Ashland BRT II – Travel Times

Updated 12 Jan 2015 at 9:00 AM:

I’ve used the week of 23 November 2014 to calculate delays on the Ashland bus, Thanksgiving week.  I’ll update this post with data from the week of November 10th later this week.

Updated 12 Jan 2015 at 2:45 PM:

I’ve updated this post with transit tracker data from the week of November 10th.  While this week does include Veteran’s Day, weekday bus service was run for the full week and travel demand would have been near normal.  The inclusion of Sunday/Holiday service in the original post significantly increased the calculated waiting time for transit customers in the original post.

Original Post:

In this post, I’ll attempt to sum the time savings and cost from the proposed Ashland BRT project and the reallocation of a vehicular lane each way on Ashland to transit use. As much as possible, conservative assumptions are made such as to reduce the apparent benefits or increase the apparent cost of the BRT project.

Let’s start with the effects on transit users. As presented in Appendix D of the Ashland BRT Environmental Assessment (the EA), an average travel time savings of 7.8 minutes should be expected compared to existing travel times on the Ashland bus. In contrast, I calculate an average in-bus travel time savings of 4.7  5.0 minutes for existing trips on Ashland. The comes from the latter using an average trip distance of 2.24 miles vs. 2.5 miles for the former, and the latter using an average bus speed of 8.7 mi/hr instead of the 10.5  10.0 mi/hr I find using bus tracker data. The numbers used for the EA’s analysis are applicable to peak times on the Phase I segment while I’m using whole day averages over the entire route.

As one of Jarrett Walker’s key points, while travel time is an important metric for the quality of automotive travel, waiting time and vehicle frequency are more important for transit customers. Appendix D of the EA uses the average additional waiting time due to bunching as an indicator of unreliability. For the existing service, an average delay of 43 seconds was calculated, which would be reduced to 22 seconds for BRT. It’s not clear to me how these figures were calculated as the reference used to evaluate BRT reliability only contains methods to evaluate BRT travel time. Based on bus tracker data, the existing average delay due to bunching is greater, somewhere between 90 and 120 seconds. However, this metric of transit reliability tries to translate reliability into the language of average speed and thus gives us a very poor intuition of the customer experience.

Ashland bus customers are making time critical trips every day where they need to arrive on time to work, school, or appointments with near certainty. Due to the severity of bus bunching on Ashland, that means being prepared for more than a 40 30 minute wait during the afternoon peak. Reliability should be measured as the maximum wait time one should normally expect before a bus arrives. During the week of November 24th 10th, averaged over the hours of the day, a customer could be confident 19 times out of 20 that a bus would arrive in the following 27 19 minutes, with even longer waits during afternoons and late nights. Improved reliability with Ashland BRT would reduce the maximum wait to around 10 minutes except late at night when service frequency would be reduced. This 17 9 minute savings combines with the lesser 4.7 5.0 minute savings expected from improved bus speeds.

The speed and reliability improvements from implementing BRT are possible by reallocating road space from either travel lanes on Ashland. Appendix B of the EA investigated the increase in travel time due to the removal of a travel lane on Ashland and the resulting reduction in automotive capacity and diversion of cars onto other streets. Traffic diversion was calculated using the CMAP Travel Demand Model. This analysis of the traffic diversion was heavily criticized by an admittedly anti-BRT biased letter to the Sun-Times editorial board. Critically, the letter pointed out that the CMAP model is not designed or validated for the type of street-by-street traffic engineering problem the EA posed. There are several important inconsistencies between the model and reality, including factor of 2 differences in vehicular volumes on various streets. CMAP’s model treats all major intersections on arterial streets as signalized intersections with capacities based on the 1994 update to the Highway Capacity Manual. The reduced vehicular capacities of all-way stop intersections^[1]All-way stops have around 2/3rd of the total capacity of a signalized intersection., delays due to pedestrian conflicts, and delays from intersections not at major cross streets are not included in the model. For example, the model predicts a far higher use of Damen Ave. than is observed. The EA models a total of 264,626 vehicle miles traveled per weekday on Ashland whereas around 407,000 were measured during traffic counts in 2010. Existing traffic speeds on Ashland are also not accurately modeled. The model obtains a speed of 18.3 mi/hr when existing traffic speeds obtained from the Chicago Traffic Tracker are closer to 15.9 mi/hr, matching those found by Syncho analysis of existing peak hour vehicular flow conditions^[2]See page 22 of the EA’s presentation boards..

How much additional delay to vehicles on Ashland should be expected with BRT? I don’t trust the EA’s trip allocation, so additional real world data would be useful. Fortunately, earlier this year the overpass of Ashland Ave. over Pershing Rd. was being demolished. During parts of this demolition, Ashland was reduced to one lane each way at Ashland with rough pavement slowing vehicles further. To estimate how lane reduction at a single point would correspond to a lane reduction along the whole corridor, I’ll compare the delay at Pershing to the expected speed over a 2.5 mile segment of Ashland around Pershing. Two and a half miles is the average length along Ashland a vehicle travels when passing Pershing. Traffic Tracker data from May 2014 shows that vehicles were slowed to an average of 11 mi/hr in the half-mile before the intersection on a segment that would normally average 21 miles per hour. During demolition, vehicles on Ashland were delayed 70 additional seconds on average through the intersection. Over a 2.5 mile trip segment at 21 mi/hr lasting approximately 7 minutes, the construction resulted in an average speed reduction of 14%. Applying the same percentage delay to the whole Ashland corridor reduces vehicular speeds to an average of 13.7 mi/hr at peak times. This happens to be faster than the 12.8 mi/hr (a 19% speed reduction) the EA’s 35% diversion of vehicles would produce. Let’s conservatively use the EA’s number with the understanding it contains significant uncertainties.

The travel time of transit and automobiles are typically measure differently. As the biggest delay in a transit trip is time spent waiting for the next vehicle, total delay per trip is the most useful metric. For vehicles, delay varies more strongly with distance, so travel speed is often the best metric. The EA compared existing to future conditions using average travel speed. Here, I’ll use travel time per trip
From the AM peak hour intersection vehicular flow survey conducted for the EA, in Appendix B-2, entering and exiting vehicles from the Ashland corridor were counted. From this, the total trip segment distance per vehicle was calculated to be 1.49 miles. In total, around 275,000 vehicular trips are taken on Ashland each weekday and Ashland BRT would delay vehicular trips that remained on Ashland by around 82 seconds each.

As a proxy for the total vehicular delay caused by Ashland BRT, I’ll multiply the existing trips on Ashland by the additional delays experienced by a fewer number of future vehicles on Ashland. The time cost experienced by vehicles diverting off Ashland will be less than for those staying on Ashland. These vehicles will cause delays on other streets. However, thanks to the street grid, there are many streets on which these vehicles can divert and delays on other streets should be significantly less than the 82 seconds expected on Ashland^[3]If someone has an idea for a reasonable conservative assumption I can make here, please comment.. In total, 6200 vehicle-hours per day in delays, with a greater than 50% uncertainty, should be expected on Ashland. This compares with a certainty of 17679 existing transit riders experiencing 21.7 14 minutes of travel time savings each, or 6400 4100 person-hours per day in time savings.

If the Ashland BRT project were only about travel time savings, it would be a wash slightly less time would be saved by transit riders than would be sacrificed by vehicles. There are other immensely important economic and social benefit that make Ashland BRT important.

• The project creates a major strategic link in the rapidly growing transit network. This will greatly expand the travel choices and opportunities for all Chicagoans far into the future. While some road network capacity is sacrificed, the only losses to the structure of the road network are the left-turn bans.
• Tens of millions of dollars of vehicle operating costs will be saved for thousands of people along with several million saved by the CTA each year.
• New economic opportunities and higher building densities will be supported along the corridor. For example, a large fraction of the seas of parking lots around the United Center and the Illinois Medical District will be developable given improved transit access.
• Citizen’s health will be improved and both local air pollution and GHG emissions reduced.

These secondary benefits more than cover the $160 million cost of implementing BRT and the potential uncertainties in vehicular travel times after implementation.

My next Ashland BRT post will explore some implementation details of the project including left turn restrictions and local bus service.