Building A Network Of Rapid Buses — Evaluating the Corridors

Originally posted on Streets.mn

As Metro Transit moves forward with its arterial bus rapid transit (aBRT) program, there’s an important opportunity to reassess the plans that will guide future investments. Those plans, which were developed almost a decade ago, are not comprehensive or inspiring. Some rethinking is needed.

Last year, I wrote about the limitations of those earlier plans. Yesterday, I illustrated the kind of scope necessary for a new aBRT study and plan that would examine nearly every arterial corridor within the transit-supportive core areas of the Twin Cities.

Having identified the universe of alternatives for aBRT expansion, the next step is to collect as much data as possible on these corridors and evaluate them against each other.

Collecting the Data

How to define a corridor? What types of analysis are important? Although we tend to talk about transit routes as operating in discrete corridors, the reality is far more messy, and it complicates any analysis in ways that are unavoidable. For instance, Broadway in North Minneapolis angles to the northwest at Girard Avenue, while Golden Valley Road continues on Broadway’s former course. Which route counts as the true Broadway corridor? Or, in South Minneapolis, bus service runs partially down Lyndale Avenue, then shifts over to Bryant Avenue. Should those streets be counted separately or together? There are not obvious answers!

It’s also important to note here that isolating individual corridors that make up pieces of a larger transit system is an approach that will generate data with significant limitations. In reality, no corridor is an island, and people are taking all kinds of trips into, out of, and along any given corridor. This research can be useful only to make broad judgments, not narrow ones.

To collect data on each corridor, I started with the 2018 Transit Stops Boardings and Alightings dataset from the Minnesota Geospatial Commons. I projected the data in GIS, and selected out all stops on individual corridors. In order to look at separate corridors, as demonstrated by the maps above, I did not select data from within either downtown Minneapolis or St. Paul. In North Minneapolis, corridors start at Plymouth Avenue at the earliest. In South Minneapolis, I used Franklin Avenue as a border. In West(ern) St. Paul, Kellogg Boulevard was the cut-off, and in Eastern St. Paul, it was Lafayette Road.

In cases where a corridor made a significant branch or jog, or crossed an important barrier, I took the issues on a case-by-case basis. In one case, I separated out the Lake Street, Marshall Avenue, and Selby Avenue corridors of the #21 bus. But I added Arlington Avenue into the dataset for Larpentur Avenue, and lumped Bryant in with Lyndale, because only one primary bus route serves that corridor and all buses take the same well-defined jog. In the case of Broadway, I analyzed the straight-line Broadway-Golden Valley corridor as one, and separated out the diagonal portion of the road.

After collecting all of the data along a single route. I analyzed I cleaned each spreadsheet by removing all stops for buses that did not travel along the corridor for a significant period. For some, that meant cutting the data down to just a single route, like the #19 on Penn Avenue N (now the C Line, but this data is from 2018). For others, that meant keeping data for both local and express routes, as was the case on Nicollet Avenue, which is served by both the #18 and the #554. For others, a number of routes traveled some portion of the route, best exemplified by Excelsior Boulevard, which carries the #12, #114, #612, and #664.

These route choices required a lot of choices for which there was no easy answer. West 7th carries a number of different routes for significant stretches, but only the #54 travels from end to end. Should a metric of ridership take into consideration the #63, #70 and #74 buses, which use West 7th or a directly adjacent road before turning down Grand, St. Clair and Randolph Avenues? I chose to use the latter two and not the former, in part because a subjective opinion of their length along the West 7th corridor. A different method of research might take a different approach. Back on Broadway-Golden Valley, both the #14 and the #30 serve significant portions of the route, but neither travels the entire way. I counted both, but the data for this corridor in particular is not trustworthy.

A map of areas of concentrated poverty in the Twin Cities metro. Serving riders from these areas could be an important goal of aBRT expansion.

A map of areas of concentrated poverty in the Twin Cities metro. Serving riders from these areas could be an important goal of aBRT expansion.

With the data organized and cleaned, I pull out several numbers. Most important was ridership. I use this term for the sake of simplicity, but because no corridor contains the entirety of a bus route, it’s not the most accurate term. This measure is the sum of all boardings at each stop on the corridor. I also summed boardings for stops that were in or along a designated Area of Concentrated Poverty. Transit is a lifeline for people with low income, and in plotting future investments, planners should not just take into account where it will be the most successful, but where it will do the most good for its individual users.

Ridership can only tell us so much, however. Some corridors have higher numbers of riders than others because they are long. As a way to account for this, I determined the number of stops on each corridor by counting the number of unique stop ID numbers in the corridor’s dataset. Some routes also have higher ridership because more bus service is run on them. As the frequency of transit service increases, the usefulness of the service increases exponentially, and ridership often follows. To account the number of buses, I took a compromise approach, and added the average number of buses for each route on the corridor. This sometimes could be tricky, as on some corridors, a given route does not run at the same frequency for the entire length. In the case of a route like the #18 bus on Nicollet Avenue, where the frequency of buses at Franklin Avenue is four an a half times greater than at the South Bloomington Transit Center, I conducted a separate analysis for the route in stages of frequency.

The final number I was interested in was ratio of peak trips to off-peak trips, for radial routes. This is important because one of the central benefits of aBRT conversion is an increase in off-peak service. Routes with high numbers of commuters are usually already well-served by an existing bus system that was, for better or worse, designed for them. But routes with a balance between peak and off-peak service might benefit more from aBRT service than others. For each route, I identified the peak direction, and divided the boardings and alightings in the off-peak direction by the boardings and alightings in the peak direction. For crosstown or circumferential routes, I divided the smaller number by the larger number. Only one route, the #61 on Arcade Street, produced an off-peak/peak ratio greater than one. That’s because the peak direction I identified was southbound to nearby downtown St. Paul. Remarkably, that direction has fewer riders than northbound to Larpentur where the bus then turns and heads towards the much further away downtown Minneapolis—illustrating the significant strength of the latter downtown in relation to the former.

Ranking The Corridors

With all of this data in hand, we can assess the strengths of each corridor relative to each other. I encourage anyone to download the data on their own and explore: I’ve pasted the summary page into a Google Sheet, for anyone who wants to sort or manipulate the information on their own.

aBRT_TotalRiders2.jpg

Total weekday ridership is the simplest metric, but the interpretation is tricky for plenty of reasons, including those outlined above, and others. For example, there is an intrinsic advantage to circumferential corridors and longer corridors, as these corridors are more likely to count the same rider on their trip to and from a destination. In contrast, shorter radial routes are more likely to see riders dropped off and boarding their return trip in downtown, which falls out of the study area.

With that caveat in mind, the numbers are still a strong showing for the kind of routes you might expect, and validate the region’s near-term transit priorities. No single corridor sees as many trips as Lake Street—an unsurprising finding, especially given the advantage described above—which is planned to carry the B Line in the coming years. Both legs of the future D Line, on Chicago and Emerson/Fremont also rank highly. Hennepin, the core corridor of the planned E Line, makes perhaps the most impressive showing, given that it is disadvantaged by being a short, radial route.

Only three corridors appear on this list that are not currently targeted for transit improvements. Franklin Avenue, a short, radial route carrying the #2 bus at the base of downtown Minneapolis, stands out. This route was recently targeted for stop consolidation, but ridership is so strong and there will soon be a natural western terminus at the future 21st Avenue LRT station. There’s a powerful case to be made for this route to be targeted for full aBRT conversion.

The other two corridors that place highly are on Nicollet and Central Avenues. The case for aBRT expansion on these corridors is strong, and they align such that a single combined route seems to make intuitive sense. Moreover, both were included on the 2012 aBRT study, and ranked well in the program’s 2018 update. But unlike any other corridors in the Twin Cities (even W. 7th), both Nicollet and Central are targeted for transit unimprovements. Both corridors are being held hostage by a zombie plan to build a useless streetcar that would serve fewer people than before and actively make the operations of the road worse. Until this relic of recession-era, know-nothing transit boosterism is permanently killed, aBRT upgrades to this important route are being held in check.

aBRT_TotalRiders_ACP2.jpg

In a metro area with indefensible equity gaps, there’s a need for government investments, whether in transit, parks, schools, or anything else, to prioritize the needs of areas and populations that have historically been ignored. By selecting riders who boarded from stations in Met Council-designated “Areas of Concentrated Poverty,” we can create a list of corridors where transit service works against the isolating effects of housing segregation. Again, the region’s near-term aBRT priorities perform well, with the just-opened C Line corridor on Penn Avenue N showing why it was highly prioritized, and the A Line corridor on Snelling Avenue also proving a useful equity tool.

Two other corridors stand out. The Rice Street corridor in West(ern) St. Paul, which was recently spared from service cuts, is a vital one for St. Paul’s north side. Transit in the Rice corridor has a lot of challenges, chief among them the murderous design of the street. But given the importance of that route to and from areas of poverty, there’s a moral imperative to work to improve transit in the corridor, instead of giving it up and making cuts.

The other new corridor in the top ten, along Lyndale N in North Minneapolis, runs entirely inside of an area of concentrated poverty. This corridor is just a quarter mile from the #5 route on Emerson/Fremont, which places even higher on the ranking, and is a reaffirmation of the importance of transit to lower income communities in this part of the city.

aBRT_Riders_Stop.jpg

The longer a corridor runs and the more territory it covers, the more people it likely serves. One way to control for this is to divide the total ridership by the number of bus stops in the dataset, to determine the number of riders per stop. This metric has several different interpretations based on the context of the route. But because the fewer stops a bus makes, the faster it goes, and because one of aBRT’s signature advantages over regular local bus operation is far shorter dwell time at stops, this ranking is useful. It shows us places where stops are already infrequent but ridership is high (Franklin), and places where there is simply a massive transit demand.

This list also doesn’t tell us much that is new. It repeats nine of the ten corridors from the total ridership list. The absurdly long Nicollet corridor drops off (although subsections of Nicollet, to 48th St, and to I-494, would both place on this list), and the absurdly short Riverside Corridor is added. Just about a mile in length from Cedar to Franklin, Riverside has eight unique stops serving Augsburg University and the University of Minnesota Medical Center, a huge concentration of ridership generators. With Franklin Avenue placing third, it’s clear why the #2 bus was targeted for stop consolidation.

The two standout performers on this metric are Lake and Hennepin. While their placement isn’t surprising, the magnitude of their lead is remarkable. Almost twice as many people are waiting for buses at stops along Lake Street as along Chicago or Emerson/Fremont. That Lake is so far ahead of even the ultra-busy Hennepin Avenue is remarkable. There should be bus lanes for the length of Lake Street.

aBRT_Riders_Bus.jpg

One of the main determinants of ridership is the frequency of service along a corridor. The more buses that run, the less a rider has to wait, and the less a rider has to wait, the more useful the service becomes. The transit consultant Jarrett Walker is famous for saying “frequency is freedom.” Given the existing frequency of buses on a corridor, is that corridor over or under performing? By dividing our total number of riders by the number of buses we estimate operate on the whole corridor, we can determine the number of riders per bus.

Ideally, this metric should show where buses are most often running full, and where more bus frequency is needed. We can sometimes make guesses about how much improvement would come from running an aBRT level of all-day ten minute frequency service on the route, but subtracting the existing frequency from the frequency achieved by the A Line (212 buses/weekday). But this also requires deeper analysis. Nicollet Avenue, as an example, averages out to (175.7 buses/weekday) along the entire corridor. This is really an undercount, because closer to downtown, the #18 bus actually has a far greater frequency (even greater than the A Line), with 268 buses/weekday arriving at Nicollet and Franklin. But south of 46th Street, frequency tails off dramatically. It’s useful to know this, and this is likely where significant aBRT ridership gains are possible, but it also muddles the data by dragging down Nicollet’s average number of buses.

The picture is far clearer on other routes such as Lowry Avenue, where the #32 bus runs at a consistent frequency (roughly 66 buses/weekday) along the entire corridor. Despite this low frequency, this corridor ranks twentieth in total riders. The numbers suggest that this corridor may not get the frequency of service that it merits.

This metric elevates several other corridors that have not previously gotten attention. The Lyndale S/Bryant corridor sees roughly half as many buses as the Nicollet corridor, especially off-peak, despite traveling through neighborhoods of similar character. With frequency like Nicollet, perhaps it would see similar ridership.

Two hybrid radial/circumferential routes north of the railroad in West(ern) St. Paul make strong showings. The Como corridor is unique, given its high student ridership, and sees packed buses throughout the day. The Larpentur-E Hennepin corridor has a similar profile to Lowry, with a nearly identical frequency and ridership. The high placement of both of those routes suggests that there is travel demand for crosstown routes along the northern end of the cities that is not being fully served.

aBRT_Riders_Stop_Bus.jpg

Combining the two metrics listed above results in a list that ranks corridors by the number of riders for every time a bus makes a stop along the route. The number of riders per stop, per bus is the closest approximation from this data to the gold standard transit statistic; passenger trips per revenue hour. Unsurprisingly, star performing corridors like Lake Street, Emerson-Fremont, and Chicago perform highly. The biggest surprises are the two corridors new to any list, which all boast surprisingly strong ridership, medium to shorter lengths, and a medium to low frequency of buses. 38th Street and the #23 bus place a stunning second, following only the crosstown corridor a mile to the north on Lake Street. Bloomington Avenue and the #14 bus place third. One early action task for Metro Transit could be to increase frequency and consolidate stops on the #23 and #14 buses and study the ridership response for these two surprising sources of concentrated ridership.

The Lyndale N corridor and #22 bus also makes a return in the top ten. This is a corridor with low-ish frequency service that sits just a quarter mile from the #5 and Emerson/Fremont corridor in North Minneapolis. When the D Line is completed on that corridor, the effects on the Lyndale N corridor should be closely studied. Why are riders on Lyndale N not using the higher frequency #5 on Emerson/Fremont, and will aBRT tempt them to switch? Or is there a need for Lyndale N to be considered for aBRT service as well?

More data is needed to answer these tough questions. But with this rudimentary analysis in hand, we can start to think about the big picture. In broad terms, what routes, and types of routes should be prioritized for future aBRT expansion? How would this network look as it unfolds? Tomorrow, I’ll lay out my crayon map of routes and my thinking behind the order of their roll out.