# What Is the Capacity of I-94?

I recently wrote about the Capacity of the Green Line, demonstrating a huge amount of underutilization under both reasonable and unreasonable assumptions.

I thought I should do the same for I-94 in Minneapolis and St. Paul. Of course a freeway and an LRT are different things, so the kinds of assumptions and the available data differ.

• 1800 vehicles per hour per lane (today, 2 second headways) vs. (with automation, 1 second headways) 3600 vehicles per hour per lane.
• 4 (12-ft) lanes today vs. 8 (narrower, 6-ft) lanes with automation
• 4 passengers per vehicles max vs. 1.5 passengers per vehicle today.
• 4 lanes in each direction, 2 directions

We also need to figure out the average length of trip, which is a bit less obvious.

Current two-way Average Daily Traffic on I-94 on the peak section (near Riverside) is about 164,000: The number of entering trips (based on sum of Eastbound entering vehicles at on-ramps between and including Hennepin Avenue (Minneapolis) and Dale Street (St. Paul) is 161,000 [A distance of about 9 miles or 14 km]). The average number of Eastbound through trips in the Lowry Tunnel is about 87,000.

Entering flows on I-94 from Lowry Tunnel to Dale Street (and final flows in St. Paul)

Average length is vehicle miles traveled divided by number of trips. If we have an average flow of 80,000 vehicles in each direction for 9 miles, this is 720,000 vehicle miles traveled in each direction for this long section. The number of entering trips from Lowry Tunnel to Dale Street (eastbound) (inclusive) is about 240,000. This implies the average vehicle which uses I-94 between the cities uses the facility for 3 miles (5 km). Obviously this is an approximation, but it is probably not too far off. There are 16 entrances over 9 interchanges over this span. Some have parts a, b, and c, and in the sequence (Exit 231 to Exit 240, inclusive). Note Exit 232 (LaSalle?) is missing. So we will go with 9 exits, or 1 mile between exits, and an average trip of 3 exits). The table shows some capacities. This assumes 4 lanes throughout, which is also not strictly true.

 A C D Automated and Unconstrained Today’s Flow Constrained Hours 24 24 10 Vehicles per Hour per Lane 3600 833 750 Vehicles per Day per Lane 86400 20000 7500 Person Capacity per Vehicle 4 1.5 1.5 Numbered Exits 9 9 9 Lanes 8 4 4 Average Trip Length (in Exits) 1 3 3 Directions 2 2 2 DAILY PERSON CAPACITY 49,766,400 720,000 270,000

In a world of automated vehicles, if everyone made short (1 exit) trips, in fully loaded (4 persons) per car, fully utilized over 24 hours per day, I-94 could carry about 50 million people per day over this stretch.  In contrast, today it carries about 720,000 people. [If we evaluate the Green Line as 2 lanes and I-94 traffic per lane, I-94 produces more person trips per lane than the Green Line]. If we were to constrain it further, so it only operated 10 hours per day (recognizing people travel only during certain hours), it would carry fewer people than it does. This is just a thought experiment to get some magnitudes. But clearly we have a lot of potential capacity in the years ahead as automated vehicle technology becomes mainstream, if we manage our roadspace and our vehicles more carefully. This argues against capacity expansion.

We conclude that Car Culture remains dominant. The number of people using I-94 on a given weekday is about 20 times larger than the number of people using the Green Line.

A \$1 billion transit investment is rounding error for the change in traffic count on the parallel highway (comparing entering vehicles between Lowry and Dale for October 2014 (244,103) and October 2013 (244,712) – average weekday traffic ). Overall trends are mostly flat, with a slight uptick in vehicle miles traveled in 2014 nationally, but the core cities (Minneapolis, St. Paul) added population faster than they added jobs last year (I believe, I don’t think the data is solid on this), so average trip lengths should otherwise have dropped slightly in the cities, independent of national trends.

Also a caution, University Avenue may be different, but the data on changes in traffic counts before and after the Green Line opening is not publicly available as far as I know.

Transit investments like the Green Line LRT serve transit users, highway investments like I-94 serve highway users, they are completely different markets and, this data suggests that at this point in history,  barely substitutes. Congestion reduction should not be a selling point for transit investments, just as reducing crowding on trains or buses is not a valid selling point of highways.

Acknowledgment:

The data comes from the Minnesota Traffic Observatory and the Minnesota Department of Transportation from their DataExtract program. All analysis is the responsibility of the author.

## 11 thoughts on “What Is the Capacity of I-94?”

1. Monte Castleman

Although I like driving, I eagerly await driverless cars because of the reduction in congestion, and being able to leave for Chicago at night, take a nap, and wake up in the morning. An additional factor is that with cars being able to stay perfectly in their lane and not slow down next to other cars, lanes can be narrower. Maybe we’d have a single 12 foot lane for semis and “manual” cars, and driverless cars can use 8 foot lanes. That and there will be a more active interface between cars and traffic control. I’m not going to have much room to go into this on my series on traffic signal controllers, but right now the interface is really dumb. A camera or loop detector tells the stoplight if there is or is not a car there. A car goes or stops depending on the color of the light. So it’s not surprising that they cause a lot of congestion and are so irritating. Imagine if stoplights knew where every car in the city was, and could adjust themselves accordingly, or a stoplight could tell a car several miles away that there was congestion ahead and to reroute.

As purely a thought experiment there’s predictions on how much congestion would increase if transit were discontinued. In Kansas City a motorist would spend an additional hour in traffic per year, in New York it’s 49. (Minneapolis is 3) So maybe the justification of subsidizing transit with bridge tolls there isn’t entirely unfounded

I may or may not make it into an article, but I plotted freeway lane miles vs time spent in traffic. There’s a sharp drop off in congestion at the high end, suggesting theoretically we can more or less build our way out of congestion. And a rise in congestion at the low end, suggesting at some point a cities roads break down. But the middle is mostly flat when you smooth out the outliers, suggesting what we all know that if roads aren’t adequate to handle demand, they’ll fill to a certain level and then people will avoid them.

Minneapolis was actually a moderate outlier with a relatively little time spent in traffic per motorist relative to our freeway lane miles. (Washington, DC was a notable high outlier, probably due to under-built river bridges and a number of government workers working office hours, Phoenix was a low outlier, probably due to lots of alternatives to freeways and the lack of geographic obstacles.)

2. Mike Hicks

The high-end case is obviously impractical, but here are a few things to consider:

Even if the highway had 8 lanes each direction, only a few of those could be used if cars were only traveling one mile. Most of the existing entrance/exit ramps only have one lane, which strictly limits the inflow and outflow at each point along the highway. Rather than having 8 lanes per direction, you only have one.

Unlike train lines, highways can’t achieve their highest capacity by having everyone only go one exit/stop at a time. Given enough lanes, it would be possible to have dedicated entrance/exit pairs per lane, but that’s also pretty impractical.

A 1-second trailing distance would be extremely difficult to achieve for a couple reasons — first, I think automated cars are going to be less likely to run with short spacing than human-operated vehicles (maintaining enough distance to safely stop if the car ahead suddenly came to a halt). Second, cars need to change lanes, which requires extra spacing.

The 1-second spacing also assumes that cars can freely enter and exit continuously, but most of them will meet surface streets with 30 mph speed limits and stop lights. I suppose many ramps end up breaking out into more than one lane at the light, but it’s still likely that the maximum capacity per lane would be closer to 1200 rather than 3600 vehicles per hour.

Plugging in the lower vehicles per hour and number of lanes would bring this down to a person capacity of something like 2 million. As I said, this would really be higher if you assume the vehicles travel more than one exit, but it still probably ends up being on par with or below the 7 million figure that you had for the high-end estimate of the Green Line’s capacity.

1. David Levinson Post author

Yes, the local roads problem is a major issue, there is no way freeway capacities with automation can be easily dumped onto local roads, even if exit ramps were expanded somehow. This is true with freeways now in many places, as we see queues backing up from local streets onto freeways at multiple sites in the Twin Cities in rush hour already. It doomed proposals for Automated Highway Systems in the 1990s (since it was the highways that were specialized as well as vehicles), it may have less effect on automated vehicles, which at least eventually, will gain some capacity on local streets.

3. Alex Cecchini

Using this methodology but for non-automated vehicles, 24 hours per day, 8 total lanes, and current vehicle occupancy, I-94’s capacity is around 1.56 million per day. Or, 195k per lane per day. Compare that to the Green Line “Scenario C” (half capacity in the train cars, 24 hour operation at current frequencies) capacity of 255k per day, or 128k per lane per day.

Both of these represent a “reasonable” best possible case scenario with current technologies operating at all hours of the day. And, as you say, they aren’t really comparable in any sense, since the Green Line operates at grade dealing with stop lights and dumps people on foot onto sidewalks rather than in cars onto city streets. Also, discussing per lane ignores the amount of buffer space required for each mode (banks, water retention, entrance ramps, interchanges, etc) as well as noise/pollution impacts. But that’s not the point of this exercise.

Comparing modes, especially between cities, is difficult. But, a better comparison to I-94 would be grade separated heavy rail. The Red Line in Chicago’s North Side (Howard to Grand Stations, 19 total and 10 miles – similar to the Green Line here) carries ~125k passengers per day. Much of that is elevated, though it does go underground for 4 of the 19 stations listed. That makes for 63k per lane per day for (this example of) heavy rail vs 90k per lane per day on I-94 today.

1800 vehicles per hour per lane (vphpl) is a low-ball. You can get up to around 2000 vphpl on freeways in uncongested conditions. And with heavy ramp metering, MnDOT has documented cases up to 2400 vphpl during peak hours.

1. David Levinson Post author

Yes 1800 is low (as might be 3600), though there is a question of the stability of the higher throughputs before queues from a downstream bottleneck puts an end to the unreasonably large upstream flows. So imagine 2 lanes at 2400 vphpl converging to 1 lane at 2400. For a while those two lanes show a flow of 2400, but eventually they will have a flow of 1200, as the shockwave propagates backwards. The lane capacity may indeed be 2400 under idealized conditions, but they can only sustain that for a short time due to other factors.

5. Rosa

I really object to the idea that these are two completely different markets – a LOT of that traffic is local and people switch between car and transit when transit is convenient and when the highway is congested.

I’m right in that corridor, taking the length of trips you’re assuming. Almost all the times I’m choosing between driving (on 94, or on Franklin during 94’s rush hour) or taking transit the main difference is “will transit run at this hour” (especially the bus connections to the train) and “how terrible is traffic on 94 at this hour). The worst traffic times for 94 are the best for transit use.

6. Tara

“Congestion reduction should not be a selling point for transit investments”

I wonder if the possible congestion reduction benefits of transit are not subject to scale? Certainly a transit system has more of an impact than a transit line. And it takes a long time to build a system, line by line.

7. Monte Castleman

I’ve been reviewing some data on how various factors do/would affect congestion. I have noticed two patterns in how discontinuing the transit system would affect congestion. The most noticeable difference is 1) Cities that already have high congestion, and 2) Cities with well developed rail systems. In New York congestion would almost double, in Kansas City which has neither it would only increase 4%.

The outliers to the first are Houston, LA, and Atlanta where discontinuing transit would have very little effect, despite the cities being congested. My data source doesn’t have rail miles in order to analyze the second.