Scheduling Regional Rail for the Providence Line

Followers of this blog will recall that nearly a year ago I started doing simulations for various MBTA commuter rail lines under plausible operating schedules for TransitMatters’ Regional Rail proposal. Since I live in Framingham, I concentrated mainly on the Framingham/Worcester Line, which is one of the easiest to schedule: there are only two intercity trains per day, and the entire line is double-tracked the full distance from Worcester to Boston. (Substantial parts of it used to be triple-tracked, and there’s a study in the works to determine whether the third track should be restored to simplify scheduling express service.)

The Providence Line is significantly more complicated. Not only are there three MBTA branch lines that feed into it — the short Stoughton Line, the much longer and heavily used Franklin Line, and the Needham Line — but there are also a dozen Amtrak trips a day which get priority access to the tracks. The Needham and Franklin Lines are somewhat easier to deal with: as I’ve written before, the Needham Line should be turned into an extension of the Orange Line, and in the mean time even Rail Vision is assuming it can be short-turned at Forest Hills peak hours; the Franklin Line was historically interlined with the Fairmount Line through Dorchester and could easily be operated that way again. (In fact, a number of Franklin Line trips each day already operate this way; all that is needed is better platforms at Readville to allow transfers between the Fairmount/Franklin and Providence Lines.) The Stoughton Line and the Amtrak trains, however, have to be integrated with the Providence Line schedule, and that creates complications for implementing Regional Rail on what pretty much everyone agrees is the line that should be done first. I’ve created a “string diagram” that shows Amtrak and Stoughton Line service during the morning commuting hours:

Stringline showing northbound Amtrak and Stoughton trains

Stringline showing northbound Amtrak and Stoughton trains

Note that the scheduled performance of Northeast Regional train 66, the only Amtrak train that operates along the line during morning rush, is pretty dismal compared even to the other northbound Regional, train 190. This will be an issue when we start trying to schedule fast EMUs around the lumbering locomotive-hauled Regional. The Providence Line, although it is built on a mainly three-track right-of-way, is presently limited to only two through tracks between Pawtucket and Readville, and even after substantial investment has geographic bottlenecks (around Route 128 in Westwood and through Mansfield) that make a dedicated express track uneconomical.

The next challenge is the southbound (return) track capacity. I did not model this in as much detail, but it presents more of a problem because the basic tenet of Regional Rail is that you don’t keep peak-service trains close to the city, you send them back to their out-of-town terminal immediately, in revenue service — and during peak hours there are substantially more southbound Amtrak trips that have to be worked around. The following diagram shows those trains:

Stringline showing AM southbound Amtrak and Stoughton trains

Stringline showing AM southbound Amtrak and Stoughton trains

The final challenge is seating capacity. I used the same simulator as I previously implemented for the Worcester Line, and the 2018 CTPS passenger counts, to estimate the 90th percentile passenger loads on various service schedules that seemed to be practical with only minor shifts in existing Stoughton Line operations. I am assuming for the purposes of this simulation that the Stoughton Line continues to operate as a diesel service with the current (2019) schedules, because actually building electrification and new stations on the Stoughton Line will be expensive and time-consuming, and might well be deferred until South Coast Rail phase 2 advances. I used the trip-time modeling provided by Alon Levy, as with the Worcester Line simulations. Alon’s trip-time model assumes that service within Rhode Island is de-linked from the MBTA service; the 49-minute trip time from Providence to South Station (making all local stops!) is close to ideal, with an 11-minute turnaround time at both terminals allowing for a single consist to complete a full round trip in exactly two hours. I assume the proposed new Pawtucket station will get built, and I also assume a new infill station at Cummins Highway in Roslindale. After doing all that, I simulated a basic all-local service with 15-minute headways, and got … a service that is either extremely crowded or physically impossible.

The problem that I ran into is that not only is the Providence Line heavily used during peak periods, but that crowding is concentrated into a small time window, and it happens mostly at the southern end of the route — rush-hour trains 806 and 808 each leave Sharon with more than 1,100 passengers, and Attleboro train 842 leaves Sharon with nearly 800 on board. The sort of single-level articulated EMUs that I favor for Regional Rail have a seating capacity of 250 to 280 seats — the Finnish class Sm5 that I simulated for Worcester is at the low end, but a train that’s 5 meters longer (267 instead of 249 feet) would hold about 280 seated and, at crush loads, about another 250 standees. The MBTA does not currently have a comfort standard for commuter rail, but in my simulations I’ve decided that the 90th percentile passenger load should always get seated (and the more emphatically so the longer riders would be required to stand if seats are all taken). The MBTA’s platforms are 800 feet long, so if you have an 80-meter articulated EMU, you can platform a train of three of them, with a total of 840 seats — we really don’t want to have a situation where we’re 20 minutes away from South Station and there are already a hundred standees on the train.

I tossed a few ideas back and forth, but they tended to run into problems when faced with the physical realities on the line. Running more frequent service — ten-minute headways or better — is a nice idea in theory, but my simulator was still coming up with numbers that don’t work. At higher headways, the coordination problem with Amtrak and Stoughton Line trains is much worse, especially with the slow Stoughton diesels, and you also run into serious capacity problems at South Station. (My 8 trains per hour Worcester service only requires two platforms, but for Providence, Stoughton, and Amtrak together you need at least five. Of course, North-South Rail Link will eventually fix this, if we can just manage to build it before the MBTA studies it to death.) So I started to experiment with different service variations that would take some running time off the route — but not too much running time, because you don’t want to catch up to the previous departure when there’s no possibility of passing. Such a variation would also need to pick up enough of the passenger load to knock down the peaks that the regular local service experiences, but not too much.

I eventually hit on the idea of a limited-stop service that picks up passengers at three of the busiest stations along the line — Providence, Mansfield, and Sharon — but leaves the other stops to be serviced by a local train. This variant takes 38 minutes to run, saving 11 minutes over the regular service making all stops. That works out nicely, because it means that a limited-stop train can leave 13 minutes after a local train and still arrive at South Station two minutes later, so it fits nicely between locals running on 15-minute headways, and at the midpoint of the run, in Mansfield and Sharon, the limited is evenly spaced between the two surrounding locals. The limited would also stop at Route 128 and Ruggles stations. This service pattern gives a string diagram that looks like the following:

Stringline showing Providence Line nortbound AM service as proposed

The diagram shows four local trains per hour, four peak-period limited trains per hour, plus existing Stoughton Line and Amtrak trains as currently timetabled.

In the simulation, the heaviest load is on the 7:39 limited (arriving 8:17 at South Station), shown as V497L in my timetable, with 774 passengers — still well below seated capacity, so there’s plenty of room for growth. The preceding two trains get about 670 each, and the next train is under 500, leaving plenty of options for riders who miss their regular train. Amtrak’s Northeast Regional train 66 remains a problem; the clockface schedule would demand a train V465, arriving at 7:45, but this conflicts with Amtrak. I’ve dropped that run, leaving two consecutive limited trains on either side of the Amtrak (assuming it’s even on time), and as you can see from the diagram, the second of those still conflicts with Amtrak. I’m going to assume that this can be managed by getting Amtrak onto the third track north of Readville, and perhaps some minor schedule adjustments. There are a few other minor conflicts with Stoughton Line trains that can be managed by timed meets at Canton Jct. and some minor schedule adjustments.

One question that arises is what to do about the 14 EMUs on the “limited” pattern once they arrive at South Station. Obviously it is necessary to get them out of the terminal as soon as possible (that will still take about ten minutes due to mandatory safety checks and the time it takes for the operator to switch ends), but there isn’t the southbound demand for eight trains per hour. The fastest reasonable cycle time, with the southbound return trip running non-revenue, would not get the first northbound “limited” consist back to Providence in time to make another trip. Probably what makes the most sense in this case is for these trains to provide additional urban service to Readville and then go out of service until needed for the evening peak. As I’ve built this schedule, frequency drops at midday to only two trains per hour, but it wouldn’t hurt to maintain four trains per hour all day; this would provide better equipment utilization and reduce the need for midday storage space in Rhode Island. The total car requirement is 30 EMUs to operate this service; at $8 million each, that’s $240 million in rolling stock, and would free up a substantial number of old diesel locomotives and coaches to improve comfort and reliability on the yet-to-be-electrified lines. (Additional EMUs would be required to operate the intra-Rhode Island shuttle service, but the cycle time is 90 minutes, which would require just three EMUs.)

The full details are available in the providence branch of my GitHub repo. If you’re not super into methodology, you can just skip to the summary spreadsheet, which gives trip times, schedules, passenger loading, equipment requirements, and the CTPS 2018 passenger boarding statistics. (Unlike with the Worcester Line, I did not complete a full equipment plan; with a two-hour cycle time it was easy enough to just count the turns and figure the maximum loading on each consist.)

Posted in Transportation | Tagged , , , | 2 Comments

Other people’s recipes: Claire Ptak’s Ginger-molasses cake

I know it’s been an incredibly long time since my last proper recipe write-up, so here’s a lovely little cake from Claire Ptak’s The Violet Bakery Cookbook (Ten Speed Press, 2014 — wow, I’ve been working through this cookbook for a really long time now). It took me an incredibly long time to get up the energy to make this, after a long, dreary and physically inactive winter; I had the cookbook open to this recipe (p. 125) on my kitchen counter for six weeks before I finally managed to get all the ingredients together and actually make it. (It had been so long that my camera wasn’t even set up properly for kitchen photography: I had the wrong lens on and no flash.) There were a few tricky bits to making this cake but it came out very good and was gone in about 45 minutes when I brought it to the office. I won’t deny that I ate a slice or two myself. Here’s how it went.

Mise en place
We start, as always, with the mise en place. The dry team is quite simple, comprising 300 g of all-purpose flour, ¾ tsp of ground cinnamon, and ¼ tsp of ground cloves. (There is no added salt, and the baking soda at front left is actually part of the wet team!) Not shown in the photo is 225 g of boiling water; the other wet ingredients are 150 g fresh ginger rhizome, two large eggs, the aforementioned 2 tsp of baking soda, 200 g vegetable oil, 250 g molasses, and 150 g of granulated sugar.

Peeled ginger
First step is to peel the ginger, which every bit of advice I’ve ever seen is best done with a spoon. Maybe with a grapefruit spoon; I found the ordinary teaspoon I used slipped on the smooth ginger skin, so peeling took quote some time. The peeled ginger is sliced into 2-mm-thick discs and then chopped (Ptak says “pulverize”) in the food processor.

"Pulverized" ginger
I felt like this was a bit drier than I was expected, and maybe not as “pulverized” as it could be — a smaller work bowl would have made shorter work here — but I dutifully set it aside and continued on to the rest of the batter. (I needn’t have worried about the size of the pieces; aside from two unchopped pieces which were easy to pull out before baking, the rest of the ginger vanished into the cake batter and no chunks were detectable in the finished cake.) It was easy enough to combine the flour and spices with a whisk, but the wet ingredients were another matter.

Molasses, sugar, and oil
Whisking together thick molasses, additional sugar, and vegetable oil was quite challenging, with the immiscible oil, layered on top of the much denser molasses, constantly threatening to escape over the sides of the bowl onto my countertop. With much effort, I was finally able to get the oil and molasses to at least partially emulsify and turned the kettle back on to bring my now-cooled water back up to temperature.

Molasses mixture after adding baking soda water
The baking soda is dissolved in the boiling water, causing it to release a substantial amount of carbon dioxide gas (and converting some of the bicarbonate into sodium hydroxide, a/k/a lye). The soda solution is then mixed with the molasses mixture, producing more bubbles (molasses is somewhat acidic) and creating a tan foam which floats to the top of the mixture.

Finished cake batter
At this point I realized that I really did need a bigger bowl, because after adding the chopped ginger to the other wet ingredients I still needed to add the dry ingredients and the eggs (beaten before adding, and yes they’re added last, after the flour mixture).

Batter deposited in 9" cake pan
The recipe calls for baking in either one nine-inch or two eight-inch cake pans, parchment-lined and buttered. If I had to do it again, I’d probably add a parchment collar around the sides of the pan, in addition to the usual disc on the bottom — but I’d probably also make two smaller cakes as described. The cake or cakes bake for an hour in a 300°F (150°C) oven, until a tester inserted into the center comes out clean.

Overflow during baking
As you can see, the cake actually rose so much in the oven that it overflowed the pan. The cake is cooled in the pan for ten minutes to allow the protein matrix to set up, before depanning and returning to the rack until fully cooled.

Fully cooled and inverted cake on stand
After cooling, I compared the two surfaces and decided that the “bottom” of the cake, which had been touching the parchment while baking, was the better surface to display, so I inverted it onto my cake stand before applying the icing. Ptak’s icing is a standard confectioners’ sugar and lemon juice icing; she calls for 250 g of the sugar and 2–3 tsp of fresh lemon juice — I found that this amount of juice did not come even close to wetting the confectioners’ sugar, and even after adding the entire lemon’s worth of juice (about two tablespoons) I still needed to add water to thin the icing enough to be spreadable.

Finished cake with lemon-sugar icing
I probably thinned the icing a bit too much; the photo in the cookbook shows a somewhat thicker icing that doesn’t completely cover the sides of the cake. It was good enough, though. The cake had a good, strong ginger flavor, with only a few people noticing the tangy lemon flavor in the icing. Recommended, would make again.

Nutrition

Nutrition Facts
Serving size: 1/12 cake
Servings per recipe: 12
Amount per serving
Calories 451 Calories from fat 152
% Daily Value
Total Fat 17​g 22%
 Saturated Fat 1.5​g 8%
Trans Fat 0​g
Cholesterol 31​mg 10%
Sodium 231​mg 10%
Potassium 374​mg 11%
Total Carbohydrate 70​g 25%
 Dietary fiber 1​g 4%
 Sugars 49​g
Proteins 5​g 10%
Vitamin A 13%
Vitamin C 4%
Calcium 9%
Iron 13%
Posted in Food | Tagged , , , ,

Two months of thinking about buses

You may be wondering why I never followed up on my promise to simulate Regional Rail ridership for the Providence Line. In part that’s because it’s actually really hard — the Providence Line is much more complicated than the Worcester Line in terms of the number of services it must integrate (Amtrak, Stoughton Line, Franklin Line), the weird set of short-turns, service into central Rhode Island, and the sheer number of passengers (it is the MBTA’s heaviest-ridership commuter-rail line). I’ll get back to that soon, I hope. But the primary reason, as the title of this post suggests, is that I’ve been doing a lot of thinking about buses.

I made a Google Map showing the nine MBTA bus routes which serve more than 10,000 passengers per weekday (counting the SL4 and SL5, which overlap for most of their route along Washington Street from Dudley station to downtown, as a single route). The busiest route in the entire system is the 28, a former streetcar route connecting Mattapan to Grove Hall and Dudley, with many trips extending to Ruggles station, serving 12,880 passengers on 233 bus trips per weekday. Based on the (2017) system-wide average operating cost information the MBTA has reported to the Federal Transit Administration and the route-specific information reported in the Better Bus Project‘s route profiles, I guesstimated that this route costs approximately $7.5 million a year to operate. Other routes connecting southern Dorchester to northern Roxbury serve similar numbers of riders and have significant route overlap, including the 22 (Ashmont–Ruggles via Columbus Ave., 8,020 riders), 23 (Ashmont–Ruggles via Dudley, 11,810 riders), 29 (Mattapan–Ruggles via Columbus Ave., 2,250 riders), 21 (Ashmont–Forest Hills via Morton St., 4,290 riders), and 31 (Mattapan–Forest Hills via Morton St., 6,100 riders) — all told, 1,168 bus trips per weekday serving 45,350 people a day. (That’s more than the entire system-wide ridership of any of the state’s 11 Regional Transit Authorities — PVTA comes closest, at 39,368 average weekday riders.)

In any other city, a ridership of more than 10,000 passengers per weekday — 2.6 million riders a year — would more than justify a shift to a higher quality transit mode. But of course this is Boston, and all of those routes serve minority communities, so they have been stuck with diesel buses since the trolleybuses (which replaced the original streetcars) were moved to Cambridge in the 1950s. (They replaced streetcars in Cambridge, too — the trolleybus routes that currently operate from North Cambridge carhouse were originally bustituted to free up streetcars to operate the then-new Riverside Line through Brookline and Newton.) Taken together, this particular set of bus routes serves on the order of 11.8 million riders per year. (This number can only be approximate because what we are actually counting is not really rides or riders but so-called “unlinked trips”, which is a single ride on a single bus, and therefore double-counts transfers.) Surely this would justify a substantial capital investment to get those people to where they need to go faster, in greater comfort, and at lower operating expense?

It was time to start seriously looking at the map. It was pretty clear how a modern tramway would fit along the route of the 29: Blue Hill Ave., Seaver St., Columbus Ave., and Tremont St. are all at least 80 feet wide and could easily support a center reservation. But the 29, again, only serves a couple thousand riders and doesn’t run frequently enough to justify the conversion on its own. The MBTA even admits that the 29 only exists to save riders the trouble of transferring from the 28 to the more frequent 22. The 31 runs very frequently, and would share part of the 29’s reservation, but again, it only serves 6,100 riders a day, which (while it would be excellent for a “new streetcar” project in most of the US) doesn’t really justify the investment here. The Ashmont routes are even worse — the streets of Dorchester are for the most part only forty feet wide, which is barely wide enough for a reservation and two-way car traffic with no parking; when these routes were streetcars before, they ran in mixed traffic (which was of course much lighter 80 years ago than it is today), and a big part of the benefit of tram conversion would be getting out of mixed traffic if not complete grade separation. So I put this speculation aside.

There is one route, however, which has no such issues — but it’s not one of the Dorchester routes, it’s the SL5, part of the “Silver Line”, which was marketed as “bus rapid transit” at a time when the Federal Transit Administration was strongly favoring such projects. The main segment of the “Silver Line”, the South Boston Piers Transitway, was built as a part of the Central Artery/Tunnel Project, and includes a dedicated bus tunnel under Fort Point Channel from South Station to South Boston; because this tunnel lacks ventilation it is restricted to electric buses only, and in practice is operated with 60-foot dual-mode trolleybuses. The SL5, however, is different. It (and its companion route the SL4) were created to respond to the demand of the Roxbury community for a service along Washington Street that was “better than a bus” after the Orange Line was relocated from the Washington Street Elevated to the Southwest Corridor to the west in 1987. The SL4/5 operate frequent service with 60-foot diesel buses, and the route along Washington St. is theoretically a reserved bus lane although the markings are poorly maintained and rarely enforced. The rest of the route is in mixed traffic, which affects the SL4 (to South Station) more than the SL5 (to Park Street and Boylston stations).

There is a subway tunnel underneath Tremont Street — part of the original Tremont Street Tunnel, in fact — which was last used by streetcars in the 1940s. It even has a flying junction at the south end to speed cars diverging to South Boston (the #9 City Point route used to run this way) and Roxbury. For a while, the state was studying converting this perfectly good rail tunnel into a bus tunnel, but ultimately decided it would be too expensive. (This was tied up with constructing a new bus tunnel all the way to South Station, parallel to the Red Line, because buses are just so awesome.) The old tunnel portal (currently blocked by a park) is barely a block from Washington Street, and pretty much every advocate for “Washington Street replacement service” since the 1980s has been demanding that the T create a new F branch of the Green Line to serve Dudley along this old streetcar corridor.

Sandy Johnston (@sandypsj) pointed out on Twitter how the Portland Streetcar, which opened 20 years ago, was able to use a modern construction technique that significantly reduced the cost of construction by reducing the amount of concrete required for the trackbed (and thus the depth of the excavation and the amount of utility relocation required). That doesn’t solve the issue with street width, but perhaps it could make the construction of a new reservation cheap enough to justify the project just on that basis. And certainly for replacing the SL4/5 it might be worth the expense.

The MBTA is currently engaged in a “Bus Network Redesign” project. The T also just approved a very unambitious 20-year capital program, and is working on replacing or renovating all of its ancient and inadequate bus garages. And it’s also working on modernizing the Ashmont–Mattapan High Speed Line, which is currently operating using five-times-overhauled PCC streetcars from the 1940s. At the MBTA board meetings I’ve attended or watched recordings of, there is considerable frustration at the fact that the T is essentially unable to increase bus service at peak periods because it has no space in its bus garages to park or maintain additional buses. I started thinking again about replacing high-ridership bus routes with tram routes, wondering if the project could be justified solely on the basis of reduced operating costs (trams have larger capacity and so could operate less frequently, alleviating bunching and congestion at major stations like Ashmont and Dudley) and freeing up buses and garage space for other routes that are less practical to restore trams on.

I spent several more hours staring at maps and trying to think of which of these bus routes, when considered together as a complete system, would make sense as part of modern tramway system, assuming the T could somehow manage to build something for a reasonable cost and in a reasonable time, with the goal of eliminating the need for an entire bus garage worth of buses and representing an overall operating cost savings to the T. The F-Dudley extension is a somewhat different case, because for schedule adherence (which matters a great deal for capacity in the parts of the Tremont St. Subway that are shared with the existing Green Line branches) you don’t really want to interline it with the other routes, except maybe off-peak on weekends and holidays when it could interline with the 28, so the assumption would be that it would take over one of the two GLX branches and its rolling stock could be stored at Lechmere where a new carhouse is being built to support GLX anyway. In any event, I came up with five distinct sets of projects:

  • Roxbury–Dorchester system: routes 22, 23, 28, and 29, including three of the highest-demand routes in the entire system (22, 23 and 28), representing 35,000 daily riders, and approximately 460 revenue bus-hours of service on today’s schedule. The 22 operates 188 trips per day, the 23 runs 250 trips, the 28 runs 233 trips, and the 29 a mere 77 trips. Nearly all of the 28 route is on streets that are wide enough for a full reservation, so with better stop spacing and signal priority, a large fraction of the 28’s nearly 13,000 daily riders should experience reduced trip times. The 22 and 23 would see less benefit, because more of their route is in mixed traffic. The overlap between the 22 and the 29 is substantial, however, and more than half of the 22’s length would be in a reservation along Blue Hill Ave, Seaver St., and Columbus Ave.
  • Roxbury supplemental routes: the 44 and 45 have substantial overlap with the the first set of routes, enough that I thought it was worth including them in my plan even though the ridership (3,450 on the 44; 3,140 on the 45) and vehicle requirements (98 bus revenue hours) hardly justify tramification. On the other hand, these routes serve mainly transit-dependent passengers and could have a real positive impact in those neighborhoods. However, they would be operating in mixed traffic on narrow streets.
  • Forest Hills system: routes 21, 31, 32, and 42 all converge on Forest Hills station in Jamaica Plain. When routes in this area were streetcars, they were served by Arborway carhouse on Washington St., which is a bus garage today. The 32 is one of the most frequent bus routes in the system, running from Wolcott Square in Readville, next to Readville commuter rail station, to Forest Hills, nearly entirely on Hyde Park Ave. Only the northern part of HPA is amenable to a reservation; the southern part (roughly, south of American Legion Hwy.) would be in mixed traffic. Roughly half of all 32 runs are short-turns from/to Cleary Square, next to Hyde Park commuter rail station, but in my estimation of operating hours and costs I have not accounted for this, so the numbers I have for operating costs on the route are not to be trusted. The 42 is another lower-volume (2,560 riders) route, but it runs straight along Washington St. between Forest Hills and Dudley, so it provides a connection between the two systems; it would however cost significantly more to operate than the current bus unless the schedule was cut back to the point of uselessness. Finally, the 21 and 31 connect Ashmont and Mattapan to Forest Hills via Morton St. and the Arborway; the 31 would operate mostly in a reservation, but east of Blue Hill Ave., Morton St., Gallivan Blvd., and Dorchester Ave. are too narrow to build a reservation so that half of the 21 route would be in mixed traffic.
  • E-Arborway restoration: route 39 was instituted to provide service along the E Line between Heath St. and Forest Hills when the E branch was truncated in 1985. Restoring the E Line was a Central Artery Project mitigation commitment which the state reneged on when business owners in Jamaica Plain objected to the loss of “their” on-street private vehicle storage that would be required to make the E fully accessible. A decade ago, the City of Boston tore up the tracks and repaved South Huntington Ave., Centre St., and South St., making any service restoration much more difficult. (The overhead electrification was trolley wire and would have had to be replaced in order to support modern trams.) In recent years, Boston has partially repented and asked for the restoration of the line as far as Hyde Square, and an anonymous guest post on Ari Ofsevit’s blog talks about the practicalities in doing that in much greater detail. For the purposes of this analysis, however, I considered complete restoration of the service, which would allow allow the E Line to again be based at Arborway.
  • F-Dudley branch: finally, as previewed above, I am most strongly advocating for a replacement of SL4 and SL5 service with a new F-Dudley Green Line branch. Moving the E Line back to Arborway would free up enough storage space at Lechmere carhouse to allow all of these trams to be stored there, and as the F would be by far the shortest route south of downtown, it would make sense to pair it with the longer of the two GLX branches, to Tufts and eventually Mystic Valley Parkway. In order to make enough capacity in the subway between Park and Government Center, one of the existing branches would need to loop at Park during peak hours; since the B Line is the longest south/west branch in running time, it makes sense for it to be the one.

In order to get any meaningful travel-time savings, it’s necessary to do something about the bottleneck along Warren St. between Grove Hall and Boston Latin Academy. Like Washington St. and Talbot Ave. in Dorchester, Warren St. is only 40 feet wide until you get north of Quincy St. — but it’s a very important and heavily traveled street, and most of the congestion-related delays experienced by the 28 happen in this section. I strongly suspect it would be highly unpopular to completely eliminate on-street parking in this stretch of road, never mind eliminating car traffic entirely, which is probably the environmentally superior option; there are no parallel surface streets, the east-west streets in the area are primarily residential, and residents of Quincy St. in particular would not welcome the additional traffic. Having ruled out any plausible surface route, I decided to propose a shallow, cut-and-cover tunnel. The state would not under ordinary circumstances make that sort of investment in an area that is predominantly inhabited by lower-income people of color with little political influence, but perhaps now with representation in Congress (Roxbury and most of Dorchester are in Rep. Ayanna Pressley’s district) the winds could shift.

In this exercise, I’ve started from the counterfactual “What if the MBTA was competent at capital construction?” As a result, I’ve assumed the use of modern, Euro-spec trams and reasonable costs for tramway construction. However, tunneling is a different story altogether. If the costs of bored tunnel construction were at all reasonable, I might propose a whole new subway line, taking over the route of the 23 — but they’re not, especially not in the US. Even cut-and-cover tunnels are still quite expensive, as it happens, and for a short bypass tunnel of ¾ mile of Warren St., that’s the low-cost construction technique. That’s also the one tunnel segment that would benefit the most people on the entire surface network.

So I collected a lot of data, traced out routes and stops using the distance-measuring tool, and tried to guess what the operating costs of these routes are today as buses and would be as tram routes, based mostly on public data. I collated all of this data in a Google spreadsheet so you can check my work, laugh at my unrealistic assumptions, etc. Unsurprisingly, since these heavily traveled routes account for a lot of bus trips per day, they also cost a lot to operate — although the 21, somewhat surprisingly, seems to be profitable. For operating costs, I first estimated the number of revenue hours and revenue miles from the Better Bus route profile (which gives a breakdown of trips by daypart and also has a section that shows the actual operating time of each route compared with the scheduled running time). Using the 2017 summary data page from the National Transit Database, I estimated the operating costs for each route as the average of the per-mile and per-hour costs — since those costs are given as system-wide averages the actual per-route numbers might well be much higher or lower than what I’ve calculated. I computed the replacement tram running time based on a formula that accounts for lower average speed of mixed-traffic operations (approximately equal to that of a bus) and a 30-second stop penalty for each stop taken (for some routes, applying a stop factor to account for not all stops being taken on every trip). I chose service levels on the tram routes to generally ensure a comparable level of comfort for current bus passengers while reducing the number of trips to account for the higher capacity of 90-foot trams over 40-foot buses.

On the basis of these calculations, I concluded that the current service on these routes costs nearly $54 million annually to operate (that’s not counting capital costs for new buses or bus garages, because transit agencies do not take depreciation charge-offs). On the basis of my tram operating schedule and the current hourly cost of Green Line operations (again taken from the 2017 NTD summary page), I calculated that a full build would save $13.2 million annually in operating expenses, nearly a quarter. If the less used 42, 44, and 45 routes were left as buses rather than converted to trams, the savings increases to $15.2 million. Of the remaining routes, the E-Arborway restoration alone accounts for $3.8 million of savings annually, and replacing the SL4 and SL5 with a new F-Dudley is worth another $1.7m. (Of course the full-system numbers need to be discounted for the issue with route 32 short turns that I discussed above.)

So what about those capital costs? That’s where it starts to get sad. Even with very optimistic assumptions about construction costs, and assuming only $350 million per mile of tunnel, the full-build scenario costs nearly $1.8 billion — and let’s be brutally honest, the MBTA is never going to spend that much money on poor brown people in the city. (It doesn’t even want to spend that much money on barely-acceptable-by-European-standards commuter rail service that mainly serves rich white people like me.) That would be $19,240 per passenger. Even dropping the tunnel doesn’t help much; assuming you could magick away the on-street car storage and build a reservation down one side of Warren St., that only drops the price by 16%, to just over $1.5 billion. I still think it’s worth doing, but it seems like a really hard slog. (And if you could actually get nearly $2 billion for Roxbury and Dorchester, again you’d probably be better off building a new bored-tunnel subway that wouldn’t need to line up with the street layout.)

But the two easiest cases I have to make would be for restoring the E Line to Arborway (eliminating the 39 bus entirely) and building the F Line. I put a bit more effort into breaking these out, because the numbers look quite favorable (although I’m not convinced they’d pass an equity analysis). Both of these projects have significant unknowns: for the E Line, it’s the cost of making the existing E Line stops fully accessible, and for the F Line, it’s the cost of rehabilitating the Tremont St. Tunnel, making Boylston Station accessible, and reopening the old portal. I don’t have a good idea what these would cost, although the accessibility improvements are already in various long-term plans. Beyond these unknowns, I also assumed the use of new Type 9 LRVs — if the MBTA actually committed to doing either project in a reasonable time (which is much sooner than they are capable of deciding to tie their shoelaces at present) then they could just tack on a few more cars to the existing Type 9 order being built by Spain’s CAF (and assembled in New York State), which is a cost that we do actually know. On that basis, I estimated the surface parts of the F-Dudley at $122.4 million, or $7,600 per passenger per weekday, and the E-Arborway at a slightly higher $115.2 million ($9,930 per passenger). These numbers seem a lot more reasonable, but still difficult to get FTA funding for.

You can see all of these numbers and the formulas used to calculate them on the “Summary” tab of my spreadsheet. That’s all I have time for now, so I’m going to put this project aside and go back to thinking about Regional Rail.

Posted in Transportation | Tagged , | 2 Comments

Updating my MBTA Commuter Rail modeling for the Worcester Line

On Friday, the MBTA quietly released the results of its 2018 in-person passenger counts. I’ve been waiting for this numbers since I started modeling service on the Framingham/Worcester Line last fall, and apparently everyone else has been waiting for them since late spring. I first learned about the release on Saturday, when I saw a retweet by an advocacy organization of a Friday tweet by Laurel Paget-Seekins of the MBTA’s Office of Performance Management and Innovation, and of course I had to take a look at the numbers and rerun my ridership model. The new figures represent a 40+% increase in F/W ridership over 2012, and include the effects of the “Heart to Hub” morning superexpress and the opening of the very popular Boston Landing station in Brighton.

My model is a kind of static demand model: it simply maps the existing ridership onto a new schedule, without any consideration for whether the attractiveness of the schedule would induce more ridership, or indeed for constraints like connecting bus schedules and parking availability at stations that might limit the willingness of riders to alter their schedules. The Rail Vision program is running a dynamic model that hopes to address some of these issues, and indicate where investments in first-mile facilities are necessary to maximize ridership. But with the new figures, I wanted to rerun the analysis I did last year and figure out what the passenger loading looks like when mapped onto an all-electric system with single-level articulated EMUs.

My preferred schedule from last time around involved running two different service patterns: an all-local short-turn pattern in shoulder, reverse peak, and midday, and a zone-express pattern in peak, with four trains per hour (two Worcester local, two Framingham local) middays and eight trains per hour (four Worcester express and four Framingham local) in morning peak. (I don’t model the PM peak, but it’s more spread out so the morning rush is the stronger scheduling constraint.) I demonstrated that this schedule can be operated with 24 75-meter, 260-passenger articulated EMUs using no more than two EMUs in a consist. With 2018 ridership, even at 8 trains per hour, there are two express trips that absolutely require three EMUs, because we don’t want to have more than 120 standees on the non-stop segment from West Natick to Boston. The problem with running “triples” is that Yawkey station, which discharges about a hundred passengers from every train in AM peak, is substandard and lacks the real estate to platform a 750-foot train. But suppose you resolved that, either by erecting signs at all stations directing Yawkey passengers to the front two cars, or (at considerable expense) lengthening the platform: does that open up any additional scheduling possibilities?

More specifically, does allowing three-EMU consists yield a lower overall equipment requirement if you slightly reduce the service frequency? I constructed a service model that runs only six trains per hour in the morning rather than eight: such a service might conceivably reduce the vehicle requirement. (I did not redo the original equipment schedule for the new ridership numbers, but I’m assuming based on the increased traffic that it will require around 30 EMUs.) The unfortunate result of the simulation is that this doesn’t help: as soon as you reduce frequency in the peak, you start requiring four-EMU consists, which are just plain impossible (they would require 1,000-foot platforms). It’s worth noting that the Rail Vision team are not planning for anything better than four trains per hour, which locks them into either extending platforms for using exclusively bilevel equipment — the former is expensive and the latter is a bad choice for passenger comfort, convenience, and speed of boarding, especially with our 48-inch high-level platforms.

What if, instead of 260 seats per train, you could get 290 seats per train? It seems plausible that an 80-m EMU (about 267 feet) — which is the longest practical single-unit length for the MBTA’s 800-foot platforms — could hold that many (figure 5 m of additional length gives you six rows of 3+2 seating for 30 more seats). So what if you take the same ridership estimates and distribute them over larger EMUs? In that case, the “6/6/4” schedule works out quite well, requiring only 26 EMUs and providing at least 50 open seats on every train to allow for ridership increases — and you still have the option of increasing frequency back to 8 trains per hour, which works given the constraints of the line. (I checked: there aren’t enough additional seats to get you back down to two-EMU consists; there really are that many more passengers.)

Just to review the major constraints: I assume full electrification and two full high platforms at every station; Alon Levy’s operating schedule for fast, light EMUs like the Stadler FLIRT; turnaround times of at least 12 minutes at each terminal; and no more than the currently available overnight layover space. Alon’s schedule requires some additional superelevation, but this is a relatively small expense compared to construction of catenary and high platforms. I assume that some midday layover space is available, although the schedule is flexible with respect to which end of the route that layover space is at. (Having some midday layover space in Boston makes it easier to ramp up for PM peak; obviously, you could also run the same service level all day, although this would result in unfavorable headlines in the Herald.) I assume that trains can follow each other with three minutes’ separation, although the proposed schedule provides somewhat more. Finally, I assume that the traffic doesn’t demand a counter-peak express service: this is important because the schedule assumes that westbound Framingham trains can lead westbound Worcester trains, which only works if both branches run local service in the reverse commute.

In developing the equipment schedule, I planned for the line to be operated with two distinct fleets of identical EMUs — one set of “Worcester” cars and separate set of “Framingham” cars — although the equipment can be swapped around at will when not in service. An actual operating service would have the same cars operating Providence and Fairmount service, so spares could be pooled and vehicles could easily be rotated through the network for preventive maintenance and washing.

All of the updated model run results, as well as the new ridership data and my spreadsheet showing the schedule, is available in my GitHub repository. This specific version of the printed spreadsheet contains the results described in this post (except for the 8/8/4 equipment schedule).

The Rail Vision Advisory Committee and the MBTA board have both expressed interest in service plans that allow for phased implementation. With that in mind, I’m going to put at least an evening’s worth of effort into scheduling a service that runs electric local service from Framingham and diesel expresses from Worcester. That’s in preparation for the real challenge, which is to figure out what the Providence Line looks like, with its challenging mix of Amtrak and MBTA service plus the single-track Stoughton branch. But first, I have to tweetstorm my notes from today’s board meeting — after some exercise.

Posted in Transportation | Tagged , , | 1 Comment

MBTA Rail Vision’s “Peer Systems Review”: Adding Helsinki to the mix

The day before Thanksgiving, while I was already away and visiting with family, the MBTA Rail Vision team published their Peer Systems Review paper, describing a number of domestic and international commuter rail systems with detailed statistics about their service area, costs, and demographics. I was of course put out that they didn’t include Helsinki, so to remedy that, I’m presenting the same data in the same format. I unfortunately don’t have the GIS setup or databases to answer some of the questions at the same level of detail, and all the caveats in the original paper for international agency statistics apply equally to Helsinki.


Helsingin Seudun Liikenne -kuntayhtymä (HSL, Helsinki Regional Transport Authority in English) is responsible for coordinating transit service, fares, and schedules in the nine municipalities of Helsinki Region (which includes, in addition to Helsinki itself, Espoo, Vantaa, Kauniainen, Kerava, Kirkkonummi, Sipoo, Siuntio, and Tuusula). Most have historic CBDs, of which Helsinki’s is the largest in terms of both resident population and employment.

HSL operates no service of its own, but contracts with private and public operators to provide bus, metro, tram, commuter rail, and ferry services; operating subsidies come from the member municipalities and the Finnish state. VR Group currently has a monopoly on regional and intercity passenger rail service and operates the commuter rail service for HSL, some parts of which are substantially integrated with intercity service; commuter rail service is being put to public tender for the first time, with the new operator contract to start in 2021. A €1 billion project to construct a new, primarily underground rail connector to Helsinki-Vantaa International Airport opened in July, 2015.

Unless otherwise stated, information sourced to HSL/HRT comes from the 2017 Annual Report

Demographics and land use

Information HSL/HRT source MBTA commuter rail HSL/HRT commuter rail
Major City Served N/A Boston Helsinki
Population within 1 Mile of Stations N/A 1,716,012 N/A
Name of UZA N/A Boston, MA-NH-RI Helsinki
Size of UZA (sq. miles) OECD 1,873 291 (2,452)1
Population of UZA OECD 4,181,019 1,041,177 (1,498,050)1
Jobs in area OECD 2,677,320 780,252
Average Wage in Area OECD $64,080 $42,7652
Time spent in congestion TomTom 29 min 27 min
Major Geographic Features Maps Boston Harbor, Charles River Baltic Sea; numerous bays, inlets, rivers, and lakes
Mode Split (Drove Alone) Deloitte 67% 39%3
Mode Split (Transit) Delotte 13% 30%3

1OECD “city area” shown; metropolitan area data in parentheses.

2OECD average wage for entire country, conversion on PPP basis in 2015 US dollars as supplied by OECD. Regional average is believed higher.

3Deloitte Mobility Index includes the four core municipalities of the Helsinki Region: Helsinki, Espoo, Vantaa, and Kauniainen.

System Characteristics

Information HSL/HRT source MBTA commuter rail HSL/HRT commuter rail
Number of Lines HSL/HRT system map 14 4 (14 service patterns)4
Length of Longest Line (miles) Wikipedia 63 approx. 30
Number of Route Miles junakalusto.fi 388 625
Number of Track Miles N/A 697 N/A
Number of Stations system map 138 506
Percent Stations That are Accessible N/A 75% N/A7
Annual Unlinked Trips HSL/HRT 33,830,904 64,800,000
Percent of Agency Unlinked Trips HSL/HRT 8% 17.3%
Number of Central Terminals System Map 2 1
Central Terminals in Relation to CBD System Map Both in CBD In CBD
On-Time Performance (System-Wide) N/A 89% (2017) N/A8
Peak Line Frequency (Most Frequent/Other) Schedules 20 minutes / 25–50 minutes 5 minutes / 30 minutes9
Off-Peak Line Frequency (Most Frequent/Other) Schedules 40 minutes / 1–2 hours 10 minutes / 60 minutes9

4The four rail lines are the Coast line, the Airport Ring Line, the main line, and the Lahti line. These lines are served by a variety of local, express, short-turn, and skip-stop services; two services provide counter-rotating local service on the 30-mile-long Airport Ring Line. Three services on the main line and one service on the Lahti line are operated substantial distances outside the HSL region by VR on its own account, and one service on the Lahti line is included in HSL system maps and timetables but is entirely outside the HSL region.

5This figure appears to include only services operated using class Sm5 low-floor EMU rolling stock, leased by HSL from Junakalusto Oy.

6Excludes 20 more regional and intercity stations outside the HSL district served by R, T, D, G, and Z trains operated by VR.

7No information about station accessibility is available; however, all stations within the HSL district are served by at least one route using class Sm4 or class Sm5 low-floor EMUs.

8No information about on-time performance is published. However, published service reliability (percent of trips operated out of scheduled trips) exceeded 99.3% in 2017, short of HSL’s service reliability goal of 99.59%.

9Comparison is difficult because of the numerous service patterns operated on the four lines; at some major stations, local, express, and short-turn trains all arrive at the same time. All commuter trains stop at Pasila station, about 32 trains per hour peak and 12 trains per hour late night. Some distant stations have peak-only service.

Operating Characteristics

Information HSL/HRT source MBTA commuter rail HSL/HRT commuter rail
Annual Operating Expenses HSL/HRT $403,654,786 $797,160,00010,11
Farebox Revenues HSL/HRT $198,331,440 $439,971,00010,11
Farebox Recovery HSL/HRT 49.1% 55.2%
Fare Range (Single One-Way Trip) HSL/HRT
fare schedule
$2.25 – $12.50 $3.57 – $8.8610,12
Operating Expenses per Vehicle Revenue Mile N/A $17.15 N/A13
Operating Expenses per Unlinked Passenger Trip HSL/HRT (derived) $11.93 $2.1610,11

10Converted at a rate of €1.00 = $1.23.

11Figures for entire HSL system, including bus, metro, tram, commuter rail, and ferry. As HSL owns no vehicles itself, operating costs include contracted service operators’ costs of capital and depreciation, as well as the cost of leasing commuter-rail trains from Junakalusto Oy.

12HSL has three geographic fare zones and fares are independent of mode (except for Helsinki trams). Fares shown are for single-use one-way tickets purchased at a ticket vending machine; lower prices apply for tickets purchased through a mobile app or from an on-board automated fare validator using a stored-value Travel Card.

13HSL reports operating expense per passenger kilometer as approximately €0.12 for the commuter rail, on par with the Helsinki Metro and significantly cheaper than bus, tram, and ferry services. In comparable US units, this would be $0.24 per passenger mile. As described in note 11 above, this includes a substantial amount of what in the US would be classified as capital costs.

Fleet Characteristics

Information HSL/HRT source MBTA commuter rail HSL/HRT commuter rail
Fleet Operator (Name, Internal/External) HSL/HRT External (Keolis) External (VR Group)
Number of Vehicles in Fleet HSL/HRT 480 11714
Percent Spare Vehicles N/A 12.3% N/A
Average Vehicle Age (Years) Wikipedia 23.0 515
Power Source(s) Stadler Rail Diesel 25 kV 50 Hz overhead catenary
Seated Capacity of Trains (Approximate) Stadler Rail 800 52016

14Includes both Junakalusto Oy-owned class Sm5 EMUs leased to HSL and VR Group-owned equipment (primarily class Sm4 EMUs) used on HSL services under contract.

15Class Sm5 EMUs only, built 2008–2017.

16On a typical peak-period train consisting of two class Sm5 EMUs with 260 seats each. Of the total, 232 seats are fixed and an additional 26 folding seats are shared with wheelchair bays and standing room. The class Sm5 is capable of operating in three-unit consists but passenger demand does not currently justify it.

Posted in Transportation | Tagged , ,

My comments on Rail Vision

I attended the meeting of the Rail Vision Advisory Committee on Wednesday, and summarized what I saw in an interminable Twitter thread, which I’m not going to repeat here. After writing all that, I spent another few hours composing an email to MassDOT’s Manager of Transit Planning, the only person whose name is given on the official Rail Vision web site summarizing my further comments (beyond what I said during the public-comment period at the meeting proper). Since I have no idea how or if any of my comments will make it into the public record, I’m posting most of what I wrote here (edited for formatting):


First, a few process concerns.

I signed up a month or two ago for “Rail Vision”-related email. I did not receive any notice of this meeting, or indeed about any other Rail Vision activities. There was no agenda mailed out that I saw; I only found out what was going to be discussed by continually refreshing mbta.com until the event listing started showing more information.

The project web site does not identify who the members of the Advisory Committee are nor their affiliations. This made it particularly difficult when I was writing up my notes this evening as many of the speakers did not identify themselves. It seems pretty fundamental for any public body, even if a mere Advisory Committee, to have a public membership list so that members of the public who attend the meeting know who is being represented.

From the agenda, it appears that you were discussing something today you called the “Tier 1 Evaluation”, which again does not seem to be public — the committee members were complaining about not having had enough time to read the documents and there’s again nothing on the official project web site. This makes it very difficult for members of the public to determine whether there is any value in attending the meeting or what particular issues they should raise with the people who represent their interests on the committee.

My overall response, despite these process concerns, is that I’m cautiously optimistic, based on the comments I heard from all the committee members today, but I am somewhat concerned that you may be, in the words of the great architect Daniel Burnham, “making small plans”. Given that the rail system is going to require a substantial public investment, it is absolutely vital that this “vision” be a truly forward-looking one, a world-class rail system for our world-class city — we the taxpayers are going to have to be convinced to pay for it, and minor tinkering around the edges while traffic and service get worse for a decade or more is not going to do that. Ask for the best system you can justify, independent of resource constraints, and if we have to make some compromises to pay for it, make sure taxpayers, legislators, and the governor know that we’re settling for less than what we deserve.

As I said at the meeting, the one way forward that actually addresses all of the committee members’ goals — which should be everyone’s goals — of fast, frequent, reliable, all-day, bidirectional rail service, is system-wide electrification with fast, lightweight Electric Multiple Units. The very least we should aspire to is “at least as good as Helsinki”, and given that the Boston-Providence CMSA has five times the population of metropolitan Helsinki (and very similar population densities in the suburbs) this really shouldn’t be that great an ask.

A few of my own notes, a little bit disorganized because I’m extracting them from the longer tweetstorm about what the committee members said:

  • [The speaker] draws the natural conclusion that what we really want is high-frequency service on *all* lines across the board. (Compare regionalrail.net or my own proposal for a revised service standard. People talking like barely-adequate 2 trains/hr would be a huge win!)
  • What’s a “vision”? Whether you call it “aspirational”, “ambitious”, or “barely adequate by the standards of small European cities”…
  • There was a lot of talk specifically about growth in and service to the 495/MetroWest corridor, perhaps because many of the committee members are from that area. I actually spent all of Veterans Day weekend researching and writing a specific proposal for this region, where I personally have lived for 17 years. The project should include in its universe of possible service expansions initiating service at least as far as Marlborough on the Agricultural Branch. See the publication list below for more information and motivation.
  • Even if Wellesley doesn’t “deserve” better service [because it’s not creating enough new housing], it may be better for both operations and for riders if the service is more uniform and the same high frequencies are provided everywhere. [My own proposal, see below, runs 8 tph peak on the line, mostly following the current zone-express pattern with 4-tph Worcester trains running express and 4-tph Framingham short-turn/Northborough/Clinton trains running local. Obviously EMUs, double-tracking, and full high-level platforms required.]
  • In response to Josh at T4MA: (Sorry, Josh, we don’t need a bar car, we need to shave half an hour off your commute from Worcester.) In all seriousness, bar cars — any distinct type of coach, especially a non-powered coach — is Bad with a capital “B”. Save those for the east-west intercity service. With only minimal upgrades, the proper EMUs should be able to cut the Worcester express’s running time to 50 minutes, while still serving all stations within Boston city limits. This then eliminates the need for the terrible “Heart to Hub” service and regularizes rush-hour schedules for the entire line.
  • Josh also calls for “flexible trainsets”: this is really an argument for modern, long (75 to 80 meter, 250 to 267 foot) articulated EMUs — not something 1850s-rail-car-sized like the Metro-North EMUs. Then you’re adding/subtracting passenger capacity in units of 250-400 passengers (depending on design and fraction of standees) rather than 160-passenger coaches. This also means you can easily, quickly, and safely make and break consists on the platform to ramp capacity up and down as needed.
  • Mike from Beverly brings up a great point about reliability targets: if, like many commuter rail commuters, you have to make a connection to another MBTA service, then suddenly you don’t have 90% reliability any more, you have 81% reliability, or even worse. (God forbid you have to get from Lowell to Logan Airport — a two-connection commute would be only 73% reliable and you’ll probably be fired for tardiness after the first month.)

While my comments were mostly in agreement with former Secretary Aloisi’s statement, I wanted to disagree with him on one point (which I didn’t raise at the meeting for time reasons): as a Framingham resident, I don’t support the concept of A/B testing service patterns with EMUs on the Providence Line and diesels on the Worcester Line. To get the track capacity required for the level of service we want and deserve, you need the power and acceleration that only EMUs give you. However, Sec. Aloisi is absolutely right on the need for Allston Viaduct mitigation, and I have argued and will continue to argue that this is an ideal demonstration environment for hybrid battery EMUs.

The gap between Yawkey and Boston Landing, during the Allston construction, is too long for a train to coast through, especially considering that the low clearances at Beacon St. will prevent energizing catenary there until all the bi-levels are gone from the South Side. A battery system with sufficient capacity to power the EMU at reduced speed for the 2½ miles from Yawkey to Boston Landing would not add too much to the weight of the vehicle — and such a vehicle would also have operational benefits elsewhere, during service disruptions and construction projects where catenary must be de-energized.

(By contrast, diesel multiple units, like the diesel FLIRT Fort Worth is buying, and which I was previously fairly high on, have a poor power-to-weight ratio and physically cannot accelerate fast enough for the close stop spacing on this line. However, DMUs purchased as part of the same family of rolling stock with EMUs have potential for short-term expansions of service in applications like East-West Rail, Worcester-centered service on the Pan Am and the Providence & Worcester, early implementation on the Old Colony lines, capeFLYER, and service to Nashua.)


I followed this up with an index to my long-form writing and analysis on this subject, which you all have presumably read because it’s all been posted here on the blog. (Any newcomers: welcome! Please be aware that my thinking has evolved as I’ve done more research into these issues and talked with some experts, so the most recent posts are more reflective of my current views than older posts.)

Aside | Posted on by | Tagged , , ,

A further examination of the Agricultural Branch for Regional Rail

About a month ago, I posted the most recent of several posts about using the former Agricultural Branch Railroad, the CSX-owned freight line which runs from Framingham to Clinton, for passenger service under Regional Rail. I made a Google Maps overlay with the route and most of the station locations shown; see the previous link for it. One frustration I had was that, since there is no existing passenger service, there are obviously no passenger counts, and I certainly have no budget to commission survey research to find out who would take such a service, nor which employers would be likely to offer employee shuttles to rail stations. I didn’t even have a good idea what the employment picture was like in the area, other than in Framingham where I live. I had seen some anecdotal evidence that the I-495 corridor is a whopper, employment-wise, but Dell EMC is a major employer and they’re in Hopkinton, south of the Worcester Main Line. And I’m no GIS nerd so I didn’t have a set of geospatially indexed databases ready to hand.

Doing a few quick Google searches, however, I was able to find two timely and extraordinarily usable data sources. The Missouri Census Data Center‘s Circular Area Profiles tool takes geographical coordinates and a radius and resamples the American Community Survey (2012–17) data to generate a broad and informative demographic profile of the people living within the circle so described. (The original census data is reported by blocks and block groups, which are designed to line up with political and physical boundaries rather than abstract geometric figures.) And in the same mode, the Census Bureau’s own Center for Economic Studies produced a fabulous interactive web app, OnTheMap, which provides data on employment from the LEHD Origin-Destination Employment Statistics 2015 data set for arbitrary geographies — and even better for my purposes, because the data set includes origin-destination pairs, it can actually answer questions like “how many people live within walking distance of point A and work within walking distance of point B”. (Both of these tools have important caveats relating to sampling error, especially at small radii, and a more GIS-informed analysis would use more sophisticated tools that took into account actual travel times on the existing roads and sidewalks rather than an arbitrary radius. Maybe someone can ask CTPS to do a real study?)

I’ve used this data to put together another slide deck. Since this presentation is heavily illustrated with Google Maps aerial photos, which I don’t actually have the rights to, I can’t release it under a liberal license for other people to give presentations from, but you’re welcome to download it, read it, and show it to other people.

Some of the key takeaways:

  • There are far more jobs in the corridor than there are people to fill them: the MetroWest/495 region is a net importer of labor — and many of the people who live in the region don’t work there.
  • With the “full build” program, comprising four phases (Framingham yard and stations; Southborough, Marlborough/495, Northborough Center, Northborough/290; Clinton; infill stations at Mt. Wayte Ave., Marlborough Jct., and Berlin/Boylston), 7,000 workers and 20,000 jobs are within walking distance of a station.
  • 79,000 jobs and 52,000 workers live within two miles (reasonable park-and-ride or corporate shuttle distance).
  • 4,650 people work within two miles of a station and live within walking distance of the Framingham Line, Orange Line, Red Line, or SL1 bus — it can be expected that nearly all of these people are currently car commuters. That’s an extraordinary reverse-commute potential, and could be significantly expanded with Regional Rail service to the region.
  • In the more traditional direction, 9,300 people live within two miles of a station and work within walking distance of the other services. These numbers would be even higher after accounting for other potential bus transfers, like the CT2 to Kendall.
  • The biggest proposed stations for residences within walking distance are Clinton (3,800), Framingham State (3,700), and Mt. Wayte Ave. (2,400) — all urban stations. For suburban small towns, Northborough and Southborough both come in respectably, at 1,700 and 1,100, respectively.
  • By far the biggest job center by walking distance is Framingham Technology Park, at over 8,000, but if you consider that most will take corporate shuttles from the station and expand the radius to two miles, that number jumps up to 16,000. It’s difficult to directly compare stations at this radius because none of the stations are more than four miles apart, so most of the commutersheds overlap in one or two directions, but the full set at this distance is worth looking at: Clinton, 4,750; Berlin/Boylston, 775; Northborough/290, 8,125; Northborough Center, 5,450; Marlborough/495, 20,800; Marlborough Jct., 12,200; Southborough, 16,500; FSU/Salem End Rd., 17,050; Mt. Wayte Ave., 25,000. (Note that all of Framingham’s CBD is within two miles of the Mt. Wayte Ave. station site, but most of them would find South Framingham a more convenient station; about 1,000 jobs are within walking distance.)
  • I estimated the “not to exceed” cost for full build at about $490 million, mostly by looking at the aerial photos and making some semi-educated guesses about what things cost. Notably, this accounts for neither the cost of acquiring the line from CSX nor potential benefits from PPPs at some station locations with high development potential, nor does it include the cost of rolling stock. This is a bit pessimistic compared to my previous post, primarily driven by an estimate of $10 million per mile for catenary, trackbed improvements, double tracking, switches, and signals. Some additional costs for RoW and station site acquisition are expected.

UPDATE: I went and drove as much of the route as is possible today, paying particular attention to the station sites. Marlborough Jct. has an aggregate mill that probably ships by rail and would need to be taken and replaced with more appropriate development. The Ken’s Foods “Corporate Headquarters” shown by Google Maps is actually a salad-dressing factory and receives tank cars (corn syrup?), but this use is probably compatible with the station if an island platform is used, but the second track here would be quite tight. Northborough Center could host a full-length platform by dead-ending Pierce St. Finally, I concluded that Northborough/290 station should actually be north of I-290, with access from Whitney St. — land use on that side of the freeway is more favorable for the sort of interceptor P&R that I envisioned, and no additional highway construction would be required to handle the traffic.

Posted in Transportation | Tagged , , , , | 5 Comments

Three Reasons the Commonwealth Needs to Commit to Regional Rail Now

After writing that very long slide deck, I thought it was worth boiling down my conclusions for policymakers to three important points:

1. There is a large investment of some sort required in commuter rail infrastructure over the next ten years. Either we can spend billions of dollars to keep the service limping along as it has been, or we can spend only slightly more to replace it with a vastly superior service. The longer the Commonwealth delays action, the more money will be wasted on keeping the obsolete equipment and the obsolete service running.

2. Riders deserve to see visible evidence of capital investment. This sounds shallow and unscientific, but public support for, and continued use of, the MBTA’s service depends to a great degree on riders seeing actual improvements happening before our eyes. We see this with the subway network: Those of us who are following things closely all know that signal system modernization and new rolling stock, if properly implemented, will alleviate many of the pain points on the Red and Orange Lines, but that’s not visible to the average rider stuck in a tunnel for half an hour with today’s broken-down trains and antiquated signals, and discourages use of the service over the long term. High platform and catenary installation are highly visible construction activities that riders can see and follow the progress of on their daily commutes, and that give evidence of ongoing work to improve service long before new rolling stock shows up or new schedules are implemented. Of course, that doesn’t excuse crumbling stairs, broken elevators, and water leaks, but riders need and deserve some visible evidence of progress. (And as soon as high platforms are implemented at a station, service immediately improves for the riders there, before any of the other investments are completed.)

3. Work on the I-90 Allston interchange is scheduled to start in 2021. While it’s unlikely that we could have full Regional Rail on the Worcester Line before then, MetroWest and Worcester commuters need mitigation — practical commuting alternatives during the anticipated eight-year construction period — which the current commuter rail service is far from being able to provide, and the sooner we get Regional Rail implemented on the Worcester Line, the more effective a mitigation strategy it will be. While it might not be ready for 2021, there is every reason to believe we could have a full implementation by 2023, only two years into construction, if the right decisions are made now.

Schedule thoughts: if we decided to do this, really seriously, with a board vote in January, then the MBTA could start issuing design-build contracts for Providence and Fairmount Line improvements and electrification late in FY19, have an RFP for rolling stock in the spring of 2019, and issue the rolling stock contract (about $2–3bn) in about a year from today (say, October 2019, so early FY20). Pilot cars (if Stadler is the chosen vendor) could be delivered by ship from Europe about a year later, call it January 2021, just in time for the completion of PTC implementation and AFC 2.0. Get a dozen pilot cars and test during winter and spring of 2021, and you can introduce them to revenue service on Providence and Fairmount lines in time for the summer 2021 rating, and take delivery of 30 more cars a year, every year, from the US assembly plant once the pilot cars are accepted.

Meanwhile, you let a single design contract for the remaining catenary and substation construction, and a separate contract for all platforms, with biddable design packages for the Worcester Line due by summer 2020 at the latest. Separate station improvements contracts for Back Bay, Newton, Wellesley-Natick-Framingham, and west-of-Framingham, so that you can start service to Framingham by late 2022 and to Worcester in early 2023. At that point, the newest commuter coaches and locomotives get redistributed to the other lines, and the oldest equipment can be sold for scrap rather than being replaced.

Oh, and you should build my Framingham-to-Northborough proposal.

Posted in Transportation, Uncategorized | Tagged , , ,

A comprehensive Regional Rail slide deck

Hard to believe it’s my 400th post on this here blog, but given recent trends it’s not hard to believe that it’s about transportation rather than computer networking, figure skating, sliding sports, travel, or (most of all) cooking. I’ve spent most of my free time over the last week putting together a slide deck on Regional Rail, although I have no plans as yet to give an actual presentation. (I would be happy to do so for any group or political body that was interested, provided the schedules can be made to work out.)

The purpose of building the slide deck was two-fold: first, I wanted to make something that would serve a broad audience with a somewhat different emphasis from the regionalrail.net report; second, I wanted to synthesize a lot of what I’ve learned over the past few months in modeling and analysis (and in reading modeling and analysis by smart people who actually know something about operating rail networks), with the intent that the slides (or an audience-specific subset of them) could be used by many people to present the Regional Rail concept to organizations and political leaders who have an influence on the MBTA’s budgets and capital priorities. (The presentation weighs in at 66 slides, because I’ve put all the information in writing, whereas a deck designed for a more specific audience would leave more of the content for the presenter to give aurally — sorry, Steve Poftak! On the other hand, this makes it more accessible to the Deaf.)

I am releasing the presentation to the public under a Creative Commons Attribution-No Derivative Works 4.0 International license, but with a special exception that excludes presentations based on these slides from the “no derivative works” clause, even if recorded or transcribed. This is basically to protect me, in case someone decided to release a maliciously altered version of the slide deck without my permission, but it’s also why there is only one photo in the presentation (because I would have had to go out after work, in the dark, and take the pictures myself). So please do feel free to use these slides, in whole or in part, to give a presentation on the subject to people you have a connection with, or if you’d be interested in having me actually give the presentation (mind: my only qualifications are as someone who wishes he could take the commuter rail), please see the contact information on slide 62.

I need to thank a bunch of people whose blog posts, blog comments, tweets, and private communications provided important information, and in some cases corrected misconceptions of mine, including Zachary Agush (@zagush), Sandy Johnston (@sandypsj), Alon Levy (@alon_levy), Ari Ofsevit (@ofsevit), David Perry (@FramWorMBTA), Ted Pyne (@Ted4P), and a number of anonymous railfans. However, all of the opinions expressed and any factual errors in the presentation are mine alone.

UPDATE 2019-10-26: Updated to respond to Alon Levy’s comments on Twitter, and correct the power and maximum acceleration for the 75m FLIRT based on the Stadler data sheet. See the GitHub repo for more details and comments regarding the limits of this simulator.

Posted in Transportation | Tagged , ,

Interlude: Physics of train acceleration

This past week I’ve been working on a slide deck that people can use to present the Regional Rail concept to non-transportation-nerd audiences, and one of the things that I wanted to include, just as an aside, was a couple of slides explaining the “stop penalty” and why it’s so bad for traditional diesel-hauled commuter rail like we have in Boston. This led me down a horrible rabbit-hole for the better part of two days, trying to figure out a closed-form expression that would allow plotting comparisons showing speed over time, or distance over time, or speed over distance. I am quite certain that there actually is such a solution, but I couldn’t come up with one that looked believable when I plotted it out, so I went back to square one and wrote a simple numerical simulator that would just generate the data points I needed to make my plots rather than trying to come up with the general formula.

In doing this simulation, there are three different regimes we have to consider, which derive from different limits (mechanical, physical, or legal) on the train system. At high speeds, there is a regulatory speed limit: the train is not allowed to go faster than a certain speed, because of track conditions, grade crossings, or the sort of signal system it has. At low speeds, the limits are mechanical: the motors can only apply so much force without the wheels spinning (or the axle shearing, or the gears stripping, or the motor windings burning up). In between these two regimes, however, physical law — Conservation of Energy — directly limits the acceleration that can be achieved. Any train will have the ability to generate (or draw from batteries or overhead wires) a certain maximum amount of power, and even without accounting for all the losses in the system (electrical resistance, inefficiency in the motor controllers and the motors themselves, friction, drag, and so on) this limits the acceleration attainable. In fact, at constant power, the conservation law tells us that

\displaystyle a_{max}(v) = \frac{P}{m\cdot{}v}.

If we assume that you always want your train to accelerate as fast as it can until it reaches the speed limit (or until it has to slow down for the next station), then a \leq a_{MAX} when we are in this power-limited regime, and a \leq F_{max}/m (where F_{max} is the maximum force the motor can exert in the forward direction, called “tractive effort”) when it hasn’t reached maximum power yet. In both equations, m is the mass of the entire train, including locomotives, coaches, fuel, and all the passengers. Note that in the energy-limited regime, acceleration is a function of velocity — in fact, it’s inversely proportional to velocity. Acceleration is the first derivative of velocity, which means this is a differential equation; luckily it’s one that has a solution, although as I said I had trouble figuring this out (because I never took Diff. Eq. in college, most likely) and couldn’t figure out how to do the integration piecewise — because what we really want to figure is speed as a function of time, and then integrate that to get distance traveled as a function of time. But just to make things clearer, here’s a plot showing the three different regimes for three different train configurations that I set up in my numerical model.

The image shows acceleration as the dependent variable and time as the independent variable, with different regimes identified depending on what the cause of the limit on acceleration is (mechanical, power/energy, or legal)

The three train configurations modeled are two trains with 500 passengers, all seated — one with my favorite articulated EMU (Electric Multiple Unit), the Stadler FLIRT, and one with a diesel locomotive of the type the MBTA uses and three bi-level coaches — and a third diesel train with the same locomotive but a nine-coach consist and a crush load of 1600 passengers (the highest load and the longest train currently operated on the MBTA). I assume that all coaches are bi-levels because I don’t have a source for the mass of a single-level coach: in actual operation, at least one single-level coach is required in every train for accessibility.

One thing is abundantly clear: that 1600-passenger “monster” load is extremely slow to accelerate. In the formula above, you’ll note that the mass of the whole consist is in the denominator of the acceleration equation, which is a consequence of Newton’s Second Law (F=ma). Naturally, the heavier the train, the slower it can accelerate, unless you have a way to add power somehow. With Multiple Units, you do have a way to add power: a single FLIRT trainset has two 1,000-kW electric motors, for a total power of 2 megawatts, and if you couple two FLIRTS together, all four motors work together at a system power of 4 megawatts. (You can couple up to three, but in most cases where you might want that many seats, you should be running more frequent trains instead.) The poor commuter-rail train has only one diesel engine with a power output of just over 2.2 megawatts no matter how many cars you couple to the end. (Yes, that does mean that a diesel locomotive on its own could accelerate faster than the electric train, because it weighs less and has more power — but that doesn’t help move any passengers!) One thing is clear: the heavy train takes inordinately long to reach the speed limit of 79 mi/h that I’ve set on this simulation. (On the Providence Line, the speed limit is 125 mi/h, but Regional Rail trains making more frequent station stops would not reach that speed, and it wouldn’t make sense to buy trains that were designed to go that fast — 99 mi/h is easily as fast as is useful on commuter trains.) This is even more clear if you look at the acceleration as a function of distance:

A line graph compares acceleration of three different train configurations; this is the same as the last graph except that the independent variable has been changed to distance traveled.

Look at that heavy train: it takes about 4.5 kilometers (more than 2¾ miles) to reach the speed limit that’s typical on the current MBTA commuter lines! Remember, that’s just acceleration; you still have to decelerate for the next station stop, and if the stations are less than 9 km (5½ miles) apart, that train is never going to reach the maximum speed allowed on the line. That’s why the current service pattern favors expresses: if you can go twelve miles without having to stop or slow down significantly, you don’t have to pay this “acceleration penalty”. (That’s only part of the full “stop penalty” — the other part is the actual time the train has to spend stopped at a station while passengers board and alight, the “dwell time”. Regional Rail aims to solve both of these issues at once with better design, EMUs, and level boarding.)

Another way to visualize this is to look at how long it take to travel a given distance. The graph below shows the same three train configurations, but the x-axis is distance and the y-axis is time, the exact opposite of how you’d normally look at this sort of physics problem, but very useful when you’re thinking about train schedules:
A line graph comparing three different train configurations accelerating from a stop, with distance as the independent variable and time as the dependent variable

The two-trainset FLIRT consist reaches the speed limit first, for obvious reasons (it has nearly double the power), at about 4500 feet of travel, and the three-coach diesel train hits the speed limit at about 6500 feet; after this point, they are traveling at the same speed and those two lines are parallel, separated by about 13 seconds that is the residual advantage of the electric train’s quicker start. The nine-coach diesel barely reaches the speed limit by the right-hand side of the plot, which is at a distance of three miles. If you assume that a train can decelerate exactly as fast as it can accelerate (which is not unreasonable for dynamic or regenerative braking, I’m not so sure about friction brakes), then the long heavy diesel shouldn’t be stopping more than once every six miles or more — which is far from what that train currently does, and that’s why even the “express” trains are quite slow (unless you’re getting on or off right before or after, respectively, the express segment). On the inner parts of many of the commuter lines, stations are spaced much more closely together — as little as a mile or two — and on lines that don’t have stops like that now, those are the exact places where Regional Rail envisions adding “infill” stops to provide better service to residential and commercial areas in the urban core. (Ideally, every Providence Line train would stop at both Forest Hills and Ruggles, for example — but this would only be acceptable to Mansfield commuters if the stop penalty were effectively mitigated so their commutes were not lengthened.)

The code for the simulator is in my GitHub repo, in the file physics.rb. The slide deck will be posted here once I finish writing it.

Aside | Posted on by | Tagged , , , | 1 Comment