Longfellow Bike Count Update

I’ve been counting bikes on the Longfellow for … a while (although apparently not in 2015, slacker). In any case, with the layout of the bike lane changed appreciably over the past several months, I decided to count again. Here’s a quick breakdown of the Longfellow’s bicycle facilities in the past few years:

  • 2013: “Normal” pre-construction travel: bike lanes on both sides, two lanes of general traffic.
  • 2013–early 2015 construction: all traffic on the downstream side, one lane of traffic inbound, inbound bike lane, outbound contraflow lane with a buffer.
  • Early 2015–Late 2015: Inbound bike lane unchanged, but sections of outbound lane routed on to the sidewalk to accommodate work on the salt and pepper shakers.
  • Late 2015–Early 2016: Inbound bike lane eliminated for approximately 100m at the Cambridge end for Red Line shoo fly trackage; outbound lane eliminated entirely, cyclists asked to walk bikes across the bridge.
  • Early 2016–present: all cyclists on upstream sidewalk, pedestrians asked to use downstream sidewalk, outbound cyclists asked to loop under bridge to access Kendall. (The netting which broke free from the barriers in high winds has partially been removed, at least.) 
Average bicycle traffic on Broadway. The westbound
Longfellow lane has been impacted since Nov 9 2015.
Back in 2014, nearly 400 cyclists used the bridge during the peak inbound commuting hour. Since then, there have been significant disruptions to the bicycling facility, so some traffic may have chosen alternate routes. When the outbound bike lane was closed in November, there was a marked drop in westbound cyclists on Broadway; this persists this spring as many cyclists seem to be avoiding the suggested loop-the-loop under the bridge. Yes, there’s data. See if you can tell when cycling west on the bridge was made more difficult?
This spring, eastbound cycling traffic in Cambridge has reached new heights, surpassing even last September’s average (although this could be due to the number of weekdays and weekend days averaged). Westbound traffic has dropped, owing to the bridge construction. Has eastbound traffic?
Yes. Slightly. The count on May 18 tallied a peak of 358 cyclists between 8:02 and 9:02. This corresponds to 392 cyclists counted at the Eco-Totem on Broadway between 8:00 and 9:15, or 314 per hour. (In other words, there are a few more cyclists crossing the Longfellow than there are at the Eco-Totem; i.e. more join the flow from Main Street and elsewhere across the bridge than leave Broadway after the Eco-Totem, or miss the counter entirely.) This drop could be due to a variety of factors, from construction to noise in the data. Hard to know.
This count was different than others since to see both sides of the bridge required sitting in an office high above the bridge. This meant, however, that I was able to see whether cyclists were using the upstream sidewalk, the roadway (sans bike lane) or, in a few cases, the downstream sidewalk (intermixed with pedestrians and some very narrow passageways under the turret reconstruction). The answer? Most cyclists use the upstream sidewalk. For Boston-bound cyclists, 95% used the upstream facility. For those coming to Cambridge, only 88% used the facility, but the absolute numbers were much lower, so that meant that only about 10 riders per hour were using the downstream sidewalk. While I wasn’t counting pedestrians, it seemed that most were using the downstream sidewalk, although this was the morning commute, which is not prime sightseeing time. Many of the upstream users seemed to be joggers, so at least their pace was better matched.
Westbound commuter counts were about even with the last count in 2014, although bizarrely the 2014 count peaked in the 8:15 range while the current count was highest around 8:45 (this could be noise in the data). There would probably be more marked differences looking at evening data; the Cambridge data suggests that many outbound commuters are avoiding the Longfellow in its current configuration.
What does this all mean? It means that most cyclists will roll with the punches as infrastructure changes, although the Cambridge data suggest that if it is too hard to use, cyclists will find other routes. It will be interesting to see how the upcoming phases change cyclist behavior as facilities are twice again shifted around the bridge prior to the final configuration. Finally, the Cambridge data is a great supplement to these counts, as it can give us a good idea of whether we counted on a high-use day or low, and such automated counts are obviously much more data-rich than simple eyes on the street, although it will take some time to build a multi-year data set to look at definitive trends. For instance: I counted more bikes in 2014, but there is no similar Cambridge data to compare that count to since the counter was only installed in 2015. 
But next year’s count, well, that will have data. And the bridge might be shifted around. Again.

Longfellow Bike Count: Year 2

It’s kind of hard to believe that it’s been more than a year since my first Longfellow Bridge bike count, but it has. I’ve posted just a few times since then about the bridge, and seen the inbound lanes deconstructed, the towers come down, and, arch-by-arch, the bridge is now being rebuilt. I even went out in the middle of winter (and by out, I mean in to someone’s office with a view of the bridge) and counted about 90 bikes per hour: 30% of the previous summer’s crossings. (I think I tweeted this during the winter but didn’t write a whole post.)

So it was high time for a new count. I waited for a morning with good weather (and when I wouldn’t miss November Project) and set off for the bridge. After chatting with the DPW workers on my street about Hubway, I didn’t make it on to the bridge until 7:45, but that meant I was there in plenty of time to hit the peak morning bike rush hour, which (still) occurs from approximately 8:10 to 9:10 on the Longfellow. While at first the bike counts seemed flat or even down, once the rush got cranking, it became clear that there are more bicyclists this year than last.

Just to review, here are the bike counts for the peak hour from last year:

Wednesday, June 19, 2013: 267 bicyclists (8:12 – 9:12)
Tuesday, July 30, 2013: 308 bicyclists (8:08 – 9:08)
Tuesday, October 15, 2013: 298 bicyclists (8:11 – 9:11)

Here’s what I found this June (on the 24th, a Tuesday). The counts today peaked from 8:07 to 9:15 (that is to say, the 8:07 – 9:07 hour and the 8:15 – 9:15 hour saw the same counts). And the number of cyclists during those 60 minute blocks?

384

That’s one bike every nine seconds for an hour. Compared with the highest count last year in July, it’s an increase of 25%. Compared with the average of the three counts last year, it’s an increase of 32%. In a single year.

I can’t think of any single factor that would have increase bicycle usage by that much, other than more people riding bikes. So, contrary to any mitigating factors, I’m operating under the assumption that bicycling eastbound across the Longfellow is up by at least 25% this year. Between 8:30 and 9:00, there were 221 cyclists crossing the bridge and only 187 vehicles: 18% more bikes than cars. There were a few moments where the bridge looked downright Copenhagenish. With more bicyclists than vehicles crossing the bridge at peak times, perhaps it’s time to revisit the design and give bikes more than 20% of the road’s real estate.

The weather is cooler. The Longfellow is the same.

Twice this summer, we counted vehicles on the Longfellow. Between June and July, when the lanes of the bridge were shifted and constricted, bicycle traffic was level (well, actually, it rose slightly) while vehicular traffic decreased. I was otherwise occupied this September and didn’t get a chance to do a comparable bike count until last week, when I eked out an hour to sit on the bridge.

And the results are so mundane they aren’t even worthy of charts and graphics. Basically, the numbers were within a thin margin of error of those from July:

(All values for peak-hour of the count, note that the Longfellow runs east-west; Eastbound towards Boston, Westbound towards Cambridge)

Eastbound Bikes: 308 (July: 298)
Westbound Bikes: 63 (July: 68)
Eastbound Pedestrians: 65 (July: 83)
Westbound Pedestrians: 191 (July: 201)
Inbound Vehicles: 411 (July: 415)

So the bridge, even after two months of people getting used to the traffic patterns, has seen no major changes. Any drop in non-motorized use might be attributable to cooler weather (in the mid 50s rather than the upper 60s) or to random variance. And assuming a normal traffic day, there has been no significant increase in traffic since the bridge has opened.

It’s the last piece that I find most interesting. It really speaks to the concept of “induced demand.” With the wider Longfellow, we say 800 vehicles per hour traversing the bridge in June. Once the bridge was narrowed, that number fell to 400. There were weeks with dozens police directing traffic, but the number of cars very quickly hit a new equilibrium. People do not seem to need a major education campaign to figure out where to go. If the new roads are gridlocked, they’ll find alternate routes. The system has not ground to a halt (although inbound at the evening rush often backs up the length of the bridge). There are too many variables to find out if people have switched to other routes or modes or just not made the trip, but traffic in the morning across the Longfellow has not been the apocalypse.

The wrong way and wrong time to do a bike count

At some point, I might get off my high horse regarding bicycle counts on the Longfellow Bridge. But for now, I’m staying on. Why? Because I did my bike count on a Wednesday morning in June, to get a “typical” commute count (although, to be fair, with colleges out for summer it may have been a bit less traffic than some of the year). And I think I’m collecting better data than the established counts. Going through this process, I’ve found out more information about the National Bicycle and Pedestrian Documentation project, which is used as a basis for counting by local transportation agencies across the country, including the CTPS locally.


And I have not been particularly pleased with what I’ve found.


I think I originally buried the lede here, so I’ll add a bit here that I reiterate in the final paragraph: the bicycle and pedestrian count methodology treats bicyclists and pedestrians more as leisure users rather than treating them as using viable modes of transportation. We don’t take typical auto counts during vacation periods, and we shouldn’t do this for bicyclists, either. We should count bicyclists and pedestrians during normal use times, and design facilities based on this usage, especially in urban areas where bicycling and walking are important not only as transportation modes but as congestion mitigation resources

So what’s wrong with the national bike counting standards? First of all, the survey methods seem to come from about 15 years ago. From the directions:

Each counter should bring counting sheets, two writing utensils, a watch or cell phone for keeping time, and a clipboard or other device to write on. Counts are conducted in 15-minute intervals, and comprise the total volume of pedestrians and bicyclists traveling in both directions passing a given point. 

Here we are, in 2013, when everyone has an iPad and an iPhone and a laptop, and we’re taking counts with paper and pencil, and then admonishing people to make sure they don’t lose the paper because it is irreplaceable until it has been entered in to the database. Great. When I went to do my count, I didn’t even consider using paper and pencil, because it just didn’t make sense. I thought about writing a python script, but settled on Excel because it was easier to write and easier to test. But for a national program, how hard would it be to have an iPad app? Nothing fancy, just something where you tap a button every time you see a pedestrian or cyclist (it would be easy to have these buttons be customizable) and then directly upload the time-stamped data. 


Is there an up-front cost? Yes. Although my only overhead was the power to charge my computer’s battery and about 30 minutes on Google to figure out how to set up the Excel spreadsheet.


Second, why on earth would you survey typical use on July 4 weekend? I can not get over this. Their explanation is that it is selected because “The 4th of July period … will afford both a typical summer weekday and what is typically the busiest holiday period and activity period for recreational facilities and activities.” Three points. First, July 4 is not a typical holiday. You are likely to wind up counting a lot of people walking or bicycling to parades and fireworks. Are there a lot of parades on Columbus Day? So with one of the four annual counts, you’re collecting data which is unusable every other day of the year. Second, and excuse the italics, but in what universe is July 5, the Friday after July 4, is going to be a typical weekday? Half of my office is taking it off. Half of the country has vacation bounce messages on their emails this week. Third, while it is a busy time, so many people are traveling that it is not representative of many other weekends. I would contend that you’d get much more useful data by surveying a typical weekend in July rather than focusing on July 4. We’re spending all of this time and money to sample bicycle use, and we’re collecting just bad data.


Third, while uniformity is sometimes good, bicycle and pedestrian use is highly localized, so what works in one city may not work in another. In other words, there’s not a very good argument for standardizing bicycle count dates across the country other than to increase awareness. Awareness is a good thing. But it shouldn’t trump good data.


Harrumph.


You’d think that the NBPDP would at least be consistent in their data collection documentation, but this is not the case at all. This current survey is planned for a Thursday, Friday and Saturday. Their instructions document (available here) states that “Weekday counts should always be done Tuesday through Thursday, and never on a holiday.” Now, I understand that this is in relation to the national count in September, but, uh, guys? July 4-6 is a Holiday, Friday and Saturday. You’re 1-for-3. That might work in baseball, but it’s not a passing grade elsewhere.


Finally, in Boston this is an especially fraught issue. Why? Because on July 4, many of the city’s highest-use bicycle facilities are being closed. The Esplanade will be closed to bicyclists all day. The Longfellow and Harvard bridges are closing at 4 p.m. So any bicyclists after that point wishing to travel from Cambridge to Boston will have to either cycle across the Charles River Dam, or go up to the BU Bridge. Does it make any sense to take bicycle counts on this day? 


I’m not saying that my methodology was perfect, but it’s hard to comprehend how the current counts make much sense. If we go around collecting shoddy data and wind up undercounting bicyclists, we have only ourselves to blame when we get underdesigned facilities. It seems like high time to assess both the overall efficacy of the NBPDP, and especially whether Boston should participate in the program.


And a larger, and perhaps more important point: by prioritizing non-commuting periods, we are treating bicyclists as recreational users, not as a viable mode of transportation. Would we count cars during July 4 week? Of course not. AASHTO even says that such periods are “atypical” and would not have valid data. By counting traffic at these times, we are doing a severe disservice to bicycle advocacy.

Longfellow Bike Count

One of the issues I’ve touched on in the Longfellow Bridge series is the fact that there as no bicycle count done at peak use times—inbound in the morning. The state’s report from 2011 (PDF) counted bicyclists in the evening, and shows only about 100 cyclists crossing the bridge in two hours. Anecdotally, I know it’s way more than that. At peak times, when 10 to 20 cyclists jam up the bike lane at each light cycle, it means that 250 to 500 or more cyclists are crossing the bridge each hour. So these numbers, and bridge plans based on them, make me angry.

But instead of getting mad, I got even. I did my own guerrilla traffic count. On Wednesday morning, when it was about 60 degrees and sunny, I went out with a computer, six hours of battery life and an Excel spreadsheet and started entering data. For every vehicle or person—bike, train, pedestrian and car—I typed a key, created a timestamp, and got 2250 data points from 7:20 to 9:20 a.m. Why 7:20 to 9:20? Because I got there at 7:20 and wanted two hours of data. Pay me to do this and you’ll get less arbitrary times.

But I think it’s good data! First of all, the bikes. In two hours of counting, I counted 463 bicyclists crossing the Longfellow Bridge. That’s right, counting just inbound bicyclists, I saw more cyclists cross the Longfellow in one direction than any MassDOT survey saw in both directions. The peak single hour for cyclists was from 8:12 to 9:12, during which time 267 cyclists crossed the bridge—an average of one every 13.4 seconds. So, yes, cyclists have been undercounted in official counts.

Are we Market Street in San Francisco? Not yet. Of course, Market Street—also with transit in the center—is closed to cars. [Edit: Market Street is partially closed to private vehicles, with plans being discussed for further closures.] And at the bottom of a hill, it’s a catchment zone for pretty much everyone coming out of the heights. They measured 1000 bikes in an hour (with a digital sensor, wow!), but that was on Bike to Work Day; recent data show somewhat fewer cyclists (but still a lot).

I also counted vehicles (1555 over two hours; 700 to 800 per hour, including three State Troopers and one VW with ribbons attached that passed by twice), inbound pedestrians (about 100, evenly split between joggers and walkers, although there were more joggers early on; I guess people had to get home, shower and go to work) and even inbound Red Line trains (30, with an average headway of 4:10 and a standard deviation of 1:50). I didn’t count outbound pedestrians, or the exact number of people I saw stopping to take pictures (at least three on my little nook of the bridge).

But did you come here for boring paragraphs? No, you came for charts! Yay charts! (Also, yay blogging at 11:20 p.m. when I should be fast asleep. Click to enlarge.)

First, bikes. Cyclists crossing the bridge started out somewhat slow—the moving average for the first 20 minutes was only one or two per minute. But the number of cyclists peaks around 8:40—people going in to the city for a 9:00 start—before tapering off after 9. I’ve written before about seeing up to 18 people in line at Charles Circle which is backed up by these data; the highest single minute saw 11 cyclists, and there were four consecutive minutes during which 36 bikes passed. At nearly 6 bikes per minute for the highest half hour, it equates to nearly 360 bikes per hour—or 10 per 100 second light cycle. Too bad we’re all squeezed in to that one little lane. (Oh, here’s a proposal to fix that.) If anything, these numbers might be low—the roads were still damp from the overnight rain early this morning.

What was interesting is how few Hubway bikes made it across the bridge—only 25, or about 6% of the total. With stations in Cambridge and Somerville, it seems like there is a large untapped market for Hubway commuters to come across the bridge. There are certainly enough bikes in Kendall for a small army to take in to town.

Next, cars. The official Longfellow traffic counts show about 700 cars per hour, which is right about what my data show. What’s interesting is that there are two peaks. One is right around 8:00, and a second is between 8:30 and 9:00. I wonder if this would smooth out over time, or if there is a pronounced difference in the vehicular use of the bridge during these times. In any case, 700 vehicles per minute is not enough that it would fill two lanes even to the top of the bridge, so the second lane—at this time of day, anyway—is not necessary on the Cambridge side.

I also tracked foot traffic. I did my best to discern joggers and runners from commuters. Joggers started out strong early, but dwindled in number, while commuters—in ties and with backpacks and briefcases—came by about once every minute. I was only counting inbound pedestrians, and there were assuredly more going out to Kendall. Additionally, I was on the subpar, very-narrow sidewalked side; the downstream sidewalk is twice the width (although both will be widened as part of the bridge reconstruction).

And I counted when the trains came by. In the time that 2500 bikes, pedestrians and cars crossed the bridge, 30 trains also did. Of course, these were each carrying 500 to 1200 people, so they probably accounted for 25,000 people across the bridge, ten times what the rest of the bridge carried. Efficiency! The train times are interesting. The average headway is 4:10, with a standard deviation of 1:58. However, about half of the headways are clustered between 2 and 3 minutes. That’s good! The problem is that the rest are spread out, anywhere from 3:30 to 8:30! Sixteen trains came between 7:20 and 8:20, but only 14 during the busier 8:20 to 9:20 timeframe. I wonder if this is due to crowding, or just due to poor dispatch—or a combination of both.

In any case, what Red Line commuter hasn’t inexplicably sat on a stationary train climbing the Longfellow out of Kendall? Well, after watching for two hours, I have an answer for why this happens!  It is—uh—I was lying. I have no idea. But I say no fewer than a half dozen trains sit at that signal—and not only when there was a train just ahead—for a few seconds or even a couple infuriating minutes. I was glad I wasn’t aboard.

Finally, we can compare the number of bikes as a percentage of the number of cars crossing the bridge each minute. Overall, there were 30% as many bikes as cars. But seven minutes out of the two hours, there were more bikes across the bridge than vehicles. Cars—due to the traffic lights—tend to come in waves, so there’s more volatility. Although bikes travel in packs, too. Anyway, this doesn’t tell us much, it’s just a bunch of lines. But it’s fun.