Or: why it matters where you draw boundaries. (Or, more specifically, the modifiable aerial unit problem. Thanks, Twitter.)

If you go to the wonderful density.website and find the census tract with, in 2016, as close to 100,000 people per square mile as you can, you’ll wind up in Los Angeles (believe it or not). If you then click the “up” arrow, you can cycle through denser and denser census tracts, and you’ll see a lot of New York City (sometimes the map in the UI doesn’t even reload, but just recenters to a nearby tract). There are a couple of tracts in San Francisco in the 100,000 to 200,000 range, and eventually you’ll get to Le Frak City (click the link to save the trouble of clicking through) which, at 250,000 people per square mile, is the densest tract in New York (about 14,000 people living in about 1/20th of a square mile, give or take).

But go ahead and click the arrow once again.

The page will reload and you’ll be transported not to Brooklyn or Manhattan or the Bronx, but to Chicago. The tract in question has about 1600 people living in 0.003 square miles: two residential towers in Edgewater, on Lake Shore Drive, several miles from the Loop. It’s a bizarrely-drawn tract almost entirely surrounded by a much larger tract, and if the two were combined the resulting tract would be home to about 35,000 people per square mile, on par for the surrounding neighborhood but nowhere near as dense as much of New York City. (The second-densest current tract in Chicago is also made up of North Side high rises at about 90,000 per square mile, but the Robert Taylor Homes, which once housed 27,000 people on 95 acres, was in the neighborhood of 180,000. Cabrini-Green may have been denser still.)

In fact, New York isn’t even the densest city in the country, with several small cities across the river in New Jersey (and one Hasidic village in Rockland County) higher. Manhattan (New York County) is the most densely-populated county in the country at 70,000 people per square mile (although at its peak population in the early 1900s, it was over 100,000; on par with Manila). Two, three and four? Kings (Brooklyn), Bronx and Queens. Of course, New York State is the 8th most densely populated state, and that’s not counting DC, Puerto Rico, Guam, the Virgin Islands and American Samoa. It all depends on where you draw the line.

Which brings us to a New York Times article about downtown areas and their post-covid resilience. They use downtown areas, as defined by CoStar here (the website is a time capsule from the mid-2000s). The issue: there is huge variation between the size of the defined “downtown” areas: in some cases they carefully clip out non-housing areas, in others, they just draw a box around the downtown. The thesis of the article is:

What remain[s] at the heart of many cities in the 20th century [are] blocks and blocks of office buildings filled perhaps 10 hours a day, five days a week — a precarious urban monoculture. […] That means there are few residents to support restaurants at night or to keep lunch counters open if office workers stay away, and few reasons for visitors to spend time or money there on the weekend.

Here is how the story shows the downtowns for Austin and Boston, with the black bar showing a quarter mile. They appear to be similar in size, however, Boston is about 0.37 square miles, while Austin is 1.7 square miles. Not only is that a nearly 5-fold difference, but the Boston metro area is more than twice the size of Austin, so we’d expect a more agglomerated Downtown area with more office space.

The real kicker is that Boston’s Downtown, as defined by CoStar, is surrounded by residential neighborhoods, none more than a 10 minute walk away. Except for the area in the direction of Back Bay, the gray areas surrounding Downtown Boston include Beacon Hill, the West End, the North End, Chinatown and Bay Village, all residential areas. If you take the neighborhoods on the Shawmut peninsula and Beacon Hill, it adds up to about to about 1.5 square miles, with about 41,000 people. Slightly-larger Downtown Austin? It’s grown significantly in recent years, but still is home to only about 12,000 people.

In the article the CBD sizes range from Boston at 0.37 miles to New York, where the CBD is defined as “everything south of 59th Street” which encompasses 9 square miles (New York’s MSA is 5 times the size of Boston’s, but the “downtown” is 24 times larger). Minneapolis is 4 square miles, Saint Paul 0.67. In many cases, the “downtown” area is defined by the highways around it. Other cities, like San Francisco, have several defined districts defined as the “CBD”; San Francisco’s “Financial District” would probably rate similarly to Boston’s in the proportion of office space, and certainly as less office-dense than their Grand Central-defined parcel in New York.

So it’s true that in the Financial District of Boston itself there aren’t many residents. Yet a similarly-drawn box carved into Austin, or most any other city noted in the article, would find the same thing. Taken to a logical extreme, a small box around a single office building would be 100% office use. Or a single tract: there’s one in Midtown Manhattan with a population density less than 1000. There are just 44 residents in the tract bounded by 5th and Park avenues, and 42nd and 49th streets. Of course, just to the east, Murray Hill and Midtown East have 100,000 people per square mile. It turns out that aggregations of this size and variability are not appropriate for this sort of analysis.

Commuting and office use patterns will certainly change post-pandemic. 9-to-5 office districts may be less valuable, and some real estate may be converted to other uses: the article notes an evolution away from single-purpose downtowns. There may be a better metric of office density, perhaps using a certain land area based on the size of the city. But comparing wildly differently-sized “downtown” areas is like saying that, because the densest census tract in the country is in Chicago, Chicago must be the most densely-populated city in the country.

If you start walking or bicycling east in Cambridge from Mass Ave in Cambridge, you can go for six miles clockwise around the Charles River basin before you have to stop for traffic at the BU Bridge. Then the final mile back to where you started at Mass Ave requires crossing both the BU Bridge adjacent to the rotary and then Mass Ave at the foot of the Harvard Bridge. Yet if you were driving along Memorial Drive, you’d be able to proceed barrier-free: there’s grade-separation for cars, but not for people.

I’ve written before about how Memorial Drive can and should be reduced to two lanes, which would allow the parkland along the river to be widened significantly and given the traffic volumes—which match Nonantum Road in Newton, which was narrowed from four lanes to two—would not have a significant detrimental effect on traffic capacity. In fact, Memorial Drive has less than half the volume of Storrow Drive, across the river. So why shouldn’t it have half as many lanes? Indeed, the roadway was narrowed to a single lane during a sewer outfall project with no perceivable loss to traffic capacity.

The question then becomes how to handle intersections, where the current roadway crosses over or ducks under the main crossroads, with turning traffic crossing the bikeways and pedways, poorly-timed light cycles, and a number of dangerous facilities for people not driving cars. These roadways are optimized for the movement of cars, not for the safe and efficient passage of people using more space-efficient modes. However, it would be possible to move vehicles to the existing eastbound lanes and, with a few new signals to allow necessary traffic movement on and off of Memorial Drive, create a corridor which would allow the uninterrupted flow of people walking and biking from Western Ave to the Museum of Science, and allow people to make a 7-mile loop around the Charles River Basin without a single traffic light in their way.

The new signals would operate much like the signal at Charlesbank Road and Nonantum Road in Newton. They would only have a single turning movement, and would generally allow traffic movement in the through direction. Ample storage would be allowed for turning movements as required. And much of the existing ramp space would be changed from impervious land to parkland, for the use of people passing through or, given the larger parkland dimensions, spending time near the river.

Here are some detailed sketches of how each intersection could be rebuilt.

The BU Bridge

The BU Bridge rotary is crossed by the Reid Overpass, a 1930s structure which carries four lanes across the rotary. The rotary itself is poorly-designed for pedestrians and people riding bikes. The pathway is narrow, and requires crossing two separate portions of the BU Bridge roadway on walk signals, which sometimes requires a two-stage crossing. East of the rotary, the pathway is squeezed onto a narrow sidewalk along the roadway, and while a buffered bikeway has been painted to allow bicycles to more easily use the roadway, the facility is unfriendly for people. Meanwhile, through cars have a traffic-free trip across the overpass.

This plan proposes converting the Reid Overpass to a two-lane roadway and a bike path on the southern side. The existing ramps between the BU Bridge Rotary and Memorial Drive would be closed, with the northern ramps converted to two-way traffic. At the ends of each of these ramps, signals would allow traffic to move safely cross where necessary, with turning lanes provided to minimize any delays four through traffic. Westbound through traffic would encounter a single signal at the east end of the bridge, where there is not room for a merge. Eastbound traffic would encounter two signals, one at each end of the overpass. Given relatively low west-to-south and south-to-east movements, these lights would primarily allow through traffic.

This would dramatically simplify the bridge rotary, which could eliminate the traffic light allowing traffic from the rotary to enter Memorial Drive east. A traffic light may be retained for bicycles and pedestrians crossing the southern end of the rotary, but the overpass would significantly reduce this volume and it’s possible that, if it were safely redesigned, this signal could be eliminated as well. This would result in some tighter turn radii on and off of the BU Bridge rotary, but only for traffic going to and from Memorial Drive, which is a route which does not allow truck traffic outside of specific cases. Buses using the roadway (the CT2) would likely continue to be able to use the new alignment, or could be rerouted to use Waverly and Albany streets to get to Kendall Square, moving a stop a few hundred feet across the Grand Junction to do so. It’s also possible that the rotary could be reimagined as a single crossing, although the position of the bridge may make this difficult.

This plan would eliminate the right-on-red movement from the BU Bridge to Memorial Drive eastbound, which often crosses conflicting pedestrian movements (but is still better than the pre-2009 slip lane). The tighter turn radii would require slower traffic for cars entering and exiting the rotary. It would also allow a redesign of the rotary to provide safer bicycle movement and a bus priority lane, as well as significant new green space.

The Harvard Bridge

A mile to the east, Memorial Drive was routed under Mass Ave, also in the 1930s, with two lanes in each direction. Traffic from Mass Ave can turn in three of four movements; there is no provision for south-to-east turns other than a downstream U-turn, which would be eliminated if the roadway were narrowed (there are other routes to reach Memorial Drive on nearby streets). This intersection has a complex series of traffic lights and phases, and busy pedestrian movements in all directions, especially parallel to the river on the busy side path system.

This intersection would behave similarly to the BU Bridge. The riverside roadway and ramp system would be closed, with two new signals to allow traffic to move between Memorial Drive and Mass Ave as necessary. The signal at Mass Ave would be relocated to a single signal closer to MIT with, perhaps, a pedestrian phase for users who do not wish to use the underpass. Since the bridge over Memorial Drive is significantly wider than the Harvard Bridge, this would allow safe bicycle movement as well as turning lanes and queue jumps for transit vehicles in an area where these facilities are lacking today. With more room east and west of the underpass, there would be ample room for turning and merging lanes, so that while westbound traffic would have to pass through two traffic lights, eastbound traffic would be able to proceed unimpeded.

Bicycles and pedestrians would utilize a new facility on the existing eastbound lanes, which would provide a safe crossing of Mass Ave. Additional pathways could be incorporated to allow bicycle and pedestrian traffic to easily move between Mass Ave and the Memorial Drive underpass. By removing the onramps, there would be significant new green space added to the area, which would result in a 120-foot-wide Esplanade along the river by freeing up the green “island” in the middle which, today, is barely accessible because of the roadway. Some parking would be lost along Memorial Drive, but this could be compensated by adding parking spaces in the existing green space south of the two-lane roadway, and by widening the roadway slightly to allow vehicles to be parked more safely. Of course, parking should be a decision made after other users, especially bicycles and pedestrians, have been accommodated.

The Longfellow Bridge

The Longfellow Bridge does not today require pedestrians to cross a roadway (or the Red Line) at grade, but there are two major issues with the roadway and parkland at this cross-section. The path along the river is very narrow on the draw span east of the bridge, and the “HAWK” signal west of there is poorly sited (also, HAWK signals are uniformly awful). The path itself lies alongside a four-lane highway which was built on stilts in the Charles River water sheet itself, so that a high-speed roadway could funnel traffic over the river and under the Longfellow Bridge, which benefits the driver at the expense of people walking, biking, and the health of the river itself.

Unlike at the BU Bridge and Mass Ave, this design would simply narrow the existing roadway to two lanes without adding any additional traffic lights (there is an existing pedestrian crossing which might be appropriate to retain along this portion of roadway, although pedestrian crossings could be handled at the existing crosswalks at either end). The existing underpass for traffic going from Memorial Drive to the Longfellow eastbound would be retained to allow traffic to pass there without requiring the use of a signal, although this could alternatively be used for bicycle and pedestrian traffic if a signal were added in front of Sloan. The roadway would be realigned on the original riverbank with a new pathway alongside, and the existing, highway-style bridges over the river would be removed, with the river restored, and potentially with a more active riverbank installed. The image above shows the extent of the existing bridges which would be restored to open river with this plan. This would allow the pre-1950s roadway plan to be restored, albeit with a wider green space along the river itself for people walking, biking, and utilizing the park as a park, not a highway.

Some Precedent

The DCR has begun a planning process for the western portion of Memorial Drive, from the BU Bridge past Harvard Square, but it has been widely panned as car-centric as it does not reduce the lane width or dramatically improve the pathway system, and does nothing about the BU Bridge intersection. However, there is some precedent for realigning roadways along the Charles River to benefit users, particularly those not using cars. In addition to the aforementioned Nonantum Road redesign from four lanes to two, Memorial Drive itself was relocated away from the river.

Well, not Memorial Drive exactly, but the eastern extension along Cambridge Parkway. The initial roadway followed Commercial Street, now Land Boulevard, inland from the river, with the area between the river wall and Commercial called “The Front.” In 1930, one of the earlier roadways was built on The Front and was labeled as the Northern Artery on some maps, which carried two lanes in each direction (albeit with bottlenecks at both ends). This was renamed Cambridge Parkway and became the main artery, with local traffic on Commercial Street. With the opening of the new roadways under the Longfellow in the late 1950s, the configuration was changed once again: Cambridge Parkway became the route for eastbound traffic, with westbound traffic moved back to Commercial Street. This configuration lasted until the late 1980s, when the industrial buildings were removed, the Cambridgeside Galleria built, and all traffic relocated onto a widened Commercial Street (which became Land Boulevard), with Cambridge Parkway reduced to a single lane westbound with a lane of parking. Thus, over the course of half a century, Cambridge Parkway changed its configuration nearly half a dozen times.

The rest of Memorial Drive has also been reconfigured. Until the late 1990s, the roadway was three lanes wide in each direction (imagine that!) and parking was allowed on both the inbound and outbound sides. The ramp from Mass Ave was two lanes, which merged with two other lanes, to create three lanes, plus a parking lane, which didn’t make sense except that DCR (well, at the time, MDC) seems to think that more lanes are always better. Parking was restricted and the park slightly widened at some point on the inbound side, but two lanes were retained, because gosh there have always been lots of cars here, right?

There’s no law that says that roadways must remain the same forever. In the case of Memorial Drive, what was intended to be a “pleasure drive” has become a riverside highway, with dangerous pedestrian crossings and high-speed traffic even if it doesn’t serve that many people in cars. With some common-sense, low-cost changes (except for some new roadway near the Longfellow, few of these changes would require more significant changes to roadways) could be made in a matter of months on a trial basis. Finally, with lower traffic and greater need for people to be outside, right now is the best time in the past 90 years to reclaim some of this road space for people.

$80 billion sounds like a lot of money for intercity rail, and it is a lot of money, but spread out across the country, it will not bring 90 minute, true high-speed trip times on the Northeast Corridor, as this Boston Globe article notes:

Upgrading the Northeast Corridor to a world-class high-speed system capable of traveling over 200 miles an hour would likely be too expensive even with the added funding, because it would require huge amounts of real estate purchases and infrastructure work to straighten the many curves through southern New England.

This is certainly true. While rail service from Boston to the Connecticut state line is reasonably fast—with trains averaging about 90 mph including stops at Back Bay, Route 128 and Providence, the latter of which also has about 10 miles of slow-running line—the speeds in Connecticut are markedly slower. Trains operate at about 70 mph through the rural eastern part of Connecticut, and then at about 55 mph the rest of the way to New York.

There are a number of reasons for this. Movable bridges require low speeds, and replacing them is expensive, especially across wider rivers in eastern Connecticut and more urban environments further west. In MetroNorth territory, the owner of the railroad—MetroNorth and the state of Connecticut—”optimize” the railroad for commuter service at low speed, which means that even where Amtrak could run faster trains, the track is not maintained to faster specifications. As Alon Levy—who is required reading for any NEC discussion—points out, Amtrak could buy out the Commuter lines, subsidize them forever, and still come out ahead of the game compared with the infrastructure it would take to bypass them.

There is a saying in the German-speaking world (probably from the Swiss): organization before electronics before concrete. It basically means that the highest marginal benefit-to-cost ratio comes from better organization of how transportation is delivered. Beyond that are changes to electronics: signals, power and similar systems. Beyond that is concrete: tunnels, bridges and new rights-of-way. In the case of the Northeast Corridor, much better organization is needed along much of the line, but is especially apparent west of New Haven, where internecine interagency squabbles and old scores means that Amtrak, as a tenant railroad, runs at slow speeds. If Amtrak could bump the average speed between New Have and New Rochelle to just 75 mph it would save 20 minutes (about 10% of the overall run time from Boston to New York); 90 mph would save an additional 10.

Much of the electronics on the railroad have been solved by the modern power system east of New Haven, although between Boston and Providence, there is a joint organization-electronics issue where the MBTA is steadfastly against any improvement to its rolling stock to use the high speed line and electrification, and runs passenger trains with slow acceleration and a top speed of 79 mph on a railroad which can accommodate trains at double the speed. Increasing the speed of the train would increase ridership while also increasing the overall capacity on the line, and reduce the length of any express bypass tracks which would be required to allow overtaking movements. This doesn’t affect Amtrak speeds significantly, but does affect reliability. (Electronics—specifically the 1930s-era overhead system—are a major issue between New York and Washington.)

Electronics and organization might also allow some improvement to the already-fast portions of the line between Boston and Connecticut, but these benefits would be marginal. Going from 150 mph to 180 mph over 30 miles of line would save 2 minutes of travel time. There are 60 miles of track in Rhode Island and Massachusetts good for more than 120 mph today (aside from a few 110 curves), even if it were replumbed for 180 entirely, it would only save 10 or 15 minutes, and would require a lot of concrete (for instance, a lengthy bypass of the Canton Viaduct, which is the oldest structure in the world with high-speed rail).

So, what about Eastern Connecticut. Could it be solved with organization? Not really, there’s only a bit of non-Amtrak traffic, Amtrak owns the railroad, and runs it about as well as they can. Electronics? The signals and power are not constraints. The issue is a combination of sharp curves, grade crossings and movable bridges, and eastern Connecticut has all three.

Luckily, concrete is cheapest far away from urban areas, and eastern Connecticut fits the bill. From Westerly, Rhode Island to Westbrook, Connecticut, the highest speed allowed is 110, although most of the line has speed limits of 75 or 90, and there are a number of much slower curves (including 25 mph through Downtown New London) and bridge crossings. Organization and electronics may be able to bleed out a couple of minutes of savings, but the fact that Amtrak averages better than 60 mph is already impressive given the geometric issues.

This sort of concrete was discussed in the past decade. As part of the NEC Future project (which is now archived), Amtrak proposed bypassing a 40-mile section of railroad from Westerly to Westbrook, and would mostly use the I-95 right-of-way to build a new high speed line. This line would allow operation at 180 mph without any interference from other traffic, and if it were extended a bit further west to near Kingston, would allow a 150 mph-or-better run for about 50 miles, and would save about 30 minutes compared to the current line: a 15% reduction in trip time between New York and Boston. Indeed, Alon extends this a longer distance to East Haven, and calculates a 38 minute trip time from Providence to New Haven. With some schedule padding, a 45 minute trip time between these cities would be 45 minutes faster than today, doubling the average speed between Providence and New Haven.

Simple, right? Put the train down the middle of the highway, build a couple of bridges, spend a few billion dollars, and save 45 minutes between Boston and New York. That means better fleet utilization and more passengers with higher speeds. Even without the improved speeds, Amtrak’s new rolling stock allows it to plan for hourly trips between Boston and New York, about 15 round trips per day, this could conceivably be increased to 18 with higher speeds. This is equivalent to 14,000 seats per day, each saving 45 minutes of travel time. With 2/3 of the seats occupied, it would lead to 2.5 million hours saved per year. The 3 extra trains each way would recoup somewhere on the order of an additional $50 million in fare revenue with relatively low additional costs (really just power and equipment wear, since they would use existing rolling stock and staff time).

This relatively sensible change ran into a big obstacle: NIMBYism. Nearly all of the direct benefits accrue to passengers on trains which do not stop in Connecticut: nearly every Acela train between Providence and New Haven runs nonstop. Even if it moves people from driving to the train, most drivers would use the inland route from Boston to New York, so local traffic wouldn’t be impacted. So when it was proposed in 2016, Southeastern Connecticut took little time to turn against the improvements, and brought out some particularly delicious NIMBY arguments against the project.

Some arguments:

One of the options would cut straight through the heart of the town

You know what else cuts through the heart of Old Lyme? I-95. Which is what the train would run alongside.

It would cut through Olde Mistick Village, Mystic Aquarium and a Pawcatuck golf course.

Olde Mistick Village is a shopping mall with poor spelling which is next to the Mystic Aquarium. Both are adjacent to the 340-foot-wide right-of-way for I-95 where the railroad would be sited. Cars in the distance? Fine. Trains? My heavens! The Pawcatuck golf course is in the path of Route 78, which was originally built to connect to the never-built Long Island-Westerly bridge. The golf course’s owner says: “I’m all for expanded rail, but we should use the right of ways that we already have.”

Guess what, that’s exactly what is being proposed! We have the right-of-way, and it connects right through your golf course. You chose to build a golf course in the direct path of a highway, what did you expect? (It’s worth pointing out that Alon suggests cutting from I-95 over to the existing right-of-way further east in Rhode Island to avoid a number of additional curves, negating this as an issue.)

But the NIMBYism goes on. Apparently, a new rail right-of-way along a highway would impact tourism.

“They’re going to have a fight on their hands,” said Joyce Resnikoff, co-owner of the 43-year-old Olde Mistick Village.

She said such a rerouting would be devastating to Mystic, which thrives on tourism.

“I can travel all over the world and everyone knows Mystic,” she said. “We all work very hard to support tourism here. It’s big business.”

First of all, people from across the world do not all know Mystic. How a new non-stop train line would devastate tourism is also perplexing. It’s not like crowds of travelers on Acela today pile off the train to go to Mystic’s big businesses: no Acela trains stop there. Before covid-19, there were five trains stopping (a sixth was planned), with an average of 78 passengers per day. If high-speed trains were shifted to a separate alignment, it would allow a regional New Haven-to-Providence service to run hourly (or better) along the existing coastal line, with a lot more service and, probably a lot more ridership.

“We already have rail lines coming into our communities. To go in a different direction seems bizarre to me … I think their bigger step is to fix what they have.”

Slow rail lines. It’s hard to say what seems bizarre other than this guy is a town council member, not an engineer. What we have is unfixable.

The Southeastern Connecticut Council of Governments sent a letter …expressing its concerns about how [the alignment] would relate to or disrupt land use and environmental resources in the region and how any of the alternative routes could result in potentially fewer train stops in New London.

There are some environmental questions, although most of the right-of-way would be adjacent to the highway and already disrupted. Then there’s just complete misunderstanding of how the railroad works:

Stonington First Selectman Rob Simmons agreed, saying the disruption, huge cost and effort of the project would only save riders a few minutes. He said that when the existing line was electrified for high-speed trains, the Acela cars were designed to tilt so they could navigate the curved tracks in the region without slowing down.

The tilt mechanism helps to wring a few extra minutes out of the curvy route along the coast. It does not allow 180 mph operation on a line from the 1840s.

But there was a public meeting, and the pearl clutching really got going. We got to hear about how long people have lived somewhere:

Jeff Andersen, director of the Florence Griswold Museum in Old Lyme. Andersen has been working in Old Lyme for 40 years. “It’s an American treasure, and it needs to be protected and preserved,” said Andersen.

Did the Florence Griswold Museum complain when I-95 was built 1000 feet away? Probably not. But now that a train might be built in the same right-of-way, my heavens. It gets worse. Senator Blumenthal represents the local population, but should know better, especially since shaving 30 to 45 minutes off (more than “a few minutes”) would be a significant benefit for his constituents in New Haven and Stamford.

Old Lyme First Selectwoman Bonnie Reemsnyder said this is not about NIMBY, “Not In My Back Yard,” but “about our cultural and environmental survival in Old Lyme.”

Old Saybrook First Selectman Carl Fortuna said the proposed bypass would shift the location of his town’s train station, which the state and town have invested in and is considered the “economic center of town.”

In East Lyme, the route would proceed through several neighborhoods, a new 400-unit apartment complex and a proposed Costco, Flanders Four Corners and possibly a recently identified site of tribal significance, East Lyme First Selectman Mark Nickerson said.

Stonington First Selectman Rob Simmons said the new segment would bypass the Westerly train station and wipe out Mystic Aquarium, Olde Mistick Village and “the gateway” to Mystic, the state’s top tourism destination.

All of this wiping out of historic town centers (and Historic Costcos) is rather preposterous, since most of the line would be built within the existing highway right-of-way, which is generally 350 feet wide, plenty wide enough for a two-track railroad and a four-lane highway. This rhetoric, however, sank the planning process, so while we could have entered Infrastructure Week in the Biden Administration with a path forward to speed the Northeast Corridor, we’ll now have to start from scratch.

Part of the blame may go to the study’s authors, who included vague maps of the corridor which were probably not clear enough about which areas would and would not require property takings. And there is a valid concern about moving the train station in New London from Downtown to somewhere a bit further afield (although there are plenty of examples abroad of new train stations outside of old towns which maintain their legacy service).

Still, the lion’s share of blame should go to NIMBYism (it’s usually the case that if you have to start your sentence with “I’m not a NIMBY” you probably are one). It will take both better communications about what is actually going to happen (no, no one is going to bulldoze the Mystic Aquarium) as well as leaders who are willing to tell the truth to their constituents. It may even take some carrots (get on a train in New London, be in Boston in an hour). It will also take a system which allows construction costs that do not escalate to a point of oblivion: a 50 mile section of highway-ROW high speed rail with two major bridges should cost $2 billion, not $20. Before concrete, we’ll have to have organization to allay the fears of the North American NIMBY.

In 1980, there were about 175 miles of what could be considered streetcar or light rail operating in the United States. This was based entirely in about half a dozen “legacy” systems which were never converted to buses (a few miles have been converted to buses since then, this figure does not include streetcars abandoned after 1980), in general because they used private rights-of-way and, in most cases, tunnels in:

• Boston
• San Francisco
• Philadelphia
• Pittsburgh
• New Orleans
• Cleveland
• Newark

There had been minimal expansion since the 1930s (most notable: the MBTA’s 1959 Highland Branch) and the systems mostly used PCC cars (or older streetcars in the case of New Orleans). The ill-fated Boeing LRV cars were only first being introduced in the late-1970s.

Over the next 35 years, light rail and streetcar mileage increased by a factor of five. This began with 100 miles in the 1980s, accelerated further in the 1990s, and further more from 2002 to 2016, with 30 miles per year being added per year from 2002 to 2012, and additional increases through 2016.

In 2016, however, LRT implementations stalled out. Funding protocols had changed, and in several cases, expensive light rail systems had been built in areas with marginal supporting land use, leading to stagnant or declining ridership in some areas. “Modern streetcars”, which do not add up to much mileage but had been some of the least cost-effective projects, had mostly fallen out of favor. 2017 marked the first year since 1989 with no new LRT mileage opening, and only 18 miles of new light rail have opened since (2020 was another year without any new mileage).

This will change in 2021 and 2022, when, over the course of a few months, five large light rail projects set to come online:

• LA’s K Line (Crenshaw)
• San Francisco’s Central Subway
• Boston’s Green Line Extension (in Cambridge, Somerville and Medford)
• San Diego’s Midcoast Trolley
• Seattle’s Northgate Link

What is notable is that these are already some of the largest and most effective light rail systems. By total ridership, they rank, respectively, 1, 2, 3, 5, and 10, and by ridership per mile, they rank (again, respectively) 10, 2, 1, 12, and 3. Together, these projects will add 30 miles of light rail service, but they are not what we saw a lot of in the 2000s, with often suburban extensions with low ridership. In each case, the new portion will have a higher ridership density than the existing system did pre-pandemic)

• K Line: 4050 (system: 2517)
• Central Subway: 42941 (4479)
• GLX: 10465 (system 7846)
• Midcoast Trolley: 3154 (system 2157)
• Northgate 5116 (system 3728)

These are really LINO projects, or LRT-in-name-only. (Or maybe LIVO; LRT in vehicle only.) The projected ridership would fall in the range of many existing heavy rail systems, and with the exception of the K Line in LA and the first few hundred yards of the Central Subway before the portal, there are no street grade crossings for any of these projects. They will all use two- three- or four-car trains to provide as much capacity as subway systems, just with light rail cars, providing travelers with frequent service at speeds which, during peak hours, are faster than driving.

If LIVO is the direction that light rail is headed, it is indeed a good one. Some earlier light rail systems are streetcar-light rail hybrid systems which can be stymied by surface congestion, or followed highway rights-of-way or relatively remote rail line corridors which reduced ridership potential. These decisions are often made in the name of reducing capital costs, but result in increased operating cost per rider. No one would accuse any of these five new projects of being built on the cheap: each checks in at well over $1 billion, with several miles of tunnel (in the case of Seattle and San Francisco, most of the project is tunneled) or viaduct. All should probably be less expensive per mile, but assuming travel patterns return to near some pre-covid normal, all should have high ridership.

In essence, they do what light rail can do best: fit a high-capacity transit system into an urban ecosystem where it something beyond the capacity of a busway is needed*, but where heavy rail or commuter rail would not be feasible. San Francisco’s Central Subway is a subway with a surface station and connection to existing transit lines. Seattle’s Link system is a mostly-elevated, sometimes-tunneled, and sometimes at-grade system (although after the initial segment, they seem to be moving away from at-grade portions). Los Angeles combines tunnels and at-grade segments, and San Diego’s is an extension of an at-grade system with grade crossings. Boston’s system mostly follows existing Commuter Rail lines and could have been built on those lines’ existing alignments, however to do so would have required completion of the North South Rail Link. Light rail, on the other hand, can allow existing vehicles to run through the city, rather than terminating and turning back at several congested 19th- and early 20th-Century Downtown terminals, which will serve to improve service across the entire line.

(* A note on busways: while most any of these projects could have been built as a BRT, all would have significant downsides. Once a project is nearly entirely grade-separated, the construction cost differential for BRT dissipates. BRT also has two major issues operating in tunnels. One is ventilation: without full electrification, buses require exhaust for diesel power. Buses also need more roadway width to operate, while rail can be fit into a more constrained environment. Buses also have lower per-vehicle capacity than rail, leading to higher labor and operating costs per passenger. Given the high ridership density for these routes, BRT would not likely be the appropriate mode.)

Light rail will probably not again see the kind of growth it did from 1990 to 2016. Some of these systems which are mostly at-grade and have lower ridership potential would have done better as bus rapid transit, such as Sacramento, San Jose, or Houston. Others, like Denver and Dallas’s sprawling systems mostly along rail rights-of-way, would be better suited for Commuter or Regional Rail, and indeed both these cities have shifted towards this mode. With the promise of new funding, light rail implementations will hopefully hew closer to the recent implementations than the “every city gets a light rail whether they need it or not” service of the 1990s and early 2000s.

Light rail is a bit of a chameleon. Depending on its situation, it can act as a streetcar, heavy rail, commuter rail or even something akin to a BRT, and can take on both the positive and negative qualities of each of these modes. It does best when it serves a corridor which would otherwise not be well-serviced by any single mode, and where it can provide high-speed, high-capacity, high-frequency service while shape-shifting as it moves through the urban environment.

I should start by stating the obvious: this isn’t really urban planning-related post. Except that shutting down a major city for a road race kind of is urban planning. And busing 30,000 people certainly is. But I’m not going to focus on that.

Some of you may know that I run marathons from time to time. Some of you may even know that I had an infamous run-in with the heat in Boston a few years ago. I have a number for this spring, and I’ve even been attempting to do some training. But with COVID-19, the whole operation appears to be grinding to a halt: the B.A.A. is planning to postpone the race to the fall, and they are rumored to be looking at dates in mid-September, when the average high temperature is in the mid-70s and one in five days eclipses 80.

Here’s my two cents on that: It’s a Bad Idea. Not the postponement: that’s probably imperative given the spread of the pandemic. September. It’s far too warm to hold a marathon in good faith.

The April version of the Boston Marathon (I kind of can’t believe this phrase, but here we are) has taken place for 124 years, just shy of the 148 years of weather data we have collected. It takes place on the third Monday in April, so the 15th to the 21st, during which time the average high temperature ranges from 55.9 to 59.6 degrees, and the average low (a decent proxy for humidity) ranges from 39.6 to 42.0 degrees. This is just about ideal temperature for a marathon race, and while there are obviously exceptions: ranging from cold and rainy to warm and hot, the average race weather is relatively amenable to a marathon. On average, the high temperature tops 80 degrees, which is really into the danger zone for a marathon, on average, once every 33 years.

September is warmer. A lot warmer. This is especially the case because, since Boston is near a large body of water, temperatures don’t cool off as quickly in the fall as they would inland. High temperatures in September start around 77.4 and fall to 67.0, so even the end of September is ten degrees warmer than the Boston Marathon in April. Minimum temperatures fall about the same amount, from 61.4 to 51.1. The chance of an 80 degree day is one in three at the start of the month and one in ten by the end, but again, much higher than April (for September 14, it’s about one in four, similar to early June).

Here are the temperatures for the past 10 years for September 14:

72 75 84 90 72 66 68 81 86 75

The 90 degree day would likely require race cancelation (like has happened in recent years in Chicago, Vermont and elsewhere). The Boston Marathon medical team is the best in the business, but holding the race when the weather is likely to be in the 70s will strain their resources. Holding the marathon in September is irresponsible and puts runners at an unnecessary risk. Note that in the chart below, mid-April temperatures don’t occur again until late October and early November. September is pretty much the same as June.

If we consider the other World Marathon Major races, a September Boston would be significantly warmer than any other race:

• Tokyo: Average high of 53˚
• Boston: Average high of 57˚
• London: Average high of 61˚
• Berlin: Average high of 63˚
• Chicago: Average high of 64˚
• New York: Average high of 58˚
• Boston (September 14): Average high of 73˚

There’s a good reason why the number of marathon finishers in the US peaks in October and November, with a secondary peak in April: that’s when the weather is best for running in the United States. September has a good deal of races, but most are small: the average size of a race is just over 200 participants. Moving Boston to September would more than double the number of marathon finishers in September nationwide.

The Boston Marathon used to start at noon, but was moved up to a 10 a.m. start several years ago to better accommodate the April heat. Can the race start any earlier in the day to take advantage of lower temperatures? Very unlikely: the logistics of moving 30,000 people to a remote start that early are next to impossible. The Chicago Marathon starts earlier in the day, but it starts (and ends) downtown, so as long as people can get on trains and buses to get to the start, an 8 a.m. start (with an earlier start for elite and para athletes) is possible.

Not so for Boston. Given the logistics of transportation (aha, I knew I would tie this post in with the blog), 10 a.m. is about the earliest possible start. The transit system’s earliest arrivals in Downtown Boston, where buses leave from is in the range of 6 a.m. Early-wave runners are told to be on buses before 7. Moving the start up an hour would move the start before the start of transit service. While trains and buses could be run earlier, the cost and complexity of doing so would be quite high, and without it, the downtown area would be overrun with parking and drop offs for 30,000 runners, plus spectators and other fans, on a day when many of the major streets are closed for the race.

So the race can’t be held in September without the high likelihood of high temperatures impacting the race. Late September would be better than early, although the last two weekends of the race are Rosh Hashanah and Yom Kippur, which have two major issues. The first is that it would be akin to telling many runners to run a race on Christmas or Easter (although the race does infrequently take place on Passover in April, and the day after Easter). The second is that the race passes through Newton and Brookline, and the course is home to no fewer than three major synagogues (with several others just off the course) and those two towns would likely veto either of those dates for the race. (The only option might be Monday September 21, but again, the average high that day is still 70.9 degrees, and one in six September 21s have exceeded 80 degrees.)

So that pushes the race into October or even November. This is prime marathon season: over the course of four weeks there are two other major marathons (Chicago, on the 11th, and New York, on November 1; the two largest marathons in the country) as well as the 4th (the Marine Corps, on the 25th) and the 10th (Twin Cities, on the 4th). Boston would probably want to avoid conflicting with either New York or Chicago, and might want to avoid the others. Also, October 4th is still significantly warmer than April, with high temperatures around 63 degrees. The 11th also conflicts with the BAA Half Marathon.

This leaves, in my mind, three reasonable dates for the Boston Marathon, assuming a Monday holiday is declared by the state: October 19, October 26 and November 9. Let’s look at the pros and cons of each:

October 19 makes the most sense with one exception: it is the same weekend as the Head of the Charles Regatta (which is basically the Boston Marathon of rowing). This would create a run on lodging in the region, and a run on resources. The weather: a high of 61.6, is reasonable. There are no other major races that week (although there are marathons in Maine, Vermont and Connecticut, but they are more local races) and it might be interesting to have a spotlight on Boston as the hub of the athletic universe for a weekend. But the logistics would be tricky with two major events (although it’s possible that some logistics could piggyback). Another option would be to see if HOCR would move the Regatta back a week.

October 26 makes sense except for the Marine Corps marathon. While a large race, it’s not a “World Marathon Major” race, so it probably doesn’t have the same level of conflict. The weather is a perfect approximation of mid-April, with an average high of 57 degrees. As long as Marine Corps didn’t mind Boston stepping on its toes, it might make the most sense.

November 9 is relatively late for the race, and can be cold and potentially even snowy or icy (although the average high of 53.8 is only slightly lower than April 15), but would have no major conflicts with other national races or local sporting events (the largest marathon in the northeast is the Manchester City Marathon in New Hampshire, with under 1000 runners). It would be late, but would be after foliage season, and at a time when lodging and other event space is most easily available in Boston.

Two other notes: first, the fall Boston Marathon might be significantly smaller than the spring version. The BAA will likely allow deferral to next year’s race, and many runners may take advantage, so the race might be a smaller affair, although the BAA could open late registration to more qualified runners to fill the field. After the 2013 bombing, about 5000 people were unable to finish the race and allowed deferred entry, but most of the field had already finished. This is a much larger population. It will also result in significant registration changes for next year’s race, with registration taking place after Boston, and including runners from more fall marathons, although this might be balanced by the number of races canceled this spring. The results of this will probably cascade for a few years.

Second, the BAA could potentially save resources by moving the BAA Half Marathon to the same weekend, or even day, as the full marathon. This would mean moving it from its course in Franklin Park to the second half of the Boston Marathon course, starting in Wellesley and finishing in Downtown Boston. With an early start, the half marathon racers could be mostly clear of the course by the time the full marathon finished, although it would not allow participants to run both races. Transportation would be quite simple, as the full field for the BAA half could be moved from Downtown Boston to Wellesley on half a dozen commuter trains, since the start line for the half marathon would be adjacent to the Wellesley Square train station. This, however, might put additional strain on the marathon resources, and holding the races a couple of weeks apart might be the best use of resources.