasonable
The town of Watford is served by three other stations; the largest of these is Watford Junction, which is less than a mile from the Met station, and sees 4 Overground trains and about 5 London Northwestern Railways trains toward London per hour
So, the Met has a few mitigating factors:
Its branches are supplemented by additional services
Despite four possible termini, really it just has three branches, one of which splits at the very end
The only stations that see 15-minute-or-worse headways are distant suburbs, over 20 miles from the core of London
Like SEPTA and the MBTA, both BART and the Metropolitan also have unusual characteristics that enable them to get away with breaking my “two branches max” rule of thumb.
When I was a kid, I wanted to send branches everywhere. As a child, it was an easy concept to understand — I knew that branches came (like the Braintree or Riverside branches) and that branches went away (like Watertown), and if the Green Line used to have five branches, maybe it could have six or seven or eight! And if the Green Line could have eight branches, why couldn’t the Orange Line?
Obviously, with age has come a modicum of learned wisdom; I stopped drawing branches everywhere and tried to focus on only sending branches where train service actually was reasonable. But there is more to the picture than that.
It is tempting to think of subway branches like roads. If you build a new road, it means that more places are connected. Providence and Hartford don’t have an interstate highway between them — you have to drive north or south first and pick up Route 84 or Route 95, and then head west. It takes longer than it would if there were an interstate directly between them, which is why you will occasionally hear proposals for one. Two disconnected places ought to have a connection between them.
Rail systems don’t work like that most of the time (and actually road systems don’t either, but that’s a different topic).
There actually is a surprising amount to discuss on this topic! Here are some rules of thumb:
Two branches max
For most systems, do not give an HRT/subway line more than 2 branches. A light rail line, or a commuter/regional rail line, can probably take more, but keep in mind the other rules of thumb below.
One evening I decided to jump down this rabbit hole and did a cursory review of pretty much every heavy rail subway system in the world. Almost none had HRT lines with more than 2 branches. Most of the exceptions were themselves exceptional systems, such as BART or London’s Metropolitan Line, both “subway” lines that act more like commuter rail lines, especially along their branches.
Branches have low frequencies, trunks have high frequencies. Trunks should only branch once they are far enough from the urban core that lower frequencies will be feasible.
Most transit systems in North America have a “full-build reach” (even if hypothetical) of a 10 mile radius from the center. Your trunk shouldn’t branch too close to downtown; every city will be different, but in general you’ll want your trunk to branch no closer than 5 miles to the core. If you need to branch closer to the core, you should try to create a second trunk line.
Remember that branching frequencies are almost always a hard-and-fast math problem: If your trunk line can take x trains per hour max, then two branches can each feed no more than 50% of x. That is often going to be enough to knock a “high-frequency” trunk service into a “mid-frequency tier” on the branches. As such, it’s worth conceptualizing a branch as something less than a “full” rapid transit line.
In the diagram below, note how the trunk line has a high frequency of 4 minute headways via 15 trains per hour, but that splitting it up into three branches very quickly drops frequencies to 12 minutes on each branch.
It’s important to recognize that branches are (usually) a game of subtraction, not addition. Your trunk line provides the pool of trains that you can send out to multiple branches. Unless you know that the trunk line is under capacity, you should assume that they are shoveling through as many trains as they possibly can, and can’t add more. So you need to think about redirecting the existing pool of trains, which is why this is a game of subtraction and not addition.
Note that I said this is usually a subtraction game – some newer/younger rail systems will not be at capacity. When that is the case, a branch proposal can work in your favor: proposing the addition of a second mid-capacity branch can combine with the existing mid-capacity service to create high-capacity service in the core (where presumably demand will be higher). Especially in newer metro systems, there may not be sufficient existing demand to justify the capital expenses of expanding the fleet to increase service in the core; adding a branch can broaden your revenue base, and enable the purchase of enough vehicles to run high-frequencies in the core. (Sometimes – these calculations are always going to be complicated.)
Look for short-turn services (and over-capacity extremities)
Transit agencies are incentivized not to run extraneous services. The further out from the core, the lower the demand for service (very much theoretically, but I digress). One solution agencies use are short-turn services, in which some fraction of vehicles do not run the full-length of the route. Sometimes this is done to reduce the number of near-empty vehicles running on the outskirts of the line, and sometimes this is done in order to increase the reliability of service on the inner sections of the line (and sometimes both). Short-turns are very common on bus routes, but do have their role in rail transit as well.
If your trunkline has an existing short-turn service, that is a very strong jumping-off point for a new branch: it points to a confirmed high-frequency trunk/mid-frequency branch demand model, and it doesn’t require additional capacity to be freed up in your trunk, since you aren’t trying to funnel a greater number of trains per hour through. (You will still need to expand your fleet size to maintain existing frequencies, although you may be able to adjust frequencies elsewhere in the network to account for the longer trips your short-turn services will now be taking on the branch).
But be careful – this technique works well if the short-turn is intended to avoid excess capacity on the extremities of the existing service. If the short-turn is intended instead to increase reliability within the core, then extending those short-turns out onto a new branch may jeopardize that reliability. And make sure to take note of the pitfalls of junctions, below.
Be cautious with reverse branching
Be cautious with reverse-branching, which is when a transit service splits into multiple branches going in to the city, rather than going out. (SEPTA’s Broad-Ridge Spur is a good example of this on-paper. In practice, it’s actually a little more complicated, but that’s a post for another time.)
Recall what we went through above: branches are lower frequency than trunks. Reverse-branching into an urban core thus means you are reducing frequency precisely in the area you need it most.
Note that I say “be cautious,” not “avoid.” I think that one of the major values of crayon maps is that they sometimes value creativity over feasibility. Reverse-branching is usually more creative than it is feasible, but sometimes it sparks good follow-up ideas. So, I wouldn’t forbid it as a hard-and-fast rule, but it’s good to be aware of its drawbacks.
Junctions are complicated
Junctions where branches come together will always have one of two drawbacks, so be careful where you put them:
If a level junction is used, then you will have capacity limits and sometimes delays, as trains need to wait for each other to cross, and may need to leave a whole signal block free (i.e. the train may not be able to wait right at the junction like a car at an intersection, but may need to be hundreds or thousands of feet away).
A flying junction avoids those problems and should be the standard for any new rapid transit junction. However, flying junctions are much more expensive, and also take up more space — horizontally and vertically — than level junctions. So, as with all things, it’s a trade-off between cost and quality of service. If you want your proposal to be taken seriously, make sure your proposal is ambitious enough to merit building the flying junction.
In conclusion, branches are not your friend
Branches are not the friend of the crayon-mapper. They look great on paper but require a lot of careful planning to be done well; when done poorly, they can be actively detrimental to the individual branches and the system overall. If you really want to use them on a subway network:
No more than two branches
Aim to branch at least 5 miles from the core
Look for segments where there already is reduced service (short-turns) or reduced demand for service (lower density)
If both branches demand full-frequency service, then both branches warrant their own trunklines through the core: i.e. they shouldn’t be branches, but should be separate lines
Post-script on LRT, BRT and mainline rail
All of the dynamics I describe here hold true on other modes of service. However, the cost-benefit calculations work out a little differently because there are different standards for these modes.
For the most part, LRT is going to have the same struggles as an HRT subway would. The main differences are all knock-on effects of light rail’s shorter and more nimble rolling stock. Shorter trains makes it easier to “fudge” capacity in a trunk line, with multiple trains stopping in a single station at once, and shorter signal blocks allowing trains to run closer together. The shorter signal blocks also mean that flat junctions – while still disruptive – can be mildly less problematic.
LRT also – by virtue of its lower capacity – is sometimes a better fit for serving lower-density regions, which can sometimes mean that the frequency cost of more than two branches is more manageable, especially if combined with a higher frequency on the trunk than you could achieve with larger trains.
Boston and Philadelphia both run a scheduled 40 light rail trains per hour through their tunnels (Philly apparently swings even more than that), with four and five branches respectively. San Francisco runs at a lower frequency (I think) in the Muni Metro subway, also with five branches. All three of these systems are plagued with reliability and other issues. (Muni recently tried to redirect two of its branches out of the subway, but now has returned them to the tunnel). That isn’t to say that these systems can’t be improved, nor am I denying benefits to their approaches. But it’s probably best to think of these as the exceptions which prove the rules.
Mainline rail faces all of the same dynamics I outlined above. The major difference is that the costs of branching don’t always end up being “disqualifying.” If you take a 20tph rapid transit line and give it 10 branches, each branch will get 30 minute headways, which no longer is “rapid transit.” But if you give a 20tph commuter rail line 10 branches, the 30-minute headways on each branch may be perfectly reasonable. The same dynamic is at play — it’s just that the requirements are different and therefore the costs are more acceptable.
On the other hand, some dynamics remain equally if not more problematic. Junctions, for example, will still be disruptive if built flat and costly if built flying. And reverse branching’s impact may be felt even more acutely: many American commuter rail lines run once per hour; split those between two downtown terminals – we’ll call them “Northtown” and “Southtown” –, and commuters now have to wait two hours between trains, which reduces flexibility for riders and may have an overall chilling effect on ridership on both branches.
Consider this fictional system below:
At first glance, it looks like a relatively robust commuter rail network. Each suburb sees 30-minute headways, and that segment between Newtown and Riveredge sees rapid transit-frequencies of 7.5 minutes. And every community sees direct service to both Northtown and Southtown.
But most commuters don’t care about going to two different workplaces – their destination is the same everyday. Let’s look at what the network looks like for a Northtown commuter:
This paints a significantly different picture. Most suburbs are reduced to hourly service, and the turn-up-and-go frequencies at Newtown are gone.
(To be fair, in this example, most American cities would be thrilled to see commuter rail service at this level – hourly commuter rail is nothing to sneeze at. But I’m keeping the numbers simple to keep the math simple; if most cities start with hourly commuter rail, then reverse branching drops the effective frequency of each branch down to every two hours, which is pretty rough. You get the idea.)
Instead of a robust trunk-and-branch network, reverse-branching reduces this network into two significantly diluted networks operating in parallel.
Some time ago (before the pandemic), I considered taking a job in New York City. Having no desire to relocate, and understanding that there would be some flexibility for how often I actually needed to be in the office, I pondered whether I would be interested in becoming a “super commuter”.
What is a “super commuter”?
According to most definitions I’ve found, a super commuter is someone who lives in one city/metropolitan area but works in another. This seems like a somewhat uselessly vague definition, at least in the Northeast Corridor, but (with apologies to Justice Potter Stewart) you definitely know a super commute when you see it.
Most super commuters do not go to the office five days a week, which is one reason they are willing to make the longer journey. For my part, I’d argue that a super commute is one which takes, let’s say, 2.5 hours or more one-way.
This definition probably encompasses more commuters than Moss & Qing’s analysis did, but it seems to me that travel time is more likely to affect behavior than raw distance or crossing MSA boundaries. Philadelphia-NYC takes 1h50m by Amtrak, but numerous Metro North journeys are comparable, such as from Poughkeepsie (1h50m), Wassaic (2h), Danbury (2h), New Haven (2h), or Waterbury (2h45m).
Amtrak schedules for super commuters
In any case, as I began to ponder this lifestyle change, I started looking at the Amtrak schedule. From what the hiring officer had told me, it would be alright for me to do some flexible hours when I showed up at the Manhattan office — for example, it’d be fine to arrive around 10:30, and then either leave at 3 on a short day, or put in the extra hours and leave around 6 or so.
Morning inbound journeys
Christopher Juckins’ Amtrak schedule archive lets us review Amtrak timetables from before the pandemic (and before Amtrak stopped publishing PDFs on their website). As can be seen on the Boston-Washington Northeast Corridor schedule, getting into Midtown from Eastern New England for a 9am start is barely doable, but additional options open up as the morning goes on:
In summary, journeys which arrive in NYC before lunch included:
5:05am to 8:47am
6:05am to 9:47am
6:10am to 10:22am
7:15am to 10:47am
I figured I would probably aim for those 6am departures, maybe with some of the 7:15’s mixed in. That seemed manageable to me.
Afternoon & evening outbound journeys
So then I took a look at the trip home. There are too many trips to took a screenshot, but in summary:
3pm to 6:46pm
3:30pm to 8:12pm
4pm to 7:40pm
5pm to 8:50pm
5:38pm to 10:10pm
6pm to 9:45pm
7pm to 10:50pm
7:50pm to 12:20am
Now, to be fair, some of those later trains really do get you home quite late. But that 4pm trip in particular struck me as perfectly fine — especially if your job is one where having uninterrupted time at the beginning and end of the day is valuable (for example, time to write or read). That could really work (and held some real appeal to an introvert like myself).
But the other thing that struck me was, “Damn. that’s a better schedule than the MBTA Commuter Rail.” Hourly departures, predictable journey times, a couple of extra trips layered in? Not bad!
And this got me thinking… there are several other corridors that feed into NYC; do they all have frequencies to support this super commute?
And by now, you surely can guess that that is true (otherwise there would be no post!). I have some further observations in the appendix below about individual routes (which I think are worthwhile reading, as they illustrate what a behemoth the Northeast Corridor is), but the main reason I’ve written all this is as prelude to a map:
This map treats Amtrak’s services into New York like commuter rail services — and I would argue that they essentially are indeed a “super commuter rail” network. All of the stations and routes marked on this map have the ability to support the kind of “super commute” I was considering for myself: leave home early, get to New York mid-morning, leave New York mid-afternoon-ish, get home late, repeat once or twice a week, depending on distance.
Scope of the network
As you can see, it’s actually quite a sprawling network — stretching from Albany to Washington, Boston to Harrisburg, close to 200 miles in each direction. Boston, with the itineraries I listed above, is actually on the extreme end of the network, with its ~4h travel times; a place like Wilmington, DE is a mere 1h40m journey, comfortably below Metro North’s longest journeys (and in significantly more comfortable seating). Hartford, CT and Lancaster, PA are both about a 3h journey.
All of the journeys depicted on this map are available via services with modest frequencies (to enable flexibility) and a short enough travel time to accommodate a same-day round trip.
Local and express tiers
One of the things that was fun about making this map was poring over the different timetables to look at which stations were frequently served. For example, the Northeast Corridor schedule for NYC-Washington lists out a whole bunch of stations, which most trains then skip. Despite the timetable listing 8 stations between New York and Philadelphia, most trains only stop at 4: Newark Penn, Newark Airport, Metropark, and Trenton.
In this way, Amtrak builds an informal tiered network akin to a local-express model. Virtually all trains stop at places like Stamford, Trenton, and Rhinecliff; some trains skip stations like Kingston, Poughkeepsie, or Downington; and then there are some stations, like Princeton Junction and Newark, DE, which may only get one train a day, or even less (spiritual successors to “whistle stops”). Acela service mainly restricts itself to the major stations, leaving the Regionals to pick up the leftovers.
The big picture
Amtrak (and/or the City of New York) would do well to publish a formal map like this, one which highlights that these are routes with high frequency service, modest journey times, and flexible schedules, and one which likewise differentiates between different service levels at each station. This network is a tremendous success story for Amtrak, and for American rail in general.
A map like this also illustrates what is possible with a strong piece of core infrastructure — in this case, the Northeast Corridor. Even communities which aren’t directly on the Northeast Corridor, such as Harrisburg, Springfield, or Albany, are able to benefit, as it becomes possible for them to “tag along” for the ride.
When advocates talk about high speed rail in places like Texas, Florida, or the Piedmont Corridor, it’s not just about connecting Atlanta to Charlotte, but about building a core piece of infrastructure that then enables branch lines to be built to Birmingham, Chattanooga, and Augusta. High speed rail infrastructure not only enables long-distance travel (for business or pleasure), it also enables daily commutes — and super commutes.
There are many ways to draw crayon maps – many more than could fit in one post. In this post I will talk about the software that I use to create maps for this blog.
Really we should be starting with something more fundamental like “pen-and-paper” or “the ideas”. But, assuming you’ve got some of the basics in mind, one of the big questions next is the software. (Although, don’t underestimate the effectiveness of a good hand-drawn diagram that is scanned/photographed and uploaded! No need for software if you feel like going old-school.)
It’s worth distinguishing two kinds of goals one can have with transit maps and diagrams. First, one can focus on elegant design solutions (especially for existing systems); TransitMap.net is a great blog that focuses on transit map/diagram design. Second, one can focus on conveying detail, particularly of proposed systems, with less focus on “looking good”. While these two obviously overlap in significant measure (elegant design also usually is particularly good at conveying detail), my focus is usually on the latter more than the former.
Put another way: I don’t worry too much about my diagrams looking pretty. In my case, this is somewhat self-serving, because I don’t have the know-how to make good-looking diagrams at a pro or even semi-pro level. But put more positively: this also means that you don’t need to worry about getting your diagrams to look good either! Don’t let that be a barrier to getting started!
Okay, but actually, what software should I use?
Okay, sorry for getting distracted. The short answer: start with what you have, and then try some of the ones I suggest here.
Microsoft Paint or Paintbrush (for Mac)
I can guarantee you that any designer who happens to read this post will physically grimace at this suggestion, so I do want to be clear that both of these programs are extremely limited; if you are going to use them, keep things simple: straight lines, 45-degree angles, simple text.
That being said, especially when you just want to sketch something out, Paint is a very useful tool, precisely because of its simplicity. With some effort and care, you can indeed use it to make clear, straightforward diagrams.
For example, this is something I threw together in MS Paint just now, illustrating a possible commuter rail network for Rhode Island, with the existing MBTA Commuter Rail line, a new RI Commuter Rail service from Worcester to Westerly, and an extra layer of local trains running between Pawtucket and TF Green:
Is it pretty? No, not really. Is it even particularly clean or neat? No – my spacing of station names is inconsistent, the size of the station markers varies, and the bends of the lines are a bit sloppy.
But does it convey the information I needed to? Yes. Station are clearly and legibly marked (yes, I got tired of labeling the MBTA stations), the service patterns are clearly indicated (at least I hope they are clear – can you tell which stops the MBTA skips?), and there are enough nods to geographic fidelity (such as the curve through Providence) to provide context for the viewer.
And while it’s true that a lot of the details are sloppy, the overall design is generally symmetrical and balanced, the stop spacing is mostly even, and the visual is streamlined and simplified – all of which means that even though the design itself is hardly inspired, it’s also not going to be too distractingly bad.
(To editorialize briefly: there is something to be said for transit diagrams and transit proposals which are straightforward enough to visualize using Paint.)
There are several downsides to using Paint, though. As mentioned above, you can get clear diagrams – “with some effort and care”. The time and effort is where you pay the price. It takes a while to do simple things nicely with Paint, and it takes a long time to do complicated things at all.
If you’re just getting started, I highly recommend playing around with Paint a little bit – get a feel for what “looks good” in the kinds of diagrams you’re making, but aim to “level up” once you can.
Metro Map Maker
This website lets you make transit diagrams from scratch, or remix those created by other users. It is very simple to learn, and can be pretty fun to use. That said, I myself don’t use Metro Map Maker, and I think there are a couple of things to bear in mind if you choose to do so.
First, I myself find the interface “fiddly” and difficult to draw things precisely the way I intend (which is my main reason for using a computer rather than sketching something by hand). I think this is visible in some of the maps in the gallery, some of which are quite clean and clear, and others of which struggle a bit. In particular, I have a hard time with the way Metro Map Maker renders Y junctions, where two branches feed into a single trunk; the way Metro Map Maker does it, it’s very easy to mistake those for three-way junctions.
The other thing to be aware of is that Metro Map Maker diagrams sometimes get poo-pooed by folks on aesthetic or “professionalism” grounds. There are some venues where such criticisms are appropriate – I would not, for example, recommend that transit advocates use this site to create visualizations of the services they are advocating for. But, especially if you’re just doing this for fun, I think it’s a perfectly reasonable tool. Just keep in mind who your audience is!
This is my application of choice these days, and the way I create most of the visualizations on this blog. (Unfortunately it is Windows-only, and I haven’t yet found a Mac equivalent that has the same “sweet spot” of features.) Paint.NET is available on its website and on the Windows store.
For me, Paint.NET has an excellent balance of ease-of-use and useful-features. Eventually I plan to “graduate” into one of the programs listed below, but for now, I’ve gotten pretty darn good with Paint.NET, so I’m sticking with it for now.
One of the biggest reasons to use Paint.NET instead of MS Paint is the ability to use layers. For example, let’s say that I’m drawing up plans for Super High Speed Rail between Boston and New York, but I’m trying to decide between an alignment via Hartford and an alignment via New London. In MS Paint, I’d have to settle for a workaround: draw both using different colors, make a copy of my file after drawing the Boston-Providence and NYC-NH segments and work on two separate diagrams, or just pick one and forget about the other. In Paint.NET, on the other hand, I can just create a new layer for each alternative, and show/hide them depending on which version I want to look at.
You can also use layers to add and hide different levels of detail. For example, in one of my big maps, I have the local bus routes drawn in a layer of their own; I keep it hidden most of the time to reduce clutter, but sometimes show it in order to see things in full context.
I also recommend isolating lines, station markers, and labels each into their own layer. This will make it easier later on to make quick changes en masse, such as changing the color of a line.
Paint.NET still has downsides. It produces raster graphics and not vector graphics, which means things are pixellated when zoomed, and you actually physically cannot create pure curves (although if you are zoomed out reasonably, no one will notice the difference aside from professional designers). And as far as I know, Paint.NET has no concept of “linked objects”, meaning that if you create a fancy icon as a station marker, you’ll have to copy and paste it each time you want to use it, and you’ll have to manually modify every single copy if you want to make changes later. (This is why I just use a simple black circle for my station markers.)
Overall, however, I’ve found Paint.NET to be extremely useful.
Inkscape and Adobe Illustrator
Now we have “graduated” into the top tier: vector graphics. These are the programs used by professionals, and damn can you make things look good with them! Vector graphics allow you to infinitely zoom without pixelation and draw perfect curves without jagged edges.
Adobe Illustrator is part of the Adobe Creative Cloud suite (along with the perhaps more famous Adobe Photoshop program). Inkscape is open-source, and available for free.
Becoming more comfortable with Inkscape is a medium-term goal of mine, but it’s not a tool that I am experienced with at this point. So, for once, I don’t have much to say on the topic. But if you get the opportunity to practice with either of these programs, I highly recommend it!
What about Google Maps?
Good question. You also may have questions about using Paint.NET specifically, or what I’m talking about in the footer of my blog about “basemaps” (and why I have that footer in the first place), or why I’ve been talking more about “diagrams” than about “maps”. All great questions! I hope to get to them in a subsequent post. In the meantime, hopefully I’ve given curious readers enough to get a running start! Have fun crayoning!
I believe extending the Blue Line from Charles/MGH to Kenmore via a Riverbank subway along Storrow Memorial Drive is the stronger “first phase” extension.
The truth of the matter is that we don’t have enough data to say which is better. As I’ve outlined above, almost any extension of the Blue Line west will be treading new ground, bringing rail transit to corridors that have never seen it before. This is virtually unprecedented in the history of Boston transit, so we are in uncharted waters.
Any construction at Charles/MGH should be done to leave open as many possibilities for future extension, including both to Kendall and Kenmore if possible; if a choice must be made, my preference is for Kenmore.
I have are four overarching reasons:
Complementing the Green Line
As outlined previously, the Blue Line and the Green Line have a long historical relationship of interdependence and interweaving. The Kenmore alignment maintains that relationship, and provides relief for the Green Line, freeing it up to adapt to the flexibility offered by LRT without needing to bear the burden of being a pseudo-HRT subway line.
The Riverbank alignment sends the Blue Line straight down the middle of the gap between Red and Orange, landing it at Kenmore – the major bus transfer hub between Central and Ruggles. This is where the Green Line’s burden would be relieved: it would no longer need to serve as the radial link between the Kenmore hub and downtown.
I have a large and expansive vision for the future of the Green Line, and many more options are opened up through the Kenmore alignment than the Kendall alignment.
Filling a gap in the HRT network
We often think of the MBTA as having 4 subway lines, plus a handful of BRT lines. Strictly speaking, that is not quite true. The MBTA has 3 heavy rail lines, a handful of light rail lines, and a handful of BRT lines. The T uses layered LRT and BRT services as stand-ins for HRT services in the Boylston Street Subway, the Tremont Street Subway, and the Piers Transitway, because the ridership along those corridors demands it, but they still are different beasts.
The future of expanded rail service in Boston lies in LRT and in frequent regional rail. These services are easier to create because they are either easier to construct or are better able to leverage existing infrastructure. With a small number of exceptions (Lynn, Arlington, Mattapan, and West Roxbury), bringing rail service to new locations will be much easier to do with LRT and regional rail (Needham, Watertown, Jamaica Plain, Waltham, Lexington, Grand Junction, Everett, Chelsea, and Dorchester).
Heavy rail rapid transit provides best-in-class service, there is no denying it. But it also requires the most up-front expense, and is rapidly becoming a form of boutique construction – a Blue Line extension will be usable by Blue Line trains alone, while new stations and improved tracks along the Worcester Line will be usable by metro Indigo Line services, local Framingham services, and expresses to Worcester (and will benefit Amtrak service to Springfield and beyond). Heavy rail expansion therefore must meet a higher threshold of viability and benefit.
As such, I think we need to reconceptualize our idea of Boston’s rapid transit system into three tiers: light metro, heavy metro, and regional metro.
Light metro: LRT and BRT services, with smaller vehicles and smaller infrastructure footprints, particularly useful for suburban radial service and urban circumferential service
Heavy metro: your standard HRT, as well as rapid-transit-frequency mainline rail, and Los Angeles-style light rail (e.g. longer trains, high-level boarding, dedicated ROWs).
Regional metro: reimagined commuter rail – 15-minute headways, limited stops within 128, service to suburbs and satellite cities.
These tiers are obviously interconnected, but do form distinct networks. And I believe we need to view them as such. As can be seen here, any Blue Line West extension will serve to fill in the large gap between the Red Line and the Orange Line.
When we exclude light metro services from the map, the gap in the heavy metro network becomes clear. The streetcar services that run into Kenmore from Beacon and Commonwealth will always prevent that stretch of the Green Line from becoming proper heavy metro, which means an alternative is sorely needed to fill in the gap.
As mentioned above, the future of rail expansion is in LRT, not HRT. Wholesale greenfield rail construction in Boston will mostly be LRT going forward. HRT expansions should therefore be sparing, and strategic, benefiting the entire network as a whole as opposed to specific corridors. Extension to Kenmore balances the heavy metro network overall, and frees up capacity on the Green Line to expand the light metro network into new areas across Greater Boston.
Put heavy metro service to its unique purpose
Let’s hearken back to the days of the El. You live on West Dedham Street in the South End, halfway between Shawmut and Tremont, a block and a half from Washington Street – you can see the El from your front door. A generation goes by, and the Orange Line is relocated out of sight, a bit less than a half mile away in the Southwest Corridor. On the balance, you’re pretty happy – the El was loud, rumbly, an eyesore, a relic of an age past.
But isn’t the loss of transit access a downside? (You are asked by your railfan friend.) Not really, you reply. Even though the El was a three minute walk away, it was nearly 10 minutes to the nearest stations, at Dover or Northampton. The El rumbled by your block every day, speeding through without making a stop. The new station at Back Bay is just as long of a walk, but now Washington Street is sunny and quiet(er). On the balance, you’re pretty happy.
(At various points through the 20th century, you might not have walked to the El at all, but instead caught a streetcar or bus at surface level and rode that in.)
The same was true in Charlestown: for nearly three-quarters of a mile along Main Street, the El expressed through, while local residents took the 92. Sullivan, with its massive hub of transfers from Everett, Malden, Medford, and Somerville, was the objective, not the northern half of Charlestown itself. Main Street suffered the drawbacks of rail infrastructure, and for the most part got none of the benefits.
To be clear, I am not claiming that the Orange Line Relocation Projects were net improvements for transit access; the sad story of Equal or Better makes clear that there were drastic downsides that persist to this day. My point is that the Relocation Projects demonstrate that the purpose of the early rapid transit lines was not to serve the neighborhoods they passed through, but to offer a bypass to enable faster journeys to downtown.
The primary role of heavy rail in Boston has historically been to whisk riders in to downtown from the Inner Belt regions and beyond. This is why the Red Line has such lengthy distance between stops in Cambridge, and why the extension beyond South Station headed down along Dorchester Ave instead of heading into Southie: the whole point was to express in to downtown.
Prior to the Orange Line relocation, the “Inner Belt” transfer hubs – with the exception of Kenmore – were at most two stops away from downtown: Dudley-Northampton-Dover-Essex, Kendall-Charles-Park, Sullivan-Community College-North Station, Maverick-Aquarium-State, Columbia-Andrew-Broadway-South Station. All those intermediate stops were themselves major transfer points or at major destinations. The whole point was to speed the journey downtown from the outer parts of the city, whence a bus or streetcar journey would be painfully long.
A Blue Line extension to Kenmore fills that gap, and provides the western feeder services (the B, the C, the 57, the 60) with a proper heavy metro express link into downtown, providing a speedy and reliable transfer consistent with similar hubs across the system. This need exists at Kenmore, but does not exist on the other side of the river.
Follow existing ROWs and generally have easier construction
To my knowledge, with three exceptions, all expansion of Boston rail infrastructure in the last 100 years has occurred on alignments and ROWs which were carved out by railroads in the 19th century. (The three exceptions are Harvard-Davis, the stretch between Chinatown and the portal east of Back Bay, and the Huntington Ave Subway.) The Blue-Red Connector would mark the fourth such exception.
It is incredibly difficult to build rail where none existed before, and it is incredibly rare. The Huntington Subway was decades in planning, Harvard-Davis was built under exceptionally favorable political circumstances, and the Orange Line connection was the lynchpin in a massive service relocation that had been in the works since the 1900s. Virtually no current expansion proposals – official or amateur – call for laying rails where none have been laid before. (The North-South Rail Link is the exception that proves the rule in this case.)
Blue Line West proposals are the only ones where we consistently see greenfield ROWs being proposed. For the Phase 1 alignments I discussed before, this is a necessity – there simply is no way to continue west from Charles/MGH without building new ROWs. Several of the Phase 2 alignments, however, do leverage existing ROWs.
The “Phase 2” alignments I discussed earlier fall into three categories: Watertown, Mass Pike, and Longwood. Kenmore offers access via existing ROWs to all three, due to its location at the confluence of the Highland Branch and the B&A (Worcester Line) ROWs. (Watertown would be the trickiest of the three, but still could utilize the B&A ROW for at least half of the journey.)
The Kendall alignments are more of a mixed bag. Longwood technically could be accessed from the BU Bridge via a subway under Amory Street, but at significant expense and with some tricky track geometry (i.e. sharp turns). Watertown and the Mass Pike could be reached via Kendall, the Grand Junction, and the BU Bridge, largely utilizing extant ROWs, but at the cost of two river crossings and the lack of access to the Kenmore transfer hub. (Circumferential service from a Kenmore hub to Longwood, Ruggles, and Nubian would be pretty easy to plan with BRT. Relocating the transfer hub to a Blue Line station at BU Bridge would be more difficult.)
One hundred years ago, the mainline gateway to Boston’s western suburbs was located in the Fenway at Brookline Junction station. Despite all that has happened in the intervening century, the ROWs that were laid out feeding into that junction have persisted over all these years. Because of that, Kenmore has remained a transit nexus, and remains the strongest “launchpoint” for any future expansion to the west.
Basically this all boils down to “serving Kenmore” vs “serving Kendall”. Both alignments can land you in Beacon Yard, setting the stage for service to Auburndale, Watertown, or Waltham. Both alignments could (at moderate-to-significant expense) lead to service to Longwood.
Kendall is a massive employment center, growing every year. Kenmore is a massive transfer hub that will only become more important as the T improves circumferential service. Kendall’s Red Line service is at capacity; Kenmore’s Green Line service is at capacity.
To a certain extent, I’m sufficiently convinced in favor of the Kenmore alignment by the added cost and expense of the dual river crossings needed for any Kendall alignment. Greenfield HRT is going to be hard enough on its own – at the very least, let’s choose the easier option.
But beyond that, I think the Kenmore alignment is simply better for the whole network overall. The Kendall alignments would benefit their immediate corridors, while a Kenmore alignment’s effect would be felt system-wide.
Consolidating Kenmore as a transfer hub makes it easier to build a flexible Urban Ring, in stages if needed.
Complementing the Green Line frees up resources to be redirected into a semi-heavy metro corridor along Huntington, and creates more flexibility for short-turning LRT services at Kenmore.
Focusing construction efforts along the Grand Junction on LRT rather than HRT creates a potentially unbroken LRT corridor from Allston to Cambridge to Sullivan to Chelsea to the Airport.
And a Phase 1 extension to Kenmore leaves open the most possibilities for Phase 2 extensions without the need for massive tunneling projects.
All of this is many years, if not decades, in the future. For now, I reiterate my original point: Any construction at Charles/MGH should be done to leave open as many possibilities for future extension, including both to Kendall and Kenmore if possible; if a choice must be made, Kenmore is the stronger option.