Design Treatments for Increasing Traffic Capacity in Road Diet Projects

DESIGN TREATMENTS FOR INCREASING TRAFFIC CAPACITY ______IN ROAD DIET PROJECTS

A Thesis Presented

______Sepehr Shekari

The Department of ______Civil and Environmental Engineering

in partial fulfillment of the requirements for the degree of

Master of Science

in the field of

______Civil Engineering - Transportation

Northeastern University Boston, Massachusetts

______August 2019

ABSTRACT

Road diets are found to be powerful paradigm in adding more functionality to roads while also improving safety and efficiency. With road diets, of which the most common type is to reduce the number of car lanes from two lanes per direction to one, right-of-way can be allocated more efficiently among road users. The feasibility of a road diet depends on the ability to provide a high traffic capacity with a limited number of lanes. This thesis focuses on design features that can contribute significantly to a successful road diet by helping preserve traffic capacity using limited space. While some of the design features this thesis addresses are well-known, it also elaborates on some less-studied features as well.

The thesis begins with a road diet case study on Cummins Highway, an urban arterial in a residential part of Boston, reducing it from two lanes per direction with a raised median to a single lane per direction, retaining the raised median. Several of the design features covered in this thesis were motivated by needs and opportunities presented by this case.

One of the features addressed in this study is auxiliary through lanes, which are lanes added at signalized intersections to increase capacity. The design considerations for them are explored in the literature, as well as in a model that was developed to be used as a simple guideline. The importance of harmony between signal timing – particularly, the length of the green interval – and auxiliary lane length is highlighted.

Informal flares, making the road wide enough for one car to pass another who is waiting to turn, but not formally marking it as a turn lane or auxiliary lane, is an important but neglected design feature for road diets. They are space efficient, providing critical functionality for traffic flow while consuming less space than a formal auxiliary lane, leaving space for other valuable functions such as crossing islands. Analysis of field observations helps highlight their critical dimensions and factors.

Shadow lines in median breaks – that is, broken lines that continue the edges of a median through an intersection – are another understudied feature that can be valuable in road diets, in particular in combination with informal flares. When left-turning drivers wait in an intersection for a gap, these lines are hypothesized to draw them deeper into the median break by reassuring them that they are not occupying the opposite lane, making it easier for cars coming from behind to pass and reducing risk of rear-end collision. While no study shows how effective they might be, field observations show that the median break space is used very inefficiently without them.

Roundabouts and mini-roundabouts have been given close attention because of their ability to carry more traffic per lane than is usually possible with signalized intersections, making them interesting for road diets. Mini-roundabouts, which are smaller and have a traversable center island and splitter islands for large vehicles, can be ideal for constrained rights-of-way. Parameters affecting their capacity are reviewed, and differences between U.S. guidelines and Dutch design are discussed.

Finally, bus stop location and type are analyzed. Previous studies emphasize the value of in-line bus stops located downstream of signalized intersections. We develop and test (using simulation) a theory for how the distance of an in-line bus stop from the stop line and the length of the green interval length affect traffic blockage. i

TABLE OF CONTENTS 1. Introduction ...... 1 1.1. Overview ...... 1 1.2. Benefits of Road Diet ...... 1 1.3. Challenges of Road Diets ...... 2 1.4. Thesis Objective...... 2 1.5. Thesis Structure ...... 2 2. Case Study of Cummins Highway Road Diet ...... 4 2.1. Existing Situation: ...... 4 2.2. Study Objective:...... 5 2.3. Traffic Volume Data ...... 5 2.3.1. Intersection Turning Volumes ...... 6 2.3.2. Balancing Volumes Along the Corridor; Turning Volumes at Minor Intersections 9 2.4. Features of the Proposed Design ...... 10 2.5. Traffic Analysis Results ...... 15 2.6. Summary of Benefits ...... 16 2.7. Cost Considerations ...... 17 2.7.1. Less Expensive Option...... 17 3. Flaring at critical intersections ...... 19 3.1. Overview: ...... 19 3.2. Literature Review: ...... 19 3.3. Concept and Theory: ...... 23 3.4. Case Study of Cummins Hwy ...... 27 REFERENCES ...... 31 4. Informal Flares ...... 32 4.1. Overview ...... 32 4.2. Existing Examples ...... 33 4.3. Proposed Examples ...... 37 4.4. Observational Study on Informal Flares ...... 42 4.4.1. Introduction ...... 42 4.4.2. Study Objective ...... 42 4.4.3. Study Locations ...... 42 4.4.4. Observations ...... 43 4.4.5. Results and Conclusions ...... 44

4.4.6. Knowledge Gaps, Applications and Recommendations for Future Research ...... 46 REFERENCES ...... 47 5. Shadow Lines at Median Breaks ...... 48 5.1. Overview ...... 48 5.2. Existing Examples ...... 48 5.3. Application in Cummins Hwy Case Study: ...... 53 5.4. Use of Median Width and Informal Flare: ...... 54 5.5. Observational Study on Use of Median Space ...... 55 5.5.1. Introduction and Objective...... 55 5.5.2. Study Location and Method ...... 55 5.5.3. Results and Conclusions ...... 57 5.5.4. Knowledge Gaps, Applications, and Recommendations for Future Research ...... 58 REFERENCES ...... 59 6. Roundabouts and Mini-Roundabouts...... 60 6.1. Overview ...... 60 6.2. Literature Review ...... 63 6.3. Existing Examples ...... 65 6.4. Proposed Design for Cummins Hwy Road Diet ...... 68 6.4.1. Design Considerations ...... 68 6.4.2. Operational Performance ...... 70 6.4.3. Benefits ...... 70 REFERENCES ...... 72 7. Bus Stops in Bays vs. In-Line ...... 74 7.1. Overview ...... 74 7.2. Literature Review ...... 75 7.3. Study on Far-Side Stops and Capacity of Signalized Intersections ...... 78 7.3.1. Introduction and Objective...... 78 7.3.2. Theory ...... 78 7.3.3. Software Simulation ...... 83 7.3.4. Results and Conclusions ...... 84 7.3.5. Knowledge Gaps, Applications, and Recommendations for Future Research ...... 86 REFERENCES ...... 87 APPENDIX ...... 89

1. Introduction

1.1. Overview

Many cities around the world are laid out from a long time ago. At the time they were planned, the objective and approach toward transportation were different from today. While in the past, motor vehicles were the primary intended users, nowadays, sustainable mobility is more the center of attention. As a result, old design most likely fails to fulfill today’s needs. But at the same time, finding space to improve service for sustainable modes of transportation in a built-up urban environment is quite challenging. To realize this goal, we need to make the most efficient use of available space, by taking it from one use and allocating it to another. In other words, we need to sacrifice, but in this path, road diets help us sacrifice the least, yet gain the most. Road diet is a well-known paradigm in transportation planning which involves reallocating roadway space to accommodate more modes as well as improve safety. The most common type of road diet applies to roads with two travel lanes per direction, reducing them to one through lane per direction. Often a central turn lane is provided as well. The one or two lanes worth of space that are freed can be spent to improve the service for other users, namely pedestrians, bikes, public transit, and to improve the beauty and environmental services of the street by adding trees and other plantings.

1.2. Benefits of Road Diet

Depending on the objectives, a road diet provides benefits in one or more of the following ways:

- Add facilities for users other than motor vehicles so they can travel safely and efficiently. An important example is adding bike lanes. - Improve the existing facilities so that users can enjoy more safety and comfort, such as turning bike lanes into cycle tracks, or widening the sidewalks. - As a result of the road becoming one lane per direction, passing will not be possible and speed will be controlled. - The available space can be used to create crossing islands for pedestrians so they do not cross both directions in one attempt. Also, one-lane-per-direction configuration allows pedestrians to cross the street safely and more conveniently as they are no longer subject to the “multiple threat” hazard, a prevalent crash type.

- Part of the space can be given to trees and landscaping strips. - Travel lanes can be divided using a median, which increases drivers’ safety. - In a commercial area, good accessibility by public transportation and bike can lead to economic growth in that area.

1.3. Challenges of Road Diets

While road diets offer many benefits, they pose challenges as well. Forcing the traffic from two lanes per direction to one necessitates that enough capacity be provided at the potential bottlenecks for the level of service to not deteriorate. Therefore, a successful road diet demands careful studies, and suitable treatments must be applied to protect capacity.

1.4. Thesis Objective

Part of the focus of this thesis is a road diet case study on Cummins Highway in Boston, Massachusetts. This case involves a significant challenge in carrying its traffic with a single lane per direction, and involves several features that help improve the capacity of that 1+1 lane general configuration. This greater part of the thesis explores those features as well as related features that are used, or could be used, in other road diet projects. Those features are studied using literature review, theory, field observations, and experiments using simulation software.

1.5. Thesis Structure

Chapter 2 presents the case study of Cummins Highway. It explains the existing safety concerns and inefficiencies in operation and shows how beneficial it would be to have a road diet. It presents a proposed design, explaining how its design features will help improve safety and efficiency in that corridor. Chapter 3 discusses auxiliary through lanes as a common treatment to protect capacity at signalized intersections. It uses theory, and presents a simple guideline that highlights the importance of harmony between signal timing and auxiliary lane length. The application of that is demonstrated in the Cummins Highway study. Chapter 4 is on informal flares, a topic which is not much addressed in the literature. An informal flare is a widening of a road that allows through cars to pass through without being blocked by a queued turning car, but without marking a turn lane. Field observations are presented and the key factors in how informal flares operate are analyzed. Cummins Highway is used as a practical example of informal flares as part of a road diet design.

A simple road feature which few have implemented in the U.S. is addressed in Chapter 5. Shadow lines at median breaks is the name given to them in this thesis. They are hypothesized to improve safety and efficiency at unsignalized intersections, and can be valuable in road diets such as in the Cummins Highway case. A software experiment and field observations are brought in to add to the analysis. Chapter 6 discusses roundabouts and mini-roundabouts as safe, efficient alternatives for intersections in a road diet. Special attention is given to mini-roundabouts, which are smaller roundabouts in which the central island and splitter islands are mountable, and which, because of their size, may be useful in road diet projects where right of way is constrained. The most influential parameters in capacity of roundabouts are discussed and applied to Cummins Highway study. Finally, Chapter 7 addresses an issue that may arise when a road diet is applied to a major bus route. The location and type of a bus stop can affect traffic capacity significantly if it leads to buses blocking the only through lane. A theory is developed in this chapter, and is tested with simulation; it can be used as a simple guideline when considering in-line bus stops downstream of a signalized intersection.

2. Case Study of Cummins Highway Road Diet

2.1. Existing Situation:

Cummins Highway, between Mattapan Square and the cemeteries just west of Wood Ave/ Harvard St, is an auto-oriented, 4-lane divided road passing through the middle of Boston’s Mattapan neighborhood. In one way, Cummins Highway serves the neighborhood by providing access to bordering towns and neighborhoods. But in many other ways, this road does a lot of harm to the community through which it passes, including:

- Speeding: Its layout facilitates speeding, even racing, with a significant proportion of traffic going faster than 40 mph. During much of the day, traffic is light. Because there are two lanes in each direction, drivers can easily pass others and easily achieve dangerously high speeds. - Dangerous crossings: Pedestrian crossings are long and dangerous, though broken by a median. Because of the multi-lane layout, motorists rarely yield to pedestrians waiting at a crosswalk. And when they do, that poses a danger known as a double-threat – the car that stops for you in lane 1 may screen you from an approaching car in lane 2. Parents do not feel safe letting their children cross to get to elementary school or other destinations. - Missing crossings: At several bus stops, no crossings are provided at all, even though it is well known that at every bus stop, 50% of the passengers getting on or off have to cross the street. - No safe place to ride a bike: There are no bicycling facilities at all. Cyclists either use the sidewalk, which has poor sight lines and is intended for pedestrians, or ride in the road, sharing a lane with fast traffic. - A particularly large and dangerous intersection: At Greenfield Rd, the oblique turning angle makes it possible for cars to turn into a neighborhood street at high speed. The corresponding left turn, from Greenfield Rd onto Cummins Hwy, is dangerous because the angle makes it difficult to see and judge approaching traffic. The intersection is huge, open paved area adds to the confusion and makes it difficult for pedestrians to cross. - It is ugly and not green: Cummins Highway cuts a wide swath of ugly, barren pavement through the neighborhood. With no grass and almost no street trees, it is unattractive as a walking route. It could be transformed into more of a parkway, where people would enjoy walking and bicycling.

2.2. Study Objective:

This study tests whether it is feasible to convert Cummins Highway into a street that serves, instead of severs, its neighborhood. The objective is to find a design that solves the six problems just described without sacrificing the road’s valuable mobility function. In practical terms, the key to meeting these six objectives is a “road diet,” meaning reducing the road from 4 lanes to 2. That will limit speeding (you cannot speed when there is a car ahead of you that you cannot pass); it will make crossing easy and safe; and it frees space for bike lanes and for street trees. But will one lane per direction be enough to carry the traffic? This study has two major dimensions: traffic capacity and street layout. It examines the traffic volumes and traffic signal settings to see whether a two-lane layout can carry the traffic, and where extra travel lanes may be needed at bottlenecks. And it explores the road layout to see whether, within the existing right of way, all the desired features can fit: sidewalks, bike lanes, street trees, parking lanes, bulbouts to shorten crossing distances, a median to keep crossings simple and safe, and bus stops with safe crossings.

2.3. Traffic Volume Data

The project area currently has one unsignalized intersection for which traffic counts are available as well as 6 signalized intersections. From the west, they are:

Table 2.1 – Cummins Hwy available traffic volume data per intersection Has little or no Existing counts: New counts: Intersection cross traffic date & period date & period August 2012, 7 AM to 6 October 2018, 4 PM to Wood Ave/Harvard St PM 6 PM 610 Cummins Hwy x - - August 2018, 4 PM to 6 Itasca St/Ridlon Rd - PM July 2018, 4 PM to 6 Woodhaven St - PM Rockdale St x - - Fairway St June 2017, 7 AM to 6 PM - (unsignalized)

Blue Hill Ave - May 2014, 4 PM to 6 PM (Mattapan Square)

Existing counts for three intersections (two signalized, one unsignalized) with significant cross traffic were available from past studies, provided by the Boston Transportation Department from their files. Analysis of existing counts showed that the P.M. peak has the greatest hourly volumes, and so this feasibility study focuses on the P.M. peak. New counts, therefore, were done for the p.m. peak only, at the three intersections indicated above. Two of the signalized intersections have little or no cross traffic, and so there is no need for traffic counts there. At Mattapan Square, where Cummins Highway is not the major arterial, this project proposes no change to the signal timing. Therefore, the only volumes that matter for this project are the volumes entering and leaving Cummins Highway.

2.3.1. Intersection Turning Volumes

The traffic model uses a set of current and consistent traffic volumes matching Fall, 2018 conditions, calibrated to the new counts done in October, 2018 at Wood Ave/Harvard St. Volumes at Wood Ave/Harvard St were found to be 5% greater than those recorded in August, 2012; therefore, existing P.M. peak volumes at Fairway St and at Mattapan Square were multiplied by 1.05. The new counts done in July and August (at Itasca St/Ridlon Rd and at Woodhaven St) were also multiplied by 1.05, a factor that brought their link counts into balance with the October, 2018 volumes. To analyze and simulate the corridor as a whole, a uniform peak hour of 4:30 P.M. to 5:30 P.M. was applied; that is the peak hour of the critical intersection, Wood Ave/Harvard St. That is, volumes for all intersections were as counted between 4:30 and 5:30, with the inflation adjustments described earlier. The following figures show the traffic volumes for each of the signalized intersections in the stretch of study after the adjustments described earlier.

Figure 2.1 – Cummins Hwy at Mattapan Sq P.M. peak traffic counts (volumes entering/departing Cummins Hwy only)

Figure 2.2 – Cummins Hwy at Fairway St P.M. peak traffic counts

Figure 2.3 – Cummins Hwy at Woodhaven St P.M. peak traffic counts

Figure 2.4 – Cummins Hwy at Itasca St/Ridlon Rd P.M. peak traffic counts

Figure 2.5 – Cummins Hwy at Wood Ave/Harvard St P.M. peak traffic counts

2.3.2. Balancing Volumes Along the Corridor; Turning Volumes at Minor Intersections

From turning counts at each intersection, the entering and exiting volumes were calculated for segments between every two intersections for which counts are available. Comparing the entering and exiting volume in each segment, a net difference was calculated. The following table shows the summary of calculations.

Table 2.2 – Cummins Hwy available traffic volume data per intersection Segment Direction Entering Exiting Net Difference Between EB 967 902 -65 Wood Ave/ Harvard St and WB 615 590 -25 Itasca St/ Ridlon Rd Between Itasca St/ Ridlon Rd EB 946 916 -30 and Woodhaven WB 655 582 -73 Between Woodhaven St and EB 902 1003 101 Fairway St WB 516 507 -9 Between Fairway St and Blue EB 901 891 -10 Hill Ave (Mattapan Square) WB 311 447 136

Net differences can random factors (counting errors and random differences between days that were counted) and / or the systematic of traffic entering or leaving Cummins Highway from side streets. If random factors are ignored, negative values in “Net Difference” column indicate that

more vehicles turned off onto the side streets than turned on from side streets onto Cummins Highway, while the positive values suggest the opposite. Moderately small negative differences, as found for almost every segment, are reasonable for a road passing through a residential area that has more homes than jobs and is therefore expected to be a net sink in the p.m. peak. In the simulation, therefore, these negative differences are accounted for by assigning them to the side streets; as a result, in the simulation, entering and exiting volumes for every segment were consistent and matched the counts. The only positive difference is the westbound segment from Mattapan Square to Fairway; the net difference there is considerable and there is no side street that can explain it. In this case, additional volume was assigned to the SBR movement from Blue Hill Ave entering Cummins Hwy; that way the exiting volume at Fairway matched the counted link volume there.

2.4. Features of the Proposed Design

Existing Basic Layout: The road’s existing basic layout has, in each direction, two travel lanes and a parking lane. The two directions are separated by a narrow raised median. There is no bike lane. There are sidewalks on both sides, usually 8 to 10 ft wide, with almost no trees. Right of way is 80 ft wide in the western part the corridor, narrower as the road gets closer to Mattapan Square. Figure 2.6 and Figure 2.7 show the existing layout between Hallowell St and Rockingham Rd, where the right-of-way width is 77 ft .

Figure 2.6 – Existing layout with 77-ft right of way

8’ 7’ 10’ 10’ 5.5’ 10’ 10’ 7’ 9.5’ Figure 2.7 – Existing cross section with 77-ft ROW

Proposed Basic Layout: As proposed, the basic layout preserves the parking lanes and median, but has only one travel lane per direction, as shown in Figure 2.8 and Figure 2.9 for the same stretch of road whose existing layout is in Figure 2.6 and Figure 2.7. The sidewalk area on each side of the street is enlarged to about 16 ft and subdivided into three parts: a tree belt or planting strip, a one-way cycle track (bike path), and a walking zone. Where right-of-way is narrower or wider than 77 ft, the difference is put into making the planting strip narrower or wider. More detailed drawings are provided in the appendix.

Figure 2.8 – Proposed layout (77-ft right of way)

6’ 5’ 5’ 7.5’ 12’ 5.5’ 12’ 7.5’ 5’ 5’ 6.5’ Figure 2.9 – Proposed cross section (77-ft right of way)

Features at signalized intersections: Turn pockets, extra through lanes, and left turn restrictions. At four signalized intersections, to get the needed traffic capacity to make a road diet work, turn pockets and, in some cases, extra through lanes are added, as listed in Table 2.3. Where an extra through lane is needed, it is roughly two blocks long on either side of the intersection to give traffic space to queue and to merge back to a single lane. For the intersection at Wood Ave/Harvard St, we propose adding a left turn prohibition westbound from Cummins Hwy onto Wood Ave. There is little demand for this left turn (only 1 vehicle every 2 minutes in the peak hour), yet the blockage it causes creates a capacity bottleneck because the road does not have the width needed to make a left turn pocket. The few vehicles with that desire line can instead turn left onto Tampa St or Seminole St. Features at unsignalized intersections: Informal flares to prevent blocking through traffic. At unsignalized intersections, the proposed design provides enough clear roadway space so that left turning vehicles can wait without blocking traffic (Figure 2.10). This clear roadway space, called an informal flare, comes about by a combination of opening the median and prohibiting parking near the crosswalks. These informal flares are critical for enabling a single lane to carry the traffic that is now being carried by two lanes.

Figure 2.10 – Passenger vehicle paths (in red) going around cars stopped to make a left turn (informal flare function). Drawn using AutoTURN.

Mini-roundabout at Greenfield Rd/Weybosset St: At the large and dangerous five-leg intersection of Cummins Hwy @ Greenfield Rd/Weybosset St/Alabama St, a mini-roundabout is proposed. With an inscribed diameter of 90 ft and a circulating lane of 16 ft wide, it will allow cars to pass through without touching the median (Figure 2.11), while at the same time forcing enough deflection to ensure that cars pass through at a speed of about 15 mph. However, school buses and large trucks (e.g., 62-ft long tractor trailers known as WB-62) will have to slow down further and let their rear wheels track over the center island and the “armpits”. For this function, the curbs for center island and “armpits” will have a 3″ vertical face, which will deter cars from using them, but which slow-moving buses and trucks with large wheels can mount (Figure 2.12). That mountable feature makes it a “mini”-roundabout. Statistically, single lane, small-diameter roundabouts like this typically reduce injury crashes by 80%. For pedestrians, the proposed roundabout has much shorter crossings (13 ft across each direction of Cummins Hwy and 20-30 feet across the other streets, versus crossings of 70 – 80 ft today). For bike safety, there is a 10-ft offset between the circulating lane and the cycle track, improving visibility and giving both motorist and cyclist time to react in case of a conflict. A roundabout provides ideal speed control for a road like Cummins Hwy – it does not force vehicles to stop, but it forces all vehicles to slow down. It eliminates the existing high-speed right turn onto Greenfield Rd and the dangerous low-angle left turn from Greenfield Rd onto Cummins Hwy.

Figure 2.11 – Large school bus (S-Bus-40) vehicle paths for left turn (in red) and through movement (in green and blue) going through the proposed roundabout (drawn using AutoTURN)

Figure 2.12 – 62-ft long tractor trailer (WB-62) vehicle paths for left turn (in red) and through movement (in green and blue) going through the proposed roundabout (drawn using AutoTURN)

Two traffic signals are removed: Traffic signals are proposed for removal at 610 Cummins Hwy and at Rockdale St. The existing signals there were installed to facilitate pedestrians crossing a 4-lane road. With the new 2-lane design with crossing islands, crossings will be safe without a traffic signal, and so the signals can be removed. This change actually improves

pedestrian service, because with a simple crosswalk, pedestrians will be able to cross with almost no wait at all, instead of having to wait for a WALK signal. Relocated bus stops with safe crossings: Some bus stops have been relocated with a goal of ensuring safe crossings. For example, a traffic signal was recently added at the Itasca St/Ridlon Rd intersection, but there is no bus stop there now to take advantage of this safe crossing; our design places a bus stop there. In some cases, closely-spaced stops have been consolidated. The proposed stop locations have a stop spacing between 600 and 950 ft, which is the right balance between easy access and efficient service.

2.5. Traffic Analysis Results

To test whether the road diet solution can carry the traffic now carried by Cummins Highway, analysis was done using two software tools: Synchro, which does formula-based capacity analysis, and VISSIM, a finer-grained traffic simulation software. The Synchro analysis shows that the proposed layout offers both sufficient capacity and good level of service – that is, reasonably short delay - at each of the corridor’s signalized intersections. Cycle lengths are relatively short as well, reducing pedestrian delay. Results are in Table 2.3. Summary of Synchro results is provided in the appendix.

Table 2.3 – Summary of intersection analysis and lane configurations

[Explanation of the last column: v/c is volume/capacity ratio. A value less than 1 indicates sufficient capacity. "Max v/c" is the worst v/c at the intersection, that is, the v/c for that intersection's most congested movement.]

Level of service at the proposed roundabout was not measured with formulas; however, simulation results show a good level of service, with average delay around 10 seconds for both Cummins Highway and side street traffic.

2.6. Summary of Benefits

Speed Control: The proposed layout will help control vehicular speeds in four ways.

- By preventing passing. Drivers who want to speed will be constrained by the vehicles in front of them. - By making the roadway much narrower – only 1 lane per direction instead of two, and only 13 feet between the median and the parking lane. - The roundabout at Greenfield Rd will reduce speed to about 15 mph there, and will help prevent speeding in its vicinity. - The short signal cycles this layout makes possible (60 s at Woodhaven St and at Itasca St, 80 s at Wood Ave/Harvard St) help reduce speeding by reducing the amount of “wide open green time” at intersections (time when the signal is green and there is no longer a discharging queue).

Safer Pedestrian Crossings: The proposed road diet has a lot of safety benefits for crossing pedestrians:

- The proposed layout adds 11 new crosswalks. There will be safe crosswalks at every bus stop. - Crosswalk length is reduced substantially. For instance, at Savannah Ave/Rugby Rd, the full crossing will be shortened from 63 ft to 40 ft overall, and the lengths of the half-crossings (between median and curb) will be reduced from 30 ft to 17 ft. - The “double threat” disappears. In the existing 4-lane layout, a car stopping for a pedestrian in lane 1 can screen a pedestrian from an approaching car in lane 2, creating a high crash risk. That double threat disappears in the road diet layout. When a car stops for you, you’re 100% safe to cross to the median, and then when the next car stops, you’re 100% safe to complete your crossing. - Vehicle compliance at crosswalks (yielding to pedestrians waiting to cross), which is notoriously poor on multilane roads, will become much better when cars are in a single lane per direction, separated by a crossing island.

Tree belts: Today’s Cummins Highway is noticeably bare of vegetation. Our layout includes a 5-ft tree belt on each side of the street, providing shade to pedestrians and cyclists and improving the beauty of the neighborhood.

Protected bike lanes: The proposed design has a one-way cycle track (also called protected bike lane) on each side of the street. Cycle tracks will be at sidewalk level, separated from traffic by a curb, a parking lane, and tree-lined buffer. Because of this separation from traffic, bicycling will be safe and attractive for all, including children.

Some traffic signals become unnecessary: Thanks to safer and shorter crosswalks, the two existing traffic signals that are currently used only to stop the traffic for crossing pedestrians (at Rockdale St and at 610 Cummins, between Kennebec St and Hebron St) can be eliminated. Pedestrians there will have less delay than now, because they will have right of way at all times. The narrow roadway layout and low traffic speeds engendered by this design make it highly likely that motorists will comply and yield to pedestrians.

Overall vehicular travel time and delay: Average delay at signalized intersections will decrease because of the lower cycle length and because two signals are removed; however, traffic will go slower between signalized intersections. Overall, our simulation model predicts that average travel time between Mattapan Square (not counting signal delay there) and a point just west of Wood Ave (and thus including delay at the Wood Ave intersection) will be 3.5 minutes, corresponding to an average speed of 16 mph. This study did not model existing conditions using simulation, and so cannot directly calculate a change in overall vehicular travel time and delay, but the change is expected to be near zero. For comparison, average speed on urban streets, including intersection and traffic delays, is typically 9 to 13 mph.

2.7. Cost Considerations

The proposed layout leaves median intact, but shifts the curbs inward. That will necessitate new drains, new sidewalks, new pavement for protected bike lanes, and a new tree belt. We have not estimated the cost of these changes. On top of that, construction of a new roundabout will add about $250,000 to the cost.

2.7.1. Less Expensive Option

A temporary, less expensive option is also possible — leaving the curbs and sidewalk intact as well as the median. Most of the features of the full design can still be retained except the tree belt. Protected bike lanes will be at street level, separated from moving traffic by a row of parked cars and a hatched buffer zone with plastic posts (bollards). At unsignalized intersections, plastic bollards will be used to create pedestrian islands within the parking lane at the ends of each block, so that pedestrian crossings will be short.

Another low-cost option — more costly that what was just described but still low-cost as road reconstruction projects go – would be to install raised pedestrian islands within the parking lanes (Figure 2.13).

Figure 2.13 – Proposed intermediate cost option

3. Adding Auxiliary Through Lanes at Critical Intersections

3.1. Overview:

Flaring at signalized intersections is a well-known treatment for a successful road diet. On many streets for which road diet is feasible, traffic demand may be relatively high (e.g., 20,000 ADT, which roughly translates into 2,000 vehicles in the peak hour). A road diet forces the same amount of traffic to be carried by fewer lanes per direction. This said, depending on the demand, minor or major congestions may occur at bottlenecks, of which the most common are signalized intersections. Unlike signal-free stretches of the road, at signalized intersections the signal acts as a gate which is open only for a certain portion of time for each movement, which depends on the green time period for that particular movement in relation to the cycle length (the rest of which is red and the gate is closed for that movement). To compensate for this gate effect, roads can be flared at signalized intersections, meaning auxiliary lanes are added. Auxiliary lanes allow traffic to discharge in more lanes, thereby serving more vehicles in a given green interval. Also, since the addition of auxiliary lanes allows vehicles to queue in parallel lanes, the reach of the queue will be diminished. This is especially important in places where blocks are short and spillback could lead to gridlock. Given the significance of auxiliary lanes in traffic flow, several studies have looked at what factors can affect their effectiveness.

3.2. Literature Review:

In study of auxiliary through lanes (ATLs), the major focus point is lane utilization, meaning how likely or unlikely it is that drivers will use an ATL compared to continuous through lanes (CTLs). HCM 2010’s method includes a lane utilization adjustment factor in equations calculating saturation flow rate (1). Lane utilization factor is defined in HCM 2010 as:

푎푐푡푢푎푙 푑푒푚푎푛푑 𝑖푛 푚표푣푒푚푒푛푡 푔푟표푢푝 푓 = 퐿푈 푁 × 푑푒푚푎푛푑 𝑖푛 푙푎푛푒 푤𝑖푡ℎ ℎ𝑖푔ℎ푒푠푡 푢푡𝑖푙𝑖푧푎푡𝑖표푛

Where: fLU = lane utilization factor N = number of lanes in a movement group

HCM suggests default values in which utilization of multilane approaches is close to, but not exactly, equal to 1, the value that would be expected if lanes were utilized equally (Table 3.1). The

default values do not include cases with auxiliary through lanes, or cases in which drivers have a structural reason to choose a lane due to, for instance, a high-volume turn a short distance downstream. HCM also mentions that as the demand reaches capacity, lane utilization factor becomes nearly equal 1.0.

Table 3.1 HCM 2010 default values for lane utilization adjustment factor Traffic in Most Lane Group Number of Lanes in Lane Utilization Heavily Traveled Lane Movement Lane Group (ln) Adjustment Factor f (%) LU 1 100.0 1.000 Exclusive 2 52.5 0.952 Through 3 36.7 0.908

Exclusive Left 1 100.0 1.000 Turn 2 51.5 0.971

Exclusive Right 1 100.0 1.000 Turn 2 56.5 0.885

NCHRP Report 178 and NCHRP Report 707 present field measurements for lane utilization in 22 auxiliary lanes at 12 signalized intersections across the U.S. with single and double CTLs (Table 3.2) (2, 3). In the results, ATL utilization is defined as:

퐴푇퐿 푑푒푚푎푛푑 퐴푇퐿 푢푡𝑖푙𝑖푧푎푡𝑖표푛 = 퐴푇퐿 푑푒푚푎푛푑 + 퐶푇퐿 푑푒푚푎푛푑

The results show huge difference between observed results and HCM default values (Table 3.2). This difference forms the basis of several research works, as they argue that HCM overestimates utilization of auxiliary lanes, and therefore overestimates lane utilization factor for the movement group.

Table 3.2 – Summary of NCHRP 178 and NCHRP 707 presented field measurements for

lane utilization on exclusive through lanes compared to HCM default values

(veh/hr)

Through

Location

Approach

Number of CTLs Number

Upstream Length (ft) Length Upstream

HCM Default Values HCM

Downstream Length (ft) Length Downstream

Average Through Volume Volume Through Average

Average ATL Utilization % Utilization ATL Average

Equivalent HCM 2010 Lane Lane 2010 HCM Equivalent Utilization Adjustment Factor Adjustment Utilization EB NC 54 at Durham, NC 1 1650 450 23 323 0.649 0.952 Fayetteville EB Walker at Beaverton, OR 1 410 220 40 424 0.833 0.952 185th NB Garrett at Durham, NC 1 320 300 19 270 0.617 0.952 Old Chapel Hill SB Garrett at Durham, NC 1 330 380 23 257 0.649 0.952 Old Chapel Hill NB MD 2 at Annapolis, MD 2 800 300 19 2011 0.617 0.908 Arnold SB MD 2 at Annapolis, MD 2 1670 1060 20 1420 0.625 0.908 Arnold EB MD 214 at Bowie, MD 2 830 510 5 2114 0.526 0.908 Kettering SB IL 171 at Melrose Park, IL 2 450 360 26 793 0.676 0.908 Roosevelt SB US 1 at New Wake Forest, NC 2 470 1040 13 1559 0.575 0.908 Falls of Neuse

HCM does not take into account ATL upstream and downstream length in calculation of lane utilization. NCHRP Report 707 recommends that an auxiliary lane’s upstream length be enough to accommodate 95th percentile queue length, and ideally be longer than adjacent CTL maximum queue length, to ensure blockage does not occur. It finds that downstream length has nearly no impact on lane utilization.

However, Tarawneh (2000) studied the effects of auxiliary lanes’ downstream length on lane utilization (4). In general, the results show that as the length of the auxiliary lane increases, lane utilization by through movement significantly increases. Moreover, in this research, the influence of increasing right-turning volume is found to be negatively affecting lane utilization, while increase in stopped delay experienced by through movement improves it. All lane utilization values found in this research are between 0.73 and 0.82, which are considerably lower than HCM default values. Hurley (1997) researched the effects of several factors on auxiliary lane utilization (5). The results show that downstream auxiliary lane’s length, right turns off the road at the intersection, right turns onto the road within 400 ft upstream, right turns off the road within 500 ft downstream, and the existence of left turn lanes downstream of the intersection influence lane utilization. Also, choice auxiliary lane users (as opposed to captive users) are found to use the continuous through lane from 48 to 81 percent in various situations, which in most cases shows a huge difference with the 52.5 percent default HCM value. In another study, Lee et al. (2005) investigated the effect of ATL downstream length on its utilization in several North Carolina cities (6). Findings of their work show a direct correlation between lane utilization and downstream ATL length. Also, the length of downstream taper, volume demand, and percentage of heavy vehicles have been found to positively correlate with lane utilization. Ring and Sadek (2011) carried out a research in Buffalo, New York similar to the one in North Carolina, but found significantly different results for lane utilization (7). The observed lane utilization in North Carolina study underestimates lane utilization in Buffalo, which was attributed to different level of aggressiveness by the drivers of each region. Utilization on the downstream section of auxiliary lane was found to be a function of total through traffic volume, right turning volume, presence of mid-block two-way left turn lanes, and land use types (density of the developments and generated trips per mile). On the upstream section, cycle length and mid-block two-way left turn lanes as well as land use types played a role. Moreover, merging behavior downstream was found to be correlated with speed, total through volume and location of the first lane drop sign. In a different approach, Bugg et al. (2013) (8) presented driver lane choice behavioral models at signalized intersections, showing lane utilization as a function of arrival phase (during green or red period) and queue length in CTL and ATL. In arrival during red, they found lane utilization can be modeled based on CTL queue length, or the difference of queue length between CTL and ATL (14-17 percent error). Arriving during green, their model also introduces lane

utilization as a function of each lane’s queue length as well as remaining green time period (18-26 percent error).

While these studies provide valuable insight into what factors are at work in ATL utilization, one might need simpler guidelines for ATLs where road diets make an auxiliary lane necessary for capacity. Also, while several researchers have investigated the effect of ATL downstream length on lane utilization, none has examined the relationship of the downstream length to the green time, which can be expected to be important since longer green times mean more vehicles queuing at the bottleneck where the ATL ends.

3.3. Concept and Theory:

When considering the addition of an auxiliary lane at an intersection, one criteria has to be certainly met: the number of lanes in all directions must provide basic capacity for the intersection. One handy yet quick method to approximate the required number of lanes Critical Sums Method (1982). This analysis results in rough estimate intersection performance (LOS) only by having movement volumes, based on the assumption that capacity of each approach is proportional to its number of lanes. For a regular four-legged intersection with exclusive left-turn lanes (Figure 3.1a), to find whether the intersection provides enough capacity for a given set of volumes, the first step would 푣 be to find the sum of volumes per lane for all the movements that cannot be run in parallel, i.e., 푁 are mutually incompatible. As shown in dual ring diagram (Figure 3.1b), there will be a pair of two such incompatible movements (a critical pair) in the E-W direction, and another critical pair in the N-S direction. For example, if the greatest sum of volumes per lane on the E-W street involve EBL and WBT, and on the N-S street involve NBL and SBT, the critical sum will be:

푣 푣 푣 푣 퐶푟𝑖푡𝑖푐푎푙 푆푢푚 = 퐸퐵퐿 + 푊퐵푇 + 푁퐵퐿 + 푆퐵푇 푁퐸퐵퐿 푁푊퐵푇 푁푁퐵퐿 푁푆퐵푇 where vm is volume for movement m (veh/h) and Nm is number of lanes for movement m.

Based on the number of phases, which in the discussed case is eight, the Critical Sums methods indicates LOS boundaries (Table 3.3). In case the LOS without auxiliary lanes is found to be unacceptably low, auxiliary lanes could be considered, with greatest priority to adding them to 푣 critical movements with highest volumes per lane . With each added lane, the calculation must be 푁 repeated to ensure the added lane increases intersection capacity to reach the desired level,

considering that adding an auxiliary lane to a certain approach (e.g., WBT) might change the critical pair in that direction.

v SBT vSBL

NSBT NSBL

S Street S

- N

vWBT NWBT vWBL NWBL vEBL NEBL v EBT NEBT E-W Street

v NBL vNBT N NBL NNBT

(a)

(b)

Figure 3.1 – Regular four-legged intersection a) layout with exclusive left-turn lanes and shared right-through lanes b) dual ring diagram

Table 3.3 – Critical Sums LOS boundaries for controls with four phases or more Maximum Critical Sum LOS for 4+ phases (veh/hr) A 825 B 965 C 1,100 D 1,225 E 1,375 F >1,375

Apart from basic capacity, other reasons can drive designers to use auxiliary lanes. Because auxiliary lanes allow for vehicles to discharge on multiple lanes, the same volume can be discharged during relatively shorter green period, resulting in a shorter cycle length. By having a shorter cycle, not only motor vehicles, but pedestrians, bikes and transit will have shorter delay as well.

Additionally, queues are shorter where there are short cycles, and the possibility of a queue becoming excessively long to spill back and create congestion in upstream intersections will be reduced. With short cycles, platoons of smaller size will be released per cycle and pulses in traffic will be less severe, which can have positive effects downstream, such as making it easier for crossing pedestrians and side-street traffic to find gaps. An upper limit on the capacity achieved by adding an auxiliary lane of a certain length on approaching and departing legs can be estimated using deterministic queuing, assuming that motorists use all lanes equally except as limited by the (limited) length of an auxiliary lane. We consider a case with a single continuous through lane and a single auxiliary lane. We assume that vehicles discharge on both lanes (continuous and auxiliary) at the same saturation flow rate s. Because the goal is to measure the impact on capacity, we assume an infinite queue on the approach. On the receiving side, before the merge point back to a single lane (Figure 3.2 – section A), absent blockage from a queue, flow will be 2s, that is, twice the rate for a single lane, vehicles after the merging point (Figure 3.2 – section B) traverse at saturation flow rate s. This causes a bottleneck, making traffic back up. The reach of the forming queue will be the same regardless of whether vehicles in the ATL filter into CTL every other vehicle, or wait for all of the traffic on CTL to discharge before they can move. While the former is closer to practice, the latter is easier for calculation. At capacity, the queue forming behind the merge point will grow at saturation flow rate (s) as long as the signal is green upstream. To prevent the reach of the queue from spilling back into the intersection, the auxiliary lane must be at least long enough to hold one green period worth of vehicles per cycle. Therefore, to prevent spillback, the minimum length of the auxiliary lane on the departure leg is proportional to the effective green period:

퐿퐷 > (푔 × 푠) × 푆퐷 ≅ 푛 × 푆퐷 Where:

LD = length of auxiliary lane on departure leg (ft) g = effective green time (s) s = saturation flow rate (veh/s)

SD = effective length (spacing) of queued vehicles on the departure leg (ft) n = number of vehicles discharged per lane during one green period

Likewise, the approaching auxiliary lane has to be long enough to hold one green period worth of vehicles per cycle in order not to starve the green period. So, on the approaching auxiliary lane, the minimum required length to prevent starvation is:

퐿퐴 > (푔 × 푠) × 푆퐴 ≅ 푛 × 푆퐴 where:

LA = length of auxiliary lane on approach leg (ft)

SA = effective length (spacing) of queued vehicles on the approach leg (ft)

A B

LA LD

Figure 3.2 – Layout of intersection with auxiliary through lane in eastbound direction

Local field measurements could show whether drivers tend to space loosely or tightly before the merging point in comparison with when stopped at the signal. Obviously enough, if the spacing is found to be equal, the minimum auxiliary lane length to avoid capacity restriction will be equal on both legs. In keeping with the assumptions made, an auxiliary lane meeting these minimum lengths will function as a second continuous through lane, with the same capacity as the through lane. If the auxiliary lane is shorter than the minimum values indicated, there will be some capacity loss, even with the assumption of full utilization of the auxiliary lane. An auxiliary lane may be shorter than ideal due to right of way limits, to avoid displacing parking spaces, or other reasons. In such a case, shortening the cycle length, and shortening the green interval correspondingly, can be a solution that increases capacity. With shorter green periods, platoons of smaller size are released per cycle, which require less space to queue behind the merging point. However, where shortening the cycle length is not practical – say, because of minimum green periods for pedestrians and vehicular movements – and the auxiliary lane length is shorter than

minimum required, to avoid capacity loss, it is of interest to find the capacity loss that the constrained auxiliary lane causes. If the number of vehicles per cycle per lane that can go through an intersection for a given green period (g) and a cycle length (C) is n, but the number of vehicles that an auxiliary lane can accommodate is only m (m≤n): 푡ℎ푟표푢푔ℎ푝푢푡 푙표푠푠 푝푒푟 푐푦푐푙푒 = 푛 − 푚

푐푎푝푎푐𝑖푡푦 푙표푠푠 = 푡ℎ푟표푢푔ℎ푝푢푡 푙표푠푠 푝푒푟 푐푦푐푙푒 × 푛푢푚푏푒푟 표푓 푐푦푐푙푒푠 푝푒푟 ℎ표푢푟 3,600 = (푛 − 푚) × 퐶 n = number of vehicles discharged during one green period m = number of vehicles accommodated in the auxiliary lane C = cycle length (s)

The above capacity loss is based on ideal conditions, but in practice, drivers might prefer CTL over ATL under certain conditions. This leads to underutilization of the ATL which has been studied as reported earlier (4, 5, 6, 7, 8). Low utilization of an auxiliary lane can be incorporated into this analysis framework. However, the results from studies of North Carolina (6) and Buffalo (7) show that the degree of underutilization can differ substantially from one place to another, which suggests the importance of relying primarily on local studies.

3.4. Case Study of Cummins Hwy

The stretch of Cummins Hwy under study has four signalized intersections, including one on either end and two within the stretch. The study and design included geometric changes and signal timing plan improvements on three. The fourth, at Mattapan Sq on the eastern end of the stretch, remained unchanged because Cummins Highway is a minor leg there, and changes would have been disruptive to the other streets. Of the three other intersections, the layout is as follow:

Cummins Hwy @ Woodhaven St: ATL only in WB direction, starting 350 ft upstream and ending 280 ft downstream of the intersection (Figure 3.3). At this intersection, providing enough capacity was not a challenge, but rather the ATL is put in place to help keep the cycle short and provide a better level of service for all movements including pedestrian crossings. During A.M. peak hour, in which the peak direction is WB (the direction with ATL), analysis with Synchro software shows a cycle length of 50 seconds is sufficient with average intersection delay of 15.6 seconds and maximum v/c ratio of 0.62 (belonging to WB through). The average queue length is

71 ft, and in only 5% of the cycles will the queue exceed 112 ft. However, eliminating the ATL makes the cycle length 60 seconds with intersection delay of 20.8 seconds and maximum v/c ratio of 0.85. Average queue length extends to 183 ft with 5% of cycles having queue of more than 357 ft, which is the distance to the first unsignalized intersection upstream. This was the case which, due to enough capacity for competing movements, all the added 10 seconds to the cycle length was added to the green. However, this becomes challenging when competing movements also demand more green per cycle. Still, in this non-severe case, the change in upstream queue storage space needed and chance of overflow is significant when ATL is removed.

N 280 ft downstream auxiliary lane

350 ft upstream auxiliary lane

Figure 3.3 – Cummins Hwy @ Woodhaven St layout with single ATL

Cummins Hwy @ Itasca St/Ridlon Rd: No ATL is provided in either direction. Signal timing plan and level of service is close to what would be at Woodhaven St intersection without flaring. Although adding an auxiliary lane on WB direction at this intersection would improve the performance, because it still provides reasonable level of service without flaring, engineering judgement has ruled against adding one at this place. Factors such as the desire to put a bus stops on far side of the intersection, avoiding left turning vehicles to traverse two lanes of traffic while turning onto the gas station, and preserving parking were key to this decision (Figure 3.4).

N gas station

bus stops

left turns onto gas station

Figure 3.4 - Cummins Hwy @ Itasca St/Ridlon Rd layout with no ATL

Cummins Hwy @ Wood Ave/Harvard St: ATL in both EB and WB direction (Figure 3.5). In EB direction, the auxiliary lane starts 410 ft upstream and continues to 450 ft downstream. In WB direction, the auxiliary lane extends from 340 ft upstream to 400 ft downstream. At this intersection, providing enough capacity yet keeping cycle fairly short was a challenge, which was overcome by adding fairly long auxiliary lanes. To show the significance of these lanes at this intersection, a quick experiment was done to see how eliminating the ATL in WB direction could change intersection performance. During A.M. peak hour in which busy direction is WB, cycle length with both ATLs in place is 60 seconds with average intersection delay of 21.5 seconds and maximum v/c ratio of 0.79. Removing ATL in WB direction, the new signal timing design that responds to the demand will have a cycle length of 150 seconds, delay of 50.4 seconds and maximum v/c ratio of 0.96. An attempt to remove both auxiliary lanes in WB and EB direction led to oversaturating the intersection.

400 ft downstream in WB direction

410 ft upstream in EB direction

340 ft upstream in WB direction

450 ft downstream in EB direction

Figure 3.5 - Cummins Hwy @ Wood Ave/Harvard St layout with double ATL

Case Study Discussion: As seen in the results of Cummins Hwy case study, auxiliary lanes are vital at intersections where providing enough capacity is the main a challenge. At Wood Ave/Harvard St intersection, applying signal timing improvements by itself would not guarantee satisfactory performance of the intersection. Flaring in both directions was essential for it to respond to traffic demand. Moreover, the experiment at Woodhaven St intersection demonstrates even with relatively short cycles, removing a ATL could extend the average and 95th percentile reach of the queue up to 2.5-3 times longer. This highlights the key role of auxiliary lanes in places where blocks are too short that the queue might spillback frequently into the upstream intersection and cause gridlock. Where neither of the discussed issues is serious, designers may or may not want to consider ATLs by taking into account various factors such as safety, performance, and priority for allocating space to other users.

REFERENCES

1- Highway Capacity Manual. Transportation Research Board, Washington D.C., 2010. 2- NCHRP Report 178: Assessment of Auxiliary Through Lanes at Signalized Intersections. Transportation Research Board, Washington D.C., 2011. https://doi.org/10.17226/22830 3- NCHRP Report 707: Guidelines on the Use of Auxiliary Through Lanes at Signalized Intersections. Transportation Research Board, Washington D.C., 2011. https://doi.org/10.17226/14617 4- Tarawneh, M. S. Utilization of Auxiliary Through Lanes at Intersections of Four-Lane, Two-Way Roadways. Transportation Research Record, Vol. 1737, 2000, p.p. 26-33. https://doi.org/10.3141/1737-04 5- Hurley, J. W. Utilization of Auxiliary Through Lanes at Signalized Intersections with Downstream Lane Reductions.Transportation Research Record, Vol. 1572, 1997, p.p. 167-173. https://doi.org/10.3141/1572-20 6- Lee, J. J., N. M. Rouphail, and J. E. Hummer. Models for Lane Utilization Prediction for Lane Drop Intersections. Transportation Research Record: Journal of the Transportation Research Board, No. 1912, 2005, pp. 47–56. https://doi.org/10.1177/0361198105191200106. 7- Ring, J. B., and A. W. Sadek. Predicting Lane Utilization and Merge Behavior at Signalized Intersections with Auxiliary Lanes in Buffalo, New York. Journal of Transportation Engineering, Vol. 138, No. 9, 2012. DOI: 10.1061/(ASCE)TE.1943- 5436.0000426 8- Bugg, Z., N. M. Rouphail, and B. Schoeder. Lane Choice Model for Signalized Intersections with an Auxiliary Through Lane. Journal of Transportation Engineering, Vol. 139, No. 4, 2013. DOI: 10.1061/(ASCE)TE.1943-5436.0000513

4. Informal Flares

4.1. Overview At intersections where right-of-way is limited, allocating distinct space to accommodate all users can pose a challenge. Sometimes, the constraint on the right-of-way could be to the extent that providing left-turn lanes (or pockets) for motor traffic deteriorates service for other users, mainly for bikes by forcing them to share a through lane with cars, which compromises their safety and comfort. At the same time, failing to provide a place for left-turning vehicles to wait where they won’t block a through lane can seriously deteriorate motor vehicle capacity and service for through traffic. For road diets in which a street is reduced to a single through lane per direction, it can be critical to protect the capacity of that single lane, ensuring that queued left turning cars do not block the through lane. What is discussed here is an option other than left turn lanes that provides the functionality of a left turn pocket: making the road wide enough at the critical point that a through car can pass by another car waiting to turn left, as illustrated in Figure 4.1.

Figure 4.1 – Informal flare function scenario

In Figure 4.1, the black car is waiting to find a gap in opposing through traffic, but the road is wide enough so that the EB thru red car coming from behind can pass easily. Widening a road at an intersection is called a flare; however, as seen in the figure, no added lane is marked, yet it functions very much like a flare with an auxiliary turn lane. This is a common situation, with through cars using parking lanes, shoulders, and bike lanes to get around a car waiting to turn left; however, to our knowledge, it does not have a name. In the traffic engineering profession, intersection capacity analysis will indicate complete blockage when a left turning vehicle is queued, and so traffic engineers will often model a situation like this as actually having an auxiliary turn

lane. Given the absence of a well-accepted and common word for this road feature, “informal flare” seems to describe it well. On multilane roads, even if a queued left-turning car blocks a lane, traffic can still get by in other lane(s). Therefore, this chapter focuses on two-lane roads (one lane per direction) where through traffic in both directions is considerable, and at least one direction has frequent left turners. In such a situation, blockage due to queued left turners can have a large, negative capacity impact, especially at signalized intersections but at unsignalized intersections as well. Obviously enough, if the road is wide enough, cars will readily pass a queued left-turner, and if it is too narrow, nobody will be able to pass. Important research questions are, how much space is needed to have the flare function? Is it all-or-nothing, or are there intermediate widths with an intermediate impact on capacity? Are there other issues involved in designing a functional informal flare?

4.2.Existing Examples

Informal flares can be observed relieving congestion at both signalized and unsignalized intersections. There are different ways to find the space to create them, including using a parking lane, bike lane (or shoulder), or median break. Following are a few of many examples in Greater Boston area:

Heath St @ Schiller St: At this signalized intersection, there is a significant demand for WB left movement, but there is no left turn lane. At the same time, WB through and EB through demand is also considerable. Without informal flare, when the signal turns green for Heath Street, WB cars waiting to turn left will block the WB lane for much of the green interval, thereby causing congestion. However, the WB lane at the intersection is effectively widened by prohibiting parking in the spot in the parking lane closest to the stop line. That way, if a left turner waits close to double yellow center line (within one ft), other cars can pass by using the parking lane (Figure 4.2). At the point of constraint, the available road width (center line to curb) is 17 ft. Field observations made at this intersection showed that the majority of left-turning drivers advance half-way into the intersection (like the black car in the figure), meaning the parking lane effectively ends almost two car-lengths upstream of the pinch point. That way, even when two left- turning cars are queued, through cars can still get by on the right.

Figure 4.2 – Informal flare at Heath St @ Schiller St

Faneuil St @ Brooks St (Figure 4.3): At this unsignalized intersection, heavy demand for left turns from Faneuil St EB onto Brooks St creates a need for the flare function. If the EB parking lane is empty, an informal flare is present with available width at the pinch point (curb to center line) of 18 ft. There is a bus stop within the intersection (see the figure), at which parking is prohibited, but there is still one car parking spot between the crosswalk and bus stop. While that spot is free for many hours of the day, when a car is parked there, the informal flare is completely blocked and delay for EB through movement spikes.

parking prohibited area (bus stop)

Figure 4.3 – Informal flare at Faneuil St @ Brooks St

A St @ West Broadway: This is a signalized intersection at which NB left demand creates the need for informal flare for NB through, which is met using the bike lane (Figure 4.4). The combined width of travel lane and bike lane is 18 ft, which provides enough space for informal flare function, just as discussed in the example of Faneuil St @ Brooks St.

bike lane

Figure 4.4 – Informal flare at A St @ W Broadway

Tremont St @ St Alphonsus St: SB through movement at this signalized intersection uses bike lane in order to make an informal flare maneuver (Figure 4.5). However, unlike the previous example, the combined width of travel lane and bike lane is 15 ft, which is still very narrow for a pair of cars. With parking in the parking lane being prohibited for almost one car-length upstream of the intersection, the combination of the through lane, bike lane, and parking lane is more than enough for a left-turning car to wait and through cars to pass by on the right.

bike lane

Figure 4.5 – Informal flare at Tremont St @ St Alphonsus St

Columbus Ave @ Rutland Sq: At this unsignalized intersection, an informal flare is made possible partly by using the bike lane and a break in the median. Columbus Ave has a 6-ft median, and queued left turners from Columbus Ave WB pull partway into the break in the median. That shift leaves somewhat of a gap for those who want to go through (Figure 4.6); that gap is further widened because cars can use the bike lane (when not occupied by a bike). In this figure, the black car is queued, waiting to turn left. If there were no median and opposing directions were only separated by a double yellow line, the position of black car would be the dotted orange line, which would make informal flare function nearly impossible. The degree to which median width and deflection angle could help the functionality of informal flare requires some considerations which will be addressed in Chapter 5 – Shadow Lines at Median Breaks.

bike lane

median

Figure 4.6 – Informal flare at Columbus Ave @ Rutland Sq

4.3. Proposed Examples

When roads are redesigned in an effort to make crossings shorter and find space for bikes, informal flares could play a key role in that they are more space efficient than formal flares (auxiliary lanes). The following proposed examples highlight their contribution to providing better service for all road users:

Longwood Ave @ Avenue Louis Pasteur: In a study done on the feasibility of adding bike facilities along Longwood Ave, informal flares were found to be a good solution at multiple intersections (1). Analysis of one intersection, the unsignalized T-intersection of Longwood Ave @ Avenue Louis Pasteur, illustrates the superiority of the informal flare design. Curb-to-curb width is 34 ft, of which 21 ft is currently allocated to EB, divided into a left turn pocket and a through lane, with 13 ft allocated to WB in a single lane (Figure 4.7a). While there are bike lanes on both legs of Longwood Ave approaching the intersection, the bike lanes have been dropped to create space for the auxiliary left turn lane, forcing bikes to share the road with cars in both directions. The proposed design, however, maintains continuous bike lanes in both directions and relies on an informal flare for WB traffic, recognizing that through cars can use the bike lane when needed to get past a waiting left-turner. The design allocates 19 ft to WB (6 ft of bike lane and 13 ft of travel lane) and 15 ft to EB (a 4-ft bike lane and 11 ft of travel lane) (Figure 4.7b).

At first, condoning cars encroaching into bike lanes might be viewed as a safety concern. However, breaking it into different possible scenarios clarifies why the proposed design is superior: Scenario 1 – A left turning car and a through car, no bike: In this scenario, no bike is present. So, there is no safety concern. Empty bike lane is used by the through car without creating a conflict with bikes. Scenario 2 – A through car and a bike, no left turner: In the proposed design, the bike has their own space and the through car stays in travel lane parallel to the bike. However, in the existing design, the only through lane is shared between the bike and the through car. This is the case in which extra conflict happens in the existing design compared to what proposed. Scenario 3 – A left turning car, a through car, and a bike even with or just ahead of the through car: In this case, both existing and proposed design are similar in that the space is shared between the bike and through car. The difference is that in the existing scenario, both compete on “equal terms” for the space close to the curb, while in the proposed design, the lane striping clearly gives priority to the bike, allowing the car to enter it only after slowing (possibly stopping) and yielding to the bike. Scenario 4 – An eastbound car and bike arrive at the same time. (This scenario is independent of the first three scenarios, which deal with the westbound side of the road.) With the existing design, either the car or bike must yield to the other; right-of-way is ambiguous. In summary, Scenario 1 functions equally in both cases, the existing design creates an unnecessary conflict in both Scenarios 2 and 4, and in Scenario 3, expected behavior is the same with both designs (the car should yield to the bike), but the existing design leaves right-of-way ambiguous while the proposed design reinforces the expected behavior by clearly showing right- of-way.

11’ 13’ 10’ 11’ 10’ (a)

11’ 4’ 11’ 13’ 6’ 10’ (b) Figure 4.7 – Longwood Ave @ Avenue Louis Pasteur a) Existing plan and cross section b) proposed cross section

Road Diet Study of Cummins Hwy: In our proposed design for Cummins Hwy, there are many unsignalized intersections where, instead of providing a left turn lane, the median is brought as close as possible to the intersection to serve as a crossing island. However, flare function was considered necessary because with a road diet, long backups could occur if a left-turning car blocks the only through lane in a direction. Therefore, the design was done in a way that the gate formed by the median, queued left-turning car, parking lane, and corner bulbout is wide enough to allow for easy and flowing maneuver of through cars. The maneuver of cars was drawn using AutoTURN software (Figure 4.8).

Figure 4.8 – Informal flare on proposed design on Cummins Hwy @ Hallowell St/Harmon St

As part of the road diet proposal, bulbouts are typically added to every corner. However, as illustrated in Figure 4.8, bulbouts that would interrupt flare function are not added and, where needed, parking is prohibited in one or two spots near the intersection. Figure 4.8 is a case where the cross streets are offset to the left of one another (left dog-leg), so that on the right side of a queued left turner is the right curb corner, creating a pinch point. Careful curb design was needed to ensure a road width that would allow through traffic to pass by a waiting left-turner. This kind of left dog-leg layout is the most constraining in terms of informal flare function. Where cross streets are not offset, such as at the intersection of Cummins Hwy @ Savannah Ave/Rugby Rd (Figure 4.9), to the right of a queued left-turner there is no curb, but rather the pavement of the cross street. The pinch points are then between the corners of the left-turning car and the nearest intersection corner. Overall, the right-hand curb becomes less constraining. As seen

in this figure, among four possible bulbouts, two are preserved (as opposed to only one at the Hallowell St/Harmon St intersection).

Figure 4.9 – Informal flare on proposed design on Cummins Hwy @ Savannah Ave/Rugby Rd (Clarification: The figure is drawn to show informal flare in both directions, which typically do not happen at the same time. So, the left-turners should not be perceived as interlocking.)

In the Cummins Hwy case, the need for flares means that bulbouts cannot be added to every corner. Still, careful design shows that many bulbouts can be added, which provide shorter and safer crossing experience for pedestrians. Of the 20 proposed crosswalks at unsignalized intersections, 11 have bulbouts on both sides, 6 have bulbouts on one side, and 3 have no bulbouts. Also, of 17 unsignalized intersections, 13 have at least one crosswalk at the intersection, and the other 3 have access to a crosswalk within 60 ft, and 1 within 200 ft. Informal flares also allow parking lanes to continue closer to the intersection, preserving more parking. If auxiliary lanes were used instead of informal flares, parking prohibition area would be much longer and no bulbouts would be provided. With wider asphalt available, speeding could also be encouraged, which would bring its own hazards.

4.4. Observational Study on Informal Flares 4.4.1. Introduction

As shown previously, numerous examples of this feature already exist and are used every day. Given the explained dynamics, the potential of this feature in congestion relief is understandable, but unless designed properly, the gains might be insignificant. Despite this potential, to the best of our knowledge, nearly no research has been done to measure the effectiveness and functionality of informal flares. One key element is width, which must be enough to provide a reasonable level of comfort for drivers to make the maneuver. Manuals and guidebooks lack in providing a standard for that. For this reason, an observational study was carried out on the effectiveness of informal flares at two locations with different widths where the informal flare scenario occurred enough times to allow for a reliable conclusion.

4.4.2. Study Objective

The study was carried out to measure the effectiveness of informal flare function on two typical urban streets with different widths by measuring the tendency of drivers to pass when a left- turning car was waiting. Available width was assumed to be an important parameter which could impact the tendency of the drivers to make the maneuver of interest. As this functionality heavily relies on driving behavior, factors other than available width may also affect the flare function, such as lateral positioning of left turning cars, psychology of drivers to perceive this maneuver as aggressive or not, and driver skill to fit in tight spaces. While providing a framework to analyze the influence of each factor requires further elaborate research, in this study, the overall influence of such variations is observed on the degree of effectiveness of informal flares.

4.4.3. Study Locations

Attempts were made at several locations to gather data, but in only two did the informal flare scenario happen frequently enough to yield a reasonable sample size. The following factors must be present at the same time in an ideal location:

- The volume of left turning cars must be considerable. - The volume of through movement in the opposite direction must be high enough to make left turners stand and wait for a while before they can complete their turn. - The volume of through movement in the same direction as left turners must high enough that while left turners are waiting, a few of them arrive to decide whether or not to go around the standing car. - Traffic in the direction of left turners must be in one lane.

- The road width must enough to physically accommodate two passenger vehicles abreast. With two fairly wide vehicles (6.5 ft each) and a shy distance of 1 ft between the two and 1 ft to the right curb, the minimum distance from curb to center line must be 15 ft. This width must be available from at least two car-lengths upstream of the studied section to ensure no frequent blockage occurs.

Taking into account the above, two locations for study are:

1- Heath St @ Schiller St, a signalized intersection, where interactions between WB left and WB through were studied in a lane of 17 ft wide (Figure 4.2) 2- Faneuil St @ Brooks St, an unsignalized intersection, where interactions between EB left and EB through were studied in a lane of 18 ft wide (Figure 4.3)

4.4.4. Observations

At each of the two selected locations, 100 observations were recorded. Each observation means one through vehicle arrives behind a stopped left-turning car, and has the opportunity to either go around or wait behind them. No heavy vehicles were included in these observations and the focus was only on passenger vehicles, including sport utility vehicles (SUVs). A noticeable behavioral difference was observed between those cars who approached at speed in a non- congested situation and those who found themselves stuck behind a left turning car. To see the difference in behavior, for each observation it was recorded whether they were first in a queue, in a queue but not first, or arrived during non-congested situation (Figure 4.10). The definition of each situation is as follow:

- First in queue: These cars were in the same platoon as the left turning car, and were following it closely at a crawling speed (less than 5 mph). Because left turners did not always use their directional signals in time, the first car behind them often found themselves stuck, and had little distance left to maneuver around the stopped left-turner without steering hard. - In queue but not first: The second or later car in a platoon behind a left-turner. When the first car in queue managed to get through, later cars had more distance between them and the left- turner and could therefore go around the left turning car with more comfort. - Free flow: A car that arrived in non-congested situation at speed. In this case, the driver saw the blocking car well in advance and could choose a trajectory that did not even require them to slow down. Because these cars had an easy-to-do maneuver that could be done without slowing down, they showed highest tendency to go around the stopped left turner without hesitation.

first in queue, must steer hard to pass

free flow, knows in queue but not well in advance to first, has more room adjust trajectory to maneuver

Figure 4.10 – Arrival types with informal flare

The number of arrivals belonging to each of the three categories is shown in Table 4.1.

Table 4.1 – Summary of observations Lane Width Location Control Type Arrival type Vehicles (ft)

First in queue 34 Heath St @ Signalized 17 In queue but not first 16 Schiller St Free flow 50

Total 100

First in queue 23 Faneuil St @ Unsignalized 18 In queue but not first 8 Brooks St Free flow 69

Total 100

4.4.5. Results and Conclusions

The observations show how likely it is that through drivers passed a stopped left-turner on the two intersection approaches studied (Table 4.2).

Table 4.2 – Results of passing tendency Vehicles Vehicles Passing Location Situation Arrived Passed Tendency (%)

First in queue 34 24 71 Heath St @ In queue but not first 16 15 94 Schiller St Free flow 50 47 94

Overall 86

First in queue 23 16 70 Faneuil St @ In queue but not first 8 5 63 Brooks St Free flow 69 66 96

Overall 87

Passing tendency was observed to be highest among drivers who arrived when traffic was moving freely. At the signalized intersection of Heath St @ Schiller St, drivers arriving at free flow where those who came during the unsaturated part of the green. While no in-depth study was done to find the reason for high passing tendency, field observations convey that having the chance to steer smoothly past the left turning car is likely to be the primary reason. Moreover, drivers approaching at speed seemed a lot more reluctant to slow down and wait, whereas many of those who were in queue behind left turners seemed to be more willing to stop and wait. The reason for this kind of hesitation might be attributed to the driver’s trade-off between time and comfort. It appeared that those who were in queue showed more willingness to keep waiting for the left turner to complete their turn so they can continue in a straight path without steering. This hesitation also introduces delay to the through movement which was not measured in this study. The passing tendency is substantially different between free flow and the other two arrival types. This is another advantage of keeping traffic flowing by eliminating signals. Signals create pulses in traffic and cause traffic to move in platoons. As the observations show, vehicles in the same platoon following a left turning vehicle may frequently get stuck behind it and show less tendency to use informal flares. Where eliminating signals is not a viable option, having short cycles helps release platoons of smaller size.

In terms of width, the results for each situation as well as the overall tendency suggest that there is little to no difference in tendency between a road with 17 ft and 18 ft available width. The only anomaly is seen in the arrival type of “in queue but not first”, which, considering the small sample size, is not significant. While approaches with 20 ft of available width were not studied, it is obvious that they could be readily subdivided into two lanes (assuming a minimum lane width of 10 ft), in which case the flare function would be near perfect. Understanding the effect of available width is important. Where full flare function is needed (e.g., at signalized intersections with a left-turn phase), an auxiliary turn lane is clearly valuable. But at unsignalized intersections, and at signalized intersections without turn phases, it may be sufficient to have a strong but limited flare function with less total available space. That can allow designers to allocate extra space to bike lanes and sidewalks instead of wasting it on widening the road for some added functionality that is not needed. In road diets where right-of-way is limited, every single foot of space becomes so valuable that might determine the feasibility or infeasibility of a design, e.g., simple bike lane vs. protected cycle track.

4.4.6. Knowledge Gaps, Applications and Recommendations for Future Research

- Passing tendency was observed to be related to delay incurred by the driver, but was not accounted for. Some cases were observed in which the delay became so long that the driver behind decided to go past despite having to make an inconvenient maneuver, whereas the same driver would probably not make the maneuver if the delay was shorter. The former case was counted as a “vehicle passed”, while in the latter was counted as not passed. - While at unsignalized intersections, occasional delays might be trivial, at signalized intersections delays are translated as capacity loss due to starvation. To measure the real effect of informal flares at signalized intersections, measuring passing tendency is not sufficient. Delays should be measured and capacity loss calculated. - There were only two study locations, which did not encompass a wide range of widths. Complementary studies on places with different widths may help in finding an optimal width. The results could also be used in developing guidelines for informal flares. - Driving behavior and passing laws differ from one place to another. Informal flare function depends on a certain level of aggressiveness among drivers and law restrictions. Therefore, results measured in Boston may not apply to locations with different passing laws or driving aggressiveness. In places where drivers are well-known for patience and courtesy, and in places where passing on the right is prohibited, informal flares may prove ineffective.

REFERENCES

1- Furth, P. G., Z. Sha, and S. Rauwolf. Bike Lanes and Informal Flares: Making Longwood Avenue Safe and Functional for All Users, 2017.

5. Shadow Lines at Median Breaks

5.1. Overview

At intersections where there is a break in the median or there are pedestrian crossing islands on both sides, broken lines that indicate the left-side edge of the road can be extended into the intersection as shown in Figure 5.1a. We name them “shadow lines.” They can help drivers waiting to turn left to see how deep they can advance into the median break while still being protected in the shadow of islands (Figure 5.1b). Without those markings, drivers are likely to advance less deeply into the median break, making them more likely to block through traffic. As discussed in Chapter 4 in the example of Columbus Ave @ Rutland Sq (Figure 4.6), using median space is one method to create informal flare function. However, uncertainty of drivers in terms of safety leads to under-utilization of that space and could limit that function. Given their potential for improving efficiency of informal flares at very little cost, shadow lines could be a valuable tool for making road diets succeed where road space is limited. However, there remains a knowledge gap as to whether shadow lines significantly draw waiting cars deeper into the median break.

5.2. Existing Examples

Using shadow lines at median breaks is common practice in the Netherlands, where many examples can be found in both urban and suburban settings. These lines are used in the U.S. as well, but are not as commonplace as in the Netherlands. Researching Manual on Uniform Traffic Control Devices (MUTCD) yielded no specific guidelines or recommendations as to using or not using shadow lines; however, they are explicitly permitted as an option within Section 3B-08:

“Dotted edge line extensions may be placed through intersections or major driveways.” (1)

From an inquiry to the Institute of Transportation Engineers Community, we learned that shadow lines are a standard treatment on highways in Florida. Following are some of examples in the Netherlands as well as in Florida:

shadow lines

(a)

(b) Figure 5.1 – Shadow lines along a break in the median on Ruys de Beerenbrouckstraat, Delft a) plan view b) left-turning vehicle waiting behind the line

Van Miereveltlaan @ Tweemolentjeskade/ Van Miereveltlaan @ Aan Het Verlaat, Delft: At these two intersections, shadow lines are marked along the distance between crossing islands on either side and the median (Figure 5.2a). The width between two shadow lines is 4.6 ft,

and the travel lane is 13.2 ft wide. The travel lane itself is too narrow to function as an informal flare. However, when left-turning vehicles pull into the median break, the space remaining on their right becomes wide enough for a through vehicle to pass easily (Figure 5.2b).

shadow lines

(a)

(b) Figure 5.2 – Shadow lines at Van Miereveltlaan @ Tweemolentjeskade and Van Miereveltlaan @ Aan Het Verlaat a) Plan view b) White vehicle waiting close to the line yielding to the silver vehicle who has just passed by

Rijswijkse Waterweg @ Van Campenvaart, The Hague: Approaching this intersection, a wide median is narrowed to accommodate the left turn lane, and is finally broken at the intersection to allow left turning movement (Figure 5.3). In this example, because there is an exclusive left turn lane, the purpose of the shadow lines is not to provide more room for through vehicles, but only for safety and driver assurance. The shadow line provides positive guidance for drivers to position their vehicles while waiting for a gap.

shadow line

Figure 5.3 – Shadow line at Rijswijkse Waterweg @ Van Campenvaart

South Byron Butler Pkwy @ West Pine Rd, Perry, FL: Like the example discussed previously (Figure 5.3), the shadow lines here are in conjunction with a left turn lane, in this case, on a divided four-lane road where the speed limit is 45 mph (Figure 5.4). So, these lines don’t have an informal flare function, but are only there for safety. As mentioned earlier, they provide drivers assurance in positioning themselves; it may also be that by promoting a more forward queuing position, which allows turning vehicles to accomplish more of their rotation while queued, they enable a left turning vehicle to clear the intersection more quickly, improving safety.

shadow line

Figure 5.4 – Shadow line at S Byron Butler Pkwy @ W Pine Rd

South Byron Butler Pkwy @ a commercial driveway, Perry, FL: In this Florida example (Figure 5.5), again on a 45-mph, 4-lane divided highway, there are no left turn lanes, because the break in the median is for an intersection only with low-volume commercial driveways. As the position of the red car in the figure indicates, the space between shadow lines is wide enough for a vehicle to stand without blocking a through lane of its own direction. The shadow lines improve safety to the extent that they help draw queued vehicles fully into the median break so that their tail does not extend into a through lane, creating a rear-end crash hazard.

shadow lines N

red turning car

Figure 5.5 – Shadow lines at S Byron Butler Pkwy @ a commercial driveway

5.3. Application in Cummins Hwy Case Study:

In our Cummins Hwy study, shadow lines are proposed within the median break at every unsignalized intersection where left turning is allowed. In places where no left turn is possible (i.e. cross street is one way), shadow lines are unnecessary and therefore, are avoided (Figure 5.6).

N shadow line encouraging left turning car fully using the available width no shadow line provided

Figure 5.6 – Shadow lines in Cummins Hwy and informal flare function

The available space is used most efficiently if waiting left-turning vehicles use the full depth of the median break. It is reasonable to assume that shadow lines encourage drivers to do so. Our study of informal flares measures the space available for cars waiting to turn and passing cars assuming drivers make use of all the available space. If the absence of shadow lines makes drivers not use the full depth of a median break, there will be less informal flare function than might be expected. In Figure 5.6, the median is 5.5 ft wide and the left turning car has gone 5 ft deep into it. The trajectory of the through red car is such that it shies away from the left turning car by 1.5 ft, and gets as close as 2 ft to the curb right afterward. This maneuver could not be made at a comfortable speed if the left turning car had advanced only halfway into the median break. The above is an example that demonstrates the importance of harmony in design, where every effort must be made to ensure the intended functions are realized in practice. Because of the crucial role of medians in providing space for informal flares, and to help use that space efficiently, it is recommended that designers consider shadow lines, which are inexpensive and easy to install.

5.4. Use of Median Width and Informal Flare:

The effective use of median width can facilitate informal flare function, as the deeper into the median a waiting left turning vehicle goes, the less its tail will occupy the travel lane. An experiment was done to see the relationship between how deep into a median a waiting left turner stops and the width that the vehicle’s tail occupies from the travel lane (Figure 5.7).

WB use of median width width occupying travel lane EB

Figure 5.7 – Use of median width and informal flare

In this experiment, the trajectory of a left turning car in at right angle intersection was drawn using AutoTURN software. The vehicle in this experiment is the AASHTO’s Passenger Vehicle (P) (2). The decrease in the width of the tail occupying travel lane compared to a base case in which there is no median is found by assuming that without a median, vehicles position themselves one foot away from the center line. At the end, a vehicle size adjustment was done as AASHTO’s Passenger Vehicle (P) is unreasonably large (19 ft long and 7 ft wide). The adjusted values are based on a fairly large, but realistic vehicle which is 16 ft long and 6 ft wide (Table 5.1).

Table 5.1 – Relationship between use of median width and width occupied by vehicle’s tail Width Occupied by the Tail (ft) Change Compared to No Median (ft) Use of Median AASHTO Adjusted AASHTO Width (ft) Adjusted Vehicle Vehicle (P) Vehicle Vehicle (P) 0 8 7 0 0 2 7.8 6.4 0.2 0.6 4 7.5 5.8 0.5 1.2 6 7.1 5.1 0.9 1.9 8 6.4 4.2 1.6 2.8 10 5.8 3.4 2.2 3.6 12 4.8 2.3 3.2 4.7 14 3.9 1.1 4.1 5.9 16 2.7 0 5.3 7 18 1.4 0 6.6 7 20 0 0 8 7

As seen in the table, if the space of a 6-ft wide median is fully used, the turning vehicle opens up 1.9 ft of space for informal flare, whereas if only 2 ft of it is used, the added space contributing to informal flare will be only 0.6 ft.

5.5. Observational Study on Use of Median Space

5.5.1. Introduction and Objective

Besides understanding how the use of median width can facilitate informal flares, it is of value to see how deep drivers typically use the median space. As mentioned earlier, shadow lines are not as common in the U.S. as they are in the Netherlands, where they are routinely used in median breaks. To our knowledge, there have been no before/after studies in the U.S. to measure the effectiveness of shadow lines, and we were not able to conduct one. However, we were able to conduct a brief observational study which aims at finding current behavior without shadow lines. This will show the potential for improvement if shadow lines are implemented.

5.5.2. Study Location and Method

The study location is the signalized intersection of Columbus Ave @ West Newton St in Boston, Massachusetts (Figure 5.8). The left turners from Columbus Ave EB to West Newton St NB are under study. The layout of Columbus Ave is rare in Greater Boston, having a median yet

only one lane per direction. That layout creates informal flare opportunities which were observed and recorded. The median – a flush median paved with rough cobbles – is 6 ft wide; the EB travel lane is 11 ft wide; and a 4.5-ft bike lane is located between the travel lane and parking lane, with parking allowed up to the corner of the intersection. When a vehicle is parked properly at the intersection corner, the width of the median break, travel lane, bike lane are available for informal flare function. This functionality depends on how deep left turners use the space in median break.

bike lane median

Figure 5.8 – Columbus Ave @ West Newton St

A camera was set in the middle of the median for video recording. To avoid driver distraction, the camera was positioned on the upstream side of the intersection and vehicles were recorded as going away from the camera (Figure 5.9). Because the camera was set in a direct angle, lateral distortions were negligible and direct perspective effect was assumed in calculations. The unused part of the median was measured; subtracting it from median width yielded the used part.

unused median width

Figure 5.9 – Camera view (Explanation: The silver car is partially using the median break while the black car uses the bike lane to make an informal flare maneuver)

5.5.3. Results and Conclusions

18 valid observations for median width use, and 16 informal flare maneuver opportunities were recorded. A valid observation is defined as a vehicle that comes to a complete stop (or nearly a complete stop) while waiting for a gap. Many left-turning vehicles were observed to slow down or coast at a low speed as they approached the intersection, expecting a gap to arrive soon; those vehicles were not counted as valid observations. Maximum recorded use of the median break was 4.4 ft, a width used by two vehicles, and the minimum was zero, used by three vehicles. Average median use was 2.25 ft (standard deviation = 1.4 ft), which suggests that the median space is less than half-effective. Informal flare maneuver opportunities were recorded for each turning vehicle by calculating the road width available for a passing through vehicle (Table 5.2). Available width was determined by adding 15.5 ft (combined width of the 11 ft travel lane and 4.5 ft bike lane) to values interpolated from the far right-hand column of Table 5.1, which indicates how much of the travel lane will be become available as a function of how much the median space is used. Also recorded was every instance of a through vehicle approaching and whether passed through or waited until the left-turning vehicle departed.

Table 5.2 – Available width for informal flare and passing tendency Cumulative Number of Opportunities at or Available Width for Median Width below Median Width Used Informal Flare Used (ft) Vehicles Vehicles Passing Maneuver (ft) Arrived Passed Tendency (%) 0.0 1 0 0% 15.50 0.5 2 0 0% 15.65 1.0 1 0 0% 15.80 1.5 1 0 0% 15.95 2.0 1 0 0% 16.10 2.5 3 1 33% 16.25 3.0 8 5 63% 16.40 3.5 10 7 70% 16.55 4.0 11 7 64% 16.70 4.5 16 12 75% 16.90 5.0 16 12 75% 17.05 5.5 16 12 75% 17.25 6.0 16 12 75% 17.40

As the sample size is low, the percentage found for passing tendency may differ considerably from reality. However, the data still show a clearing rising trend in passing tendency with the median being used deeper, as expected.

5.5.4. Knowledge Gaps, Applications, and Recommendations for Future Research

The sample size in this study is small. So, drawing reliable conclusions which match the real life will require additional observations. Still, they confirm a general trend between the use of median space and its effect on informal flare function. The study also provides a framework for similar future studies. With large data sets, the informal flare behavior in this study could be incorporated into the study that was presented in Chapter 4 by breaking the behavior into three categories of first in queue, following in queue, and free flow. As mentioned in the previous chapter, the delay caused by left turning cars should also be measured, which is of significant value in determining capacity loss at an intersection.

REFERENCES

1- Manual on Uniform Traffic Control Devices. Federal Highway Administration of the U.S. Department of Transportation, Washington, D.C., 2009. 2- A Policy on Geometric Design of Highways and Streets. American Association of State Highway and Transportation Officials (AASHTO), Washington D.C., 2011.

6. Roundabouts and Mini-Roundabouts

6.1. Overview

Modern-day roundabouts, or simply roundabouts, are well-known type of intersection design in which the traffic flows in a circulatory path (Figure 6.1). Roundabouts have certain common features which set them apart from other circulatory intersection designs, of which the most widely-used ones are old-style rotaries and traffic circles. These features are explained briefly as follow:

- Yield on entry: At roundabouts, vehicles entering must yield to circulating traffic. In rotaries and traffic circles, the entry rules and dynamic of traffic can be different. While some relatively small rotaries require yielding on entries, large ones may allow merging at high speeds, similar to merging from an entry ramp onto a freeway. Traffic circles can be signalized (categorized as signalized traffic circles), stop-controlled or have no control. The last two are common in neighborhood traffic signals. - Deflection on entry: Roundabouts are meant to slow down traffic, not to provide high-speed passing opportunities. So, they are designed in a way that deflects the traffic on the entry, unlike many rotaries with tangential entries that allow vehicle to enter at high speed. - Small diameter: Roundabouts keep traffic going slow in the circulating lanes by having a small diameter, unlike large, old-style rotaries. - Splitter islands: These islands make it easier for pedestrians to cross each leg, and separate exit legs from entry legs, which helps reduce false conflicts (cars yielding to a car they think is circulating but is actually exiting). - Crossable legs: crosswalks are provided on every leg for pedestrians, passing through splitter islands. - Inaccessible center: Pedestrians do not cross to the center. - Apron (and armpit): Circulatory lanes are designed wide enough for passenger vehicles to navigate with comfort. For large vehicles, there is a mountable apron designed in a way that discourages cars from using it, but which heavy vehicles can comfortably use. Where needed on the corners, rough paving in the armpits can also facilitate heavy vehicle movement while discouraging cars, so that cars are forced to deflect as designed. - Counterclockwise rotation only: This is unlike some small neighborhood traffic circles where large vehicles are allowed to turn left clockwise.

Figure 6.1 – Roundabout example in Braintree, Massachusetts

The discussed features earn roundabouts several safety and operational benefits compared to other intersection alternatives such as traffic signals. Following are the most important advantages of roundabouts:

- Yielding of entry movement to circulating traffic avoids congestion within the roundabout when traffic is heavy, yet incurs little to no delay to vehicles when traffic is light. Yielding ensures that circulating traffic remain fluid even when the demand is so high that the intersection is oversaturated. Moreover, when no circulating traffic is present at the roundabout, yield law requires no stopping. So, no delay will be experienced by a driver who arrives at the roundabout when no conflicting traffic is there. - Vehicles need to slow down as they arrive at a roundabout and maintain low speed while circulating. Also, vehicles all travel in the same direction within a roundabout. As a result, the possibility and severity of crashes decrease significantly. Contrary to signalized intersections, the chance of having right-angle or head-on crashes will be eliminated. Also, roundabouts slow down traffic all the time, as opposed to traffic signals which control speed only when the signal is red. - Splitter islands provide safe pedestrian crossing, allowing pedestrians to cross one direction of traffic at a time. Also, pedestrians will have right-of-way all the time and can cross as soon as they find it safe to do so.

- The apron facilitates the movement of heavy vehicles within the roundabout, while at the same time is made rough and bumpy enough to deter passenger cars from going over it. Apron is designed for accommodating heavy vehicles without making the circulatory lane excessively wide for passenger vehicles. This feature ensures the center island still effectively deflects the traffic and prevents speeding.

At the same time, roundabouts pose two major challenges which in an urban environment:

- Roundabouts typically need more space than other alternatives, of which the most commonplace one is signalized intersections. Where space is constrained, mini-roundabouts, a more compact version of roundabouts can be considered. - While pedestrian crossings are safe and easy at single lane roundabouts, crossings at multilane roundabouts – which always include at least one multilane exit – present a multiple threat hazard. Many cities will not use multilane roundabouts without traffic signals (which detract from the roundabout advantages). That limits the application of roundabouts to places the capacity of a single-lane roundabout is sufficient.

NCHRP Report 672 divides roundabouts into three categories: mini-roundabouts, single-lane roundabouts, and multi-lane roundabouts (1). Table 6.1 shows a highlight of the most important differences of the three categories.

Table 6.1 – NCHRP 672 roundabout category comparison Mini- Single-Lane Multi-Lane Design Element Roundabout Roundabout Roundabout Desirable maximum entry design 15-20 20-25 25-30 speed (mph) Maximum number of entering lanes 1 1 2+ per approach

Typical inscribed circle diameter (ft) 45-90 90-180 150-300

Fully Raised (may have Raised (may have Central island treatment traversable traversable apron) traversable apron) Typical daily service volumes on 4- leg below which may be expected to <45,000 (for two- <15,000 <25,000 operate without detailed capacity lane roundabout) analysis (veh/day)

In the study of road diets on urban arterials, where the goal is to reduce the lanes of traffic to one per direction, single-lane roundabouts can be a helpful intersection type. Roundabouts can support higher flow per lane than signalized intersections, so that a road that would need 4 through lanes (plus turn lanes) with traffic signals can have only two through lanes using roundabouts. Compared to stop-controlled or signalized intersections, roundabouts reduce delay, help traffic flow steadily (as opposed to pulsing it), and are safer for all road users. Mini-roundabouts are reported to have nearly all the benefits of single-lane roundabouts with a more compact layout that more often can fit within the available urban right-of-way (2), but they have less capacity. Given all their advantages and the crucial role they can play in urban settings, many studies have been done and guidelines are available that assist designers in understanding capacity analysis and modeling of roundabouts.

6.2. Literature Review

Implementing a roundabout in most cases requires capacity analysis to be done to make sure it responds to the demand. For this purpose, HCM 2010 presents the following capacity formula for an entry in a single lane roundabout, which is found using simple empirical regression models (3):

−3 (−1.0×10 )푣푐,푝푐푒 푐푒,푝푐푒 = 1,130푒

Where: ce,pce = capacity of entry lane (adjusted to equivalent passenger car) (pc/h) ve,pce = volume of conflicting traffic (adjusted to equivalent passenger car) (pc/h)

Figure 6.2 shows the same HCM formula plotted in a graph. NCHRP Report 572 compares other analytical and regression models against the HCM’s, and finally recommends the same HCM model (4).

Figure 6.2 – HCM entry lane capacity against conflicting flow rate graph

Lochrane et al. (2012) investigated the validity of above HCM capacity model for two size of mini-roundabouts with 50 ft and 75 ft of inscribed circle diameter (5). They used VISSIM software simulation and developed new capacity models for the two mentioned size of mini- roundabouts by calibrating the results to U.S. field data in terms of headway, speed and gap. Their findings show that HCM model for single-lane roundabouts significantly overestimates the capacity of mini-roundabouts. Capacity on the entry correlates with drivers’ gap acceptance for entering the roundabout, a topic being vastly studied around the world. In the concept of gap acceptance, a critical headway (gap) is defined as a headway (gap) longer than which the driver accepts the gap to enter the main traffic stream, and otherwise rejects it (6). HCM’s capacity model has a theoretical basis in gap acceptance, whereas this is strongly influenced by driving behavior, which may vary significantly in one place compared to the other. Follow-up headway, being the headway between the queued vehicles, is another important parameter in gap acceptance. HCM’s default value in capacity analysis of single-lane roundabouts corresponds to a critical gap of 5.19 seconds and follow-up headway of 3.19 seconds. Many studies in the U.S. and other countries show noticeable difference with those values.

The most prevalent methods followed in studies on finding gap acceptance parameters are Raff’s Method (1950) and Maximum Likelihood Method (1999) (7, 8). Abrams et al. (2013) studied gap acceptance behavior at roundabouts in Amherst, Massachusetts (9). They concluded that HCM values underestimate capacity. Their study was based on video analysis of over 1,500 vehicles on a typical suburban single-lane roundabout. Using Raff’s Method (1950), they found a temporal gap of 2.2 seconds, which corresponds to a spatial gap of 42 ft. Giuffre et al. (2016) did a comprehensive review of distinctive past research works on gap acceptance parameters across the world (10). The summary of their work is brought in Table 6.2, which includes U.S. and the Netherlands studies on single-lane roundabouts.

Table 6.2 – Summary of studies on gap acceptance parameters in single-lane roundabouts Critical Headway (s) Follow-Up Headway (s) Method Study Name Country Mean SD Mean SD Applied Min Max Min Max Min Max Min Max Abrams et Raff’s 2.20 ------al. (9) Method Tian et al. 3.80 5.50 1.00 2.00 2.30 3.80 1.00 2.60 (11) Maximum Rodegerdts US Likelihood 3.90 5.90 0.70 1.80 2.60 4.30 0.80 1.50 et al. (4) Method Xu and Tian 4.50 5.30 0.90 1.10 2.30 2.80 0.30 1.00 (12) Mensah et 2.50 2.60 ------al. (13) Not Fortuijn Available NL 3.16 3.28 0.19 0.28 2.10 - - - (14)

As shown in Table 6.2, a wide range of variability can be observed among different studies, and almost all of them show huge difference from HCM’s default values. So, for capacity to be determined with more accuracy, HCM parameters should be calibrated based on local studies.

6.3. Existing Examples

Considering the compactness of mini-roundabouts, which makes them the most viable roundabout alternative for constrained urban right-of-way, the focus of the examples in this section is on mini-roundabouts, meaning roundabouts whose entire central island and splitter islands are

traversable. Many successful examples of mini-roundabouts exist in the U.S. and other parts of the world. Following is one example in the U.S., and another in the Netherlands:

Country Rd 79 @ Vierling Dr East, Shakopee, Minnesota: This mini-roundabout was built in place of an all-way stop controlled intersection. The inscribed circle diameter is 77 ft, with circulatory lane of 16 ft wide (Figure 6.3). The center island is fully traversable with a diameter of 45 ft. The entries are tangential and the conflict points of entry and exit are almost in the same spot. Before construction, the old intersection is reported to have had persistent queues of as long as 6 minutes on the northern leg during PM peak (15). After construction, however, long queues on the northern leg have been vanished.

77 ft

Figure 6.3 – Roundabout in Shakopee, Minnesota

Westlandseweg @ Aletta Jacobsstraat/Zaagmolen, Delft, The Netherlands: In the U.S. context, this roundabout is the size of a mini-roundabout. The inscribed circle diameter is only 82 ft and the circulating lane is 18 ft. However, unlike U.S. mini-roundabouts, the center island is not fully traversable (Figure 6.4). The diameter of center island is 46 ft with a mountable apron strip of 6.5 ft wide, which is meant to be only used by heavy vehicles.

82 ft

Figure 6.4 – Roundabout in Delft, The Netherlands

Light trucks, buses and articulated farm tractors (with carriers attached) were observed to navigate easily within the roundabout with little to no need to mount the apron (Figure 6.5). Despite U.S. guidelines may categorize this design as a mini-roundabout on account of its size, the functionality of it is more similar to a typical single-lane roundabout. In this Dutch roundabout design, entries are straight, which separates the conflict point at the entry from the one at the exit. Also, as markings suggest, all entry and exits are slightly raised as tables to grab driver’s attention better to yield to pedestrians and bikes, where applicable.

Figure 6.5 – Bus navigating roundabout in Delft, The Netherlands

6.4. Proposed Design for Cummins Hwy Road Diet

Within the stretch of study on Cummins Hwy, a mini-roundabout was designed at a five- leg intersection (Figure 6.6). The inscribed circle diameter is 90 ft, with a 16-ft circulating lane and a 58-ft diameter fully mountable center island. A cycle track is accommodated all around the roundabout with a 10-ft offset from the circulating lane. In the design, it was not determined in advance whether it would be a miniroundabout (with a mountable median and splitter islands) or not; that would be determined as part of the design process depending on whether intended large vehicles could negotiate the roundabout without going over the central and splitter islands.

right-of-way N constraints

90 ft

right-of-way constraints

Figure 6.6 – Mini-roundabout design on Cummins Hwy

6.4.1. Design Considerations

The design features of this roundabout are based on the following considerations: - The right-of-way is constrained in four points (Figure 6.6). Given the desire to provide 5-ft cycle tracks with a uniform 10-ft offset from the travel lane across every leg, and maintain at least 6 ft of sidewalk at pinch points, the inscribed diameter is limited to 90 ft. - The width of circulating lane (which determines the diameter of center island) is chosen based on assuring desirable deflection for cars in all movements. Special attention was given to ensuring enough deflection to control speed for cars traveling along Cummins Hwy in the WB direction, and turning right from Cummins Hwy EB onto Greenfield Rd. Using AutoTURN,

travel paths were drawn for passenger vehicles based on the assumption that drivers tend to smoothen their path to pass through the roundabout at highest possible speed (Figure 6.7). The widest radius at which a car can navigate the roundabout is 40 ft (performed by vehicles going WB), which ensures the speed for all the cars is limited to roughly 15 mph or lower. Also, the design allows a 40-ft school bus to navigate with little to no need for using the center island.

Figure 6.7 – Fastest path trajectories for passenger vehicles

- The design vehicle for this roundabout is 62-ft long tractor trailer (AASHTO’s WB-62) (16). The turning path for this vehicle was drawn using AutoTURN (Figure 6.8). To turn left from Cummins Hwy EB onto Weybosset St (red trajectory), and to travel through the roundabout on Cummins Hwy going EB (green trajectory), WB-62 mounts over a large part of the center island, which necessitates that it be fully traversable.

Figure 6.8 – Turning path trajectories for WB-62

6.4.2. Operational Performance

Synchro software analysis shows a total delay of 15 seconds and maximum v/c ratio of 0.84 during P.M. peak. The roundabout was also modeled in VISSIM microsimulation software. The parameter primarily controlling the roundabout capacity in VISSIM is minimum gap, which is equivalent to critical gap. As discussed earlier, critical gap is highly dependent on driving behavior, and could differ significantly in different places. In terms of geographical location and driving behavior, the closest study found on this topic was Abrams et al. (2013), which found a critical gap of 2.2 seconds (9). For this study, the critical gap in VISSIM was set to 2.5 seconds. During P.M. peak, during which traffic was found the heaviest, the roundabout works with no congestion, i.e., no persistent queues on any leg.

6.4.3. Benefits

This mini-roundabout has several safety and operational benefits: - The existing 5-leg intersection has no speed control. Through vehicles on Cummins Highway traverse the intersection at high speeds, as can vehicle turning right from Cummins Highway EB onto Greenfield St. The mini-roundabout controls speed for all of these movements. - Presently, left turners from Greenfield Rd and Weybosset St have to advance half-way into the intersection and wait for a gap there. This puts them in a hazardous situation as they might be

involved in a right-angle crash with vehicles going EB on Cummins Hwy. The mini-roundabout allows left turners to safely enter Cummins Hwy. - Pedestrian crossings become much shorter. Crosswalks across each direction of Cummins Hwy will be 13 ft (26 ft in total), whereas today they are longer than 70 ft. In addition, vehicle speed near the roundabout is much slower and drivers are more likely to yield to pedestrians, which results in less pedestrian delay. - Cummins Hwy currently has no bike facilities. With the proposed road diet, cycle tracks will be added all along and the roundabout provides a safe place for bikes to turn from Cummins Hwy onto residential streets and vice versa. - Located between two signalized intersections, the roundabout provides a metering effect for the traffic going EB on Cummins Hwy. Relatively large platoons released in two lanes by the 80-second cycle (during P.M. peak) at Wood Ave/Harvard St intersection will gradually discharge from the roundabout in one lane at a steady flow to the Itasca St/Ridlon Rd intersection, where the short 60-second cycle prevents queues from spilling back into the roundabout. - The capacity of roundabout is sufficient for the queue on the EB entry (which is the busiest direction during P.M. peak) to discharge every cycle. When the platoon arrives from Wood Ave/ Harvard St intersection, cars form a temporary and flowing queue at the entry. However, the queue discharges completely before the next platoon comes. - The roundabout allows Cummins Highway to operate with a single through lane in each direction, while a traditional intersection would probably require it to have two lanes per direction. By enabling one lane per direction, the mini-roundabout helps bring speed control, safety, crossability, and bicycling benefits to the entire corridor.

The above case study shows the feasibility of a roundabout in a constrained urban environment, using the form of a mini-roundabout. It shows the potential of mini-roundabouts to create substantial safety and operational benefits in the context of a road diet. Given the advantages in safety and operation, it is recommended that mini-roundabouts be considered as a part of road diet projects.

REFERENCES

1- NCHRP Report 672: Roundabouts: An informational Guide. Transportation Research Board, Washington D.C., 2010. 2- Mini-Roundabouts. Federal Highway Administration of the U.S. Department of Transportation, Washington, D.C., 2010. 3- Highway Capacity Manual. Transportation Research Board, Washington D.C., 2010. 4- NCHRP Report 572: Roundabouts in the United States. Transportation Research Board, Washington D.C., 2007. 5- Lochrane T. W. P., W. Zhang, and J. Bared. Mini-Roundabouts for the United States and Traffic Capacity Models. Institute of Transportation Engineers Journal 82(11):20–24, 2012 6- Tupper, S. Knodler, M. and Hurwitz, D., Comparative Analysis of Critical Gap Analysis Methods. University of Massachusetts Amherst, 2011. 7- Raff, M.S., and J. W. Hart, A volume warrant for urban stop signs. The Eno Foundation for Highway Traffic Control, 1950. 8- Tian, Z., M. Vandehey, B. W. Robinson, W. Kittelson, M. Kyte, R. Troutbeck, W. Brilon, and N. Wu. Implementing the Maximum Likelihood Methodology to Measure a Driver’s Critical Headway. Transportation Research Part A, Vol. 33-3-4, pp. 187–197, 1999. 9- Abrams, D. S., C. D. Fitzpatrick, T. Tang, and M. A. Knodler. Spatial and Temporal Analysis of Driver Gap Acceptance Behavior at Modern Roundabouts. Transportation Research Record: Journal of the Transportation Research Board, Vol. 2388-1, 2013, pp. 14-20. https://doi.org/10.3141/2388-03 10- Giuffre, O., A. Grana, and M. L. Tumminello. Gap Acceptance Parameters for Roundabouts: A Systematic Review. European Transport Research Review 8: 2, 2016. DOI: 10.1007/s12544-015-0190-4 11- Tian, Z., R. Troutbeck, M. Kyte, M. Vandehey, W. Brilon, W. Kittelson, and B. W. Robinson. A Further Investigation on Critical Gap and Follow-up Time. 4th International Symposium on Highway Capacity, Transportation Research Circular E-C018. Maui, Hawaii. 2000 12- Xu, F., and Z. Tian. Driver Behavior and Gap-Acceptance Characteristics at Roundabouts in California. Transportation Research Record: Journal of the Transportation Research Board, No. 2071, 2008, pp. 117-124. https://doi.org/10.3141/2071-14

13- Mensah, S., S. Eshragh, and A. Faghri. A Critical Gap Analysis for Modern Roundabouts. 89th Annual Meeting of the Transportation Research Board, Washington, D.C., 2010. 14- Fortuijn, L. G. H. Turbo Roundabouts: Estimation of Capacity. Transportation Research Record: Journal of the Transportation Research Board, No. 2130, 2009, pp. 83–92. https://doi.org/10.3141/2130-11 15- Stein, W. Mini-Roundabouts in Minnesota: Benefits of Roundabouts with a Smaller Footprint and Lower Cost. North Central Section Institute of Transportation Engineers (NCITE) Summer 2017 Newsletter, 2017. 16- A Policy on Geometric Design of Highways and Streets. American Association of State Highway and Transportation Officials (AASHTO), Washington D.C., 2011.

7. Bus Stops in Bays vs. In-Line

7.1. Overview

Road diets pose a two-way challenge involving public transit, which in most places in the US is offered by buses in mixed traffic. One is to prevent congestion from harming the quality of transit. Sometimes that can be done by reallocating space to bus lanes and queue jump lanes. Cummins Highway, the road in our case study, is not an important transit corridor; it is served by a single bus route running twice an hour, and, being a crosstown route with almost no destinations on it, is not likely to develop much of a transit market. Therefore, special efforts to improve transit service quality were not part of this study, and are left for other researchers. The other challenge is that bus stops, if buses stop in line, can significantly impact traffic flow. Between stops, buses and cars have the same goal – to move as quickly as possible. But at bus stops, the intended functionality of cars and buses contradict; while buses need to stop, other motor vehicles need to keep flowing. This becomes more crucial when a road diet limits motor traffic to one lane per direction, in which case having a bus block a lane can create a large impact. Many researches have been carried out on bus-car interactions to identify the contributing parameters and quantify their effects in this matter. From the point of view of operation, these interactions are strongly influenced by bus stop location and type. The most comprehensive guideline book available on bus stop design in the U.S. is TCRP Report 19, which classifies bus stops into three location categories and five types (1). Bus stop location refers to whether the stop is upstream of a traffic signal (near-side), downstream (far side), or located far enough away from a signalized intersection (mid-block). The five bus stop types are defined based on how buses enter a stop, and whether they serve passengers while occupying a lane of traffic or they pull over to a designated area off the travel lane (bus bay) (Figure 7.1). While serving passengers, curbside or nub type interact exactly the same with cars. So, we can put the two in the same category and call them in-line stops. Also, an open bus bay and queue jumper have essentially the same effect on cars and we can cluster them together. The first two types depicted in Figure 7.1 have had the widest applications, and therefore have been given the most attention in previous studies. In line with the past works, this chapter also discusses those two more than others. A drawback of bus bays is that they require more space than in-line stops. Where part of the purpose of a road diet is to create protected bike lanes, bus bays may consume so much space that there isn’t enough left for a sidewalk, a bus shelter, and a protected bike lane, especially considering the accessibility requirement of an 8-ft deep landing for bus passengers. Thus, from

the viewpoint of making the best use of limited space, in-line stops are of interest for road diet projects where they can be shown to not have a serious capacity impact.

Figure 7.1 – Five different types of bus stops identified in TCRP 19

7.2. Literature Review

Past researches give an insight into the advantages and disadvantages of each type of bus stop and its location, which allow for a more informed decision in planning. As said, the focus will be on the two most common types (in-line and simple bus bay) as well as on near-intersection alternatives. TCRP Report 19 lists advantages and disadvantages for each location and type alternative qualitatively (1). For in-line stops, ease of use by bus drivers, minimal bus delay and low cost are the main benefits, while disruptions to through traffic is the biggest downside. For bus bays,

through traffic was mentioned to experience minimal delay whereas the bus becomes more likely to be delayed when re-entering the traffic. The most important conditions provided in TCRP Report 19 for considering the use of bus bays are:

- Traffic in the curb lane exceeding 250 vehicle per hour. - Traffic speed being greater than 40 mph. - Bus volumes being 10 or more per peak hour. - Passenger volumes exceeding 20 to 40 boarding per hour. - Average peak-period dwell time exceeding 30 seconds per bus. - Potential for auto/bus conflicts warranting separation of transit and passenger vehicles.

Fitzpatrick and Nowlin (1997) studied bus stop design on traffic operations around the stop on multilane roads (2). The study encompasses all different locations and design alternatives of Figure 7.1. The results suggest that for far-side in-line bus stop, dwell time had nearly no effect on vehicular traffic. Also, for mid-block and far-side locations, for traffic volumes above 350 veh/hr/ln, advantages of bus bays over in-line stops were significant. Wong et al. (1998) investigated the delay effects of an in-line bus stop located upstream of a signalized intersection (3). With their developed model, they found that average delay incurred per vehicle is directly related to traffic demand, bus frequency, average dwell time, proportion of time the bus stop is blocked, and cycle length. However, it is inversely related to the distance of bus stop to the intersection and effective green time. Their findings suggest that delay is inversely proportional to square root of the green ratio (green time divided by cycle length, or g/C), which is valuable in signal design. Furth and SanClemente (2006) studied the impact of bus stop location and grade on bus delay by developing theoretical models (4). They did not study delay to traffic. Far-side stop was found to add almost no delay for buses compared to a case where stops are located far enough away from an intersection. As for near-side stops, delays were significant in cases where queue was long enough to block the bus stop. Keeping cycles short with short red intervals, and increasing bus stop setback all contribute to alleviate the queue interactions issue. Moreover, if bus stops are located on steep uphill grades, they found significant delay being imposed to buses when accelerating. Compared to a level stop, for instance, a 6% uphill grade increases acceleration delay by more than 8 seconds for a bus to reach 30 mph. Meng and Qu (2013) developed a probabilistic model to estimate bus dwell time at bus bays on multilane roads (5). What findings show is that traffic flow on the travel lane adjacent to bus

bay can significantly increase bus dwell time, which introduces a high degree of uncertainty to dwell time. The reason is attributed to the bus waiting for a gap to merge back into the traffic. Arhin et al. (2016) studied the total bus stop time (TBST) in Washington D.C., to compare near-intersection results with mid-block (6). TBST is the summation of dwell time and other time losses such as time lost in acceleration and deceleration and time for doors to open and close. The results of their research found that TBST is longer at stops near intersections compared to mid- block, which was interpreted as lowering the service reliability. No comparison was made between near-side or far-side stops. Impact of bus stop location to both bus and traffic were studied by Gu et al. (2014) (7). This study finds that under most circumstances, near-side stops are favorable for cars, while far-side stops almost always yield less delay for buses. It was also demonstrated that for a given ratio of bus to car occupancy, as we move away from an intersection, near-side stops become more attractive from a perspective of total person-delay, whereas for distances in close proximity to an intersection, far-side has less overall delay impact. Liu et al. (2017) studied the effects of bus stop type, in-line or bus bay, on bus delay in Singapore using statistical analysis (8). The goal was to find whether acceleration and deceleration delay as well as boarding/alighting time is different between the two type options. While no significant difference was found in deceleration delay, acceleration delay was longer and more variable in bus bays, which was reported to be due to re-entering the traffic. Also, boarding and alighting per person was quicker on with in-line stops. The observations found that with bus bays, drivers are often unable to pull up close enough against the curb, which makes passengers take an extra step down onto the road and up again. In a study which considers both cars and buses at the same time, Luthy et al. (2016) carried out a theoretical analysis of overall delay effects incurred by buses and cars on a single-lane road based on whether bus stops are near or far from the bottlenecks, and whether they are in-line or in a bay (9). With minimizing the overall delay experienced by all the users being the main objective, they presented guidelines to be used in the decision making about the type and location of a bus stop. Cvitanic (2017) studied the joint effect of bus stop location and type on the performance of a signalized intersection (10). The parameters of study are bus headway, traffic demand, and dwell times. For the case of a single lane road, the upstream bus bay case has the longest bus delay in almost all headways, traffic volumes, and dwell times. For traffic volumes under 800 veh/hr and headways of over 600 seconds, in-line stops are recommended. Below this critical volume, car delay is insignificantly higher in the case of in-line stops. These findings suggest that in-line stops have advantage over bus bays on urban streets where traffic volume is under capacity.

In summary, previous studies reflect a consensus that under most circumstances in an urban environment, in-line stops have more operational benefits and less overall delay (5, 10). Also, most studies agree on that far-side stops are superior to near-side (4, 7, 10).

7.3. Study on Far-Side Stops and Capacity of Signalized Intersections

7.3.1. Introduction and Objective

As discussed, in-line bus stops located downstream of a signalized intersection are a superior option in most urban settings. However, when considered on a road diet in which two lanes’ worth of traffic in each direction have been confined to a single lane, in-line stops might pose challenges, especially if they hurt the capacity of signalized intersections. Several studies have measured the effects of bus stops on cars and buses, which mostly focus on delay. Although delay is an important parameter to consider, protecting intersection capacity is also crucial in road diets. Therefore, to find the effects of an in-line, far-side bus stop on the capacity of a single-lane per direction road at the intersection, a theory has been developed and validated against software simulation. The model shows the effects of bus stop setback, dwell time, and bus frequency on a signalized intersection capacity loss.

7.3.2. Theory

The impact of a bus, stopping at a far-side, in-line stop, on the discharge capacity of a through movement at an intersection depends on both its position in a discharging platoon and the setback, that is, the distance of the stop from the intersection stop line. If the setback is great enough that the full platoon released during a green interval can fit, then an in-line bus stop won’t affect intersection capacity. But if the stop is closer to the intersection, then a stopped bus could result in the queue spilling back to the intersection, reducing capacity. However, this effect depends on the bus position in the queue – if the bus is the first vehicle in the discharge platoon, the spillback effect will be greatest, while if the bus is the last vehicle in the platoon, there will be no spillback effect. Using spatial queuing, the time-space diagram is drawn for a cycle in which a bus passes through during an effective green interval (g) as vehicle number n (Figure 7.2). A few assumptions are made: - Because the objective is to measure capacity impact, the lane used by the bus is assumed to operate at capacity, meaning that during the whole period of effective green, vehicles

(including the bus) discharge at saturation headway (hsat). - All the traffic is going in the same direction as the bus. No vehicle turns off during the bus effective green phase, and no vehicles turns onto the street during other phases.

- The road is one lane per direction, and passing the bus is not allowed. - Instantaneous start and stop is assumed for all vehicles, together with a short fixed lost time that accounts for deceleration and acceleration delay. - The bus is equally likely to pass through the intersection in any position in the queue during a given cycle. - The bus blockage time (b) is a constant value, and is the summation of dwell time and other lost time components, such as acceleration and deceleration lost time, waiting for doors to open and close, and time to start rolling. - The blockage time is shorter than the effective red. So, the red period helps the intersection recover from the blockage effect for the following cycle so that it does not propagate. - The distance of “do not block the intersection” is zero. This assumption will later be adjusted.

The start of diagram is the start of effective green. At time zero, the queue including the bus starts to discharge. The bus moves forward to the stop, which is located at setback S downstream. It stops in-line to serve passengers and blocks the only travel lane, delaying all the cars that are behind it. That blockage functions as a red light downstream of the intersection and creates a forming queue shock wave that moves backwards toward the original intersection.

bus b S hsat Lveh

t g TB TA

Figure 7.2 – Time space diagram with bus blockage

The reach of the forming queue behind the bus depends on how many cars have been able pass through the intersection during the effective green after the bus. In other words, the position of the bus in the discharged platoon determines the number of cars in queue. Capacity loss occurs when the queue reaches back to the intersection, at which point no additional cars can get through.

Based on whether the entire period of blockage occurs during effective green or not, three different scenarios may happen:

Scenario 1 – The reach of the queue does not extend to the intersection during effective green: No blockage will occur in this scenario. This happens when the bus goes through the intersection toward the end of the effective green, or the setback is far. In Figure 7.2, this translates into the case where the projected forming shockwave reaches the intersection (TB) after the green time ends. So, the boundary condition for this scenario is when the projected shockwave reaches the intersection just when the effective green ends: 푆 푇퐵 = 푛 × ℎ푠푎푡 + × ℎ푠푎푡 = 푔 퐿푣푒ℎ where: n = position of the bus in the platoon hsat = saturation headway (s) S = setback distance of far-side bus stop (ft)

Lveh = spacing of the vehicles while in queue (ft)

Solving for n, we find the position of the bus that creates this scenario: 푔 푆 푛퐵 = − ℎ푠푎푡 퐿푣푒ℎ Obviously, the position of the bus cannot be any value less than 1, and cannot be more than nmax, where nmax is the position of the last vehicle that goes through in a cycle:

푔 푛푚푎푥 = ⌊ ⌋ ℎ푠푎푡

However, later we will show that negative values for nB translates as no capacity impact is possible, regardless of the position of the bus. So, the value of nB is not limited to non-negative values.

Scenario 2 – The queue reaches the intersection and the blockage ends while still in effective green: Maximum possible blockage occurs. This scenario happens when the bus is near the head of the platoon, or the setback is very close. In Figure 7.2, this occurs when both the forming shockwave and recovery shock wave reach the intersection during effective green.

The time when recovery shockwave reaches the intersection is TA, the point at which the congestion will be relieved. The boundary condition will be TA occurs just when effective green ends: 푆 푇퐴 = 푛 × ℎ푠푎푡 + × ℎ푠푎푡 + 푏 = 푔 퐿푣푒ℎ where: b = bus blockage time (s)

At this boundary condition, the bus is in the following position: 푔 − 푏 푆 푛퐴 = − ℎ푠푎푡 퐿푣푒ℎ

Same as what mentioned for nB, negative values for nA are physically meaningless.

However, they imply that maximum blockage cannot occur. So, we allow nA to take negative values. Conditions will later be applied when calculating lost time.

Scenario 3 – The queue reaches the intersection, but the blockage ends after the end of effective green: In this scenario the blockage only partially impacts the effective green. The position of the bus must be anywhere between Scenario 1 and Scenario 2 for partial blockage to happen.

As the position of the bus moves away from nA toward nB. the capacity effect linearly decreases from maximum blockage (b) to minimum, which is zero. So, for a given effective green and setback, and positive values for nA and nB, the lost time – that is, green time that cannot be used due to blockage from a stopped bus – changes based on the position of the bus in the platoon according to Figure 7.3.

Lost time

n nA nB nmax

Figure 7.3 – Blockage vs. bus position diagram

To find the overall lost time effect of the bus stop with respect to variability in position of the bus (n), the expected value must be taken over all possible scenarios in cycles which a bus arrives. Based on the set of nA and nB calculated, one of the three following conditions will hold true, which gives the expected value of lost time given a bus arriving:

푛퐴 푛퐵 − 푛퐴 ( + ) × 푏 , 0 ≤ 푛퐴 ≤ 푛퐵 푛푚푎푥 2푛푚푎푥 퐸[퐿표푠푡 푡𝑖푚푒|퐵푢푠] = 푛퐴 (1 + ) × 푏 , 푛퐴 < 0 < 푛퐵 푛퐵 − 푛퐴 {0 , 푛퐴 < 푛퐵 ≤ 0 and 퐸[퐿표푠푡 푡𝑖푚푒] = 퐸[퐿표푠푡 푡𝑖푚푒|퐵푢푠] × 푞 where: q = bus frequency per cycle (1/cycle)

For practical cases, q is expected to be sufficiently less than 1 that there is scant possibility of two buses in the same cycle. The capacity loss can be expressed as a ratio, between 0 and 1 (no loss, 100% loss). Because discharge capacity is proportional to effective green time, and lost time due to bus blockage takes away from effective green, capacity loss can be expressed in terms of how much effective green time is lost. We indicate capacity effect as a factor indicating how much capacity is not lost: 퐸[퐿표푠푡 푡𝑖푚푒] 푐푎푝푎푐𝑖푡푦 푟푎푡𝑖표 = 1 − 푔 Next, two adjustments can be applied easily to the above formulas, which widen the application of this theory to cover more realistic conditions:

- The setback value can be easily adjusted to effective setback Seff for the case in which a “do not block the intersection” restriction is followed. In that case, lost time occurs when the queue reaches a distance d, which is intersection width, from the stop line. The adjustment will work as if the discharge start and end time are projected forward in the future to account for that distance. In other words, we can imagine the control point moved to the downstream edge of

the intersection and use the exact same formulas, replacing S with Seff , given by:

푆푒푓푓 = 푆 − 푑 - If the effective green is shorter than the blockage time (b), then the maximum possible blockage will be limited to effective green (g). To avoid meaningless negative results, b can be replaced everywhere by g in such a case.

7.3.3. Software Simulation

To validate the theory developed, the results of the theory was compared against VISSIM software simulation. To model the simulation, the following were considered:

- The demand volume was set high enough to ensure the intersection always has a full queue behind the stop line. With this setting, discharge flow rate equals capacity. - Other than transit buses, all vehicles are passenger vehicles. No heavy vehicle was modeled. - The speed for buses and passenger vehicles was set to 25 mph. - The bus was introduced to the model from an exclusive link, then merged into the traffic far enough upstream that the merge did not affect discharge at the stop line. - Buses are generated at regular intervals, but before merging into the traffic they dwelled at a dummy stop whose dwell time follows a uniform distribution from zero to cycle length (C). That ensures that buses are equally likely to arrive at any time in a cycle. - The dwell time was constant and equal to 15 seconds in all circumstances. - Bus acceleration and deceleration rates are set to 4 ft/s2 and 7.5 ft/s2, respectively. - Bus frequency is 1 every 5 cycles.

The simulation was run with different values for green interval and effective setback. Values for green interval were 30, 40, and 50 seconds, and effective setback ranged from 0 to 600 ft in 50- ft increments. For each set of values, the simulation was run for 10.5 hours continuously with one random seed. With frequency being 1 every 5 cycles and cycle length being 80 seconds, a total of 95 buses passed through the intersection. As the blockage effect ended every cycle, each of the buses had independent effect on the system. Also, the following parameters were measured in simulation and inserted in formula to ensure consistency:

- Blockage time: The dwell time was set to 15 seconds, but as mentioned, blockage time in the formula consists of other lost times as well including effects of deceleration, acceleration, and door opening and closing. Total blockage was measured in the simulation to be 23.5 seconds. This value of blockage was used in the formula. - Saturation flow rate: The simulation was first run without interference of bus blockage. The saturation flow rate was measured to be 1.8 seconds. - Effective green: For each value of green time tested, effective green was measured as the proportion of throughput in the presence of signal in relation to the throughput without signal.

7.3.4. Results and Conclusions

A strong agreement was observed between the results from formula and the simulation (Table 7.1 and Figure 7.3). It is observed that for a given blockage time and setback, longer green time means less capacity loss. Also, the capacity loss for the input values tested is not severe. Within the range of values tested, the maximum loss is limited to 10%, which occurs for green interval of 30 seconds and effective setback of 0. This is very promising result in the application of these bus stops in a single-lane road. Some reasonable values that may apply in many places in a city like Boston are bus frequency of 1 every 5 cycles, 15 seconds of dwell time, cycle length of 80 seconds, and green interval of 30 to 50 seconds. That means the use of far-side, in-line bus stops near those intersections is not likely to cause severe disruptions. For the case of 30-second effective green, the bus stop has no impacts on the capacity if it is more than 400 ft away from the intersection. (This value is 450 ft for 40-second green, and 600 ft for 50-second green.) To limit capacity loss to 2%, the critical bus stop setback is roughly 200, 300, and 375 ft, respectively, for green intervals of 30, 40, and 50 second. Table 7.1 – Comparison of capacity factor from theory and from VISSIM simulation green interval = 30 s green interval = 40 s green interval = 50 s Setback (ft) Theory VISSIM Theory VISSIM Theory VISSIM

0 0.907 0.916 0.916 0.915 0.925 0.936

50 0.926 0.929 0.928 0.928 0.933 0.939

100 0.943 0.940 0.939 0.947 0.941 0.943

150 0.960 0.959 0.951 0.960 0.948 0.945

200 0.975 0.979 0.959 0.968 0.956 0.952

250 0.986 0.980 0.972 0.981 0.964 0.962

300 0.994 0.990 0.981 0.990 0.971 0.968

350 0.998 0.994 0.989 0.993 0.978 0.976

400 1.000 0.999 0.994 0.995 0.985 0.985

450 1.000 0.999 0.998 0.998 0.990 0.991

500 1.000 0.999 1.000 1.000 0.995 0.995

550 1.000 0.999 1.000 0.999 0.998 0.996 600 1.000 1.000 1.000 1.000 0.999 0.998

g=30 s, b=23.5 s, q=1/5cycle 1.000

0.980

0.960 Theory 0.940 VISSIM

0.920

0.900 0 100 200 300 400 500 600

(a)

g=40 s, b=23.5 s, q=1/5cycle 1.000 0.990 0.980 0.970 0.960 Theory 0.950 VISSIM 0.940 0.930 0.920 0.910 0 100 200 300 400 500 600

(b)

g=50 s, b=23.5 s, q=1/5cycle 1.000 0.990 0.980 0.970 0.960 Theory 0.950 VISSIM 0.940 0.930 0.920 0 100 200 300 400 500 600

(c) Figure 7.3 – Comparison of capacity factor from theory and from VISSIM simulation a) green interval = 30s b) green interval = 40s c) green interval = 50s

7.3.5. Knowledge Gaps, Applications, and Recommendations for Future Research

This study assumes that all the traffic is carried on the same lane and direction as the bus (we call that main street here). In other words, when the signal for the main street turns red (which in reality means the cross street receives a green), no car will be added to the back of the queue, which does not reflect real-life situation because vehicles from the cross may turn on to the main street. However, the assumption that no car will turn off from the main street compensates the unrealistic effect of the previous assumption to some extent. With this in mind, if this model is used for an intersection where more cars turn onto the main street that turn off from it, this model may overestimate the capacity. Conversely, if more cars turn off than turn onto the main street, this model underestimates the capacity. While this model may provide a good understanding of how intersection capacity can be affected, it might be possible to develop more comprehensive and detailed models that account for vehicles turning onto and off of the main street. At a certain point, however, efforts to make a model more general make it so complex that it becomes simpler to simply model and intersection of interest realistically. For the Cummins Hwy case study, only two bus stops are in-line. Both are located within 50 ft downstream of Wood Ave/Harvard St intersection on either direction. At this intersection which flares are provided for both EB and WB traffic, the bus blocks the auxiliary lane when serving passengers, and the traffic on the continuous through lane can still flow. In other places, bus bays are provided by prohibiting parking, and the bus does not interfere with car flow while stopped.

REFERENCES

1- TCRP Report 19: Guidelines for Location and Design of Bus Stops. Research Board, Washington D.C., 1996. 2- Fitzpatrick, K., and R. L. Nowlin. Effects of Bus Stop Design on Suburban Arterial Operations. Transportation Research Record: Journal of the Transportation Research Board, No. 1571, 1997. https://doi.org/10.3141/1571-05 3- Wong, S. C., H. Yang, W. S. Au Yeung, S. L. Cheuk, and M. K. Lo. Delay at Signal- Controlled Intersection with Bus Stop Upstream. Journal of Transportation Engineering, Vol. 124(3), pp. 229-234, 1998. 4- Furth, P. G., and J. L. SanClemente. Near Side, Far Side, Uphill, Downhill: Impact of Bus Stop Location on Bus Delay. Transportation Research Record: Journal of the Transportation Research Board, No. 1971, 2006, pp. 66-73. https://doi.org/10.1177/0361198106197100108 5- Meng, Q., and X. Qu. Bus Dwell Time Estimation at Bus Bays: A Probabilistic Approach. Transportation Research Part C. Vol. 36, 2013, pp. 61-71. https://doi.org/10.1016/j.trc.2013.08.007 6- Arhin, S., E. Noel, M. F. Anderson, L. Williams, A. Ribisso, and R. Stinson. Optimization of Transit Total Bus Stop Time Models. Journal of Traffic and Transportation Engineering, Vol. 3(2), pp. 146-153, 2016. https://doi.org/10.1016/j.jtte.2015.07.001 7- Gu, W., V. V. Gayah, M. J. Cassidy, and Nathalie Saade. On the Impacts of Bus Stops Near Signalized Intersections: Models of Car and Bus Delays. Transportation Research Part B. Vol. 68, 2014, pp. 123-140. https://doi.org/10.1016/j.trb.2014.06.001 8- Liu, X., Y. Yang, M. Meng, and A. Rau. Impact of Different Bus Stop Designs on Bus Operating Time Components. Journal of Public Transportation, Vol 20, No. 1, 2017, pp. 104-118. DOI: 10.5038/2375-0901.20.1.6 9- Luthy, N., S. I. Guler, and M. Menendez. System Wide Effects of Bus Stops: Bus Bays vs. Curbside Bus Stops. Transportation Research Board 95th Annual Meeting, Washington D. C., 2016

10- Cvitanic, D. Joint Impact of Bus Stop Location and Configuration on Intersection Performance. Promet - Traffic & Transportation, Vol. 29, No. 4, 2017, pp. 443-454. DOI: 10.7307/ptt.v29i4.2338

APPENDIX

Cummins Highway Road Diet 90 7 from Mattapan Square to Wood Ave/Harvard St Intersection (Index Map) 6 Wood Ave/Harvard St Intersection 5 4 3

Mattapan Square N 91

Median and Sidewalk Level 1 Cycle Track 0 100 ft N 92

Median and 2 Sidewalk Level

Cycle Track 0 100 ft N 93

3 Median and Sidewalk Level

Cycle Track 0 100 ft N 94

Median and Sidewalk Level

Cycle Track 0 100 ft N 95

Median and Sidewalk Level

Cycle Track 0 100 ft N 96 6

Median and Sidewalk Level

Cycle Track 0 100 ft 7 N 97

Median and Sidewalk Level

Cycle Track 0 100 ft Timings 98 3: Cummins Hwy & Woodhaven St 02/13/2019

Lane Group EBL EBT WBT SBL Lane Configurations Traffic Volume (vph) 114 801 450 101 Future Volume (vph) 114 801 450 101 Turn Type Prot NA NA Prot Protected Phases 5 1 6 3 Permitted Phases Detector Phase 5 1 6 3 Switch Phase Minimum Initial (s) 5.0 5.0 5.0 5.0 Minimum Split (s) 10.0 20.0 20.0 20.0 Total Split (s) 11.0 37.0 26.0 23.0 Total Split (%) 18.3% 61.7% 43.3% 38.3% Yellow Time (s) 3.0 3.0 3.0 3.0 All-Red Time (s) 1.0 1.0 1.0 1.0 Lost Time Adjust (s) 0.0 0.0 0.0 0.0 Total Lost Time (s) 4.0 4.0 4.0 4.0 Lead/Lag Lead Lag Lead-Lag Optimize? Yes Yes Recall Mode Max Max Max Max Act Effct Green (s) 7.0 33.0 22.0 19.0 Actuated g/C Ratio 0.12 0.55 0.37 0.32 v/c Ratio 0.62 0.87 0.42 0.52 Control Delay 41.5 24.1 14.6 11.1 Queue Delay 0.0 0.0 0.0 0.0 Total Delay 41.5 24.1 14.6 11.1 LOS DCBB Approach Delay 26.3 14.6 11.1 Approach LOS CBB Intersection Summary Cycle Length: 60 Actuated Cycle Length: 60 Offset: 0 (0%), Referenced to phase 2: and 6:WBT, Start of Green Natural Cycle: 60 Control Type: Pretimed Maximum v/c Ratio: 0.87 Intersection Signal Delay: 20.2 Intersection LOS: C Intersection Capacity Utilization 67.0% ICU Level of Service C Analysis Period (min) 15

Splits and Phases: 3: Cummins Hwy & Woodhaven St

Cummins Hwy @ Woodhaven P.M. 5% inflated 05/10/2018 Baseline Synchro 10 Report Page 1 Timings 99 3: Cummins Hwy & Itasca St 02/13/2019

Lane Group EBL EBT WBT SBL Lane Configurations Traffic Volume (vph) 66 836 540 106 Future Volume (vph) 66 836 540 106 Turn Type Prot NA NA Prot Protected Phases 5 1 6 3 Permitted Phases Detector Phase 5 1 6 3 Switch Phase Minimum Initial (s) 5.0 5.0 5.0 5.0 Minimum Split (s) 10.0 20.0 20.0 20.0 Total Split (s) 10.0 38.0 28.0 22.0 Total Split (%) 16.7% 63.3% 46.7% 36.7% Yellow Time (s) 3.0 3.0 3.0 3.0 All-Red Time (s) 1.0 1.0 1.0 1.0 Lost Time Adjust (s) 0.0 0.0 0.0 0.0 Total Lost Time (s) 4.0 4.0 4.0 4.0 Lead/Lag Lead Lag Lead-Lag Optimize? Yes Yes Recall Mode Max Max Max Max Act Effct Green (s) 6.0 34.0 24.0 18.0 Actuated g/C Ratio 0.10 0.57 0.40 0.30 v/c Ratio 0.41 0.88 0.87 0.36 Control Delay 33.1 24.3 32.5 13.5 Queue Delay 0.0 0.0 0.0 0.0 Total Delay 33.1 24.3 32.5 13.5 LOS CCCB Approach Delay 24.9 32.5 13.5 Approach LOS CCB Intersection Summary Cycle Length: 60 Actuated Cycle Length: 60 Offset: 0 (0%), Referenced to phase 2: and 6:WBT, Start of Green Natural Cycle: 60 Control Type: Pretimed Maximum v/c Ratio: 0.88 Intersection Signal Delay: 26.3 Intersection LOS: C Intersection Capacity Utilization 61.1% ICU Level of Service B Analysis Period (min) 15

Splits and Phases: 3: Cummins Hwy & Itasca St

Cummins Hwy @ Itasca P.M. 5% inflated 05/10/2018 Baseline Synchro 10 Report Page 1 Timings 100 3: Wood Ave/Harvard St & Cummins Hwy 02/13/2019

Lane Group EBL EBT WBT NBL NBT SBL SBT Lane Configurations Traffic Volume (vph) 99 775 467 80 223 180 280 Future Volume (vph) 99 775 467 80 223 180 280 Turn Type Prot NA NA Perm NA pm+pt NA Protected Phases 5 1 6 4 3 7 Permitted Phases 4 7 Detector Phase 5 1 6 4 4 3 7 Switch Phase Minimum Initial (s) 5.0 5.0 5.0 5.0 5.0 5.0 5.0 Minimum Split (s) 10.0 20.0 20.0 20.0 20.0 20.0 20.0 Total Split (s) 12.0 32.0 20.0 28.0 28.0 20.0 48.0 Total Split (%) 15.0% 40.0% 25.0% 35.0% 35.0% 25.0% 60.0% Yellow Time (s) 3.0 3.0 3.0 3.0 3.0 3.0 3.0 All-Red Time (s) 1.0 1.0 1.0 1.0 1.0 1.0 1.0 Lost Time Adjust (s) 0.0 0.0 0.0 0.0 0.0 Total Lost Time (s) 4.0 4.0 4.0 4.0 4.0 Lead/Lag Lead Lag Lag Lag Lead Lead-Lag Optimize? Yes Yes Yes Yes Yes Recall Mode Max Max Max Max Max Max Max Act Effct Green (s) 8.0 28.0 16.0 24.0 44.0 Actuated g/C Ratio 0.10 0.35 0.20 0.30 0.55 v/c Ratio 0.63 0.79 0.83 0.84 0.78 Control Delay 52.3 28.1 41.0 47.1 21.4 Queue Delay 0.0 0.0 0.0 0.0 0.0 Total Delay 52.3 28.1 41.0 47.1 21.4 LOS DCD D C Approach Delay 30.5 41.0 47.1 21.4 Approach LOS CD D C Intersection Summary Cycle Length: 80 Actuated Cycle Length: 80 Offset: 0 (0%), Referenced to phase 2: and 6:WBT, Start of Green Natural Cycle: 75 Control Type: Pretimed Maximum v/c Ratio: 0.84 Intersection Signal Delay: 33.0 Intersection LOS: C Intersection Capacity Utilization 76.8% ICU Level of Service D Analysis Period (min) 15

Splits and Phases: 3: Wood Ave/Harvard St & Cummins Hwy

Cummins Ave @ Harvard St/Wood Ave P.M. Oct 05/10/2018 Baseline Synchro 10 Report Page 1