Severe congestion continues to plague the intersections on either side of the Charlestown Bridge, also called the North Washington Street Bridge, as reported in a WHDH Channel 7 News story. The intersection on the North End side is Keany Square; on the Charlestown side, it’s City Square. It’s been at least 2 months since the local North End newspaper reported on the severity of this problem, and more than 7 weeks since my 10/31/2022 analysis of the problem which I provided to the City of Boston and MassDOT, authorities who own and control the intersections and the bridge, respectively. Since then, they have made one small change to traffic signal timing at City Square that I recommended; unfortunately, that change by itself has had almost no effect, and so the problems persist.

The problem is urgent in several dimensions.

  • For pedestrians in Keany Square, it is a safety disaster, as others have documented. We have all seen the videos of pedestrians, including a large number of children, walking through a maze of cars, trucks, and buses that are stopped, blocking the crosswalk – and sometimes those “stopped” vehicles start to move when a gap opens up ahead of them.
  • For the MBTA and its riders, it is a service disaster. Buses leaving Haymarket for Charlestown (Route 92, 93) are in congestion from the moment they pull out of the station; it takes them an extra least 15 minutes to get to Charlestown, and buses trying to return from Charlestown likewise suffer delays. These delays make it impossible for the buses to keep to their schedule; as a result, bus service is extremely slow, less frequent – and therefore more crowded – than it’s supposed to be, and extremely unreliable.
  • And for general traffic, including delivery vehicles as well as personal automobiles, it’s a congestion disaster. Yesterday, at 3:30 p.m., the backup approaching the bridge on N. Washington Street, which then continues as Cross Street, reached all the way to Christopher Columbus Park. It’s a wait of about 15 minutes to reach Keany Square. Commercial Street in the North End was backed up for blocks as well, forcing drivers to wait through multiple light cycles.

The congestion problem has a single source that spreads congestion in almost every direction.

Because one can see congestion almost everywhere on both sides of the bridge, it can be hard to detect the source. The bridge is fully backed up in both directions. But as I wrote in October, the problem stems entirely from one source: poorly timed traffic signals at the City Square intersection leading to insufficient capacity for northbound traffic. At that intersection, there is capacity to discharge only about 720 vehicles per hour; however, many more vehicles are trying to get onto the bridge. As a result, the northbound lanes of the bridge back up, and by 3 pm, earlier on some days, the backup spills back into Keany Square. The queue doesn’t stop there; it keeps moving back on N. Washington Street all the way to Haymarket, and then keeps growing down Cross Street all the way to Christopher Columbus Park.

The spillback into Keany Square also causes secondary queuing, as drivers trying to get onto the bridge but unable to fully get into their receiving lane “block the box,” creating a blockage that stops other traffic movements, including southbound left, that is, traffic turning from the bridge onto Commercial Street. When they get the green, they can’t move because their lane is blocked by cars that turned left from Causeway Street but can’t fully make it into their receiving lane because the receiving lane is jammed. And so southbound left traffic backs up and, because the left turn lane is only long enough to store about 5 cars, it quickly blocks one of the bridge’s two southbound lanes. That means that when southbound gets the green at Keany Square, discharge is cut in half. By 3:30 (earlier on some days), the blockage at Keany Square creates a queue in the bridge’s southbound lanes that spills back all the way back to City Square, slowing vehicles (including buses) trying to get onto the bridge from Chelsea Street and Rutherford Ave. You can visualize it as a boomerang – congestion starts at City Square in the northbound lanes, spreads back to Keany Square, then infects the southbound lanes, where it spreads back to City Square.

Commercial Street likewise suffers from secondary congestion – during much of their green period, they are blocked by the gridlocked vehicles in Keany Square.

Critical traffic facts

Getting counts of traffic volume during a period of severe congestion is difficult. The standard way of doing traffic counts is to count as vehicles depart, because only then can one observe whether which way they are turning. When there is no persistent congestion (i.e., when the queues clear each cycle), it is fair to say that the volume counted departing an intersection equals the demand arriving at the intersection. But where there’s persistent congestion, demand can be substantially greater than what any counts will reveal. Therefore, traditional traffic counts are of limited value.

Because demand is hard to pin down and the critical problem is northbound capacity at City Square, the only counts that are really needed to find a way to increase northbound capacity are two things related to northbound traffic flow:

  • The proportion of northbound traffic turns right versus going through. When I observed the City Square intersection in October, about 75% of the bridge traffic was turning right onto Chelsea Street. However, later observations by myself, my students, and others indicates that normally, only 25% to 30% of bridge traffic turns onto Chelsea Street, while the majority continues thru onto Rutherford. This, in turn, alters right-turn related aspects of my 10/31/2022 recommendations – an extra right turn lane isn’t really needed, and while allowing northbound right turns during the Chelsea Street phase (a change the City made) is still a good idea, it doesn’t improve things much.
  • The number of northbound vehicles per cycle that use the “auxiliary lanes,” the extra lanes provided to northbound as they approach City Square. From my counts there on 12/20/2022, I found that it averages 12, of which 3 turn right during the Chelsea Street phase and the other 9 discharge during the first 10 to 12 seconds of the northbound green.


Why standard signal timing solutions don’t work

As I’ve explained before, all the congestion problems at Keany Square stem from insufficient northbound capacity at the City Square intersection. The traditional solutions to increase capacity are.

  • Add more lanes. The roadway layout already does this, as the northbound roadway flares from a single lane to 4 lanes. However, the added lanes, known as auxiliary lanes, are short, able to hold between 3 and 5 vehicles as most, and queued cars often block access to the auxiliary lanes before they’re full.
  • Give more green time. The current signal timing does this, too – there is a very long green for northbound, lasting more than 40 seconds. Anyone can see how futile this is – for the last 20 seconds of green, the flow is sparse, with less than one vehicle every 2 seconds, and often with gaps of up to 5 seconds with no vehicles.

Those traditional solutions don’t work here because this intersection, with its short auxiliary lanes, is a non-standard geometry which demands a traffic flow model that correctly accounts for it. Unfortunately, nearly all traffic signal timing nowadays is done using software like Synchro that is based on methods and formulas of the Highway Capacity Manual, which does not have formulas or methods for dealing with short auxiliary lanes. It’s easy to see that treating the northbound approach as though it has 4 full lanes, is wrong, but it’s equally wrong to treat it as though it has only 1 lane is also wrong. Interestingly, treating it as through it has some intermediate number of lanes, such as  2 or 1.5 lanes, is also wrong. The right way to model short auxiliary lanes – a theory that I teach my students in my course in Advanced Traffic Control – is to model traffic discharge as having two components: one that is proportional to green time (that’s for the continuous lane(s)), and one that is constant for every time the green starts (for the auxililary lanes). The first component is standard; the second component can be modeled using negative lost time.

[Technical note for traffic modelers:  Ideally, you should model northbound as a single lane, and reduce lost time by 2 seconds for every vehicle that stacks up in the auxiliary lanes. So if normal lost time is 4 seconds and 9 vehicles stack up in the auxiliary lane, that means lost time should be set to 4 – (2 * 9) = -14 seconds. Unfortunately, it appears that Synchro won’t allow large negative lost times like that. So there’s no way around it – if you want an accurate result, you’ll have to do manual calculations.]

A simple signal timing solution: two short greens instead of one long one

Once traffic flow from short auxiliary lanes is understood, then one can see that the solution is simple:  instead of one long green period for northbound traffic at the City Square intersection, there should be two short green periods, separated by red periods between that allow queued traffic to stack up in the auxiliary lanes.

As I wrote earlier, during red periods, 9 vehicles, on average, stack in the auxiliary lanes. As a result, when northbound gets the green, there is strong flow for the 10 to 12 seconds while the auxiliary lanes discharge. After that, the flow of cars crossing the stop line diminishes to a trickle because from then on, the intersection is fed by only a single lane. In the first 12 seconds, on average, 15 cars are discharged (9 from the auxiliary lanes and 6 from the continuous lane); in the next 12 seconds, only about 6 cars are discharged. As a result, with the 42-second green offered today, total discharge northbound, including right turns during the Chelsea Street phase, is only 28 vehicles per cycle; with an average cycle length of 140 seconds, that’s a flow rate of 720 vehicles per hour.

If that long green northbound period is replaced with two short green periods, each lasting 16 seconds, the number of vehicles that can be discharged per cycle will rise to 35. At the same time, the cycle length will fall (there will obviously be less time for northbound, but then the shorter red that results for other traffic movements means they will need shorter green periods, too). Without a full set of traffic counts, it’s hard to determine exactly how much shorter the cycle will be, but from the traffic volumes I observed, I estimate that the cycle length will fall from 140 seconds to about 120 seconds. With more discharge per cycle and more cycles per hour, northbound capacity will rise to about 1,040 vehicles per hour, a 45% capacity increase over what’s offered now. That increase in capacity will immediately end all the congestion problems – no more gridlock at Keany Square, no more backups on the bridge either northbound or southbound.

Here’s the phase sequence the traffic signal cycle should follow, with phase lengths, at the City Square intersection.You can see that is has two northbound phases (phases 2 and 12) per cycle, as well as two southbound phases.

Ring diagram for proposed traffic control at City Square

My analysis assumes that there are working detectors in the southbound left turn lanes and in the westbound lanes (Chelsea Street), which will allow the controller to end their phases when no more traffic is detected. If those detectors aren’t there or aren’t working, they be installed / repaired immediately.

Are signal control changes needed at Keany Square?

At the Keany Square intersection, from what I’ve observed, the signal timing pattern that’s in place now is mostly fine, and would work well if there were no congestion spilling back from City Square. However, some improvements are still possible.

  • The lane assignments on Causeway Street should be changed so that left turns can be made from one lane only, since the bridge has only one receiving lane (not counting the very short auxiliary receiving lane which should be used only by buses coming from the N. Washington Street bus lane).
  • The slip lane for right turns from Commercial Street onto the bridge should be controlled by a Stop sign, not a traffic signal, so that pedestrians will have priority at all times crossing to and from the delta island. As it is now, pedestrians who comply with the signals have to wait a long time on that delta island, and many don’t comply.
  • Once congestion on the bridge is relieved, it will be possible to see whether some of the phases are unnecessarily long, and those phases should be shortened, reducing the cycle length and thus reducing delay to pedestrians and buses. The phase most likely due for shortening is the N. Washington Street phase.

Are the City and MassDOT working on this?

Yes. Hopefully they’ll fix the problem soon.