Sunday, December 27, 2015

Rethinking MBTA weekend schedules

There is not much good that can be said about MBTA's weekend Commuter Rail schedules.

A few lines have been graced with two hour, clockface headways (Worcester, Lowell, Franklin)—not frequent, but at least predictable—but most others have service only every three hours, and sometimes less frequently. Combine that with variable headways and a lack of off-peak fares, and it's not surprising that ridership is less-than-stellar.

I suggested adding a station at Walden Pond, and today I'm going to go a step further. Rework the schedules, adding a small amount of service, to provide two-hour, clockface scheduling on all North Side lines. (I'll leave the South Side to someone else, for now, although boy could the Providence Line use better service!) A caveat: I'm not sure about how crew rules work, and that may be the cause of some of the schedule irregularities here, but we should design a schedule for passengers first, and crews can be shuffled between lines at layovers to shorten or lengthen shifts as necessary.

Newburport/Rockport is relatively easy. The current schedule works relatively well, but has some bizarre gaps. For instance, Rockport has inbound trains at 7, 10, noon and 2, and then a 3:10 gap before the next train. Going in to Boston for the afternoon? You can arrive at 3 or 6. At the beach and want to get back to the city? You can go home at 2 or 5, but not in between. Even at Salem, where the two branches interline, there is a nearly two-and-a-half hour gap between 3:20 and 5:46. For a city with a significant tourist draw (the witch museum, the PEM), poor highway access and limited parking, this schedule is severely lacking.

The two branches of the line have run times of 64 minutes (Newburyport) and 71 minutes (Rockport), so it is relatively easy to interline the two with three train sets:

This does add some service, with one extra Rockport trip and two extra Newburyport trips each day, adding about 22% to the current service. But the utility increases dramatically, with predictable train schedules and the elimination of service gaps. In terms of service hours, this schedule adds 6:38 to the schedule. The marginal operational cost per service hour, as calculated pertaining to the CapeFlyer, yields an extra cost of about $4500 per day. The line currently carries 5600 passengers on the average weekend day. Assuming an average fare of $7.50, just 600 additional riders—an increase of 11%—would cover the operational costs of these additional trains.

Fitchburg and Haverhill, however, are harder. Haverhill has a run time of 65 minutes, long enough that a single train set shuttling back and forth (see Lowell) can't make a round trip in two hours, so the current schedule has service every three hours (or so), which is hardly useful. The Fitchburg Line's current schedule is awful, with three-hour-plus service gaps and then short-turns running within a half hour of other trains.  The line is long enough that even with the new schedule accounting for higher speeds, the line would not even be able to attain two hour headways with two sets. The schedule is better than the current one, it's nowhere near clockface, with service about every 2:25, or so.

But there's no rule that says a train on the Fitchburg Line has to run the Fitchburg Line all day and all night, right? Imagine Fitchburg and Haverhill as two branches of the same line, with a train running first to Fitchburg and then to Haverhill, kind of like the Rockport and Newburyport lines. The relatively short Haverhill Line interlines well with the long Fitchburg Line, and the same three train sets can, combined, provide service every two hours on both lines.

I think that's pretty elegant. This requires just 4:12 more running time each day, or $3150 of marginal operational costs. The two lines have a combined ridership of 4500, and would require just 420 additional riders—a 10% increase—for fares to cover the costs. And the benefit would be high: the frequency of trains on the Haverhill Line would increase by 50%, from every three hours to every two.

One final idea would be to have the trains operate on a "pulse" system at North Station to enable transfers, which is attainable if every line has clockface scheduling with the same headway. Every train would arrive at a similar time, allowing passengers going from, say, Lynn to Waltham, with an easy transfer at North Station (assuming trains are on time). With the current schedules, this isn't possible. With a couple adjustments, the above schedules, with the Lowell Line, would work quite well for that.

Saturday, December 26, 2015

Why is there no train station at Walden Pond?

Photo from here.
When I was hiking the Appalachian Trail through New York, I passed MetroNorth's Appalachian Trail station. Every weekend, a couple of trains stop there and drop off hikers, who can hike in either direction on the trail. There's no parking lot at the station, nor road access (although Route 22 is only a couple hundred feet away). It is scheduled specifically for hikers, and while it only serves a few hundred passengers a year, the small station is sufficient for the crowds. Breakneck Ridge is a similar station, although not on the AT, as is Manitou, although it also has very limited rush hour service.

While the Appalachian Trail was built in New York in the 1920s and 1930s (meant to serve as a regional resource for city dwellers), the Appalachian Trail station is much newer. It was built in 1991 at the behest of MetroNorth and the New York-New Jersey Trail Conference, which maintains the trail in the area. The total cost of the station? Just $10,000, with negligible additional costs for a few trains to stop. Even if only a few hundred people use the station each year, their fares add up to more than that every year.

The Bay Circuit Trail runs 200+ miles around Boston,
including around Walden Pond. Full section map here.
Which brings us to Concord, Mass. While not on the Appalachian Trail, Walden Pond is one of the most heavily-visited outdoor sites in the region, both for swimmers and beach-goers in the summer, and walkers and hikers year-round. The Fitchburg Line goes right by the pond—Thoreau rode the train there, although he may not have liked the new technology—but the nearest stations are about two miles away (Concord, which is closest, is along a busy road with a crossing of Route 2, hardly conducive to pedestrian travel). While the pond abuts the railroad, and the Bay Circuit Trail—the 200-plus mile trail around Boston—comes within a stone's throw, the trains roll through at 60 or 70 miles per hour, weekdays, and weekends.

And even if you choose to drive, Walden Pond can be hard to get to. There's an $8 per car fee, but the 300-space lot routinely fills to capacity on weekends in the summer, with a line of cars idling on Route 126, waiting for an open spot (there is minimal other parking nearby on the narrow roads in the area). And while some people certainly come by train or bicycle, many more are put off by the relatively long walk from the local train station. But what if the train stopped at Walden?

The $10,000 MetroNorth spent on its station in 1991 would be about $18,000 today, with more cost, perhaps, because the site is more remote than the Appalachian Trail Station. And two platforms might be required since the railroad there has two tracks, so the cost might be upwards of $50,000 for a simple station. (I don't know if you could get an ADA waiver given the remoteness of the station; if not a mechanical lift may suffice for infrequent wheelchair-bound visitors.) Stops could be made on weekends year-round (or at least during non-snowy months) and perhaps even on some afternoon and evening trains in the summer for late-day swimmers and picnickers.

From there, the Bay Circuit Trail provides hiking in either direction, and the station sits in the middle of one of the largest trail networks and conserved areas in the region. (We can have an argument about the merits of conservation land encouraging sprawl and high housing prices, but this at least is a large, mostly-contiguous portion.) For pond visitors, the swimming beach is a 15 minute walk along a wooded path, rather than a narrow roadway shoulder. This is about as long as the walk from Manchester-by-the-Sea to Singing Beach, at trek made by hundreds of beachgoers on warm summer days. At $15 per roundtrip, it would only take 83 passengers per day for 20 weekends of the year to make back the construction cost of the station—in one year.

Photo from here.
This could be sold as a partnership, or even partially funded through private sources. The MBTA can provide access to a DCR property and other outdoor resources. A station could even provide a safer crossing location for the tracks, which several trails cross and which now have a speed limit of 80, up from 60. It may also help tourism: the train I rode in late December had British tourists headed for Concord looking out the window at the pond, an international destination. Perhaps we should make it easier to access.

The Bay Circuit Trail, Appalachian Mountain Club and The Trustees of Reservations—two large, longstanding Boston-based outdoors advocacy organizations—could work to promote transit-friendly excursions for city dwellers. There would be more options for Walden Pond visitors to get to the reservation (and fewer cars), more revenue for the T with minimal expenses and more opportunities for people from the city who may not have a car to get outside and active. There are questions of erosion: perhaps such a partnership could put a portion of each ticket sold to and from Walden towards the maintenance of the trails there. Other than slightly longer trips for Fitchburg Line weekend riders (and given the padding currently built in to the schedule, I'm sure a stop could be built in without any delay), everyone is a winner.

This should happen.

Boston Transportation Data

Sometimes, people ask for transportation data that they assume doesn't exist but, au contraire, it does. There are some other data I have which I can provide on request, but aren't posted publicly (and some of those are based on calculations I've made which might not be entirely perfect). Here are some of the resources I know of. Most are large PDFs, because that's where data exists, right?
  • Massachusetts traffic volume counts. This has most vehicular traffic counts in the Commonwealth.
  • MBTA Blue Book has a lot of detailed information on MBTA operations. It does not have everything, but a lot: bus and transit ridership, some historic data (older versions of the Blue Book have data going back further; at some point I would like to compile these older data in to one document beyond Commuter Rail, which I have).
  • Commuter Rail counts by train. These data are outdated (2012) but give a good idea of how many people rider specific trains, at least on the day the count was completed.
  • Commuter Rail system survey. These data are even more outdated (2008) but even richer, including access and transfer data. It's pretty dense.
  • Changes to MBTA service since 1964. A very comprehensive look at every bus line in the system, and every change to service in the past 50 years. 
  • Summary of Commuter Rail operations, including the number of train sets required.
  • MBTA salaries, in a very not-useful PDF document. I have an Excel somewhere. 
  • Keolis bid document, with salary information for various positions and job descriptions.
  • NTD Database, with transit system comparisons. You can also download the whole database as a database or in excel.
  • Monthly data by mode, 2007-2014
  • Archive of Commuter Rail schedules (Dave Weebly).
I will update this post frequently as needed. 

Wednesday, December 16, 2015

Go to a Public Meeting … Tonight!

In the past, this page has implored people to go to public meetings. There's one tonight that you should definitely put on your calendar:

The City of Boston and MassDOT is rebuilding the North Washington Street / Charlestown Bridge between Boston and Charlestown. It is designed with top-notch cycling and pedestrian facilities, but the designers seem to have forgotten about transit. Five bus lines cross the bridge carrying 20,000 passengers per day, including the 111 to Chelsea—a lifeline for that community—yet these riders are subject to the same delays as traffic.

There is plenty of room as the bridge is design for transit priority as it is designed, and it is most important to have a lane on the inbound side. At rush hours, 30 MBTA buses (as well as many private, last-mile shuttles) carry 1200 passengers across the bridge per hour. At the same time, the two lanes of traffic are used by about 2000 vehicles, or 1000 per lane per hour. So a lane of buses is more efficient at moving people than a lane of car traffic, even if there's only a bus every 2 minutes.

What's more, this will dramatically improve operations and reliability. Approximately 5000 people cross the bridge inbound by transit during peak hours inbound, at times when there are often five to ten minute delays. If a bus lane saves an average of five minutes of travel time on weekdays, this will amount to an annual savings of 100,000 hours for passengers on the buses serving this route. For bus operations, a five minute savings amounts to 10% of the route, which would allow 10% more riders to be carried on the same number of buses, increasing the efficiency of the route.

It's also low-hanging fruit. The entire bridge is being redesigned, so it's easy to add in a few feet necessary for a bus facility. There are no parking spaces, so there are neither impacts to the neighborhood losing parking nor issues with designing roads to accommodate bus lanes and parking spaces, one of many of the problems the Silver Line has. It should be a slam dunk. If we can't get transit priority here, how ever are we going to get it elsewhere?

The details:

6:30 PM 
The West End Museum 
150 Staniford Street 
Boston, MA 02114
We need as many people to come and speak out for transit priority as possible! See you there!

Can't make the meeting? Send public comments by Dec 25 to: attn Project File No. 604173


Patricia Leavenworth, P.E., Chief Engineer, MassDOT, 10 Park Plaza, Boston, MA 02116, ATTN: Bridge Project Management, Project File No. 604173.

Wednesday, December 9, 2015

Saving GLX by (temporarily) cutting Fitchburg?

The Commonwealth of Massachusetts is mostly finished with a $300 million rehabilitation of the Fitchburg Line, which will yield more capacity on the line and faster speeds on the longest in-state line in the system. At the same time, the Green Line extension project has seen exponential growth in cost, mostly due to bizarre contracting methods and poor oversight. But it's still a complex project. And now an idea is being floated to cut back a branch of the line to Union Square in Somerville—a branch which parallels the Fitchburg Line—which is in dire need of better transit.

But maybe this presents an opportunity for some out-of-the-box thinking. Both branches of the GLX are slated to be built alongside existing Commuter Rail lines, and building next to an active railroad is not particularly easy. There are safety concerns, FRA issues, and even the cost of staff to insure the safety of workers—and the safe passage of trains—comes at a significant cost. If you could cut all parallel rail service, you could save a lot of time and money: rather than having to rebuild in a constrained corridor, you could more quickly build the project in a much easier work environment.

This, of course, would require cutting Commuter Rail off from the the terminal in Boston. For Lowell, this is a non-starter: busing from Winchester or West Medford would be subject to the whims of traffic on I-93, as would buses from Anderson/Woburn. And the Lowell Line also serves Amtrak and some Haverhill trains which bypass traffic on the single-tracked portions of the Haverhill Line. At busy times on this line, there is a train every ten minutes.

But for Fitchburg, this presents an opportunity. Before getting to Boston, Fitchburg Line trains stop at Porter Square. Already, between 30 and 50% of Fitchburg riders begin or end their Commuter Rail trip at Porter, transferring for a trip to Harvard, Kendall or even in to Boston. What if, for a short period of time, you closed the line inbound from Porter to allow for reconstruction, and had everyone transfer at Porter?

While this would be inconvenient for some riders, and put a bit more of a load on the Red Line, it might save a lot of time, and a lot of money, in the construction of the Union Square portion of the Green Line extension, as well as the portion near Lechmere where the branches meet and cross over the Fitchburg Line. Shuttle bus service could be provided between North Station and Porter (a 20 minute ride) but most passengers would take the Red Line. Considering that nearly half the riders already get off, and most jobs downtown are located near or south of the Red Line (in the Financial District and Back Bay) this would be only a minor inconvenience for them. With faster track speeds on the Fitchburg Line, in fact, it might actually be a wash for many commuters.

Using GTFS, I've tested out a few test trips from South Acton:

+16 minutes to North Station
42 minutes via North Station
58 minutes via Porter, Red Line and Green Line.

+6 minutes to Seaport
66 minutes via North Station and #4 bus
72 minutes via Porter, Red Line and Silver Line

+5 minutes to Copley
56 minutes via North Station and Green Line
61 minutes via Porter, Red Line and Green Line

+3 minutes to Park Street
52 minutes via North Station and Green Line
55 minutes via Porter and Red Line

0 minutes to LMA
72 minutes via North Station and Green Line
72 minutes via Porter, Red Line and #47 bus

0 minutes to South Station, Kendall and Harvard
58, 48 and 42 minutes via Porter already faster than via North Station

It's a wash for most commuters other than those traveling to North Station, and most commuters' final destination is not at North Station, but somewhere to the south (since North Station is mostly surrounded by highway ramps and water). In the long run, running trains to North Station makes sense. But if service could be curtailed at Porter for a year to save millions of dollars, I think it is a worthy sacrifice.

In addition, service on the Fitchburg Line could increase in frequency. Each train terminating at Porter would save 20 minutes of round trip running time. This could be translated in to an extra trip in each rush hour to help spread the load, and an extra midday and evening trip as well using the same equipment. It would also be a good opportunity to, once and for all, construct high-level boarding platforms at Porter to allow faster boarding of trains at the station. The two-track station would be more than adequate for the current schedule on the Fitchburg Line, and during the midday, trains could be stored on the tracks beyond the station well shy of construction in Union Square.

The five train sets currently stored in Fitchburg would be enough for full service on the line; the first train arrives in Porter at 6:40 and could easily make the outbound run to Littleton for the 8:20 service back to Boston in the morning. In the evening, the first outbound train to Littleton could easily turn back to Boston in time for the late local departure at 6:20 (which could be pushed back a few minutes with no ill effect on passengers).

With the route mostly cut off from the network, light maintenance would have to be established somewhere along the line (perhaps at the maintenance of way facility near Alewife). For heavier maintenance, trains would have to be shuttled to Lowell across the Stony Brook line from the Willows. This railroad is slow and would need improvement for anything but occasional moves (if it were faster, it could host passenger service from Fitchburg to Boston via Lowell) but would likely be adequate for short-term moves. The extra crew costs would be offset by the savings of rebuilding the inner part of the line quickly and economically.

Is there precedence for this? There is. From 1979 to 1987, the Southwest Corridor was rebuilt below grade between Hyde Park and Back Bay for the relocated Orange Line. The issue was that the Needham Line was only accessible via the corridor. Rather than keeping a track in service and continually moving it around for the rest of the rebuild, they shut down Needham service, replaced it with express buses (which encountered less traffic on the Turnpike than they would today) and rebuilt the corridor in place, That project is obviously larger than the Fitchburg project, and necessitated a longer shutdown, but there are certainly similarities which could yield similar cost savings. See page 202 here.

Tuesday, December 8, 2015

The "Profitable" CapeFlyer: Innovation? Or Accounting?

When a teenager borrows his parent's car for a cruise around town, he assumes that driving is cheap. He drives 20 miles, put a dollar's worth of gas in the tank, and voila, it only costs 5¢ per mile. Never mind that mom and dad plunked down $20,000 for the car, paid sales and excise taxes, and take it in for maintenance (and probably pay for the increased insurance cost that come from having an inexperienced driver as well). Those are all fixed costs, whether Junior drives it or not. Junior only pays the marginal cost of driving—a few cents per mile—which is a small percent of the actual cost of owning and operating a car, so to him, it seems pretty cheap.

The Frontier Institute makes exactly this point in a recent blog post (I even get a mention!). Now lets extend this logic to the recent, inaccurate report that the CapeFlyer train makes a "profit", while weekend Commuter Rail service operates at such a loss that the MBTA Fiscal Control Board wants to cut it. The Control Board unfairly compares the two services, using the same “logic” that makes Junior think that driving is a great bargain.

If someone else pays for every aspect of driving except for the gas, then, yes, driving seems cheap. If someone else pays for the all of the fixed costs of operating a passenger rail service except for the crew and the fuel—as is the case with CapeFlyer—it seems cheap as well. Comparing CapeFlyer's marginal costs against the overall cost of Commuter Rail is as silly as saying that a teenager's topping up the tank of their parents' car pays the entire cost of car ownership.Yet that's exactly what the MBTA Control Board does in claiming that the "innovative" CapeFlyer is a model for weekend Commuter Rail funding. It's not.

A bit of background: in 2013 MassDOT and the MBTA started running the long-overdue CapeFlyer service from Boston to Hyannis (in about 2:20, just 35 minutes slower than 1952!). The train makes three round trips per week, one each on Friday, Saturday and Sunday, from Memorial Day to Labor Day. It has decent ridership—nearly 1000 per weekend—which is not bad given its limited schedule. And according to every published report, it turns a profit.

Wait, what? A train turns a profit three years in a row? If it's that easy to turn a profit, then we should have Commuter Rail popping up all over the place, right? Private companies should be clamoring to run trains every which way. If it is profitable to run a few round trips a week to the Cape (with mostly-empty trains back; I can't imagine there are a lot of Cape Codders coming back to Boston on a Friday evening), then surely full rush hour trains—some carrying upwards of 1000 passengers—must make a mint for the MBTA.

Of course, the reason the CapeFlyer makes a profit is because its finances are accounted for very differently than the rest of the MBTA: it makes a profit against its marginal operating costs, but this misleading calculation does not account for the significant fixed costs that come from running a railroad. There’s nothing wrong with this – the CapeFlyer runs on the weekend and uses equipment, crews and trackage which would otherwise sit idle. It’s a good idea to offer people more rail options. That’s not the issue.

The issue is that when the MBTA Control Board proposes cutting Commuter Rail weekend service, they compare that service against the CapeFlyer, without accounting for the fixed costs that CapeFlyer doesn’t have to worry about. If we analyze the cost of weekend Commuter Rail service the same way we analyze the cost of CapeFlyer service, the Commuter Rail service would look a whole lot cheaper than the T’s Control Board makes it out to be. But when the preconceived agenda is to cut service, a fair comparison quickly becomes an inconvenient comparison, and therefore they resort to the accounting equivalent of comparing apples with oranges.

Let's do some math.

The CapeFlyer runs 15 to 16 (16 in 2015) weekends a year, plus extra runs on Memorial Day, July 4 (sometimes) and Labor Day. That's a total of 47 to 51 days—let's say 50 for the roundness of the number. The distance from Boston to Hyannis is 79.2 miles (about 80), and it makes a round trip each day, so the total distance traveled is about 8000 miles. The operating costs are about $180,000, yielding an overall operating cost of $22.50 per train mile (or about $750 per train hour; note also that a Northern New England Intercity Rail Initiative report had very similar operating costs for a five-car Commuter Rail consist: $22.97 per train mile or $793 per train hour). Just the cost of the crew's hourly wages and fuel adds up to about $150,000 [1]. If you ignore all the other costs of running the railroad (capital costs and depreciation, vehicle maintenance, maintenance facilities, stations, signals, crew benefits, maintenance of way, overhead costs, etc.) then, yes, the CapeFlyer makes a profit.

This graphic contains a footnote which reads:
"Average cost, including an allocation of fixed costs 

which may not vary if service is reduced."

In other words, the report even goes so far as
to admit that most of the costs of operating
Commuter Rail are fixed costs, by a 3:1 margin.
Note how they bury this slightly important
information deep in a footnote that no one is
going to read.
But what are the actual costs? According to the National Transit Database, in 2013, the T's commuter rail operated at a cost of $15.92 per vehicle revenue mile. Not train revenue mile. Vehicle revenue mile. The Cape Flyer is generally made up of nine vehicles: eight coaches and an engine (often two for redundancy; a rescue train would be hours away from Hyannis). So the actual cost of the CapeFlyer is $15.92 times 9: $143 per mile. That's more than six times the direct marginal operational costs which are used to show the "profit" that the train apparently turns.

If the CapeFlyer had to cover all of the fixed costs as well as the marginal ones, it would be bleeding red ink, just like weekend Commuter Rail service.

And there's the rub. The MBTA Control Board wants to nix MBTA Commuter Rail service, because they ascribe to that service the full costs of running the trains, while pointing to the CapeFlyer as an example of a "profitable" service when it comes nowhere near to covering the costs that weekend Commuter Rail is asked to cover. It is not a fair or principled comparison.

Most of the costs in running weekend Commuter Rail are fixed. The Control Board even admits this (see the screenshot from their report to the right). We already have the trains. We already have the stations. And the track, and the signals, and the crew benefits and all else. According to the Control Board documents, the average weekend subsidy per ride is $23.52. Assuming the average fare paid is in the $6 range, fares cover only 20% of the costs of running the trains on the weekends (overall, fares pay about 50% of the cost of Commuter Rail). But let's assume that weekend Commuter Rail trains are accounted for at the rate of the CapeFlyer: $22.50 per train mile [2]. If we only count those costs, the total cost per passenger drops from $29.52 to between $4.52 and $10.33 [3]. That's a lot less than $30, and in some cases, less then the average fare. So, using the same "innovations" that the CapeFlyer uses to turn a profit, weekend Commuter Rail service might well be able to turn a profit, too. Even if you assume that weekend service would use shorter trains (and therefore cost less) it would certainly require less than a $30 subsidy for each passenger. [4]

By not accounting for the fixed costs, CapeFlyer appears three to seven times cheaper to operate than regular Commuter Rail service. But perhaps instead of using these numbers to kill weekend Commuter Rail, we should use them to enhance it. If the CapeFlyer can show a profit at an operating cost of $22.50 per train mile, better Commuter Rail service on other lines likely could as well. Right now, with high fares and infrequent, inconvenient service (on some lines, only once every three hours!), weekend Commuter Rail ridership is low. What if we had hourly service (and perhaps lower off-peak fares) to and from Gateway Cities like Lowell, Worcester, Lawrence and Brockton? What about convenient, hourly service from Providence to Boston (which is generally faster than driving), with tourist attractions at either end of the line. And certainly hourly service to the beaches on the Newburyport and Rockport lines; lines which often fill trains in the summertime, despite the anemic current schedule.

We already pay for the track, the stations, the signals and myriad fixed overhead costs. The marginal cost of running the trains themselves—as the CapeFlyer shows—is relatively low. If we apply the accounting "innovation" of the CapeFlyer to weekend Commuter Rail service, it would be an argument to run more service, not less. If we're going to uphold the costs of Commuter Rail with the CapeFlyer, perhaps we should try that.

A note before the footnote calculations:

While CapeFlyer has a higher fare than any other Commuter Rail service, it is not really a "premium service." The $22 fare for 79.2 miles works out to 28¢ per mile (cpm). For comparison, the 49.5 mile trip from Fitchburg to Boston costs 10.50 (21 cpm), but shorter trips cost a lot more—some coming in at double the cost per mile of the Cape Flyer. So higher fares do not come anywhere close to account for the difference between the "profitable" CapeFlyer and other services. There are savings available for monthly pass-holders, but this is only in the range of about 20%. Even with these discounts, any trip inside Zone 6—approximately I-495—is more expensive than the CapeFlyer.

From Keolis bid documents. 3/4 of the cost categories are fixed costs.
As is some of the fourth.
[1] This assumes 4 crew members for 8 hours including report time and layover time per day, 50 days of operation. $43/hr for an engineer, $35 for a conductor, $34 for each of two Assistant Conductors. Total direct wages of $58,476. See page 10 here for exact numbers, and note that there are a lot of other staff listed there beyond these which are not included in the calculation (nor are benefits, which account for an additional 40% of pay, but are fixed costs). A Commuter Rail train uses approximately 3 gallons of diesel per mile of operation, so 8000 train miles per year use about 24,000 gallons of diesel. Diesel prices have ranged in the past three years from $2.50 to $4 per gallon, for a cost of $60,000 to $96,000. Thus, the total crew and fuel costs range from about $120,000 to about $160,000.

[2] And, no, comparing hourly costs does not make this any different: the CapeFlyer and regular weekend Commuter Rail service both average about 35 mph.

[3] $4.52 if you compare a nine-car train like the CapeFlyer, $7.67 for a five-car train, and $10.33 according to the Control Boards own documents comparing the net marginal costs and the total costs. In theory the T should only be reporting costs for open cars during operation, but that doesn't seem to be the case.

[4] Let's look at this another way. A train from Boston to Worcester runs 44.5 miles. At $22.50 per hour, that means it's direct CapeFlyer-accounting costs is almost exactly $1000. If the cost per passenger is actually $30, it would mean that only 33 people were on the average weekend train. I have taken the Worcester weekend service many times, including the (at the time) 7:00 a.m. outbound train. And there were quite a few more than 33 people on the early train outbound from Boston (more than I expected). The average weekend train carries about 100 passengers per train according to the report's numbers, closer to 150 per train according to the T's data (although ridership has dropped in recent years, perhaps owing to more elasticity on weekends when the train is less convenient and parking costs are lower). For what it's worth, a CapeFlyer train only carries, on average, 130 passengers.

Finally, the T pays Keolis $337 million each year to run Commuter Rail on average—$308 million in 2015—and trains run approximately 4 million miles per year and 22 million coach miles per year. This works out to an average train length of 5.5 cars (about right), a cost per coach mile of $15 (about right) and a total cost of $84 per train mile—about four times what the "profitable" CapeFlyer costs, and that's at the low end of the estimates.

Sunday, November 29, 2015

Hyperloop Test Costs

It's been nearly a year since I posted on the HYPErloop, but since I've revisiting old topics this weekend, here's a quick update. The Boston Globe, as part of their rail day (infrastructure day?) had a column from the president of one of the groups building the Hyperloop (supposedly) who is not named Lyle Lanley. It's going to be great, kids! They're breaking ground on the test track soon!

Now, when we last left off, the whole of the Hyperloop was going to cost $7.5 billion from San Francisco to Los Angeles—a cost of $18 million per mile. That includes everything, apparently, stations, vehicles, maintenance facilities, land acquisition, the whole lot. So the expected cost of the test facility (free land, rural, pancake flat, no stations, etc) would be a lot less than that right?

Wrong. The five mile text track will cost $150 million, or $30 million per mile. It is slated to be built in Quay Valley, a fanciful solar-powered city in the middle of the Central Valley (with stifling summer heat and pollution, to say nothing of the scenery), so the land acquisition costs are, in all likelihood, zero. It's pancake flat there with nothing to go over or under, so there's another zero item on the budget. And still, a test track is going to cost more per mile than the overall project is slated to cost.

As it happens, it will also cost significantly more than parallel stretches of the actually-feasible California High Speed Rail track costs. The 65-mile phase 2-3, which happens to run in the neighborhood of the yet-unbuilt Quay Valley, came in well below the expected budget: $1.2 billion. That's $18 million per mile, barely half what the cost of the Hyperloop's test track will cost on a per-mile. That includes three dozen grade separations, by the way, and it's a firm number; as a design-build contract the bidder will be responsible for (most) cost overruns. And when complete, it will actually be able to carry people somewhere useful (by 2018 it will speed existing Amtrak trips along the same corridor where empty Hyperloop pods may be going around in costly circles)

I'm reminded of an early Seinfeld episode in which Kramer wants to rebuild his apartment with levels. Kramer proposes his plan—on Youtube here—and Jerry's reaction is that it will not happen ("I know that you can't, and I'm positive that you won't!"). Kramer proposes a bet, to which Jerry agrees. The script takes it from there:
MORTY: … So, how are your levels coming along?
KRAMER: Oh, well … I decided I'm not gonna do it.
JERRY: (Sarcastically) Really? What a shock.

JERRY: So, when do I get my dinner?
KRAMER: There's no dinner. The bet's off. I'm not gonna do it.
JERRY: Yes. I know you're not gonna do it. That's why I bet.
KRAMER: There's not bet if I'm not doing it.
JERRY: That's the bet! That you're not doing it!
KRAMER: Yeah, well, I could do it. I don't want to do it.
JERRY: We didn't bet on if you wanted to. We bet on if it would be done.
KRAMER: And it could be done.
JERRY: Well, of course it could be done! Anything could be done! But it only is done if it's done. Show me the levels! The bet is the levels.
KRAMER: But I don't want the levels!
JERRY: That's the bet!
That's about how I feel about the Hyperloop. Could it be done? Of course it could be done! Anything could be done! But it's only done if it's done.

And an over-budget test track is not going to inspire a lot of confidence.

Wednesday, November 25, 2015

More ITDP truth-stretching

I wrote a lot last spring about the ITDP and their biased reporting on bus rapid transit. I thought I was done. I wanted to be done. And then I was looking something up and found this. More specifically, I found pages 23 and 24 and I just don't know where to start. First on page 23:
BRT, however, has a distinct operational advantage over LRT: A BRT vehicle can operate in mixed traffic on normal streets and then enter dedicated BRT infrastructure without forcing passengers to transfer to another vehicle. LRT, by contrast, can only operate where there are rail tracks, and passengers coming from locations not served by the tracks must transfer to and from buses, or to space-consuming park-and-rides, in order to use the system. 
This is just not true at all! A subway system can't operate in mixed traffic. But light rail systems can—and do—all the time! Here's an example from Boston. Here's Pittsburgh. Here's Philadelphia, San Francisco and Sacramento. And all of Toronto. Their big hit against light rail is belied by examples in many of the large cities in the United States! It's just wrong.

And then there's page 24. Oy, page 24. Here, they have a list of average speeds for different transit systems. The idea being, I think, is that look how fast some of those BRT systems are! Look, in Ottawa, the average speed of a bus is 50 mph. 50! That's really fast. That's basically faster than any transit service in the country. I'm really not sure what conclusions they're trying to draw from these data.

And the data seem specious, and I looked in to where they got their data from. Which was from a footnote (well, really an endnote) 129 pages later in the document. They know no one is scrolling back that far (Al Franken—yes, Senator Al Franken—makes this point about Ann Coulter; see page 19). Here's where the data is from:
12. Speed data is from the following sources: Ottawa, Interview with Colleen Connelly, OC Transpo, 2012; Cleveland: Interview with Michael Schipper, Greater Cleveland Regional Transit Authority, 2012; Las Vegas, ITDP, Recapturing Global Leadership in Bus Rapid Transit, 2011; Pittsburgh, Interview with David Wohlwill, Port Authority of Allegheny County, 2012; Eugene, Interview with Tom Schultz, Lane Transit District, 2012; Boston, ITDP, Recapturing Global Leadership in Bus Rapid Transit, 2011; Portland: Interview with Jillian Detweiler, TriMet, 2012; Phoenix, Interview with Abhishek Dayal, Valley Metro, 2012; Charlotte: Interview with Tina Votaw, Charlotte Area Transit System (CATS), 2012; Los Angeles: Interview with Gayle Anderson, Metro, 2012; Kansas City: Interview with Randy Stout, Kansas City Area Transportation Authority, 2013
Let me get this straight. They got speed data—data!—from a bunch of reports they commissioned themselves and from interviews. If they got accurate data, it would be one thing. But these numbers are wrong! First of all, there's no way that the average speed for the Ottawa transitway is 50 mph. And they asked people for an average speed, and people gave it to them, and they probably didn't understand what an average is because the ITDP itself doesn't understand what an average is.

Here's a trip on buses which serve the Ottawa Transitway. It takes 55 minutes on the vehicle. The distance is about 25 km. The average speed? 17 miles per hour. Even an all-on-the-transitway trip takes 20 minutes to go 15 km. That's 30 mph, which is better. But still nowhere near 50. Maybe they got their metric conversions confused?

Let's go to the Charlotte Lynx. Does it average 35 mph? It runs 9.6 miles in 26 minutes. It averages 22 mph. Not 35. Pittsburgh? 9 miles in 22 minutes (25 mph).

And some of the low numbers are too low. They claim Denver's light rail averages 14 mph. Even through the city center, it runs 14 miles in 38 minutes. That's 22 mph. Phoenix? 23 miles in 65 minutes (21 mph). The Orange Line in LA—advertised as 11.2 mph—actually runs 18 miles in 57 minutes, at 19 mph. They're even sandbagging the mode they want to push!

In any case, it turns out that a fixed guideway transitway—light rail or bus—will run at about 20 mph, stops included. The Silver Line in Boston is susceptible to traffic, but at low-traffic times it even makes 12 mph. How did I find these numbers? The Internet, from actual schedules and times. It would behoove the ITDP to actually do some research as opposed to just making numbers up.

Friday, November 20, 2015

Four things the T has done well

I throw a lot of shade at the T. They deserve it. But every so often it's good to remember that while there are lot of things the T does poorly, there are certainly things they've excelled at, things that are done well, and things that are way ahead of other transit agencies. Here are four:

  • In-Tunnel cell service.
    Go to New York. Get on the Subway. Try to send a message to someone. Unless you're at the brand new Fulton Street Transit Center (which cost, uh, a lot) there's basically no service. Maybe once in a while you get a little through a subway grate, and a few stations have wifi. But the T has had service in the tunnels for years (I think it was in 2012 when the Red Line in Cambridge got service) and there are only a few short segments (the E Line at Prudential and Symphony, the Silver Line) left to go. And I've seen some of the strongest signal ever in T stops, standing right by (read: getting zapped by) the antenna.

    The T also had the first Commuter Rail wifi service, but that's never been terribly reliable (although it launched in 2008, which was the dawn of the iPhone). Now everyone can tether anyway.
  • Mobile Commuter Rail ticketing.

    It's now been three years since the MBTA introduced mobile ticketing. As far as I can tell, it works pretty darned well. You can buy single tickets pretty easily, and even get a discount on a monthly pass (since you don't get a subway/bus pass as well). The T was the first to have it, and other cities are catching up. Chicago just launched their app for Metra and it's, well, not perfect. The mTicket app was inexpensive for the T to launch (free, I think, but it takes a cut of each transaction, which makes sense) and works well. And ahead of other agencies. It's hard to argue with that.
  • Countdown Clocks.

    No, the Green Line doesn't quite have countdown clocks everywhere. But they're coming; and the T has done a pretty good job of getting the data and the infrastructure for a relatively low cost. But before you complain, go to New York. Except for the A Division (the IRT lines, or the lines with numbers) and the L train, none of the trains have countdown clocks. The reasons why are certainly very complex, but the Green Line also lacks infrastructure to tell where a train is (except on a much smaller scale). New York has sort of taken a "wait and see" approach, adding the data when signal upgrades are made. The T realized that people might want this information, and made it happen.
  • Open data.

    The T is really, really good about open data. Back in the dark ages (7 years ago, when the DOT was EOT) transit agencies were very protective of their data, wanting licensing fees to release it (really). No one wanted to tell anyone where a train might actually be, and when it might actually get there. Credit to Jim Aloisi for calling everyone out and basically opening up all the data. Within hours of the open data, there were apps.

    Now pretty much everyone has open data, but the T's is actually pretty good, with bus and train locations for all vehicles in the system, useful data aggregation, and a ton of third-party apps. Turns out, it's cheaper to give the data away for free than to try to make your own apps. Shocker.

    Plus, you get some amazing visualizations.

So when you get pissed off at the T because their train schedules are a mess, or their bus schedules are a mess, or they procure untested technology for high prices, remember that there are at least some things that they get right.

And if all else fails, remember: it took the T 120 years to wreck the transit infrastructure. DC has done it in 35.

Tuesday, November 17, 2015

New North Side Schedules: Good, Bad and Ugly

State of the art computer modeling? Yes.
Common sense? No.
Better service? Time will tell.
UPDATE! Schedule changes postponed!

The MBTA's new schedules are out for the North Side Commuter Rail. The changes are … interesting. There are good parts, especially for Fitchburg and Newburyport riders. There are bad parts (things people won't like), especially for people used to less-crowded short turns, and some bizarre service gaps which impact the shoulder periods of rush hour. And then there are ugly parts—parts that just don't make any sense—that show that the process was not fully thought out. Unfortunately, it will probably take some time before this is changed, and require some public outcry for this to occur. A consultant and a computer model were used for these schedules, but common sense seems to have been left by the wayside.


Good news for 
  • Newburyport branch riders
  • Outer Haverhill riders
  • Fitchburg line except Lincoln, Concord, West Concord
Bad news for
  • Rockport branch riders used to every train being express
  • Rockport branch reverse commuters
  • Swampscott, Lynn and Chelsea
  • Inner Haverhill riders, especially Wakefield
  • Winchester, Wedgemere, West Medford
  • West Concord, Concord, Lincoln

Visualizing changes at North Station
The most blatant example of this is the timing of the schedules: there are no trains—on any line—arriving at North Station between 9:00 and 10:00 a.m., and none which depart North Station between 6:30 and 7:25 p.m. So for anyone who works a 10 to 6 shift is completely hosed. This needs to be fixed. If no one rode these trains, I could see the logic. But people do. MassDOT counts these sorts of things. Some trains in the 6:45 range carry 400 customers. Let's not make them all drive, mmkay? Each line should have a "catch-all" train: a train between frequent peak service and infrequent midday and evening service that arrives North Station around 9:30 and departs around 7:00 to fill those gaps. Just because the T defines rush hour as ending at 9:00 and 6:30 doesn't mean demand shuts off entirely at that point.

Here's another thing: right now, arrivals and departures are somewhat spread out, so that passengers on connecting services (read: the Orange and Green lines) aren't all riding at peak hours. In case you haven't noticed, the Orange and Green (and Red and Blue, for that matter) lines are, uh, full. These schedules will mean that more passengers will funnel in to North Station between 8:00 and 9:00, when the trains have less capacity, rather than spreading service out later in the morning when there is some more space on the Green and Orange lines.

And weekend service is still anemic: most lines have gaps between trains of three hours. Why would anyone take the train? Hourly service should be a goal.

Let's also take a quick look at some of the T's stated goals with this schedule change, and whether they are being met, in particular, page 7. (The PDF can be found here.)

* Increased Peak Train Service. In general, this is attained.

* Additional Peak Express Service from Outer Points. This is attained, but often at the expense of service to inner core stations.

Does this look more evenly spaced to you?
Current: Average service gap 25 mins, standard deviation: 13 minutes
Future: Average service gap 35 minutes, standard deviation: 30 minutes
Evenly Spaced Peak Service from Inner Core Stations. No! Not only is this not attained, and do most inner core stations lose service, but some of the spacing gets much more uneven. For instance, service gaps at Swampscott go from 14-28-20-21-31-16-28-9 to 13-47-50-14-38-10. Currently, the range of gaps is 9 to 31 minutes; with the new schedule it ranges from 10 to 50!

The average time between departures goes from 25 to 35 minutes—a significant service cut, and the standard deviation in the gap goes from 13 minutes to 30! (If you include only arrivals at North Station before 9 a.m., the current average/stdev is 20/7, the future is 26/15. In no way is that evenly spaced.

Not only is this goal not attained, but it is downgraded significantly from current service. Again, they may have used computers, but lack some common sense.

* Optimized Peak Arrival and Departures from North Station for Key Job Start and End Times. Key Job Start Times is completely undefined, but it seems to mean "everyone works 9-5." This is not the case.

The T claims a goal is to reduce crowding.
But the new schedule has arrivals bunched at North Station.
Efficient Train Movements at North Station to Mitigate Crowding. Guess again. There are actually more concentrated arrivals and departures at North Station with the new schedule. This means more people crowding on to connecting services—the Green and Orange lines, and shuttles like EZRide—which are already at capacity.

* Standardized Peak “Box”: AM Peak North Station Arrival: 6:00 -9:00 PM Peak North Station Departure: 3:30 - 6:30. This is a solution looking for a problem. Passengers aren't concerned about standardization of schedules between lines. They are concerned about having a train which serves their needs. And this doesn't address the fact that the span of rush hour service, and the location of service gaps, is worse with the new schedule.

* Regular Off-Peak Departure Times. In general, this is attained.
Line-by-line, here are some

The Fitchburg Line (New Schedule Here)

Fitchburg wins. So do North Leominster, Shirley, Ayer, Littleton and South Acton. Beyond there …

The Good. After investing a lot of money in the Fitchburg corridor in the past three years, cutting back weekend service during that time while the improvements were made. Now that it's mostly finished, top speeds go from 60 to 80 mph, and with some high level platforms to speed boarding. The result? Travel times drop from 4 to 14 minutes, with the greater speeds on express trains. The fastest "super-express" that runs from Porter to South Acton drops from 32 to 24 minutes—a 25% improvement—giving that segment an average speed of 55 mph: the fastest single run of any train on the T's Commuter Rail system. (Yes, even the Providence Line: the T refuses to spec equipment to run over 80 mph even though that line supports 125 mph.) There are more express trains, more local trains, and an earlier arrival at North Station. It's a real win.

Train 410 boardings, 2012.
From here
The Bad. The new express services are nice, but there are some issues. First, West Concord, Concord and Lincoln lose express service all together. This is a reduction in service for these communities, which will have fewer trains and longer trips. In addition, parking at South Acton is usually full before 7 a.m., so even though later express trains will stop there, they may have fewer riders as no one will be able to get to the station. Running the later express trains local to Lincoln would make too much sense, apparently, to help spread the ridership.

With the new track work, the schedules for these express trains will still be very competitive, but they'll have more ridership. In the evening, one of the two express trains should have service added back at Lincoln and the Concord stations. Despite all the improvements, Concord, West Concord and Lincoln go from 5 departures to 4 in the morning. On train 410, Concord is the busiest stop, and more than 200 passengers board at these three stations (see right, from here). There is very little logic in cutting these stations. If anything, they should get a second express. How this counts as providing better service is questionable.

Weekend schedules get slightly better with all trains running to Fitchburg. But there are only trains every two-and-a-half hours, which, well, sucks. This kind of schedule means that the train is a mode of last resort on a weekend, and will attract few riders. This is an issue for the MBTA as a whole.

At some point, I'd heard that the plan was for the first train to arrive at Fitchburg before 8:00. 8:35 is certainly an improvement over the current 10 o'clock arrival, but could be earlier. The 45 minute headways the line currently has inbound between 7 and 9 p.m. shift to the 4 to 6 p.m. range, great for reverse commuters.

The new schedules should attract more commuters, especially to the express trains from Acton and Littleton. Driving Route 2 will never be time-competitive at rush hour with a 35 minute trip from South Acton to Boston. The issue there is that the parking lots have limited capacity and fill early. Increasing parking may be necessary, and in the longer range a station at West Acton would help spread the demand.

The Ugly. The last Fitchburg train leaves at 7:13—earlier than the current schedule—and there's not another train until 9:00. The last rush hour train leaves Littleton at 7:51. In the current schedule, there's another train at 8:58, but the new schedule has a gap until 9:35. So there's no train in to North Station between 9 and 10 a.m. Want to drop the kids at school, grab a train at 8:30 and arrive downtown at 9:30? Tough luck. The T decided that rush hour trains should arrive between 6 and 9 a.m. and doesn't seem to understand shoulder scheduling. Many other Commuter Rail schedules go from 30 minute rush hour to 60 minute midday headways (or longer) with a 45 minute headway in between. The same thing happens on either side of the evening rush. The T apparently doesn't think that anyone works a non-9-to-5 schedule. If you want to grab a drink after work, you better grab two; you'll be waiting a while.

There's also the issue of what I call the "last trip fallacy." Often the second-to-last service a route runs will be relatively busy, and people will see the last service as a safety net. So it appears that the final service doesn't get high ridership, but it is the reason that the run before gets better ridership. If an agency cuts the final service—or in this case, creates a gap between the final rush hour train and the first evening train—people traveling at that time will simply abandon the system. There is no excuse for this type of poor service planning.

* Push trains 410 and 492 back 10 to 20 minutes for a slightly later departure.
* In the evening, push train 421 back 15 or 20 minutes to better fill that service gap.
* Serve Lincoln, Concord and West Concord with trains 406, 410 and 417.

The Lowell Line (New Schedule Here)

The Lowell Line is currently the best-scheduled line the T runs. It has hourly, clockface schedules (which this page has long advocated for): show up midday at Winchester at 42 minutes past the hour and there's a train to Boston. It's what most other cities do, but the T can't be bothered to run this sort of schedule on any other line. Lowell suffers from other issues, notably poor highway access to the Lowell and North Billerica park-and-ride lots (an extension to a park-and-ride lot at the New Hampshire border would be key—but should be paid for by New Hampshire), and a long gap between stations along the Billerica and Tewksbury Line.

But otherwise, this schedule is … ugly.

The Good. There's not much improvement to report. I guess at least you don't lose the hourly midday headways. That's nice. And the bizarre and silly 322/216 pairing mid-afternoon goes away (two inbound trains ten minutes apart). Outbound service gains an express, and gets a nice gap-plugger 3 and 4 p.m.

The Bad. The weekend service. The T demonstrates five days per week that hourly service is feasible on the Lowell line, and that ridership supports it. Lowell, with the National Historic Park and Folk Festival, as well as a dense downtown, could easily support weekend headways. Yet it only sees a train every two hours.

In the morning, inbound service before 10 a.m. actually decreases significantly, from from 11 trains to 8! Neither of the Haverhill line trains stops at Wilmington or Anderson, reducing service on the inner part of the line. Right now, there are trains inbound from Anderson at 8:30, 8:45, 9:15 and 9:35. The new schedule has no train departing between 8:36 and 9:35. There's plenty of room at the park-and-ride at Anderson, and today, if traffic on 93 is worse than usual, you don't have to wait too long for the next train. With the new schedules? Good luck: miss 8:36 and you're waiting an hour. The trains at 8:45 and 9:15 aren't empty. Maybe both of those trains need not run, but a service around 9:00 seems like it would be necessary.

While the outer end of the line doesn't lose service, the inner part gets killed. Winchester, Wedgemere and West Medford go from 9 trains arriving North Station before 10 a.m. to 6—a 33% service cut!

Outbound, the same issue arises: there's no train between 6:25 and 7:25. Right now there is a train at 6:25, 6:55 and 7:30. The new schedule just kills off the 6:55 run. That's just plain silly.

The Ugly. Let's start with trains 304 and 306. Train 304 runs local out of Lowell at 6:40. Train 306 runs express ten minutes later and is scheduled in to North Station just four minutes after 304. Even without any issue on 304, this would mean that even without any delay, the express would be riding the local's signals. This makes no sense. It could be easily solved in two separate ways: 1) Swap the local and express runs for 304 and 306. 2) Run 306 five minutes later, so that instead of a 10 minute gap and then a 24 minute gap, you get more even headways.

As commenter Tommy points out, the longest inbound gap in service at any time in the Lowell schedule is between—drumroll please—8:30 and 9:45 in the morning. Who goes to work between 8:30 and 9:45 in the morning? Anyone? Bueller? The answer is: more than 213 people from Winchester, Wedgmere and West Medford alone! The T touts "consistent, hourly local off-peak service" but doesn't bother to provide this at the shoulder period of rush hour. Please.

Then there's the Downeaster. The T claims they used computers to plan for constraints like Amtrak services. So then how would they explain the following:

* MBTA 337 departs North Station 6:00, arrives Anderson 6:25
* Amtrak 687 departs North Station 6:05, arrives Anderson 6:23

Now some of you may look at that and say "ooh! ooh! midline overtake!" Nope, there's an inbound scheduled to leave Anderson at 6:05 and arrive North Station at 6:30. Sorry, guess again.

This is simply a glaring oversight. The T hired a contractor, ran a bunch of simulations, and completely borked Amtrak's schedule. That's just sloppy. Come on.


Oh, boy, there are a lot here:

* Fix trains 304 and 306. This is not hard! Push 306 ten minutes later (it won't affect the following Downeaster) to fill the gap.
* Stop each of the Wildcat trains at one of either Anderson or Wilmington, and maybe inner stations as well. Those trains aren't full at Ballardvale, and the extra time to make one stop will not drastically affect the schedule (especially at Wilmington, where the line speed coming off of the Wildcat is 20 mph so the stop penalty is minimal). In the long run, running more trains via Wildcat and having skip-stop service on the inner part of the line would make a lot of sense; this can be a separate post all together.
* Push train 312 back to at least 8:25, its current departure time. Run it local; the two Haverhill Line trains will provide express service to Anderson and Wilmington.
* Good god, fix the Downeaster conflict. That's just painful.
* And run service between 6:25 and 7:30. Reinstating the 6:55 Wildcat train would make a lot of sense.

The Haverhill Line (New Schedule Here)

The Good: Line speeds get faster; train schedules improve by 5 to 6 minutes for many trains. That's good, and might mean that some trains running ahead of schedule will no longer have to wait a couple of minutes to maintain schedules. Train 206 will run from Haverhill to Boston in 58 minutes via the Wildcat Line, which is really quite good, although some of this time is saved with an express run in on the Lowell Line; this will result in a reduction in service on the Lowell Line. Another change is that midday short-turns are mostly eliminated, including the Anderson turn (which then ran duplicative service via the Wildcat), with more full-line service. This is good.

The Bad: Same as above. No trains inbound between 9 and 10, but that's actually the current case for this line. And since this schedule is only being tweaked, it might be better planned when it's complete.

But for some outbound trains, there's some weirdness. Right now, there are outbound trains at 2:20, 3:10, 3:50 and 4:30. The new schedule has a big gap between 2:00 and 3:30 for the outer stations (vial the Wildcat Branch) and until 3:45 for the inner local. The 5:15 express, which currently makes its first stop in Wakefield, now will skip Wakefield—the third busiest stop on the line—but won't save any time. This stop should be added back in. (More than 1000 people—more ridership than the Wakefield station!—have signed a petition calling for just this.)

The Ugly. Like other lines, there's a big gap—70 minutes—after 6:30. The current schedule then has an 8:40 train; that disappears. So there's a two-hour gap in the early evening where there wasn't one before. This could adversely affect ridership in this area.

We'll see if they can further disrupt the line once the bridge maintenance in Haverhill is done.

* See if there is a way to account for the long service gap between the 2:00 departure an the first evening departures at 3:30 and 3:45. Maybe running the 3:30 as a local train would free up the other set to run another local run earlier in the afternoon (or reinstate the 6:55).
* Add Wakefield back in to the afternoon express run.
* Add some stops to Wildcat trains (see Lowell Line)
* Add service back to the evening schedule, which loses service with the new schedules.

The Newburyport/Rockport Line (New Schedule Here)

The Good: Newburyport gets some express trains. They haven't had those in a long time (maybe not since the '60s). Rockport loses some express trains but gets a later morning departure—currently the last train inbound on the Rockport branch departs at 7:22; the new latest train leaves 28 minutes later. That's good! Newburyport's last train gets a little earlier; 7:42 instead of 7:55.

Afternoon reverse-peak service on the Newburyport line is improved dramatically, with three inbound rush hour departures, no three-plus hour service gap after 5:55, and hourly departures all evening. This serves Ebsco Publishing, which has several dozen Commuter Rail riders from Boston and the North Shore owing to a generous transit benefit, limited on-site parking and a short walk from the station to the office.

The Bad: Service on this line is squeezed dramatically, and stops closer to the city lose service.

There is significant reverse-peak ridership on this line to Ipswich, where Ebsco Publishing has a parking crunch and transit benefit. Yet there's no outbound service between 7:15 and 9:00. The new schedule doesn't address this, but it could, and improve late-morning commute times as well:

The 7:05 Beverly turn train could be extended to Newburyport. This would give Newburyport a later departure in the range of 8:05, with an Ipswich arrival of 7:45. It could then depart at 8:20, depart Salem around 8:50, and arrive at 9:20. What happens to the current short turn? There are trains inbound from Beverly at 7:53, 8:00 and 8:10. The middle train is the current turn. An extra car on the 7:53 and 8:10 would take the necessary capacity. We don't need Orange Line-level headways on the inner part of the Eastern Route, at least not at the expense of later service.

There's also the question of why there is a 50-minute gap between Rockport service during the height of evening rush, and two Newburyport trains running 20 minutes apart. These could be better scheduled. And the current Salem transfer for the 6:45 evening train—which allows it to serve both lines—is gone. This is a good idea poorly executed: there is really no explanation why the transfer should take 23 minutes. Other Commuter Rail systems with transfers are able to run them much more quickly; the transfer train could sit north of Salem, wait for 171 to pull out, get backed in by the conductor, load, and leave 5 or 10—at most—minutes later. But instead of fixing this, it is axed. Great. (This would be a good place for DMU service, allowing one train to run the trunk route from North Station to Salem and set of DMUs to pick up one of the branches, especially for non-rush service.)

And for Swampscott, Lynn and Chelsea? Sorry, guys, you lose two trains per rush hour, and while you currently have service every 30 minutes or better, get ready for some 50 minute service gaps.

The Ugly:

Train 120. Express from Rockport to North Station. I just can't even. I don't have words. It makes my eyes bleed. This introduces a three hour gap for the reverse commute from Gloucester inbound. This is insane. People use these trains. I've been on these trains. They're not full, but they serve the general public. Run this train with stops. If it means that some trains in the 6:00 range are run a bit later, that's probably good. Just make the goddamn stops. This schedule is just stupid.

The huge gaps in late rush hour service both morning and evening is also so bad it belongs in Ugly, not just Bad. There are six trains scheduled in to North Station between 8:06 and 9:00—that's one every 9 minutes. Then there's a 90 minute service gap. Do you run the Red Line every four minutes at rush hour and then have a 40 minute service gap? No! Don't do that here.

The current schedule, while not much better, has six trains in 1:06, and then a 54 minute gap.  We're taking something that is a problem, and making it worse. This is asinine. The above schedule change, which would enhance reverse-peak service, would also plug this hole quite nicely, with a catch-all train leaving Beverly and Salem around 8:50, and arriving in Boston around 9:20. Subway-level service is great, but it should not come at the expense of shoulder-peak times.

The afternoon is similar. Currently, the 6:45 departure from North Station carries more than 400 passengers, and it is unceremoniously brought out back and shot. There are trains leaving North Station at 6:10, 6:20 and 6:30, and then nothing until 7:45. Just pushing each of these trains back by a few minutes—departing 6:10, 6:30 and 6:50—would make a huge difference. If one of these is the turn of 120, then the extra time will give it sufficient time to make inbound stops.

* Train 120. Fix it. It should make inbound stops. Otherwise it's a brutal waste of public funds, running a revenue train with maybe half a dozen passengers and creating a three-hour service gap.
* Fix the service gaps in the later morning and later evening rush periods. Two plans to do so are outlined above.
* A train every 9 minutes is nice, but with any hiccough will cause cascading service issues. Better to have a train every 15 minutes, and get rid of the 90 minute later-morning gap.
* Look in to an improved Salem transfer for a later PM rush hour train.

Monday, October 26, 2015

Fixing the 70 Bus (at least part of it)

I posted about a year ago about the 70 bus, with its bizarre schedule and uneven headways between Waltham and Boston. I have a friend who has taken the 70 daily for a decade, outbound in the morning and inbound in the evening. And she's not alone. After another discussion about the uneven headways, poor schedule adherence and endemic bunching, I decided to ask her (and some of her friends) to keep track of scheduling and crowding. That spreadsheet is growing. If you take the 70, I'm happy to have you help out.

I'm particularly interested in crowding, because it seems the MBTA, in the afternoon, has taken the initiative to have even outbound headways (a laudable goal; morning inbound could use some help, too), but pay no attention to inbound traffic. They may assume that because it runs from Cambridge towards Waltham, everyone takes it in that direction. And plenty do. But plenty also use it for the "reverse" commute; my friend regularly shares the bus with 50 or 60 of her closest friends. The corridor is growing; she reports the bus is getting busier. It needs to be scheduled to provide service to this area, lest potential customers vote with their feet and drive instead.

If you look at the afternoon inbound schedule (part of it is shown to the right), it is scheduled to have bus bunching. There are 16 buses departing Watertown Square between 4 pm and 7 pm, on average, one every 11 minutes. However, twelve of these buses are scheduled to come in pairs, three minutes apart or less. So in reality there is only service 10 times during those three hours, or once every 18 minutes. That's a big difference. To put it in a math-ier way, the current schedule has an 11 minute headway with a standard deviation of 9:25. That's not good. (The goal, with even headways, should be zero.)

With some complex routing schemes, buses come in to a route from one line and leave the other end on to another. But the 70 extends out to the edge of the system, and a bus at the end of a line had little choice other than to turn around and run back to Cambridge. It's particularly complex with three separate termini—Cedarwood, the North Waltham figure-8 (the 70A) and several short turns in Waltham Square—but it's pretty easy to reverse-engineer the schedule and figure out which outbound bus runs inbound, and how much schedule recovery each bus has. By doing this, we can see whether it might be possible to make some small changes to the schedule in order to have buses scheduled at more regular intervals and provide better service overall. Right now, with the T out of space for new buses, we can't really add service. But we can certainly make the service we have run better.

Scheduled times in Watertown Square.
So here's my proposal: by slightly changing two early departures out of Cambridge (in the 3:00 range) and changing layover times by no more than five minutes (and leaving at least 7 minutes for schedule recovery at the Waltham end of the route) we can go from 11 minutes with a standard deviation of 9:25 to the same 11 minute average headway, but with a standard deviation of 2:45. The longest current headway is 23 minutes, with five headways over 20 minutes, and the current shortest headway is 0 minutes (two buses scheduled at the same time). The proposed schedule has all headways between 7 and 17 minutes. They're not perfectly even, but they're a whole lot better.

Take a look at the conditional formatting to the right. Muted, middle colors are less variable. Right now, the schedule seems optimized for even layover times at the expense of even headways. So a bus is somewhat more likely to leave on time, but then scheduled to be part of a bunch. The proposed schedule varies the layover times somewhat more (stdev of 2:19 vs 3:02), but it is far less variable for headways. Note how all the dark blue and bright red colors disappear from the chart on the right, and while there is some variability, there is far less than the current schedule. This means that buses aren't scheduled to run in groups of two, and there are no longer scheduled 20-25 minute service gaps (which, with traffic fluctuations, can easily turn in to waits of half an hour or more, on a route which, nominally, has a bus every 10 to 12 minutes).

The T needs to analyze the patronage of the route, and make sure that they aren't robbing Peter to pay Paul. Service reliability for normal peak ridership is very important, but it should not come at such a great expense to reverse peak riders. They constitute a large percentage of the passengers on the route—likely higher than many other comparable MBTA routes—and the T should balance the needs of all passengers, not just create a schedule which works for one direction at the expense of the other.

This afternoon, I watched the 70 on Nextbus as buses bunched—some as scheduled and some on their own. The schedule for the 70A in particular does not seem to have enough time for all buses on the route, and the figure-8 service pattern in North Waltham is pretty special. The 70 is a very complex route and in the long run the right answer is probably to cleave off the 70A and operate the route as a single route to alleviate bizarre headways and bunching. But in the short term, a few minor changes could be made for the next schedule rating, which would alleviate some of the problems experienced by riders today.

For the reference of current passengers, here's an afternoon schedule which shows the originating point of various buses. Note that Waltham buses are likely more timely than Cedarwood buses, and both are more reliable than the 70A from North Waltham. Note that both this schedule and the one below are sorted by Watertown inbound, so the outbound trips are not in order. (If you're going outbound, it's basically a bus every 10 minutes at rush hour, inbound, it's a bit more difficult to comprehend)

And here is the proposed schedule, with changes highlighted (and the number of minutes the layover is changed shown as well). It should be entirely feasible to implement this schedule, making very minor changes to the early part of the pre-evening rush outbound, and dramatically improving conditions for riders in the evening coming inbound.