JOURNAL OF TRANSPORTATION

JOURNAL OF TRANSPORTATION A

JOURNAL OF TRANSPORTATIONof the Institute of Transportation Engineers

Advancing transportation knowledge and practices for the benefit of society

VOLUME 6, ISSUE 1

Senior EditorDr. Lester A. Hoel, P.E., FITEL.A. Lacy Distinguished Professor of Engineering Emeritus University of Virginia

Assistant Senior EditorDr. Dennis L. Christiansen, P.E., FITEDirector Texas A&M Transportation Institute Texas A&M University System

Dr. Max DonathProfessor of Mechanical Engineering and Director, Intelligent Transportation Systems Institute University of Minnesota

Dr. Adib KanafaniProfessor of the Graduate SchoolUniversity of California, Berkeley

Dr. John M. Mason, Jr., P.E., FITEVice President for Research and Economic Development Auburn University

Dr. Kumares C. Sinha, P.E., FITEOlson Distinguished ProfessorSchool of Civil EngineeringPurdue University

Dr. Joseph M. SussmanJR East Professor of Civil and Environmental Engineering and Engineering Systems Massachusetts Institute of Technology

Dr. C. Michael Walton, P.E., FITEProfessor of Civil EngineeringErnest H. Cockrell Centennial Chair in Engineering Civil Engineering Department University of Texas at Austin

Marianne SaglamManaging Editor

Courtney DayEditorial Assistant

Dr. Ardeshir FaghriProfessor & Director of Delaware Center For TransportationUniversity of Delaware

Dr. Jon D. Fricker, P.E. (FITE)ProfessorPurdue University

Dr. Peter G. FurthProfessorNortheastern University

Dr. James L. Gattis, P.E.ProfessorUniversity Of Arkansas

Dr. Bruce N. JansonProfessorUniversity of Colorado

Dr. B. Kent Lall, P.E.ProfessorPortland State University

Dr. Thomas Mulinazzi P.E., FITEProfessorUniversity of Kansas

Dr. Errol C. Noel, P.E., FITEProfessorHoward University

Dr. Martin T. Pietrucha, P.E., FITEDirector, The Larson InstitutePennsylvania State University

Dr. Mitsuru Saito, P.E., FITEProfessorBrigham Young University

Dr. William J. Sproule, P.Eng., FITEProfessorMichigan Tech University

Dr. William C. Taylor, P.E.ProfessorMichigan State University

Rhonda K. Young, P.E.Associate ProfessorUniversity of Wyoming

Dr. William Young, P.E., FITEProfessorMonash University

ASSOCIATE EDITORS

STAFF

EDITORIAL BOARD

The Journal of Transportation of the Institute of Transportation Engineers is circulated electronically yearly by the Institute of Transportation Engineers, 1627 Eye Street, NW, Suite 600, Washington, DC 20006 USA, phone: 202-785-0060. ©2014 Institute of Transportation Engineers. All rights reserved except for brief quotation with attribution. The views expressed in the articles are those of the authors and do not reflect official ITE policy unless so stated. ISSN 2162 1616. Journal of Transportation of ITE is distributed free electronically to ITE members. Issues may be downloaded at www.ite.org/JOT.

INSTITUTE OF TRANSPORTATION ENGINEERS 1627 Eye Street, NW, Suite 600, Washington, DC 20006 USA P: 202.785.0060 F: 202.785.0609 www.ite.org

VOLUME 6 | NUMBER 1 | JULY 2014

JOURNAL OF TRANSPORTATIONof the Institute of Transportation Engineers

Comparing Perceived and Actual Travel Time 1 Savings on Freeways with Managed LanesBy Prem Chand Devarasetty, Ph.D., Mark Burris, Ph.D., P.E., and Huang Chao, Ph.D., P.E.

Transforming the Washington, DC, USA, 15

Parking Meter Program Using a Lean Six Sigma Improvement ProcessBy Soumya S. Dey, P.E., PMP

Public Perception of Double Crossover 33 Diamond InterchangesBy Kristy Jackson, Christopher Cunningham, P.E., Chunho Yeom, Joseph E. Hummer, Ph.D., P.E., and Adam Kirk, P.E., PTOE, AICP

A Methodology Using GPS to Inventory 55 University Campus ParkingBy Brian Maleck, Wayne A. Sarasua, Ph.D., P.E., Jennifer Ogle, Ph.D., P.E., and Kweku Brown

The Effect of Additional Lane Length on 71 Double-Lane Roundabout OperationBy Samuel Hammond, Christopher Hunter, Ph.D., and Kevin Chang, Ph.D., P.E.

CONTENTS

JOURNAL OF TRANSPORTATION 1

Respondents to a 2010 revealed preference survey were asked questions related to their recent travel on Houston’s Katy Freeway (Houston, TX, USA). One question asked if they experienced any travel time savings by using the new Katy Managed Lanes (MLs). This study examined any difference between their perceived and actual travel time savings. This study found that travelers overestimated the travel time savings they experienced by traveling on the MLs by a factor of 4. The magnitude of misperception varied with individuals, but the average individual saved approximately 3 minutes and thought they saved nearly 12 minutes. Linear regression models were fit to model the misperception of the travel time savings and found that both trip characteristics and respondent socioeconomic characteristics had an effect on the magnitude of misperception of travel time savings. Respondents’ trip purpose, gender, and income were found to be significant predictors of how well they estimated their travel time savings.

Comparing Perceived and Actual Travel Time Savings on Freeways with Managed LanesBy Prem Chand Devarasetty, Ph.D., Mark Burris, Ph.D., P.E., and Huang Chao, Ph.D., P.E.

IntroductionThe emergence of managed lanes (MLs) in the United States provides transportation planners and researchers an opportunity to better understand how travelers value their time. Since MLs offer an uncongested, faster, paid alternative to general purpose lanes (GPLs), it is possible to observe the amount of toll travelers are willing to pay to save travel time on the MLs. This comparison has the advantage of both sets of lanes being in the same highway corridor with similar entry and exit opportunities. Even with this relatively straightforward comparison, there are still difficulties in determining the value travelers place on their travel time savings. Two of the larger issues include (1) the value travelers place on the added reliability of the MLs and (2) the fact that the amount of time

2 THE INSTITUTE OF TRANSPORTATION ENGINEERS

the travelers think the MLs save them may differ from reality. This paper takes an in-depth look at the second issue.

This research compared data collected on actual travel conditions along Katy Freeway to data obtained as part of a survey of Katy Freeway travelers. The survey data included questions regarding the respondent’s most recent trip on the Katy Freeway along with the amount of time they thought the MLs saved them plus several stated preference questions.

Literature ReviewUnderstanding traveler behavior, and particularly travelers’ willingness to pay for travel time savings, has been an important area of research going back to the late 1960s.1–3 Some of the early efforts to determine travelers’ value of time focused on revealed preference data, including the choice between modes.2,4 However, those comparisons clearly have many additional, and critical, differences other than just cost and time.

More recent efforts have frequently used stated preference data from surveys to estimate value of time, controlling for as many other factors as possible.5–9 In the stated preference approach, several hypothetical travel scenarios or travel alternatives with different travel times and trip costs are presented to respondents, and they are asked to choose the alternative that best suits their travel. From the stated preference data, discrete choice models are fit to estimate the respondents’ willingness to pay.

Both revealed preference and stated preference approaches have their strengths and weaknesses.10

Since the revealed preference approach relies on the revealed responses, it is impossible to get revealed preference data on a facility that does not exist, whereas in the stated preference approach, since the scenarios can be hypothetical, it is possible to find travelers’ preference among choices that include a nonexistent facility. However, this is also a major weakness of the stated preference approach, since the stated preference approach includes hypothetical choices and not actual choices. Making the actual decision is more complex compared to answering a hypothetical scenario, since the consequences of the decision (benefits or loss) are experienced.10

There is abundant literature with both revealed preference and stated preference approaches used to estimate travelers’ willingness to pay for travel time savings.4,5,11–14 Value of the estimated willingness to pay varied by the type of the approach used. There were a few willingness-to-pay studies conducted on MLs, but some of those found that the median willingness-to pay-values estimated using stated preference data were approximately half the values estimated using revealed preference data.5,13 Two possible explanations were given by Brownstone and Small for this underestimation of willingness to pay by stated preference approaches.15 One explanation is that people have time inconsistency in their actual behavior, but not in their hypothetical behavior. In actual behavior, people may generally prefer to choose the free lane, but other time constraints may come into play and they end up needing to take the toll lane. However, they overlook those constraints while answering the stated preference questions and thus choose the toll road more often in real life than in hypothetical stated preference questions. In other words, in the stated preference question the traveler answers based on their typical trip. But in real life some of the trips may be hurried or pressed for time. On those few trips they do choose the toll lane.6

The other possible explanation is one of the major criticisms against the revealed preference approach. Travelers have difficulty reporting their actual travel times and travel time savings. Rather, they tend to over- or underestimate the travel time savings they experienced. In the studies conducted on SR 91 and I-15, when travelers were asked to report the travel time savings they obtained when using the express lanes, their responses were typically about twice as much as the average travel time savings


in these lanes.15,16 This seems to indicate that people tend to overestimate their travel time savings. However, these previous studies used the average time saved compared to the reported time saved. The studies did not have information on how much time was saved during the exact date and time the trip was taken. It is possible, although unlikely, that the travelers reporting their travel time savings did indeed save as much time as they thought they did, and they just happened to travel during some pockets of extreme congestion. Our research removed this potential explanation by comparing the actual travel time of travelers on the Katy Freeway during the time of their trip to their estimate of time saved.

Steimetz found that travelers give a value for congested travel time that is twice as much as for noncongested travel time.17 This may be a possible explanation for overestimation of the travel time savings. Psychologists also argue that the misperception of time depends on the magnitude of the duration. Fortin and Rousseau stated that short intervals of time tend to be overestimated and long intervals of time tend to be underestimated.18 Since the revealed preference approach is a common method of estimating the willingness to pay for travel time savings, there is a need to better understand the travel time misperception among the travelers.

ObjectivesThe main objective of this research is to examine the difference between travelers’ perceived travel time savings and their actual travel time savings. An effort will also be made to understand if the magnitude of the misperception (over- or underestimation) of travel time savings is dependent on trip characteristics and/or traveler characteristics.

This study uses data collected from a 2010 survey conducted on Katy Freeway travelers in Houston and detailed traffic data collected on the roadway. Details on the survey data and actual ML usage are described in the next section.

DataThe Katy Freeway is an eight-lane road with a three-lane one-way frontage road in each direction (see Figure 1). In addition to these lanes, a portion (12-mile stretch) of the Katy Freeway near downtown was designed with two MLs in each direction.19 The 12-mile Katy Freeway MLs extend from west of SH6 to the I-10/I-610 interchange (see Figure 2). The MLs are open to both single-occupant vehicles (SOVs) and high-occupancy vehicles (HOVs). The SOVs pay a toll to use the MLs at all times, while HOVs only pay a toll during off-peak hours.


Figure 1. Katy Freeway (a crash occurred downstream causing this congestion, which also allowed for easy identification of the lanes).

Figure 2. Katy Freeway managed lanes and automatic vehicle identification sensor locations.

Source: TxDOT. Katy Freeway Website, Texas Department of Transportation. Available: www.katyfreeway.org. [Accessed May 15, 2009.]


2010 Katy Freeway Survey

The survey was posted on a Texas A&M Transportation Institute server and was made available for public access through the Katy Freeway website. The data collection process started on June 1, 2010, and continued until July 15, 2010. Residents of Houston who use the Katy Freeway on a regular basis or have used it recently were encouraged to participate in the survey. The existence of the survey was advertised to the public through online and news media. To increase participation in the survey, two gas cards worth $250 each were given to two randomly chosen respondents. In addition to the website ads, the Harris County Toll Road Authority added a brief note regarding the existence of the survey to its monthly account e-notices. E-mails were also sent to the 3,077 respondents from a previous (2008) Katy Freeway survey who had indicated an interest in participating in a follow-up survey.20 A total of 4,919 responses were obtained from the survey. However, only 3,325 of those 4,919 responses were completed to a point where they were useful for analysis.

The 2010 survey questionnaire consisted of five sections. The first section asked the respondents details about their most recent trip on the Katy Freeway. About half of the respondents were asked about their trip toward downtown Houston and the other half about their trip away from downtown. Respondents were asked what day of the week the trip occurred on, when the trip began, when it ended, where the respondent got on and off the Katy Freeway, the type of vehicle, etc. Since the survey was administered online, the date the survey was taken was also known. Knowing the date the survey was taken, combined with the day of week and time of day the respondents indicated their last trip occurred, means we know the most likely date and time of their most recent trip. There is no guarantee that this is the correct date of the trip, since the questionnaire asked travelers to indicate the day of the week of their most recent trip. For many, that would be the most recent day of the week indicated in the survey (for example, the most recent Monday). But for infrequent Katy Freeway travelers that may have been 2 or 3 weeks prior. Based on their frequency of use of the Katy Freeway (see Table 1), more than 96 percent of respondents traveled on the Katy Freeway during the most recent week, and therefore their day and time of travel was known.

Table 1. Survey respondents’ frequency of Katy Freeway travel.

Number of Trips on the Katy Freeway during the Last Full Week

Percentage of Respondents

0 3.3

1 3.9

2 4.7

3 2.9

4 0.1

5 19.6

7 0.1

8 1.6

10 23.4

12 1.4

15 21.9

20 or more 17.1

In the second section, respondents were introduced to the new MLs and were asked about their use of those lanes.14 The third section was intended to identify the risk-taking behavior or preferences


of the respondents with one risk-aversion question presented in the survey. In the fourth section, the respondents were presented with stated preference questions, and the last section consisted of questions regarding socioeconomic characteristics of the respondents.

Of the 3,325 respondents, 1,004 indicated that they had used the MLs on their recent trip. Although 1,004 respondents indicated that they used the MLs, 277 respondents’ entry and/or exit locations did not match ML entry or exit locations. Twenty respondents did not provide some of the information (entry and/or exit location, day of the most recent trip, start time of the trip) required to extract the actual travel time savings. Those respondents were removed from further analysis, leaving 707 responses from travelers who used the MLs on their most recent trip.

These 707 survey responses were then weighted to better reflect both the frequency of travel on the Katy Freeway and the percentage of trips that were on the MLs. Using the data from the automatic vehicle identification (AVI) sensors (see Figure 2), it was possible to get an estimate of how frequently drivers with transponders are observed using the Katy Freeway and what percentage of their trips are on the GPLs versus the MLs (see Table 2). This is only an estimate, since the AVI sensors do not capture 100 percent of the trips on the GPLs. Note that this sample of survey respondents comprised only those respondents who used the MLs. Therefore, only those travelers who actually used the MLs at least once during the period of this survey were included in the control totals for the weights.

In addition, we found 57 respondents who indicated 0 trips on the Katy Freeway in the prior week but did indicate trips on the Katy MLs. These respondents did not read or pay attention to the instructions in the question that indicated for them to count all trips on the Katy Freeway (GPLs or MLs). Another 7 respondents did not include their number of trips, and 28 respondents indicated 0 trips on the MLs. To get an accurate weight, these respondents also had to be removed, leaving 617 responses. Despite losing some data and the inherent problems with AVI sensors on the GPLs, weighting the survey sample to better reflect the observed (AVI) data should improve the survey’s ability to represent all ML travelers based on their frequency of Katy Freeway use and percentage of ML use. However, note that the same analyses were run on unweighted data, and the results were nearly identical to the results shown in this paper that are based on weighted data.

Table 2. Weighting of survey responses.

Observed Percentage/Survey Percentage of Travelers = Weight

Katy Freeway Trips on the MLs

Katy Freeway usage (All Lanes) Low (1 trip/week) Medium (2 to 4 trips/week) High ( 5+ trips/week)

Low (1 or 2 trips/week) 53.8/12.0 = 4.49a 0.4/7.3 = 0.05 0.00/0.2 = 0.00

Medium (3 to 9 trips/week) 29.3/3.6 = 8.22 7.3/17.8 = 0.41 2.2/10.9 = 0.20

High (10+ trips/week) 2.9/1.6 = 1.78 2.1/8.1 = 0.25 2.0/38.8 = 0.05

aAs an example, 53.8 percent of Katy Freeway travelers used the Katy Freeway (any lane) one or two times per week and the Katy MLs once per week. The survey included 12 percent of respondents who indicated that frequency of Katy Freeway use. Therefore, that group of survey respondents had a weight of 4.49. Remember, for both the survey and for the real-world data we have limited the sample to those travelers who used the MLs at least once.

Observed Traffic Data

Two types of vehicle sensors—Wavetronix and AVI—are installed along the Katy Freeway by Texas Department of Transportation (TxDOT) (see Figure 2 for AVI sensor locations). These sensors


collect data on the speed and volume on all the lanes on the Katy Freeway. These data were used to estimate the actual travel time savings offered by the MLs.

Travel Time

Time taken to travel various sections of the Katy Freeway on the MLs and GPLs was calculated using the AVI data. AVI sensors are located along the MLs and the GPLs in each direction of travel on the Katy Freeway (see Figure 2). Each AVI sensor identifies each transponder-equipped vehicle based on the vehicle’s unique transponder ID and records the time when the vehicle is identified. The vehicle IDs recorded at an AVI sensor are matched with the adjacent AVI sensor data, and the time difference each vehicle required to cover the distance between those sensors is known. There are more than 1 million instances per day recorded in Houston and approximately 170,000 observations per day on the Katy Freeway—more than sufficient to obtain reliable travel times. For each 15-minute period, the average travel time for all identified vehicles between those two sensors was calculated and recorded. These travel times were used to calculate the actual travel time savings offered by the MLs.

ResultsThe final data included the responses from the 617 survey respondents, and all analyses were weighted based on the weights shown in Table 2. In the survey, the respondents were asked how much travel time they thought they saved by traveling on the MLs. Their perceived travel time savings are presented in Table 3. Approximately 97 percent of the respondents reported that they had experienced some travel time savings. From the survey question regarding their most recent trip, we obtained their Katy Freeway entry and exit locations. This information was used to determine the total length of their trip and the length of their trip in the portion of the freeway that had MLs.

Table 3. Respondents’ perceived travel time savings.

Perceived Travel Time Savings Percentage of Respondents

0 min. 3.3

1 to 2 min. 8.6

3 to 5 min. 22.6

6 to 10 min. 25.1

11 to 15 min. 23.3

16 to 20 min. 8.1

21 to 25 min. 1.2

26 to 30 min. 6.6

More than 30 min. 1.2

Weighted actual averagea 11.8 min.

aNote that this weighted average uses the actual responses prior to the responses being grouped in bins as in this table.

As a comparison, from the observed travel time data, the average travel time savings (over all non-holidays for all of 2009) were estimated for the 11.4-mile section of the Katy Freeway for both the eastbound and westbound directions (see Figure 3). The average travel time savings varied by time of day, but the maximum average travel time savings was approximately 8 minutes in the westbound direction and approximately 6 minutes in the eastbound direction. Average travel time savings outside of the peak period was considerably less than 6 to 8 minutes.


Figure 3. Average travel time savings on the MLs by time of day for all of 2009.

This research assumes that these values are a representative sample of drivers’ true perceptions regarding the amount of time saved when using the MLs. However, when using survey data, there are always potential biases. In this survey there may be a sample bias consisting of travelers who really like using MLs because they think they save so much time—thus making them more likely to take the survey. There could also be a respondent bias, as respondents are justifying their own use of MLs by indicating an impressive amount of travel time savings. In an effort to minimize these potential impacts, (1) we examined the data for any responses that appeared to be from the same person (none were found), (2) the results were weighted to better reflect ML travelers (as discussed above), and (3) any respondents who indicated a time savings greater than 30 minutes were removed.

Next, the perceived travel time savings for respondents’ most recent trip (Table 3) (which were obtained from the survey) will be compared to their actual travel time savings. The actual travel time savings for that recent trip was obtained by using the respondents’ recent trip information (survey date, day of week of the most recent trip, entry and exit locations, and start time of the trip) provided in the survey and then looking up the actual travel time savings obtained from sensor data. For example, if a respondent had taken the survey on June 16, 2010, and indicated that their most recent trip was on a Friday and it started at 8:00 a.m., then the most likely date of his/her recent trip would be the previous Friday, that is, June 11, 2010. This was based on our assumption that since most of the respondents were frequent users of the Katy Freeway (see Table 1), their most recent trip reported on the survey likely occurred within the seven days prior to their taking the survey. In addition, many travelers did not travel the full length of the MLs. Only the travel time savings between the respondents’ entry to and exit from the Katy Freeway was included in the actual time savings.

The average actual travel time savings was only approximately 25 percent of the average perceived time savings. This is similar to what has been found in previous studies of MLs, although previous studies found the estimated actual time savings to be closer to 50 percent of perceived time savings.5,13

However, with the previous studies there is the possibility that those travelers happened to travel when the GPLs were particularly slow and thus may have saved as much time as they perceived.


Previous studies did not have access to actual trip times, so our study is able to investigate this difference in perceived versus actual time savings in greater depth.

A scatter plot of the perceived and the observed average travel time savings was created (see Figure 4). It can be seen that the perceived travel time savings is much higher than the average observed travel time savings. This plot shows the magnitude of the overestimation of the travel time savings and raises an interesting question: Can revealed preference responses be used for policy analysis? It can be seen that few respondents underestimated travel time savings and a majority of the respondents overestimated the travel time savings. The weighted average difference between perceived and observed travel time savings was around 8.8 minutes when all data are included (8.5 minutes when responses greater than 30 minutes are excluded; see the following discussion). (Note that the result for unweighted data excluding values greater than 30 minutes was 9.6 minutes, slightly larger than with weighted data.)

Figure 4. Perceived vs. average observed travel time savings.

To examine whether the magnitude of over- or underestimation of travel time savings is related to any of the respondents’ trip characteristics or their socioeconomic characteristics, a linear regression model was estimated (see Table 4 and Equation 1). A small number of respondents (approximately 1.2 percent) indicated that they saved more than 30 minutes by traveling on the MLs. However, more than 30 minutes of travel time savings for a 12-mile section seems extreme, and it can also be seen from Figure 4 that 99.9 percent of the time the observed travel time savings was less than 20 minutes. Therefore, those respondents who indicated that they saved more than 30 minutes were excluded from the model, leaving 597 respondents in the model.


The dependent variable for the model is the difference between the perceived and observed travel time savings. Trip purpose, distance traveled on the MLs, age, gender, income, direction of travel, education, and number of trips on the Katy Freeway in the previous week were considered as independent variables. However, many of the variables were found not to be significant in predicting the magnitude of the difference in perceived travel time savings. Misperception of the travel time savings was higher for female respondents than male respondents. Respondents on recreational trips perceived travel time savings that were more accurate than those respondents with other trip purposes. This might be because travelers who are traveling for recreation are generally not pressed for time and can more clearly internalize the difference in travel speeds. Or it may be because they more often travel at off-peak times, when the speeds on the two types of lanes are closer, making comparison easier. It is interesting to see that lower-income (annual household income less than $25,000) respondents did not overestimate the travel time savings as much as mid- and high-income respondents. A possible explanation for this might be that lower-income respondents are more carefully judging what value they are receiving (in time savings) for the toll that they pay, since it is a larger portion of their spending money.

Table 4. Linear regression model for difference between perceived and observed travel time savings.

Parameter Estimate p Value

Intercept 8.0 0.00

Annual income

< $25,000

-3.0 0.00

Trip purpose is recreational -1.1 0.05

Female 2.5 0.00

Perceived Travel Time Savings – Actual Observed Travel Time Savings=8.0 – 3.0×Annual Income<25,000 – 1.1×Trip PurposeRecreational+2.5×Female (Equation 1)

Where: Annual Income < $25,000 = 1 if annual household income was less than $25,000; 0 otherwise

Trip PurposeRecreational = 1 if the traveler was on a recreational trip; 0 otherwise

Female = 1 if the traveler was female, 0 otherwise

ConclusionsThe objective of this research was to examine the difference between travelers’ perceived travel time savings and their actual travel time savings, and to better understand the magnitude of any misperception of the travel time savings. This research used data from a 2010 survey on Katy Freeway travelers. In the survey, respondents were asked for information on their most recent trip on the Katy Freeway, including the date, time, and how much travel time they think they saved by using the MLs. This provided their perceived travel time savings and enough information to identify their (most likely) actual travel time savings using sensor data from the Katy Freeway.

The perceived travel time savings varied considerably across the respondents. Nearly 97 percent of the respondents indicated that they experienced some travel time savings. Very few respondents underestimated travel time savings, and a majority of the respondents overestimated the travel time savings. On average, respondents perceived they saved approximately 12 minutes when they really saved closer to 3 minutes, estimating that they saved nearly four times as much as they actually did. These results are higher than in the limited literature in the area, where perceived travel time savings are approximately twice the average maximum savings in the peak period.


Linear regression models were fit to model the magnitude of over- and underestimation of the travel time savings. Among the trip characteristics, only trip purpose (work or recreation) was found to be a significant predictor of the degree of misperception of travel time savings. Respondents’ characteristics, including gender and income, were also found to be significant in predicting the degree of misperception of travel time savings. This study shows that there is considerable difference between perceived and actual values.

This research has shown that travelers do have difficulty estimating the time they save while using MLs. They greatly overestimate the amount of time saved. However, exactly how to incorporate this understanding in mode choice models or traffic revenue estimates is unknown. It may well be that even though travelers are saving a small amount of time, they value that time savings (and avoiding congestion) much higher—possibly similar to their amount of perceived travel time savings. This is an area that needs additional research. However, with empirical ML data now available to examine, we are gaining a better understanding of how drivers perceive and value travel time savings.

AcknowledgmentsThe authors recognize that support for this research was provided by a grant from the U.S. Department of Transportation, University Transportation Centers Program to the Southwest Region University Transportation Center, which is funded, in part, with general revenue funds from the state of Texas. The authors would also like to thank Harris County Toll Road Authority, Houston-Galveston Area Council, and Houston Transtar for their help in spreading the word about the survey. The authors would also like to thank Richard Trey Baker for all the support he provided in hosting the survey. Three anonymous reviewers from ITE also helped to make this paper better through their comments. Any errors and omissions are the responsibility of the authors.

Works Cited1. Becker, G.S.A. “Theory of the Allocation of Time.” The Economic Journal, Vol. 75 (1965): 493–517.

2. Beesley, M.E. “The Value of Time Spent in Travelling: Some New Evidence.” Economica, Vol. 32 (1965): 174–185.

3. Oort, C.J. “The Evaluation of Travelling Time.” Journal of Transport Economics and Policy, Vol. 3, No. 8 (1969): 279–286.

4. Calfee, J. and C. Winston. “The Value of Automobile Travel Time: Implications for Congestion Policy.” Journal of Public Economics, Vol. 69 (1998): 83–102.

5. Ghosh, A. “Valuing Time and Reliability: Commuters’ Mode Choice from a Real Time Congestion Pricing Experiment.” Dissertation, University of California at Irvine, 2001.

6. Patil, S., M. Burris, and W.D. Shaw. “Travel Using Managed Lanes: An Application of a Stated Choice Model for Houston, Texas.” Transport Policy, Vol. 18, No. 4 (2011): 595–603.

7. Patil, S., M. Burris, W.D. Shaw, and S. Concas. “Variation in the Value of Travel Time Savings and Its Impact on the Benefits of Managed Lanes.” Transportation Planning and Technology, Vol. 34, No. 6 (2011): 547–65.

8. Devarasetty, P., M. Burris, and W.D. Shaw. “The Value of Travel Time and Reliability-Evidence from a Stated Preference Survey and Actual Usage.” Transportation Research Part A, Vol. 46 (2012): 1227–1240.


9. Devarasetty, P., M. Burris, and W.D. Shaw. “Do Travelers Pay for Managed Lane Travel as They Claimed They Would? A Before and After Study of Houston Katy Freeway Travelers.” Transportation Research Record: Journal of the Transportation Research Board, No. 2297 (2012): 56–65.

10. Whitehead, J.C., S.K. Pattanayak, G.L. Van Houtven, and B.R. Gelso. “Combining Revealed Preference and Stated Preference Data to Estimate the Nonmarket Value of Ecological Services: An Assessment of the State of the Science.” Journal of Economic Surveys, Vol. 22, No. 5 (2002): 872–908.

11. Hensher, D.A. “Measurement of the Valuation of Travel Time Savings.” Journal of Transport Economics and Policy, Vol. 35 (2001): 71–98.

12. Black, I.G. and J.G. Towriss. Demand Effects of Travel Time Reliability. London: UK Department of Transportation, Her Majesty’s Stationery Office, 1993.

13. Small, K.A., C. Winston, and J. Yan. “Uncovering the Distribution of Motorists’ Preferences for Travel Time and Reliability.” Econometrica, Vol. 73 (2005): 1367–1382.

14. Burris, M., P. Devarasetty, and W.D. Shaw. Managed Lane Travelers-Do They Pay for Travel As They Claimed They Would? College Station, TX: Texas A&M Transportation Institute, The Texas A&M University System, 2011.

15. Brownstone, D. and K.A. Small. “Valuing Time and Reliability: Assessing the Evidence from Road Pricing Demonstrations.” Transportation Research Part A: Policy and Practice, Vol. 39 (2005): 279–293.

16. Sullivan, E., K. Blakely, J. Daly, J. Gilpin, K. Mastako, K. Small, and J. Yan. “Continuation Study to Evaluate the Impacts of the SR91 Value-Priced Express Lanes: Final Report.” California Polytechnic State University at San Luis Obispo, December 2000, [Online]. Available: http://ceenve.ceng.calpoly.edu/sullivan/SR91.

17. Steimetz, S.S.C. “Defensive Driving and the External Costs of Accidents and Travel Delays.” Working paper. Department of Economics, University of California at Irvine, April 2004.

18. Fortin, C. and R. Rousseau. “Interference from Short-Term Memory Processing on Encoding and Reproducing Brief Durations.” Psychological Research, Vol. 61 (1998): 269–276.

19. TxDOT. Katy Freeway Website, Texas Department of Transportation, [Online]. Available: www.katyfreeway.org. [Accessed May 15, 2009.]

20. Burris, M. and S. Patil. Estimating the Benefits of Managed Lanes. College Station, TX: Texas A&M Transportation Institute, The Texas A&M University System, 2009.


Prem Chand Devarasetty received his Ph.D. from Texas A&M University. He is currently working as a senior risk analyst at PayPal Data Services in Austin, TX, USA. He has published several research articles and reports on the value managed lane travelers place on travel time and travel time reliability. His research interests are in toll road pricing and in understanding travelers’ behavior toward toll road pricing.

Mark Burris, Ph.D., P.E., is the Herbert D. Kelleher Professor in the Zachry Department of Civil Engineering at Texas A&M University and a research engineer with the Texas A&M Transportation Institute. His main area of interest is traveler behavior in response to pricing, particularly congestion (or value) pricing. He has been, and continues to be, involved in many projects in this area and has published extensively in this area of

research. He is a member of ITE.

Chao Huang, Ph.D., P.E., is a transportation system modeler with C&M Associates in Dallas, TX, USA. His main area of interest is traffic and revenue (T&R) forecasting and congestion pricing in the toll facility. He has been involved in various T&R studies with C&M and research projects with a focus on toll road/managed lane operating policy and toll facility users’ response to price at the Texas A&M University. He is a member of the

TRB Committee on Transportation Economics, and his experience includes stated preference survey design, benefit–cost analysis, transportation investment analysis, and traffic demand forecasting.


The District Department of Transportation (DDOT) is responsible for maintaining and operating more than 18,000 metered on-street spaces in Washington, DC, USA. The program went through significant changes in 2009 and 2010 including two rate adjustments, reintroduction of meter enforcement on Saturdays, and extending hours of meter operation to 10:00 p.m. in high-demand areas. These changes caused operational problems for the department and frustration for the customers. This paper describes how DDOT applied Lean Six Sigma (LSS) processes and techniques to dramatically transform its on-street parking meter program. The paper introduces the concept of LSS and discusses how some of the analytical techniques and concepts were applied. Techniques such as root cause analysis, process capability, mean testing, Pareto analysis, and process mapping were used to identify fundamental problems with the program and assets. Once the problems were identified, DDOT quickly developed a strategic vision for the future and aggressively implemented the vision. Applying LSS techniques has reaped significant rewards for DDOT within a very short period of time. These benefits include higher customer satisfaction through enhanced payment options, a lower number of service requests, better system uptime, more proactive management of assets, better executive visibility, and increased revenue. Washington, DC’s parking program is now recognized as one of the most innovative, forward-thinking programs in the country. The success of applying LSS in parking has encouraged DDOT to apply this concept in other program areas as well.

Transforming the Washington, DC, USA, Parking Meter Program Using a Lean Six Sigma Improvement ProcessBy Soumya S. Dey, P.E., PMP


BackgroundThe District Department of Transportation (DDOT) is responsible for developing and maintaining a cohesive, sustainable transportation system that delivers safe, affordable, and convenient ways to move people and goods—while protecting and enhancing the natural, environmental, and cultural resources of the nation’s capital. As an agency, DDOT is unique—it has the attributes of both a state and municipal department of transportation. As part of this mission, DDOT manages and operates all the transportation assets in Washington, DC, USA. These assets are valued at $44 billion.1 Over the past few years, DDOT has gone through a paradigm shift in how it manages its assets. These techniques include performance-based contracting (with incentive/disincentive and liquidated damages), application of data mining and process enhancement techniques such as Lean Six Sigma (LSS), and participating in constructive partnering forums with the private sector for more cost-effective service delivery.

DDOT is responsible for maintaining and operating more than 18,000 metered on-street spaces in Washington, DC. Until late 2010, these spaces were controlled by two basic asset types: (1) traditional single-space meters (SSMs) that each cover one space and (2) multi-space meters (MSMs) that cover approximately an average of eight metered spaces. The MSMs are networked while the SSMs are mechanical, non-networked assets. Moreover, the SSMs are more than 10 years old and at the end of their useful life. From a customer’s perspective, credit cards were accepted as a payment option (in addition to coins) only at the MSMs. For the rest of the assets, only coins were accepted. Up until late 2010 customers could pay by credit cards at only 23 percent of the metered spaces.

Between March 2009 and January 2010, the program underwent significant changes including two rate adjustments, lifting of the Saturday moratorium on parking meter fees, and extending of the duration during which meters are operational.2–4 These operational changes put significant stress on a system that had assets that were beyond their useful life.

The changes also caused significant frustration among customers. Due to the rate adjustments, the number of coin transactions on the system increased. This increased failure rates for the meters. Consequently, customers frequently encountered broken meters.5 The rate increase also required customers to carry more change. This caused additional frustration.6 Parking-related service requests through Washington, DC’s centralized 311 system became the highest service request in DC. In 2010, the agency received approximately 200,000 parking meter–related service requests.7 Just to put things in perspective, in 2009 DDOT received more parking meter–related service requests than the Department of Public Works received as an entire agency.8 Also, the perception gap (the difference between the percentage of on-time service delivery and customers’ perception about the level of service delivery measured through a sample survey) for parking meters was at 55 percent.9 This was the highest among all DDOT service categories. The press picked up on the negative customer sentiment as well.10,11 The program became a burning issue for the department and the District of Columbia. There was significant pressure on the agency to enhance the program.

This paper discusses how DDOT adopted an LSS-based approach to identify fundamental problems with its assets and the program, developed a strategic approach to enhance the program, implemented the recommendations, and achieved significant enhancement. It discusses some of the analytical tools that were used in the process.

What is Lean Six Sigma?Lean and Six Sigma are process improvement methodologies. Lean is a set of tools that is geared to improving process flow while primarily focusing on the elimination of waste or non-value-added


activities from all business processes. Lean arose as a method to optimize auto manufacturing (initially at Ford and Toyota). Lean focuses on eliminating seven sources of waste:12

1. Overproduction (manufacturing items ahead of demand);

2. Inventory (excess material and information);

3. Defects (production of off-specification products);

4. Transport (excess transport of work-in-process or products);

5. Motion (human movements that are unnecessary or straining);

6. Overprocessing (process steps that are not required); and

7. Waiting (idle time and delays).

Six Sigma specifically focuses on process variation. It is a methodology driven by understanding of customer needs and the disciplined use of data, facts, and statistical analysis to improve and reinvent organizational processes by reducing variation. Six Sigma evolved as a quality initiative to reduce variance in the semiconductor industry (initially Motorola). From a purely statistical standpoint, a Six Sigma process has 3.4 defects per one million opportunities.13 The statistical roots of the term Six Sigma have become less important as Six Sigma has evolved into a comprehensive management system.

LSS is a combination of Lean and Six Sigma approaches. Despite their disparate roots, Lean and Six Sigma share several fundamental commonalities including a focus on customer satisfaction, continuous improvement, identification of root causes, and comprehensive employee involvement.14 LSS uses a data-based approach to enhance workflow and business processes.

Though used primarily in the manufacturing sector, the concept of Lean and Six Sigma combined has been applied successfully in the service sector as well.15 LSS for services is a business improvement methodology that maximizes value by achieving the fastest rate of improvement in customer satisfaction, cost, quality, process speed, and invested capital. LSS is a five-step process as shown in Figure 1.16 Each of these processes has sample tools and procedures, as shown in Table 1.17

FIGURE 1. Lean Six Sigma process.


Table 1. Sample tools in Lean Six Sigma approach.

Define Measure Analyze Improve Control

Project definition

Process mapping

Cause and effect matrix

Failure modes and

effects analysis (FMEA)

Basic statistics (Measure of central tendencies, measure of variation, distributions, etc.)

Graphical techniques

Confidence intervals

Measurement systems analysis (MSA)

Capability analysis

Statistical process control (SPC)

Hypothesis testing

Means testing

Power & sample size

Regression analysis

Analysis of variance (ANOVA)

Design of experiments (DOE)

2k factorial design

Control plan

Project closure

Literature reviewLSS has been used extensively in the private sector. The author reviewed, among others, literature on LSS application in the construction industry in Abu Dhabi, UAE,18 various service organizations (hospitals, consulting, hotels, etc.) in Singapore,19 a production facility in Sweden,20 and a printing factory in Ohio, USA.21 There are several documented uses of LSS across different levels of government as well. Internationally, Randor discusses a framework that assesses tools and processes that are most relevant and impactful when introducing Lean into a large government department in the United Kingdom.22 In the United States, federal agencies such as the Department of Defense, Environmental Protection Agency, and Department of Energy; state agencies such as the Florida Department of Revenue; and municipal agencies such as the City of Fort Wayne (Indiana), City of Hartford (Connecticut), and City of Irving (Texas) have applied Six Sigma techniques to various components of their programs.23–25 Most of the documented applications of LSS in the transportation industry have been in the area of logistics and supply chain for humanitarian relief, airline, rental car, and rail operations.26,27 Lean techniques have been used for the delivery and construction of infrastructure projects.28,29 However, as a concept, LSS has not made significant inroads into the mainstream transportation engineering profession.

Change FrameworkThis section discusses the framework used by DDOT to transform its parking meter program. It was a three-step process including (1) application of LSS to identify fundamental issues with the parking program, (2) developing a vision for the future program and pilot testing state-of-the-art assets and solutions to achieve those goals, and (3) implementation. The following three sections in this paper discuss each of these three steps in the framework (shown in Figure 2).


Figure 2. Change Framework.

Application of Lean Six Sigma TOOLSThis section discusses how some of the tools in the LSS framework were applied to gain valuable information and knowledge about DDOT’s parking meter assets.

Root Cause Analysis

There are several techniques that can be utilized for root cause analysis such as fishbone diagrams, cause and effect mapping, and failure modes and effects analysis (FMEA). DDOT utilized cause mapping to identify the “root causes” of the problems associated with the program. During this process, it was important to distinguish between symptoms and causes. During the analysis it became apparent that symptoms such as increase in the number of broken meters, broken meter call volumes, and customer dissatisfaction were driven by four fundamental causes: aged assets, non-communicating meters, increased coin transactions (because of the rate increases), and limited payment options for customers (shown in Figure 3). Identifying the root causes was fundamental in charting a future course for program enhancements.


Figure 3. Results of root cause analysis.

Control Chart for Call Volumes

The control chart is a tool that helps detect any extraordinary variation in the process—variation that may indicate a problem or fundamental change in the process. Figure 4 shows a control chart for the number of broken meter 311 calls received between February 2007 and April 2011. The control chart shows the overall process mean (x) as well as the upper control limit (UCL) and lower control limit (LCL). The UCL and LCL are three standard deviations from the mean. The chart also annotates policy changes that likely impacted call volumes: the rate changes in April 2009 and January 2010 and installation of networked SSMs in November 2010. It appears that the process has four distinct means:

■■ Between February 2007 and March 2009, the mean appeared to be between the overall mean (x) and the LCL.

■■ Between March 2009 and January 2010, the mean jumped to between the overall mean (x) and the UCL.

■■ Between January 2010 and November 2010, the mean jumped to above the UCL.

■■ After new assets were introduced in November 2010, the mean dropped again to between the UCL and the overall process mean (x).

Individual mean testing was conducted for each of the four time slices and indicated that the means during the four periods were statistically different. This shows that external influences such as policy changes and asset refresh affected the process.


Figure 4. Control chart for call volumes.

Pareto Analysis/Bad Actor Report

All service request calls to DC’s 311 system are fielded by the Office of Unified Communications, entered into a service request system, and routed to appropriate agencies. DDOT analyzed parking meter–related service request calls from the 311 system and performed a Pareto analysis. The analysis (referred to within DDOT as the “bad actor report”) revealed that the worst 1,500 meters (8 percent of assets) accounted for 40 percent of the call volume. (The worst meter is defined as the meter that had the maximum number of 311 calls/service requests.) The 1,500 worst performing assets averaged 25 calls per year. This implies technicians were dispatched to investigate these meters an average of 25 times during the course of 1 year. The worst performing meter was “called-in” 95 times. Ninety-five percent (1,425 out of 1,500) of the assets in the bad actor report were the oldest asset types (SSMs). Figure 5 shows the distribution of call volume for the 1,500 worst performing meters.30

From a business and customer service standpoint, it made more sense to systematically replace these aged assets with newer assets. The findings from this analysis were one of the major criteria for prioritizing the asset refresh schedule.

Figure 5. Call distribution for 1,500 worst performing meters.


Mean Testing

Based on the Pareto analysis and the findings of the pilots, DDOT started replacing its worst performing assets with newer, networked assets. Figure 6 shows how the call volumes for specific groups of meters dropped after the old assets were replaced with new, networked assets. Hypothesis testing of the mean call volumes before and after the installation revealed that the means were statistically different.

Similar before and after mean testing was conducted on revenues for routes where new assets were installed. The analysis revealed that asset refresh resulted in higher revenues as well.31 This was most likely driven by two factors: (1) newer meters had better uptimes and were less likely to break down, and (2) the newer, networked assets accepted credit cards, and hence customers had two options to pay at these meters: coin or credit/debit card.

Figure 6. Mean testing for routes with new assets.

Process Capability

DDOT manages and operates the parking meter program using a performance-based asset management contract. The contract requires the contractor to fix SSMs within 72 hours and MSMs within 24 hours. It also requires an operability rate of 99 percent for MSMs and 97 percent for SSMs. The contract allows incentives for exceeding the performance metrics and liquidated damages for failing to meet the performance measures.

A capability analysis was conducted to ascertain whether the process was capable of meeting the 72 hours requirement. As shown in Figure 7, the contractor’s mean repair time was significantly below


the upper specification limit for SSMs (72 hours). The contractor was also meeting the 97 percent operability requirement. Yet the program had a negative perception.

Figure 7. Capability analysis for meter repair process.

In examining the issue more closely, the team realized that there was a “measurement error” built into how performance was being measured.

Measurement Error

Given that most of the assets are non-networked, operability and repair time were based on a work order/service request system. This is largely a reactive system in that broken meters factor into the operability equation only if they are “called in.” A broken meter that is not reported does not factor into the operability calculation.

DDOT conducted some field observations based on a statistical sampling. The 95 percent confidence interval for meter operability was between 82 percent and 90 percent, with a mean of 86 percent. This is significantly lower than the 97 percent operability calculated through the work order system. DDOT also conducted two independent “touch a meter” programs. During the program, DDOT staff assessed the operability of its entire installed meter inventory (13,000 to 14,000 parking meters) within a 1-week period. The operability of the meters using this effort was calculated to be between 80 percent and 85 percent. The measurement error helps explain the difference between the “ground truth” and work order–based operability and helps explain the public frustration.

Process Flow Charts

DDOT also spent a considerable amount of time mapping out the various processes related to parking and curbside management. The parking function in DDOT is very fragmented, with individual groups responsible for policy, operations, signage, permitting, and enforcement. Process mapping techniques revealed inefficiencies such as “hidden factories,” bottlenecks, and non-value-added activities. It was also apparent that there were delays whenever there was a hand-off


of responsibilities between different groups in DDOT. Process mapping allowed DDOT to get a common understanding of the process and responsibilities and identify inefficiencies.32

Development of the Strategic VisionBased on the findings of the LSS analysis, DDOT started the process of developing a strategic plan and long-range vision for its metered on-street parking program. This section discusses the high-level program goals and the pilot tests that were conducted to identify assets and programs that would enable DDOT to reach the established goals.

Establishment of High-Level Program Goals

DDOT developed three high-level, simplistic goals for the parking meter program. Each goal was subdivided into more detailed subgoals. These included:

■■ Improved customer service

■● Multiple payment options

■● Maximized convenience

■● Real-time parking availability

■● Fewer broken meters

■■ Enhanced operational efficiency

■● Dynamic pricing

■● Real-time operational status

■● Better uptime

■● Lower operating cost

■■ Better revenue management

■● Minimized coin transaction

■● Real-time auditing

Parking Pilots

In the summer of 2010, DDOT evaluated the state of the art in meter hardware, payment options, and real-time occupancy sensing through a series of pilots. The pilots were conducted using a competitive procurement process. DDOT tested two different configurations of MSMs (pay by space and pay by license plate), credit card–accepting SSMs, occupancy sensors, and new payment options such as pay by cell. The pilots were evaluated against the program goals, as shown in Table 2.33


Table 2. Assessment of pilots against program goals.

Program Goals Pay by Cell Smart SSM Smart MSM Space Occupancy

Improved Customer Service

Multiple payment options X X X

Customer convenience X X X X

Real-time parking availability

X

Fewer broken meters X X

Enhanced Operational Efficiency

Dynamic Pricing X X X X

Real-time operational status

X X

Better uptime X X X

Lower operation cost X X X

Better Revenue Management

Minimized coin transaction X X X

Real-time auditing X X X X

Implementation Based on the LSS analysis and the findings of the parking pilots in 2010, DDOT implemented a series of changes to its parking meter program, with dramatic results. These included:

■■ Asset refresh—DDOT significantly increased the number of networked assets by procuring and installing new, networked credit card–accepting SSMs and increased the coverage provided by MSMs.

■■ Launched a citywide pay-by-cell program—Following two very successful pilots, DDOT launched a citywide pay-by-cell program in July 2011. The program turned out to be the most successful pay-by-cell program in the country. As of May 2014, the program has 750,000 customers and has resulted in more than 15 million transactions accounting more than 40 percent of the annual parking revenue.34

■■ Issued a new contract for the next generation parking asset management—DDOT selected a vendor for maintaining and managing its parking meter assets for the next 5 years. The contract is structured as a performance-based contract with liquidated damages and incentive/disincentive clauses. It is also structured to make sure that the city is not locked into paying for services that it does not require. Recognizing that the asset mix and revenue mix would change substantially over the next 5 years, the contract has fixed price elements, fixed unit prices by asset type for maintenance, and a percentage fee for items such as coin collection and credit card processing. It also requires the use of LSS techniques for program management and an integrated back-end system for asset monitoring and management.35 The integrated back-end system would aggregate all the data (operational, revenue, maintenance, etc.) from disparate assets (such as multiple meter types, pay-by-cell transactions, parking occupancy detection devices, etc.) into a common platform so that DDOT can make smart operational and strategic choices in managing and operating its on-street parking meter program.


Parking Program TransformationThis section discusses some of the trends that started to evolve as DDOT started infusing smart assets and alternative payment options into the parking program.

Non-Communicating Assets to Networked “Smart Assets”

DDOT started a systematic process of phasing out non-communicating meters with networked assets. The changeover was done in a systematic manner based on the findings of the bad actor report. MSMs were written into the standard specifications and incorporated as part of capital improvements such as streetscape projects. Table 3 shows how the asset mix has changed over the last 4 years. The number of spaces that are covered by credit card–accepting, networked meters jumped from 23 percent in 2009 to 45 percent in 2013.

Table 3. Washington, DC’s changing asset mix for parking meters (2009 and 2013).

Asset Type NetworkedPayment Options

2009 Asset Mix 2013 Asset Mix

# of Meters

# of Spaces

% of Spaces

# of Meters

# of Spaces

% of Spaces

Non-networked single-space meters

No Coin 13,234 13,234 77% 10,238 10,238 55%

Networked multi-space meters

YesCoin

Credit514 3,923 23% 638 5,104 28%

Networked single-space meters

YesCoin

CreditNone 3,197 3,197 17%

TOTAL – – 13,748 17,157 14,073 18,539

Meter/ space ratio

– – 0.80 0.75

% of spaces covered by networked credit card–accepting assets

– – 23% 45%

Coin Transactions to Cashless Transactions

With the introduction of credit card–accepting, networked meters and the launch of pay by cell, DC’s revenue mix has undergone a significant transformation. In 2009, 80 percent of DC’s parking meter revenue was in the form of coins. Currently, that percentage stands at only 30 percent. Therefore, a majority of DC’s revenues are through virtual transactions, which have a lower cost structure than coin revenues.36 Virtual revenue also provides an easier audit trail and real-time visibility into revenue streams and meter usage.


Reactive to Proactive Maintenance

Operability of networked meters can be monitored in real time through a back-office system (as opposed to non-network assets, where meter malfunctions have to be reported). Hence, with the increase in networked assets, the meter maintenance strategy can be shifted from reactive to proactive. This increases system uptime, revenue potential, and customer satisfaction. It also enables real-time assessment of asset condition and a more realistic assessment of operability, repair time, etc., making it easier to manage and monitor a pure performance-based contract. Proactive maintenance also reduces the cost associated with fielding parking meter–related 311 calls.

“Asset Lite” Solutions

As pay-by-cell adoption rates increase, DC can start considering removing meters from a block face in areas with high pay-by-cell usage. These meterless blocks would be available to customers using pay by cell only. This can result in significant cost savings for DC, since an asset-lite solution such as pay by cell costs significantly less than meter-based payment methods (such as coins or credit cards). DDOT will be testing some of these asset-lite strategies in a federally funded pilot project in the Chinatown area.37

ResultsProgram performance can be measured across various dimensions. The parking meter program has gone through a transformation as a result of LSS analysis and follow-through. The following section discusses some of the accomplishments.

■■ Increased payment options for customers—In 2009, people had one mode of payment (coin) at 70 percent of the metered spaces. Now people have at least two options (coin and pay by cell) to pay at all locations. At 50 percent of the locations, people have three payment options (coin, credit/debit, and pay by cell).

■■ Increased revenues—In 2009, DC’s parking meter revenue was $20 million. The last set of rate adjustments occurred in January 2010. DC’s revenue in 2010 was $25 million. DC’s 2012 and 2013 parking revenues were $38 million and $40 million, respectively. This represents a 60 percent revenue increase with the same set of parameters or policy framework. This is a function of better management of the assets, more options to pay for metered parking, and better system uptime.

■■ Drop in call volumes—Parking meter–related call volumes can be considered as a surrogate for customer satisfaction. Call volumes have dropped significantly since an all-time high in 2009 and 2010. 2013 call volumes were 40 percent lower than 2010 levels.

■■ Public perception about the program has increased—If the sentiment in the press is a reflection of public perception, the program has undergone a sea change over the last 4 years. From headings such as “Parking Meters a Bane to Mankind,” the press sentiment has turned to “Government at Its Best” after the launch of the successful pay-by-cell program.38,39

■■ Enhanced executive visibility—Recognizing the success of the program and the full potential, the executive office of the mayor allocated $25 million in capital funds for asset refresh and modernization over the next three fiscal year budget cycles.

■■ National recognition—DC’s parking meter program is now widely acknowledged in the industry as a forward-thinking program. In a recent survey conducted by the International Parking Institute, DC was ranked fourth among innovative parking programs in the country.40

Figure 8 summarizes some of the performance trends in the parking meter program.


Figure 8. Program performance metrics.


Lessons LearnedHilton and Sohal examine the relationship between the successful deployment of LSS in Australia and a number of key explanatory variables that essentially comprise the competence of the organization, the competence of the deployment facilitator, and the competence of the project leaders.41 Laureani and Antony discuss critical success factors for LSS deployment in organizations.42

The success factors/lessons learned for DDOT are well aligned with the findings in the literature and are discussed below:

■■ Get executive buy-in—For a program such as LSS to work, there needs to be a commitment and buy-in at the executive level. At DDOT all key executives (the director, deputy director, associate directors, and chief of staff) went through Six Sigma black belt training. This “signaled” the agency staff about the importance of LSS and set the tone for the program within the agency. The executives in turn have deep appreciation of the process and its potential.

■■ Develop a core group of champions—Build a core group of LSS champions throughout the agency. This core group needs to comprise individuals who can adapt to change, believe in the benefits of data-based decision making, be enthusiastic about applying the concept to their program areas, and mentor other people in the organization. At DDOT, this core group got subsequent training in a Six Sigma train-the-trainer course.

■■ Institutionalize LSS—Develop training program(s) and provide infrastructure support for the concept to promulgate throughout the agency. LSS is now part of DDOT’s standard curriculum through the training office.

■■ Apply LSS to the “right projects”—It helps if LSS is applied to a project of critical interest. This ensures executive support, continuing momentum, and a sense of urgency. For DDOT, the parking meter program was one of the highest priority problems that needed to be fixed. In addition, black belt certification requires participants to work on a real-life project. Having executive leadership trained ensured that LSS was being applied to projects of critical interest to the agency.

■■ Share success stories—Ensure that success stories are shared throughout the organization. DDOT (and the DC government) has a performance/data-based culture. Sharing tangible benefits achieved through application of LSS is part of the natural performance reporting process.

■■ Be patient; maintain momentum—Agency transformation through LSS is a marathon, not a sprint. Success requires patience, discipline, and sustained focus over a period of time.

ConclusionApplication of LSS helped transform the parking meter program that used to be a source of frustration for DDOT and its customers. LSS helped identify fundamental issues with the program and helped DDOT to aggressively chart a path forward using a fact-based decision-making process. Following the success of the parking program, LSS has been applied to other asset classes and programs at DDOT as well. The principles and tools from LSS have been successfully applied to streetlights, urban forestry, traffic control officers, and safety service patrols.

The current trends in public sector transportation agencies toward increased use of performance management/measurement, a higher level of citizen engagement, and operating in a financially constrained environment can serve as a catalyst for adopting process improvement efforts such as LSS. Applied correctly, it has the potential to provide the framework for achieving sustainable improvements to business processes and service delivery.


AcknowledgmentsThe author would like to acknowledge the contributions of Mark Derrick of Xerox Corporation and Stephanie Dock of DDOT and their help in identifying, synthesizing, and analyzing the data used for this paper. Their input, suggestions, and analysis were invaluable in crafting the message in this paper. The author would also like to thank the reviewers and editors for their valuable input and suggestions, which helped improve the quality of the final product.

Works Cited1. District Department of Transportation. “Action Agenda—DDOT Delivers.” 2010. [Online]. Available:

www.dc.gov/DC/DDOT/About+DDOT/Who+We+Are/Action+Agenda/Action+Agenda+-+2010. [Accessed October 23, 2013.]

2. DDOT Press Release. “DDOT Announces DC Parking Meter Rate Increase.” (February 26, 2009), [Online]. Available: http://newsroom.dc.gov/show.aspx/agency/ddot/section/2/release/16237. [Accessed October 23, 2013.]

3. Council of the District of Columbia. Fiscal Year 2010 Budget Support Act of 2009. May 2009.

4. DCMR. Title 18. Chapter 18-24. Section 18-2404. “Parking Meter and Parking Meter Zones.” Effective December 26, 2011.

5. Neibauer, M. “More Money, More Problems for District Parking Meters.” Washington Examiner, March 1, 2009.

6. Bridge, T. “The Difficulty with Parking Meters.” WeLoveDC, September 23, 2010, [Online]. Available: www.welovedc.com/2010/09/23/the-difficulty-with-parking-meters/. [Accessed October 25, 2013.]

7. Dey, S.S., J. Thommana, and S. Dock. “Public Agency Performance Management for Improved Service Delivery in the Digital Age.” Journal of Management in Engineering (accepted for publication).

8. DC CapStat Session. 911, 311, and Customer Service. October 29, 2009.

9. DC Office of Unified Communications. 2009 Customer Service Survey Data. 2009.

10. Neibauer, M. “Area Drivers File a Record Number of DC Parking Meter Complaints.” Washington Examiner, November 26, 2008, [Online]. Available: http://washingtonexaminer.com/area-drivers-file-record-number-of-d.c.-parking-meter-complaints/article/104732. [Accessed July 28, 2013.]

11. Newell, J. “Parking Meters Out to Destroy the Human Race.” NBC4 News, February 24, 2009, [Online]. Available: www.nbcwashington.com/news/local/Local-Parking-Meters-Develop-Artificial -Intelligence-Will-Destroy-Humans.html. [Accessed July 28, 2013.]

12. United States Environmental Protection Agency. The Environmental Professional’s Guide to Lean and Six Sigma. EPA-100-K-09-006. August 2009.

13. Snyder, K. and N. Peters. Lean Six Sigma in the Public Sector. White Paper. Xerox Corporation, September 2004.

14. American Strategic Management Institute. Six Sigma Excellence Summit Course Materials, July 2010.


15. Tjahjono, B. and P. Ball. “Six Sigma: A Literature Review.” International Journal of Lean Six Sigma, Vol. 1, No. 3 (2010): 216–233.

16. The Performance Institute. Lean Six Sigma Train-the-Trainer Course Materials, May 2011.

17. The Performance Institute. Lean Six Sigma Black Belt Certification Course Materials, 2010.

18. Al-Aomar, R. “A Lean Construction Framework with Six Sigma Rating.” International Journal of Lean Six Sigma, Vol. 3, No. 4 (2012): 299–314.

19. Chakraborty, A. and K.C. Tan. “Case Study Analysis of Six Sigma Implementation in Service Organizations.” Business Process Management Journal, Vol. 18, No. 6 (2012): 992–1019.

20. Assarlind, M., I. Gremyr, and K. Backman. “Multi-faceted Views on a Lean Six Sigma Application.” International Journal of Quality & Reliability Management, Vol. 29, No. 1 (2012): 21–30.

21. Roth, N. and M. Franchetti. “Process Improvement for Printing Operations through the DMAIC Lean Six Sigma Approach—A Case Study from Northwest Ohio, USA.” International Journal of Lean Six Sigma, Vol. 1, No. 2 (2010): 119–133.

22. Radnor, Z. “Transferring Lean into Government.” Journal of Manufacturing Technology Management, Vol. 21, No. 3 (2010): 411–428.

23. Maleyeff, J. “Improving Service Delivery in Government with Lean Six Sigma.” IBM Center for the Business of Government, 2007.

24. City of Fort Wayne, Indiana. “Mayor to be Keynote Speaker at Six Sigma Event in California.” 2006, [Online]. Available: www.cityoffortwayne.org/news-archive/974-mayor-to-be-keynote-speaker-at-six-sigma-event-in-california.html. [Accessed October 25, 2013.]

25. “City of Irving Innovates Services with Lean Six Sigma.” [Online]. Available: http://www.tmac.org/node/102. [Accessed October 25, 2013.]

26. “Lean Six Sigma Success Stories in Transportation and Travel Industries.” [Online]. Available: http://www.goleansixsigma.com/lean-six-sigma-success-stories-in-transportation-and-travel-industries/. [Accessed October 25, 2013.]

27. Helferich, K. “Application of Six Sigma to Humanitarian Relief Logistics.” 2013 TRB Annual Meeting (2013), Session 383, paper P13-5944.

28. Hanna, A., et al. “Applying Lean Techniques in the Delivery of Transportation Infrastructure Projects.” 2011 TRB Annual Meeting (2011), Session 388, paper 11-1405.

29. Han, S.H., et al. “Six Sigma-Based Approach to Improve Performance in Construction Operations.” Journal of Management in Engineering, Vol. 24, No. 1 (2008): 21–31.

30. SAIC. “Parking Meter Performance Audit Report, Task 1—Program Assessment Report.” Prepared for DDOT, December 2009.

31. Dey, S.S., S. Dock, and E. Patterson. “Economic Implications of Changing Dynamics for Metered On-Street Parking.” Case Studies in Transportation Policy (under review).

32. SAIC. “Strategic Parking Plan.” Prepared for DDOT, September 2010.

33. Dey, S.S. and T. Viteri. “Feedback & Progress—Washington, DC Pilots New Meter Programs to Enhance Its Parking System.” The Parking Professional, July 2011.


34. Dey, S.S. and A. Rao. “Dialing and Thriving—Two Years After Its Introduction, Pay-by-Cell Enjoys Tremendous Success in Washington, DC” The Parking Professional, August 2013.

35. Office of Contract & Procurement, DDOT. Asset Management Services for Parking Meter Assets Citywide. DCKA-2012-R-0018. November 2011.

36. Dey, S.S. “Asset Lite Solutions to Metered Curbside Parking.” ITE Journal, Vol. 84, No. 6: 30–35 (2014).

37. DDOT. “Multimodal Value Pricing for Metered Curbside Parking, 2012 Discretionary Grant Program—Value Pricing Pilot Grant Application.” January 2012, [Online]. Available: http://s3.documentcloud.org/documents/407820/ddot-value-pricing-pilot-program- grant-proposal.pdf. [Accessed July 31, 2013.]

38. Alpert, D. “An Example of Government at Its Best.” The Washington Post, October 18, 2011, [Online]. Available: www.washingtonpost.com/blogs/all-opinions-are-local/post/an-example-of-government-at-its-best/2011/03/22/gIQA332RuL_blog.html. [Accessed July 29, 2013.]

39. Jouvenal, J. “DC Rolls Out Pay by Phone Parking Meter System.” The Washington Post, April 20, 2011, [Online]. Available: www.washingtonpost.com/local/dc-rolls-out-pay-by-phone-parking-meter-system/2011/04/20/AFg9pSEE_story.html. [Accessed July 29, 2013.]

40. International Parking Institute. “2013 Emerging Trends in Parking.” May 2013, [Online]. Available: www.parking.org/media/emerging-trends-in-parking.aspx. [Accessed July 28, 2013.]

41. Hilton, R.J. and A. Sohal. “A Conceptual Model for the Successful Deployment of Lean Six Sigma.” International Journal of Quality & Reliability Management, Vol. 29, No. 1 (2012): 54–70.

42. Laureani, A. and J. Antony. “Critical Success Factors for the Effective Implementation of Lean Sigma.” International Journal of Lean Six Sigma, Vol. 3, No. 4 (2012): 274–283.

Soumya S. Dey, P.E., PMP, is the director of research and technology transfer for the District Department of Transportation. He has more than 20 years of experience in the transportation profession, spanning the public and private sectors. He has a B.S. in civil engineering from the Indian Institute of Technology, an M.S. in civil engineering from Purdue University, and an MBA from the University of Maryland. He received ITE’s Past

Presidents Award and the Cafritz Award for Excellence in Public Leadership. He is a certified black belt in Lean Six Sigma. He is a Fellow of ITE.


A double crossover diamond (DCD) interchange (a.k.a. diverging diamond interchange) is a new unconventional diamond-type interchange. The purpose of this interchange design is to accommodate left-turning movements onto arterials and limited-access highways while eliminating the need for a left turn bay and signal phase at the ramp terminals. During a large study sponsored by the Federal Highway Administration, the research team examined how users perceived new DCD interchanges using focus groups formed at three DCD sites with 20 total participants and attitudes summarized from surveys conducted at two DCD sites with a total of 1,649 participants. The team further explored the results of one of the surveys, comparing opinions expressed before and after the installation of a DCD. The results from both the focus groups and surveys revealed that DCD interchanges operated properly, but there is room for improvement in how these interchanges might be more readily accepted by the public. In regards to general knowledge and comfort with the interchange and safety (real or perceived), users generally thought the interchange was an improvement over the existing facility. The most frequently reported issues are summarized for further consideration in outreach, education, planning, or design. Specific changes could be made in regards to the scope of the project to include adjacent interchanges, adjustments with signal operations, and the inclusion of facilities to accommodate both bicyclist and pedestrian travel to increase user satisfaction with DCD interchanges.

Public Perception of Double Crossover Diamond InterchangesBy Kristy Jackson, Christopher Cunningham, P.E., Chunho Yeom, Joseph E. Hummer, Ph.D., P.E., and Adam Kirk, Ph.D., P.E., PTOE, AICP


IntroductionA double crossover diamond (DCD) interchange (also known as a diverging diamond interchange, or DDI) is a new unconventional diamond-type interchange. This interchange design accommodates left-turning movements onto and off of the freeway while eliminating the need for a left turn bay and signal phase at the ramp terminals.1 While the ramp configuration at the freeway is similar to a traditional diamond interchange, traffic on the cross route moves to the left side of the roadway for the segment between signalized ramp intersections.2 Pictured in Figure 1 is a typical DCD interchange diagram oriented with an east-to-west arterial road. Directional movements are presented by origin–destination methods, so, for instance, “W to E” means vehicular movements from the west to east (shown in red).

Figure 1. Typical double crossover diamond layout showing vehicular movements through the interchange

The first DCD interchange was installed in France in the 1970s, and the Missouri Department of Transportation (DOT) implemented the first U.S. DCD interchange in Springfield, Missouri, on June 21, 2009.2,3 The total number of DCD interchanges has increased rapidly since the initial one in Springfield, and many more DCDs are under construction in the United States.

Due to the lack of experience with DCD interchanges, the Federal Highway Administration (FHWA) executed a project (Project DTFH61-10-C-00029), Field Evaluation of Double Crossover Diamond Interchanges, to investigate the first seven DCD interchanges installed nationwide. As part of the project, the research team evaluated attitudes through focus groups and surveys at five of the seven sites to gain insight into the public perception of the new DCD interchanges.

Source: Map data ©2014 Google Imagery ©2014 DigitalGlobe, USDA Farm Service Agency.


Literature ReviewShumaker et al. conducted a random survey of Institute of Transportation Engineers members to document barriers inhibiting the implementation of unconventional intersection and interchange designs (UIIDs), which included the DCD interchange.4 The survey responses from 103 UIID projects rated both before and after opening showed that the public opinion generally improved after opening, although the results did not specify which type of UIID projects were evaluated. Respondents identified the largest barrier to widespread UIID use was lack of public and professional support, although a range of other responses were given. Respondents were also asked to rank several identified barriers such as public acceptance, professional support, and political barriers. For public acceptance, the largest barriers were the potential for driver confusion and fear of the unknown. Lack of proof of design function and safety concerns were the largest professional barriers. Political barriers included public opinion and reiterated the lack of proof of design function as an issue. Respondents overwhelmingly indicated that proof of the benefits offered by UIID designs is important and noted that education should not be overlooked when attempting to overcome implementation barriers.

The North Carolina DOT studied benefits and capacities of superstreets, also known as restricted crossing U-turn intersections, to develop a level-of-service estimation program.5 As part of the study, they conducted mail/email surveys of residents and commuters in parallel with in-person surveys of businesses to determine the general perception of the intersection for each user group type. The results of the surveys showed mixed perceptions among user groups.

■■ Residents: More than half (52 percent) agreed there was a positive effect on safe navigation, while 23 percent said the superstreet has had a negative effect. In contrast, for the same resident survey, 38 percent reported navigation through a superstreet more confusing, whereas 35 percent reported it easier/less confusing.

■■ Commuters: Of those identified as commuters in the survey, approximately half of commuters reported they experienced difficulty/confusion driving through the superstreet. Nearly half (44 percent) reported that the superstreet had not affected their ability to safely navigate the roadway compared with the previous design, while 33 percent and 22 percent reported positive and negative impacts, respectively.

■■ Businesses: In-person surveys of surrounding businesses were completed at two superstreet locations. Sixty percent and 42 percent at locations 1 and 2, respectively, indicated that the superstreet had no effect or an improved effect on customer satisfaction with access to the store. Business respondents reported some confusion by drivers, with 30 percent and 42 percent reporting some confusion at locations 1 and 2, respectively. Responses varied on the reaction to traffic safety, with 40 percent and 53 percent reporting improvement at the two locations, respectively.

FHWA developed a simulator experiment of a DCD interchange for the public before DCDs were field implemented in Missouri.6 With the help of the Missouri DOT, they simulated three types of DCD interchanges: a DCD interchange with full signs and lane markings, a DCD interchange with minimized signs and lane markings, and a conventional diamond interchange. After 74 participants had completed the evaluation experiment in the driving simulator, they concluded that safety concerns of wrong ways at the crossover were not problematic, as no drivers looked at or attempted to drive down the wrong approach. Other types of driver errors were found to be no more likely with the DCD configurations than with the conventional diamond interchange. As a result of the simulated experiment, they recommended a DCD interchange as an attractive alternative due to a prospective safety benefit of less conflict points and more efficient signal operation.


ObjectivesThis paper presents the results from the focus groups and surveys conducted at the five new DCD sites. The focus group and survey results should help agencies thinking about DCDs in several ways. First, they should help agencies understand the general public perception for the unconventional interchange. Perception is important and can assist with targeting educational materials that can explain unique features of DCDs, such as driving on the opposite side of the road. Agencies looking to implement a DCD project should be aware of perceptions regarding this design. Policymakers or highway designers should consider the results of the focus groups and surveys when deciding whether or how to implement DCDs for a candidate location. Second, it would be informative for agencies to use survey data to design or calibrate DCDs to operate more efficiently. For example, there are many comments regarding signal timing, which provide opportunities for traffic engineers to investigate and optimize a DCD signal system based on public input. Lastly, issues brought up through the focus groups and survey may influence an agency’s decision to alter a project so that it receives more public support. For example, a project may receive less scrutiny if it includes considerations for adjacent intersections within the project scope or has provisions for integrating bicycle and pedestrian facilities. The “before” and “after” survey approach is also provided as a useful supplement to traditional public involvement. In understanding the general issues and giving the public another opportunity to express their opinion, transportation professionals trying to implement these designs can gain a better understanding of and work to address some of the issues (both real and perceived) with the unconventional design.

ScopeFrom the seven new DCD installations studied for the FHWA project, a focus group was conducted at three sites including Alcoa, TN, USA; Maryland Heights, MO, USA; and Kansas City, MO, USA. Surveys were conducted at the sites in Lexington, KY, USA, and Rochester, NY, USA. The focus groups and surveys were conducted after construction and at least 5 months after opening of the interchange, with the exception of Rochester, where a “before” survey was also conducted prior to construction. Table 1 shows the five DCD sites studied, the method (focus group or survey), the dates when the public perceptions were gathered, and the overall number of participants/respondents.

Table 1. DCD interchanges studied for the focus group and survey.

DCD Site [Freeway @ Arterial]Date DCD

Opened Method

Focus Group/

Survey DateParticipants / Respondents

Alcoa, TN [US129 @ Bessemer St.]

Dec. 12, 2010 Focus group June 21, 2011 7

Kansas City, MO [I-435 @ Front St.]

Nov. 7, 2011 Focus group Aug 15, 2012 7

Maryland Heights, MO [I-270 @ Dorsett Rd.]

Oct. 17, 2010Focus group (interviews)

Aug 9, 2011 6

Lexington, KY [KY 4 @ US 68]

Aug. 14, 2011 SurveyJan 20, 2012–Feb 24, 2012

1,418

Rochester, NY [I-590 @ Winton Rd.]

Sept. 11, 2012Survey (before) May 2012 98

Survey (after) June 2013 133

Source: Details on the designs and operations of the five interchanges studied for this paper are available in the FHWA project technical report: Cunningham, C.M., B.J. Schroeder, J.E. Hummer, C. Yeom, C. Vaughan, and K. Salamati. Field Evaluation of Double Crossover Diamond Interchanges. Washington, DC: Transportation Research Board, 2013 (report in progress).


The research team employed focus groups and surveys to ask standard questions and gain insight into public perceptions regarding DCDs. The focus groups were led by a facilitator to develop a deeper discussion of questions related to the DCD with a small number of participants. Focus group discussion provided participants the opportunity to engage with and build upon each other’s responses. In contrast, surveys allowed the team to reach a larger number of participants, but with no opportunity to build on other responses. Survey respondents and focus group participants had the opportunity to be more expressive through open-ended questions provided within each topic category, with the exception of the Lexington, KY, survey, which only provided one opportunity to give open-ended comments. A comparison of “before” and “after” surveys conducted online gives insight into public attitudes regarding DCDs prior to and postconstruction, with the understanding that limited inferences can be drawn due to a high likelihood of sampling bias.

Surveys

An analysis of responses from two online surveys is presented in this paper. The first was a User Satisfaction Survey conducted by the Kentucky Transportation Center for the DCD constructed in Lexington, KY.7 The second survey was conducted by the research team for the DCD in Rochester, NY.

The Lexington, KY, survey was completed by participants over a 1-month period following construction of the DCD. Three methods were used for survey outreach. First, attendees that had participated in a public meeting prior to the survey were provided the survey in online and traditional mail formats. Second, the survey was advertised on variable message boards in the corridor. Third, a link to the survey was given to a variety of government, news, and social media pages. In total, 1,421 responses were received, with 954 unique open-ended responses.

The surveys of the Rochester, NY, DCD consisted of a “before” and “after” analysis to allow comparisons of public perceptions prior to and following construction of the DCD. The survey was distributed to public meeting attendees via email and posted to the project website, where respondents could provide feedback to questions with regard to safety, congestion, avoidance, travel behavior, complexity of the interchange, and user comfort when walking, biking, or accessing transit. The online survey was open for a period of 1 month in the “before” and “after” periods, with the latter survey taking place 6 months after the official opening of the DCD. Surveys given in the “before” and “after” periods did not track opinions of the same respondent, although “before” survey respondents who elected to receive the “after” survey were notified when it became available. A total of 98 and 133 survey participants responded to the “before” and “after” surveys, respectively, and 159 and 184 unique comments were provided, respectively.

Respondent demographic information from the Rochester “before” and “after” survey is provided in Table 2. No demographic information was collected with the Lexington-based survey. Respondent gender type was fairly consistent before and after construction of the DCD. The vast majority of respondents during both time periods were between the ages of 36 and 65 years. Most respondents were from professional, scientific, and technical service backgrounds, with retired individuals and the health care and education professions following close behind.


Table 2. Rochester survey respondent demographic information.

Respondents’ Information

Before After

Respondents % Respondents %

GenderMale 47 52.8 66 60.0

Female 42 47.2 44 40.0

Age Group

16–25 3 3.4 6 5.4

26–35 10 11.2 11 9.9

36–50 26 29.2 35 31.5

51–65 37 41.6 44 39.7

66–75 8 9.0 12 10.8

Over 75 5 5.6 3 2.7

Profession

Transportation and warehousing 4 4.5 5 4.5

Information 6 6.7 7 6.3

Finance and insurance 5 5.6 4 3.6

Professional, scientific, and technical services

27 30.3 39 35.2

Educational services 6 6.7 9 8.1

Health care and social assistance 14 15.7 7 6.3

Other services (except public administration)

5 5.6 8 7.2

Retired 14 15.7 13 11.7

Others 8 9.2 19 17.1

Focus Groups

For focus groups, flyers and open-ended phone calls were utilized to gather a group of individuals from the community surrounding the DCD. Where possible, subjects were recruited by corresponding transportation agencies from contact lists made during public involvement meetings. In the focus group meetings, primary questions for each category were asked first and supplementary categorical questions were asked if time allowed for a more in-depth discussion of the topic. Since focus groups require face-to-face meetings, a conference room was utilized, where a member of the research team facilitated the discussion among the participants. Figure 2 illustrates the room setup and group setting.

Figure 2. Discussion at a DCD focus group.


Participants generally lived or worked in the immediate vicinity of the DCD. All participants were 18 years or older and possessed a valid driver’s license. At the focus group meeting, each participant provided baseline information including location of residence, where they were employed, how often they utilized the DCD interchange, and for what purpose they used the DCD. The types of participants represented at a typical focus group are presented in Table 3, which shows the breakdown of attendees at the Maryland Heights Focus Group.

Table 3. Typical focus group participation.

Respondent Work Type Type of Work Quadrant

Primary Reason for Driving through

Interchange

Approximate Distance from Interchange

(miles)

R1 Receptionist Public NW Work 0.3

R2 Receptionist Public NW Private, lives in area 0.3

R3 Manager Retail NW Work 0.5

R4 Cashier Retail NW Work 0.5

R5 Manager Hotel NE Work 0.1

R6 Manager Hotel SE Work 0.1

Topic CategoriesIn the focus groups and surveys, questions were centered on five specific topic categories. Questions addressed general knowledge of the DCD, opinions related to safety, opinions related to its design and operation, and finally concerns over bike/pedestrian and transit issues. Categories were utilized to ensure consistency in questioning and reporting and to ensure the appropriate themes were covered. These categories and related questions included:

■■ General knowledge and perception of the DCD

■● Have you ever heard of a DCD before this meeting?

■● What is a DCD? How is it intended to work? What is the purpose of a DCD (what problem does it solve)?

■● What are your reasons for driving through the interchange? Why use it instead of other roads?

■■ Opinions related to safety

■● How safe do you feel while driving through the DCD?

■● What kinds of roadway design, marking, signage, and roadway maintenance situations do you feel make the DCD safe and easy to navigate?

■● What driving behaviors do you feel are major causes of crashes at the DCD? Which ones do you think are a bigger problem on the DCD than on other roads?

■■ Operations and design

■● Do you feel the DCD is kept in good repair?

■● Do you ever encounter problems seeing lane markings and other pavement markings at the DCD? Under what conditions?


■■ Walking, bicycling, and accessing transit

■● A question about the safety of bicyclists and pedestrians was asked at the focus groups if time permitted, but the quality of responses depended on the level of discussion.

■■ Solutions

■● Participants were encouraged to give comments that built upon an idea or improved it, starting with the problem that most participants identified as being important.

Focus group and survey methods were used to obtain public opinion; therefore, some questions were slightly modified to fit the format. For instance, for the two surveys conducted in Lexington and Rochester, many questions were presented in a 1 through 5 format instead of an open dialogue as in the focus groups. In addition, survey respondents from Rochester were provided an opportunity to offer solutions to perceived problems both verbally and through a specific open comment section. No opportunities to offer solutions or specific questions regarding walking, bicycling, or accessing transit were given in the Lexington survey, as this survey was developed prior to the funding of this research effort. Even so, some comments from Lexington regarding those topic areas were recorded in the one open-ended section that was provided.

Before And After Survey Comparisons: Rochester, NYThe team conducted “before” and “after” surveys at the Winton Road/I-590 interchange in Rochester, NY, to gauge respondents’ opinions of the interchange project in regards to the topics previously presented in Table 3. The “before” survey was conducted while the interchange was still in a standard diamond configuration. The “after” survey was conducted 6 months after the DCD interchange opened to allow users time to acclimate. Since this was the only public outreach effort that spanned the “before” and “after” periods of the construction of a DCD, it will be discussed in more detail than previous considerations noted only after construction of the facility.

Demographic information was requested at the end of the online survey but is presented at the beginning of the discussion to establish a comparison of the survey samples (see Table 2). As the research team understands the issues with self-selection and the potential for stakeholder bias from comparing two convenience samples, the following comparisons are not intended as generalized results. These results are important, however, as this is the first type of survey of its kind regarding the comparison of public opinion of a DCD in the “before” and “after” conditions of a construction project. There is interest from professionals in understanding what type of shift in perception might occur between existing conditions (traditional diamond interchange) and the subsequent unconventional design (DDI). An effort has been made to ensure that outreach was conducted in a similar manner in each condition.

The New York DOT advertised the survey using contacts from a list developed from previous planning and preconstruction meetings with the public. Most of the responses were collected by an electronic survey system, and paper copies that were submitted were recorded in the online format. A total of 98 participants were surveyed in the “before” condition and 133 in the “after” condition. Of the 98 who participated in the survey distributed during the “before” condition, 55 provided their email addresses and were contacted with an opportunity to participate in the survey distributed in the “after” condition. While there is a potential for survey bias typical with online surveys, respondents had similar characteristics in terms of gender, being primarily male, and had a similar stratification of age (see Table 2). Of those who responded, 30 percent (“before” survey) and 35 percent (“after” survey) were in the professional, scientific, and technical services, which may indicate a large amount of interest from professionals related to the project (Question 18 [Q18] on the survey). Retired persons were the second highest group of respondents in both surveys.


General Knowledge and Perception of the DCD

Each “before” and “after” survey asked four questions to document overall understanding of the interchange before and after implementation and to see how frequently the interchange was used, any changes in general travel patterns through the interchange, and reasons for potential avoidance behavior. The four questions are summarized in Table 4.

Table 4. General knowledge and perception survey results

Q1. How often do you travel through the interchange at Winton Rd. and I-590?

Answer Options Pre-DCD [Diamond] After [DCD]

Never 1 1.0% 0 0%

A few times 10 10.2% 38 28.6%

Once or twice a week 23 23.5% 44 33.1%

Most days 12 12.2% 13 9.7%

Almost every day 52 53.1% 38 28.6%

Total answered 98 100% 133 100%

Q2. How often do you travel through the interchange during rush hour?


Never 6 6.1% 19 14.3%

A few times 27 27.6% 55 41.3%


Most days 10 10.2% 24 18.1%



Q3. How often do you take a longer alternate route (avoid the interchange at Winton Rd/I-590) to get to your destination?


Never 48 49.0% 90 67.6%

A few times 35 35.7% 29 21.8%


Most days 9 9.2% 3 2.3%


Total answered 98 100 % 133 100%

Q4. If you sometimes take an alternate route to avoid the interchange at Winton Rd/I-590, what reason is most applicable?


Rush hour traffic 41 41.8% 21 15.8%

Collision 8 8.2% 5 3.8%

Not applicable 41 41.8% 86 64.6%

Other (please specify) 8 8.2% 21 15.8%


Generally speaking, respondents were more frequent users in the “before” period than the “after” period, with 88.8 percent using the existing diamond interchange at least once or twice a week, compared to 71.4 percent who used the new DCD with the same frequency. When asked how often they used the interchange during rush hour (Q2), 76.3 percent and 44.4 percent said they used the


interchange at least once or twice during the “before” and “after” periods, respectively. It is unclear to what extent behaviors have been altered by the construction or the exact routes taken by those surveyed, but respondents surveyed indicated that a lower percentage of them used the DCD with some frequency after construction (Q2). Further, fewer reported taking alternate routes to avoid the interchange postconstruction (Q3). After DCD construction, 67.6 percent of respondents said they never avoided the interchange, compared to only 49 percent with the old design. Lastly, respondents were much more likely to take an alternate route due to rush hour traffic prior to construction of the DCD. Of the 21 who responded that they did take an alternate route in the “after” condition for a reason not listed, 13 respondents (approximately 62 percent) specified that it was due to the design or operation of the new interchange (Q4).

Opinions Related to Safety

Safety at DCDs is one of the primary concerns for designers and the driving general public and is regularly brought up in public meetings. The research team surveyed how drivers in Rochester recognized the overall safety of the interchange before and after DCD implementation during different times of the day and for different turning movements. The results are presented in Table 5.

Table 5. Safety survey results.

Q5. Using a scale of 1 to 5, with 5 being safest, rate the safety of the Winton Rd./I-590 Diverging Diamond interchange during the following times. Select N/A if you never drive through the interchange at that time.


Average Standard Deviation Average Standard Deviation

Morning rush hour 3.3 1.2 4.0 1.1

Lunch rush hour 3.6 1.2 4.1 1.1

Afternoon rush hour 2.7 1.3 3.9 1.1

Weekend 4.2 0.9 4.4 1.0

Total answered 98 117

Q6. Using a scale of 1 to 5, with 5 being safest, rate the safety of each of the following movements at the interchange. Select N/A if you never drive through that part of the interchange.



Driving through the interchange on Winton Road

3.67 1.16 4.13 1.03

Turning left from Winton Road onto I-590 ramp

2.95 1.23 4.34 1.06

Turning left after exiting I-590 onto Winton Road

3.35 1.14 4.31 1.10


Respondents rated the safety of the diamond and DCD interchanges on a scale of 1 to 5, with 5 being the safest. The results indicate that, overall, the DCD interchange was rated safer than the previous diamond interchange. When looking at specific periods of the day (Q5), the afternoon rush hour showed the greatest improvement in safety. Respondents generally reported that the morning and lunch periods were safer than before, though not as much as the PM peak.

Respondents rated the safety of the through movement and two left turn movements on a scale of 1 to 5, with 5 being the safest. Comparing “before” and “after” periods, the team found that the overall


safety perception related to each movement changed between the two interchange configurations. In the “before” survey, of the three movements, through movements were rated the safest. This response was likely due to the fact that vehicles turning left onto the freeway queued up significantly under the bridge at the second interchange signal. For the left turn movement off the freeway, traffic had to cross a conflicting movement at the downstream signal. In contrast, responses after construction of the DCD interchange rated the through movement as the least safe compared to the left turn movements. This may be attributed to the fact that the left turn onto the freeway is now a free-flow movement and is therefore perceived as safer. Overall, respondents rated all movements as safer in the after condition than the before condition.

Lastly, with regard to safety, respondents were provided the opportunity to express written opinions of the old and new interchange designs (Q7). The most common response in the “before” survey for the conventional diamond interchange was difficulty with traveling northbound on Winton Road and making a right turn onto the ramp to I-590 eastbound. A total of 14 comments described a stressful merging situation when attempting to complete this right turn maneuver. In some instances, comments described drivers not respecting a right turn only lane upstream of the interchange, using it as a right turn lane for a turn onto the ramp at the interchange. Also, the heavier right turn movement was required to yield to a lighter left turn maneuver onto the freeway from the northbound approach. The low volume of conflicting traffic from this left turn often led right-turning drivers to forget to yield, resulting in conflicts. Another seven comments referred to issues at an adjacent intersection, with a few comments regarding aggressive drivers, and a wide variety of other operational considerations.

In the “after” survey, the most frequent responses were the 22 comments related to driver confusion or lack of comfort—most reporting observations of others being confused (including wrong-way maneuvers) and several reporting that they experienced confusion traveling through the interchange. There were 14 generally positive comments regarding the design, 10 comments regarding issues at an adjacent intersection, and 7 comments regarding a desire for a change in signal timing. There were two comments for each of the following: issues with right turns onto the freeway, lack of safe bicycling infrastructure, and desire for a reduction in speed limit.

Operation and Design

Improved operation is often a primary motivation for considering a DCD. Table 6 presents results for the Rochester survey questions related to congestion, maintenance, and perceived complexity of driving through the interchange before and after implementation of the DCD.


Table 6. Operation and design survey results.

Q8. The current interchange is generally kept in good repair and does not encounter maintenance problems.


Strongly agree 8 8.3% 61 52.1%

Agree 51 53.1% 39 33.3%

Neutral 24 25.0% 16 13.7%

Disagree 11 11.5% 1 0.9%

Strongly disagree 2 2.1% 0 0%


Q9. Using a scale of 1 to 5, with 5 being least congested, rate the ease with which you are able to drive through the current interchange in the following scenarios:


AverageStandard Deviation Average

Standard Deviation

Driving through the interchange on Winton Road

3.3 1.3 3.9 1.5

Turning left from Winton Road onto I-590 ramp

3.0 1.3 4.1 1.5

Turning left after exiting I-590 onto Winton Road

3.2 1.3 4.1 1.5


Q10. Using a scale of 1 to 5, with 5 being least complex, rate your understanding of the interchange in its current configuration when you drive through it to get to your destinations.


AverageStandard Deviation Average

Standard Deviation

Understanding of driving through interchange

4.2 1.7 4.3 1.7


With regard to maintenance, survey respondents generally agreed that the new interchange was in good repair (Q8), with 85.4 percent of respondents agreeing the interchange was well maintained. Although the previous conventional diamond interchange received a lower rating of 61.4 percent in these two categories, this would be expected as the previous interchange was decades old compared to the new interchange.

Survey participants were asked to rate the level of congestion for each of three movements on a scale of 1 to 5, with 5 being the least congested (Q9). Each movement was rated as less congested in the “after” condition than in the “before” condition. There was a shift in which movement was regarded as the most congested of the three listed. In the before survey, through movements were regarded as the least congested, whereas in the after survey, turning movements were rated as the least congested.

Similarly, the survey asked users to rate the complexity of driving through the interchange (Q10) on a scale of 1 to 5, with 5 being the least complex. The average complexity rating for the existing diamond interchange was 4.21, whereas the average rating for the DCD interchange was 4.25, indicating little difference in the perception of complexity with the new DCD interchange.

Respondents were given an opportunity to provide additional comments regarding operations at the interchange (Q11). Of the 19 comments provided in the “before” survey, issues at an adjacent


intersection received 5 comments, followed by 2 comments each for topics covering lane configuration, drainage, issues with the survey itself, and no improvements needed at the interchange. A single comment was also given for issues with sight distance, backups, and hesitation regarding the new interchange.

Following DCD implementation, of the 35 comments given regarding operations in the “after” survey, signal-related concerns were the chief issue expressed, with 9 comments regarding signal phasing adjustments at the DCD and 7 comments that discussed issues at an adjacent intersection (primarily due to signalization issues), where 7 were primarily positive comments and 4 were chiefly negative comments; 2 comments were related to each of the following: additional signage, driver confusion, and inadequate bicycle facilities. Other comments given were either neutral or related to avoidance of the interchange.

Walking, Bicycling, and Accessing Transit

Accommodations for bicyclist and pedestrians can also be a concern when implementing an unconventional design. The survey asked respondents to rate how comfortable they were traveling on Winton Road during three activities: walking, bicycling, and accessing transit. The results are shown in Table 7.

Table 7. Walking, bicycling, and accessing transit survey results

Q12. Using a scale of 1 to 5, with 5 being the most comfortable, please rate your comfort when traveling on Winton Road during the following activities:



Walking 2.0 1.1 2.5 1.4

Bicycling 1.8 1.0 2.7 1.5

Accessing transit 2.5 1.2 3.2 1.5


The team asked respondents to rate their comfort level during the three activities (Q12) on a scale from 1 to 5, with 5 being the most comfortable. The overall comfort perception for all activities—walking, bicycling, and accessing transit—improved following DCD construction; however, it should be noted that facilities for pedestrians were incomplete during both the “before” and “after” conditions. Rating averages ranged from 1.75 for bicycling to 2.48 for accessing transit in the “before” survey, indicating the majority of respondents were not comfortable. In the “after” survey, averages ranged from 2.72 for bicycling to 3.17 for accessing transit. Although the range shows some improvement, the results suggest that there is still a general lack of comfort when bicycling or walking.

Respondents were given an opportunity to provide additional comments regarding walking, biking, or accessing transit at the interchange in both the “before” and “after” conditions. Although response rates were higher after DCD installation, a review of comments provided in Q13 appears to show little shift in public opinion regarding the safety or provision of bicycle and pedestrian facilities at the interchange. In the “before” survey, of the 14 comments that were given, there were 10 comments noting a dangerous or unsafe environment for bicyclists and/or pedestrians, 2 comments about inadequate facilities, and 2 comments about issues at the ramps. In the “after” survey, of the 27 comments that were given, 8 were to the effect that they did not use the facilities in this way or that they had no additional comment to provide. Of the remaining 19 comments, 11 comments noted a dangerous or unsafe environment and 5 comments noted inadequate facilities for bicycling and/or


walking. Other comments discussed inadequate maintenance (snow removal), failure to yield at the on-ramps, and circuitous/long routes for pedestrians to cross the roadway. There was one positive comment stating that the interchange looked like a “bicycle-friendly touring route.”

Suggested Solutions

In Q14, respondents were asked to provide “any additional comments about your experience at the interchange or suggestions on how to improve the interchange.” Due to the potential for multiple discreet comments in one response, these numbers do not always add up to the total. In the “before” survey, 19 comments were given on a very wide variety of topics, 9 of which did not offer suggestions for improving the interchange. The most common solution (4 comments) was related to making improvements at an adjacent intersection. Three comments about planning or making improvements for nonmotorized traffic were also given. Other comments included installing a DCD interchange, reducing the speed limit, and improving sight distance. In the “after” survey, after removing those that offered no additional comment, 28 responses were given, 13 of which offered solutions. Seven respondents reported positive feedback and offered no changes. Conversely, four respondents reported negative feedback and/or requested a “normal” interchange. Of the comments that were given offering solutions or pointing out issues, six were related to changes regarding timing of the lights through the interchange, five were related to adjacent intersection issues, four were related to improved signage, and two were related to driver confusion. A comment was also given on making a crossing improvement for pedestrians. These responses suggest general satisfaction by motorists with the new DCD interchange, with the majority of changes suggested pointing toward an interest in operational improvements.

Analysis of Public Outreach Responses from Focus Groups and Supplemental SurveysThe following sections provide summaries of the general findings from focus group and survey efforts based on the topic categories outlined above. These findings, especially those from the focus groups, often explain inconsistencies in results from the more quantitative surveys, along with providing more details about specific positive and negative outcomes of the DCD interchange. The general findings are reported by the four primary topic areas discussed earlier in the paper.

General Knowledge and Perception of the DCD

Participants were typically knowledgeable about the DCD. Many respondents indicated that the interchange was intended to assist with the flow of traffic and aid in traffic congestion, although some indicated little understanding of the specific details surrounding the design or operation of the interchange.

The general perception of the DCD obtained through varied outreach methods was mixed; however, the vast majority of the discussion and comments were cautiously positive. Concerns were expressed for a wide variety of reasons; however, individuals most frequently expressed reservations about:

■■ Driver confusion, inattentiveness, and unfamiliarity with the interchange;

■■ Wrong-way maneuvers (fear of meeting drivers head-on in the interchange);

■■ Need for exposure/experience with the interchange; and

■■ Concern for queuing and signal timing at adjacent intersections.

A few of the general respondent perceptions from the sites are discussed in the following section.


The large sample size in the survey conducted in Lexington, KY, allowed more insight about how local travelers versus commuters felt about their experience traveling through the DCD. In Lexington, daily drivers of the DCD were more likely to identify benefits in safety and operations from the project, while infrequent users were more likely to remain undecided (though more than 50 percent still identified benefits). When asked if it was easy to understand how to drive through the interchange, 86 percent of frequent (everyday) drivers agreed it was easy to navigate, while 73 percent of infrequent drivers agreed.

The majority of DCDs studied implemented signal timing that prioritized peak directional movements, such as left turns to and from the freeway. For this reason, it is likely that negative impacts to travel were more often cited by individuals traveling locally to destinations on the minor roadway as opposed to those commuting through the area to and from the freeway. Respondents living in and around the DCD often expressed that the new design had a negative impact on trips that required turns at adjacent roadways.

Focus group participants in Alcoa, TN, perceived an overall increase in problems after the installation of the DCD. The participants were particularly concerned with a lack of consideration for access to neighborhoods adjacent to the DCD, which resulted in limited access to right-in, right-out movements at the neighborhood entrance.

Respondents in the Rochester, NY, “before” and “after” survey indicated driving through the interchange was not complex, even when compared with the previous diamond interchange. Survey responses from both Lexington and Rochester indicate that traveling through the DCD interchange is somewhat of a learned task and requires exposure before drivers are truly comfortable, especially at those sites that have significant populations who do not use the facility regularly.

Opinions Related to Safety

The majority of participants agreed that the interchange was safer following construction of the DCD. Some expressed some concern when the interchange was first installed. Many comments were given regarding safety, and although there is no way to distinguish perceived versus real safety concerns, for the purposes of reporting user opinion, primary safety concerns most often provided by respondents include:

■■ Wrong-way maneuvers;

■■ Illegal U-turns at the crossovers;

■■ Red light running; and

■■ Sight distance issues.

Comments regarding wrong-way maneuvers were due to apprehension about driving through the DCD and encountering a vehicle head-on. Focus group participants felt this issue was likely more prevalent with visitors to the area or infrequent or distracted drivers. With regard to personal driving concerns, many participants mentioned that increased experience driving through the DCD interchange alleviated hesitation.

Respondents from two of the sites noted that red light running was a serious problem at their interchanges after DCD construction. At one focus group, this was discussed for a significant period of time before the moderator finally changed subjects. In the Rochester survey, respondents expressed that red light running was due to the duration of the two lights for the through traffic on the DCD being too short or timed improperly. It was reported that motorists could not make it through the DCD in one signal cycle, so frustrated drivers often ran the second light.


During the focus group in Alcoa, TN, participants discussed concerns with the height of two fill sections (berms), which caused sight distance issues: 1) on the southbound off-ramp turning left onto the cross street (Bessemer Street), where traffic coming straight through the preceding crossover was difficult to see, and 2) at the northbound off-ramp where right-turning vehicles merged with through traffic along Bessemer Street. As shown in Figure 3, material from the berms was removed after the opening of the DCD interchange to allow for better sight distance.

Figure 3. Sight distance issues caused by dirt fill were remedied after construction of a DCD in Alcoa, TN.

At the Dorsett Road DCD in Maryland Heights, right-turning vehicles at the freeway off-ramps were allowed to turn right on red. Based on feedback, there seem to have been quite a few conflicts involving these movements due to the close proximity of the off-ramp right turn movement to the DCD crossover. This effect was also noted during a parallel field data collection effort, where the ramp alignment, along with the raised median barrier, provided poor sight distance under the bridge.

Operation and Design

With the exception of the Alcoa site, participants’ comments generally stated that the interchange operated better than the previous diamond configuration. This may be due to the fact that the Alcoa DCD was constructed in anticipation of future build out and its original design was not a diamond interchange.

Concerns and areas for improvement in regards to operation and design were noted most often in responses from the focus groups and surveys in the following areas:

Source: Map data ©2014 Google Imagery ©2014 DigitalGlobe, U.S. Geological Survey, USDA Farm Service Agency.


■■ Ease of through movement at DCD;

■■ Coordination of signals at adjacent intersections;

■■ Queues and congestion at adjacent intersections;

■■ Opportunities for right turn on red (RTOR); and

■■ Positioning of advance signage.

Rochester survey respondents provided feedback on the complexity of overall driver paths with regard to congestion and noted that the through movement was slightly more difficult than the left turns on and off the freeway ramps. Several survey respondents expressed dissatisfaction with the signal operation limiting protected left turns to a side road movement adjacent to the DCD.

Several participants at all sites noted that coordination of signals seemed counterintuitive. Because the primary movements at four of the five sites surveyed are left turns off and onto the freeway, the downstream signal coordination provided less than normal bandwidth for through movements. This often caused drivers from upstream through movements to stop at the second downstream signal of the DCD. As noted in the safety issues, this may be the cause of red light running and has been reported to contribute to driver frustration.

Survey respondents from Lexington noted a significant problem with the closely spaced intersection to the south at Beaumont Centre Parkway. Respondents noted severe congestion at this particular intersection during the AM and PM peak periods, with queues backing up upwards of “a mile or more” through the DCD in the AM peak hour. In Rochester, respondents clearly noted problems with the two closely spaced intersections to the south. Respondents’ comments generally said, “It would be nice if we had better signal timing or synchronization with adjacent signals,” meaning they recognized that there were queue issues at the adjacent signals that had less mainline capacity than the DCD. Some participants indicated systemic issues, notably that while the capacity of the DCD interchange has improved, the adjacent intersections have now become problematic.

Survey respondents at the Lexington, KY, site noted that the no-RTOR condition on the eastbound off-ramp caused significant queuing at the site during the PM period. Respondents noted that the original design allowed a free-flow right turn movement. According to survey respondents, this restriction now caused significant queuing back to the freeway. Right turns at the freeway off-ramp were discussed at all sites, particularly as RTOR was prohibited at the majority of sites. The Kansas City site had a particularly interesting scenario because it was signed as no RTOR during the transition period immediately following construction. However, during the focus group session, which took place 6 months after the opening, the southbound right turn configuration was changed to a channelized right turn. Participants favored the free-flow channelized right turn lane but called out issues with weaving as vehicles tried to access driveway entrances on the opposing side of the road after making this right turn.

Participants were generally satisfied with signage at the sites. Respondents noted that the advance signing was instructive, but in some cases it should have been installed further back. Queues during peak periods along the arterials often spill back far enough that some drivers were not able to pre-position by the time they saw the signing—especially drivers wanting to turn left onto the freeway. This left turn movement sometimes leads to heavy left lane utilization at the first DCD signal. For instance, it was noted in Kansas City, MO, that trucks sometimes had problems with pre-positioning and stopped under the bridge, blocking through traffic as they tried to merge into the left turn lane to access the freeway on-ramp.


Walking, Bicycling, and Accessing Transit

Concerns for pedestrians were brought up at two of the focus groups. Feedback on bicycling (including transit) was generated primarily from the surveys, and feedback was primarily negative with the exception of the Kentucky DCD, where a paved bicycle and pedestrian pathway provided on the outside of the roadway was well received by the majority of commenters. It should be noted that the layout of the pedestrian facilities differed among the sites and that the Kentucky site is the only site that provided full access through the DCD on a separate facility. Issues that were highlighted in the surveys include:

■■ Placement and use of the walkway in the median;

■■ Long distances to crosswalks;

■■ Drivers disregarding pedestrian right-of-way at on-ramps or stopping in the crosswalk;

■■ Concern for bicyclist safety through an underpass facility due to space constraints;

■■ Snow removal from sidewalks; and

■■ Placement of “trail-sized” stop signs at crosswalks.

At the Kansas City site, pedestrians use the center median walkway to travel through the DCD. Participants discussed issues with the pedestrian facilities at this site in some detail. They noted that most pedestrians use the center walkway provided under the bridge but leave this facility prior to the intersection, choosing to negotiate walking along the side of the roadway rather than using the sidewalk that continues down the middle of the travel lanes to the next adjacent signal. They seemed to believe that a crosswalk location across the “y-line” at the actual DCD intersection might have been a better idea, as these users seemed to want to cross back over meeting pedestrian expectations. In Alcoa, there were no pedestrian facilities installed with the DCD, which focus group participants noted as a safety concern.

In Lexington, the Kentucky Transportation Center survey yielded 956 open feedback submissions on the DCD interchange. Among those, 17 comments were related to walking and bicycling conditions. Ten comments provided favorable feedback, mostly praising the new asphalt path providing access for bicyclists and pedestrians where none existed before. On the other hand, there were five unfavorable comments, which included three basic areas of concern. The primary shortfalls, according to respondents, were short pedestrian walk phases, drivers encroaching into the crosswalk, and issues regarding the use of trail-sized stop signs at crosswalks where pedestrians have the right-of-way.

In Rochester, 27 survey respondents provided specific feedback about pedestrian, bicycle, and transit facilities. Most respondents indicated that bicycling or walking through this interchange would be uncomfortable or dangerous, and comments were typically critical. One respondent noted motorists often ignore pedestrian right-of-way at the on-ramps to the freeway where pedestrian facilities are provided along the outside of the cross street. Areas of significant concern noted by bicyclists were the limited amount of space though a crossover next to a barrier wall and that a bicycle lane ends abruptly at a splitter island at the entrance to the interchange (Figure 4).


Figure 4. Issues with bicycle facilities at the Rochester DCD. Left: Bike lane ending at splitter island on the approach to the DCD interchange. Right: Barrier wall and inside shoulder limits bicycle operating space.

Conclusions and RecommendationsThe team drew several conclusions from the focus groups and surveys conducted at the five DCD interchanges regarding changes to DCD interchanges as a result of public perception through the focus groups and survey strategies used. At four of the five sites, surveys and focus groups were conducted following DCD implementation while a “before” and “after” survey was conducted at the Rochester, NY, location.

Scope of Project

When designing a DCD, the impact of the design on the surrounding intersections, commercial centers, and neighborhoods should be fully considered. Many focus group and survey participants noted that local access at all hours may have been sacrificed to improve the experience for commuters, particularly in the peak hours. It was suggested that a broader scope of impact should be considered prior to final design and construction. Although the general reception of the DCDs by motorists was positive, the design may need context-specific modification elsewhere given concerns related to access, bicycle/pedestrian accommodations, signage, traffic signal timing, number of lanes, signal treatment at adjacent intersections, education, and solutions to aid potential driver confusion, which can cause wrong-way maneuvers.

Operation and Safety

Participants typically clustered safety and operations comments together, which is understandable considering the complexity of many of the traffic scenarios described. Overall, participants were generally supportive of the DCD at each site location, and it was generally agreed that overall safety and operations had improved, although numerous respondents indicated that the interchange was confusing, especially for infrequent drivers. Many of the comments pointed out issues that persisted at adjacent intersections for turning motorists and frustration with signal timing both in the interchange and at adjacent intersections. Education and enforcement were recommended as the treatment for most problems. Law enforcement was cited as a way to increase driver compliance with speeds and could possibly deter illegal maneuvers.

Driver Understanding

Respondents generally reported that driving through a DCD interchange was no more complex than a conventional diamond interchange, although the comments from the surveys indicate


that observations of driver confusion were common. This finding suggests that the experienced commuters responding to the survey were not likely to experience driver confusion; however, novice drivers with less exposure may initially have problems. Many concerns related to driver confusion are often touted, and more research may be needed to understand to what extent this is true.

Effects on Nearby Intersections

The majority of negative comments given by survey or focus group participants were related to nearby intersections. Left-turning issues at adjacent signals were identified as creating long waits and queuing at some approaches of the adjacent intersection. Issues with adjacent intersections were identified at almost every location surveyed, suggesting that this problem was not as prevalent with the previous design. Field observations found this was a noticeable problem at some DCDs due to the efficient two-phase signal operation at the interchange providing additional thru traffic that adjacent intersections were not able to readily process.

Turning Movements

The problem with right turns at the off-ramps was brought up multiple times at all sites. Participants noted at four of the five sites that they could no longer turn right on red and that queues at many off-ramp locations backed up much farther than at the conventional diamond interchange. At the one site studied that allowed RTOR, participants noted a serious safety concern that needed to be addressed: that right-turning vehicles were frequently making poor decisions about when to turn on red. Focus group participants recommended moving the right turns at the off-ramps farther away from the primary intersection or providing channelized right turns with an acceleration lane to alleviate RTOR conflict and add additional capacity. In theory, these ideas sound promising; however, these movements will require additional right-of-way and some consideration on how to address the quick weaving maneuver to get to the left turn at an adjacent intersection in close proximity to the interchange. Safety for bicyclists and pedestrians would also need to be considered at any location where facilities for nonmotorized travelers may cross a free-flow ramp or channelized right turn lane, creating additional conflict points or the potential for more serious crashes.

In Rochester, left turn movements on and off the freeway were considered much safer than at the previous conventional diamond interchange. Respondents’ views on the safety and operation of the three primary movements (a through, a left onto the freeway, and a left off the freeway) switched after installation of the DCD. Instead of left turn movements being the problem, respondents now perceived through movements as the most congested.

Pedestrian and Bicycle Facilities

The majority of participants in the focus groups didn’t use the pedestrian and bicycle facilities, but the comments from the surveys expressed some concerns and problem areas in the design of facilities. If the pedestrian walkway is designed to use the center of the road, designers should recognize that pedestrians will immediately want to go back to the left or right side of the road once they are past the crossover. Care should be taken to make sure that these facilities have good sight distance for pedestrians and drivers, and if sufficient gaps are not available, signalization or some other treatment may be necessary to make sure pedestrians and bicyclists who use the sidewalk are safely accommodated. Lastly, if bicycle lanes are added to the roadway, adequate operating space must be provided adjacent to any barrier on the right-hand side of the roadway between the two DCD intersections. Bicycle lanes should not end abruptly at the crossovers.


Public Outreach

It is highly recommended that agencies installing a DCD conduct similar focus groups or surveys. This provides the public an opportunity to express their opinions, and these viewpoints often provide real issues and practical opinions that should be considered, especially as designs are vetted. Focus groups take more effort to organize due to the need to find a location and participants. As the research team learned through this effort, willing participants do not always attend. Conducting a similar comparison survey prior to and after the installation of unconventional interchanges is a potential way for designers and engineers to qualitatively estimate how new types of interchanges are perceived. A variety of free services exist for agencies to create and disseminate a public survey online. It is recommended that an outreach plan be established to increase response rates within a community and gain a wider sample of public opinion. If transit stops are located near an interchange, a targeted strategy may be needed to get a response from transit users.

Suggestions for Future Research

In order to break down the barriers to the use of DCD interchanges, public support and acceptance is an important area of future research. Similar before and after studies—using a method that can show a statistically significant result using the same respondents, or a survey that obtains a larger convenience sample—should be conducted to determine the success of a DCD project as perceived by the public and obtain results that can be more readily generalized. It is important to collect demographic information in order to understand the types of respondents who complete the survey (and to see if they are representative of the general population) and understand the bias that might be prevalent in the users’ opinions if a convenience sample is used. To gain a deeper understanding of respondent opinions, any method of outreach should give an opportunity for guided open-ended responses. These methods indicate areas in which to improve the DCD to gain public acceptance. These public outreach methods indicate that specific changes could be made to increase support for DCD interchanges and increase user satisfaction with the project. These include broadening the project scope to include adjacent interchanges, making timing adjustments with signal operations, and including facilities to accommodate both bicyclist and pedestrian travel. Additionally, further research could be conducted to determine the real and perceived effect of driver confusion and potential for wrong-way maneuvers at a DCD interchange.

AcknowledgmentsThe study on which this paper is based was funded by the Federal Highway Administration under project DTFH61-10-C-00029. The authors would like to thank FHWA for their support throughout this project. The authors also thank all of the participants in the focus groups and surveys, along with the officials at the five DCD sites. The views and opinions in the paper are those of the authors only. The authors are fully responsible for any errors or omissions.


Works Cited1. Hughes, W., R. Jagannathan, D. Sengupta, and J.E. Hummer. Alternative Intersections/Interchanges:

Informational Report (AIIR). FHWA-HRT-09-060. Washington, DC: U.S. Department of Transportation Federal Highway Administration, 2010.

2. Missouri Department of Transportation. Missouri’s Experience with a Diverging Diamond Interchange—Lessons Learned. Report OR 10-021. May 2010.

3. Chlewicki, G. “Learning from Six (Plus) Operational DDIs in the US.” In 2012 ITE Midwestern District Conference and TRB 4th Urban Street Symposium, 2012.

4. Shumaker, M.L., J.E. Hummer, and L.F. Huntsinger. “Barriers to Implementation of Unconventional Intersection Designs: A Survey of Transportation Professionals.” Public Works Management & Policy, Vol. 18, No. 3 (2013): 244–262.

5. Hummer, J.E., R.L. Haley, S.E. Ott, R.S. Foyle, and C.M. Cunningham. Superstreet Benefits and Capacities. 2010.

6. Bared, J.G., T. Granda, and A. Zineddin. Drivers’ Evaluation of the Diverging Diamond Interchange. Federal Highway Administration, 2007.

7. Kirk, A. US 68 & KY 4 Double Crossover Diamond Evaluation. Lexington, KY: Kentucky Transportation Center, University of Kentucky, 2012.

Kristy Jackson is a research associate in the Bicycle and Pedestrian Program at the Institute for Transportation Research and Education at North Carolina (N.C.) State University. She has provided support on a wide range of multimodal projects and has experience in research on bicycle and pedestrian modes, development of educational materials for bicycling and walking, and facilitating public involvement. She contributes to the advancement of bicycling and

pedestrian data collection in GIS and traffic monitoring. She holds a master’s degree in urban planning from the University of Wisconsin-Milwaukee and a B.S. from the University of Wisconsin-Madison.

Christopher M. Cunningham, P.E., is a program manager in the Highway Systems Group at the Institute for Transportation Research and Education (ITRE) in Raleigh, NC, USA. He specializes in operational analysis of transportation systems, applications of traffic analysis tools, safety analysis, and custom traffic engineering studies. He holds a bachelor of science and a master’s degree in civil engineering from N.C. State University.

Chunho Yeom is a research assistant at the Institute for Transportation Research and Education (ITRE) in Raleigh, NC, USA. He is studying for a Ph.D. at N.C. State University, participating in a diverging diamond interchange and freeway work zone capacity project. He holds a master’s degree in civil engineering from N.C. State University.

Joseph E. Hummer, Ph.D., P.E., is professor and chair of the Department of Civil and Environmental Engineering at Wayne State University in Detroit, MI, USA. He has specialized in research on alternative intersections and interchanges for 25 years. He has a Ph.D. from Purdue and an M.S. and B.S. from Michigan State University. He is a member of ITE.

Adam Kirk, Ph.D., P.E., PTOE, AICP, is a principal research engineer with the Kentucky Transportation Center at the University of Kentucky. In that role he manages the Center for Advanced Transportation Solutions Laboratory (CATSLab), a state-of-the-art traffic engineering laboratory used to direct academic and professional engineering learning and applied research. He holds a B.S. and a Ph.D. from the University of Kentucky and an M.S. from the University of

Washington. He is a member of ITE.


Global positioning systems (GPSs) have been used by both public and private entities to collect the locations of transportation assets and other spatial data including traffic signs, bus stops, bridges and culverts, and incident locations. In combination with geographic information systems, GPS data enable asset managers to track changes in assets and conduct “what if ” assessments to aid in improved system management. This research explores the use of GPS to inventory individual parking space data, and subsequently generate a geospatial parking management system for Clemson University. On a growing college campus, with new buildings now occupying former surface parking lots and peak parking utilization rates reaching 95 percent, accurate and aggressive management of this commodity is a priority.

While handheld GPS data collection devices provide opportunities to locate individual parking spaces for development of comprehensive space-by-space parking inventory maps, most published parking inventory studies have been conducted with pen and paper, limiting data to counts by lots or street segments. Collecting GPS locations for individual parking spaces provides a unique challenge due to the relative proximity of one space to another. Even small spatial errors that are inherent in GPS data collection not only are apparent but could compromise effective use of the data. This research combines the use of Wide Area Augmentation System–enabled GPS with digital aerial/satellite imagery and computer-aided design (CAD) as-built drawings to locate and inventory individual parking spaces. A methodology was developed to enhance the GPS locations to match the underlying CAD and aerial maps.

A Methodology Using GPS to Inventory University Campus ParkingBy Brian Maleck, Wayne A. Sarasua, Ph.D., P.E., Jennifer Ogle, Ph.D., P.E., and Kweku Brown


IntroductionClemson University is a land grant institution located in the South Carolina, USA foothills, with more than 19,000 students and growing. Along with the recent growth in the student population, new building construction has replaced some of the existing surface lots, and the demand for parking has become a significant issue—especially for major permit holders including student commuters and residents and employees. As these lots have been taken for new construction, the occupancy rates on campus have. Most parking managers agree on the 85 percent occupancy rule of thumb: “When parking occupancy exceeds 85 percent capacity, parking supply becomes constrained.”1 According to the 2012 Clemson University parking utilization study, the overall parking occupancy during weekday morning hours is 86 percent of available capacity. The highest occupancy rate on campus was 95 percent, and was observed for both student commuter and resident permit holders. Employee lots were not far behind at 89 percent occupancy. With this overall level of parking utilization, it is imperative to ascertain if there are any underutilized spaces on campus. Thus, the campus parking manager can ensure efficient use of all resources before making capital investments in new parking facilities. Not only are parking lots expensive, but they take up significant green space on campus at roughly 300 square feet per space.2 If parking garages are required, capital investments increase substantially, with an average per-space price tag of $30,000.2

Given the plethora of construction activities on campus over the last few years, the Parking and Transportation Services Office has been placed in a reactive position. The parking situation is further exacerbated by monumental demand for parking during Clemson home football games. The loss of spaces—particularly those central to campus—has necessitated significant shifts in permit designations for numerous lots along with changes in transit services and the adoption of several transportation demand management strategies (i.e., car-sharing, carpool, bike programs, etc.). The reconfiguration and re-marking of many parking lots has wreaked havoc on the campus parking inventory, because there was no formal method in place to track the individual space changes. In some instances, the permit type for whole lots has changed; for other lots, only a portion of the spaces have been redesignated. In some lots there were even additional spaces added to the lot during re-marking. Regardless, the actual number of parking spaces by permit type for a number of lots was uncertain going into this study. In addition to the sheer quantity of spaces by permit type, Clemson Parking and Transportation Services Office was interested in knowing precise details about all of the spaces on campus including:

■■ Exact location (latitude and longitude);

■■ Unique identifier (lot/street, row, space number);

■■ Space dimensions (length and width);

■■ Parking type (perpendicular, parallel, or angled);

■■ Permit/space designation (commuter, resident, employee, motorcycle, service vehicle, Americans with Disabilities Act [ADA], low emission vehicle [LEV], WeCar, carpool, etc.);

■■ If designated ADA, ADA compliance (cross-slope, running slope, ramp access to adjacent sidewalk, signage, markings, etc.); and

■■ Comments on problems (drainage issues, potholes, faded/missing paint or markings, etc.).

A detailed spatial inventory of spaces would allow for efficient work orders to be processed by lot or by problem type. However, with more than 12,000 parking spaces on campus, inventory data collection was viewed as a daunting task. Further, researchers were uncertain if the spatial aspects could be accurately collected with pen and paper, and the likelihood of data entry errors increases


with each successive processing step. Thus, researchers decided to attempt to use a mapping-grade global positioning system (GPS) receiver with a touch-screen data entry interface to conduct the inventory. This technology was chosen as the primary geocoding tool because the research team had experience using the device to assist in the collection of transit data and campus bike rack locations. This paper discusses the development and implementation of a methodology for using GPS to conduct the Clemson campus parking inventory. Various issues of concern related to GPS data collection are covered, along with recommendations for successfully adopting these methods at other sites. The final system is envisioned as a tool for future campus planning activities as well as for aiding in better management of the parking assets.

Literature ReviewAsset management is a strategic method of allocating resources to operate and maintain infrastructure for optimal performance.3 Strategic implies that there are goals, objectives, and performance targets to be met with respect to the infrastructure, and that all decisions for funding distributions would refer to this guidance to ensure the most efficient use of resources to attain the goals. Asset management systems improve efficiency, increase productivity, add accountability, and increase the ease of use of assets. These benefits come from a data-focused approach to managing systems.4

Advancements in technology have made data collection easier and more efficient (i.e., paper and pen–based collection to electronic collection, survey to GPS, etc.).5 More efficient collection allows more comprehensive and accurate data to be collected, but if the agencies are not prepared to use the data by linking them to decision-making at all levels, the data may not be worth the investment. In some cases high-level performance data such as overall parking occupancy rate may be enough data to make a decision at the system-monitoring level. However, in a capacity-constrained environment, such as the one that Clemson University is operating in now, more data are needed to best utilize all available parking infrastructures and select projects that will extend the life of the existing infrastructure. An example is capturing the occupancy rate for specific transportation demand management spaces (i.e., carpool, motorcycle, and even bicycle parking).

Many departments of transportation (DOTs) have realized the capabilities of GPSs to collect the locations of numerous transportation assets such as signs, bridges, transit infrastructure, and roadway conditions. North Carolina, USA, and Louisiana, USA, DOTs have employed GPSs to collect sign inventory data to comply with regulations.6,7 These states cite that GPS was highly effective in combination with a geographic information system (GIS) to map signs in the state. GPS accurately provides the location of the signs, and GIS can display the locations along with associated attributes, such as retro-reflectivity, condition, and type of sign. GPS has also been used to collect information on bus stops in Atlanta, Georgia, USA, prior to the 1996 Summer Olympic Games. In this study, handheld GPS devices combined with postprocess differential correction for enhanced spatial accuracy were used to collect location information. A user interface on a notebook computer was used to collect bus stop attributes such as marker type, whether or not a shelter was present, and if access to the bus stop was ADA accessible.8 GPS and GIS have also been used to inventory large sections of pavement for condition assessments, as well as siting bicycle parking facilities.9,10

While handheld GPS data collection devices have been used since the mid-90s for various asset management applications, they have not been widely adopted for parking asset management. Most published parking inventory studies only record the location and type of parking facilities down to lot/garage level for off-street spaces, and down to street block level for on-street spaces. While much of this is collected by pen and paper, some GISs have been developed to visualize and analyze the zonal data.11,12 A variety of parking studies have been conducted in the last few years that focused on downtowns.1,2,13–23 The typical goal of these studies is to determine the supply and demand for parking in the downtown area to ensure adequate parking is available for all users. It should be noted that all


of these studies were conducted using paper forms, and inventories were usually collected just prior to the utilization data collection—indicating that most cities do not have a formal parking asset management system. There are a number of differences between downtown parking management studies and those conducted for college campuses. For example, university campuses are often self-contained and all parking is on university property, so ownership data are not necessary. Further, universities usually provide permits to regular users (employees and students), and thus only visitors may need to park in pay lots, so the cost of the parking is typically tied to the permit. In downtowns, parking is usually inventoried by lot or street; however, on a university campus with broad zone parking, users buy a certain type of permit and park in spaces designated for that permit. This is the type of parking system that Clemson University employs. From an inventory perspective, this can be complicated because there may be 2 service vehicle spaces, 20 faculty spaces, 2 ADA spaces, 80 student spaces, 1 carpool space, and 2 LEV spaces in one parking lot. Thus, inventory data and tracking at the individual space level are important.

Some cities, such as Bloomington, Indiana, USA, actually have GIS layers with individual parking spaces identified, but the data were digitized from existing drawings and design files.24 So individual space data exist, but they were not collected via GPS. Today, new technologies exist that can provide real-time parking space occupancy on a space-by-space basis using an embedded sensor. An added benefit of these technologies is that they are tied to GISs and can wirelessly transmit data to populate online map applications used by drivers to navigate to available parking.25 Unfortunately, there is limited application of the new technologies, and most have been used for special parking situations like accessible parking spaces and visitor parking spaces.

Collecting GPS locations for individual parking spaces provides a unique challenge due to the relative proximity of one space to another. Even small spatial errors that are inherent in GPS data collection are not only apparent but could compromise effective use of the data. In addition, vehicle occupancy within the parking space precludes a data collector from obtaining the centroid of the space using GPS. The accuracy of GPS points can be absolute or relative. Absolute accuracy compares the data that a GPS collects to the actual location on the earth. Relative accuracy refers to the relative positioning of collected points in reference to one another. Using the GPS coordinates of two collected points, the distance is easily calculated. If the actual distance is equal to the calculated distance, then the relative accuracy of the two points is good. It is possible for points to have good relative accuracy but have poor absolute accuracy if the cluster of points is systematically shifted from their true location. Relative accuracy is just as important (and maybe more so) than absolute accuracy because of the relative proximity of parking spaces.

A standard GPS device such as a smartphone or GPS device in a car can have an absolute accuracy of about 15 meters. This is accurate enough for many applications; however, this would not be sufficient to identify individual parking spaces. Extra accuracy can be achieved using a process called differential correction. This method of improving accuracy relies on two receivers, one at a known location and another collecting data. The station at a known location calculates ranging parameters based on satellite signals and compares this information to actual ranging parameters between the known location of satellites and the known location of the station. Differences in ranging parameters are calculated and applied to the data being received by the second device. This will correct for a variety of errors and increase accuracy to within a few meters.26

A very convenient type of differential correction that only requires one receiver is provided through the Federal Aviation Administration’s Wide Area Augmentation System, or WAAS. WAAS works via 25 ground stations that monitor satellite data. These stations compare the data to those of reference stations and correct for orbit, atmospheric, ionospheric, and clock drift errors. This corrected message is sent to one of two satellites in a fixed position over the equator. These satellites broadcast the corrections, which are processed by compatible GPS receivers in real time. The result of this


correction is accuracy of less than 3 meters.27 Actual performance measurements of the system at specific locations have shown that it typically provides better than 1.0 meter laterally throughout most of the contiguous United States and large parts of Canada and Alaska, USA.28 WAAS correction is available without cost to anyone using a WAAS-enabled GPS receiver.

Real Time Kinematic (RTK) GPS can enhance GPS quality even further. This method of improving accuracy also relies on two receivers, one at a known location and another collecting data; however, these units use precise measurements of the phase of the GPS signal’s high-resolution carrier wave, rather than the low-resolution GPS code contained in the signal. Surveying-quality RTK GPS can provide sub-centimeter-level accuracy in real time; however, these receivers can cost many thousands of dollars and at least two are needed to achieve the specified accuracy.29

This research combines the use of GPS with digital aerial/satellite imagery and computer-aided design (CAD) as-built drawings to locate and inventory individual parking spaces. A methodology was developed to enhance the GPS locations to match the underlying CAD and aerial maps. The GPS receiver used in this research is WAAS enabled. There are several other potential errors associated with GPS data collection that will be discussed further in the Results section of this paper.

Methodology The GPS device chosen for the parking inventory was acquired from PinPoint GeoTech, LLC, specifically designed for public works inventory applications. The device is a mapping-grade unit with enhanced accuracy through WAAS real-time differential correction. This device was chosen over a surveying-grade unit because of its ease of use and relatively low cost (~$3,000). Another benefit is that PinPoint’s offices are located in Clemson, close to campus. This provided timely local customer support for any problems with the GPS device. Further, the PinPoint device’s customizable interface facilitated creating a user interface that would simplify data collection.

Initial Field Test

One of the initial steps in preparing for the inventory was to determine if the PinPoint GPS would be practical. A test was conducted on a sample row of parking spaces to ensure that the GPS would give acceptable accuracy for the collection and be suitable for collecting the related parking space attributes. Parking spaces provide a unique challenge due to the relative proximity of one space to another. Even small spatial errors that are inherent in GPS data collection are apparent. Prior to conducting the test, the interface was edited to simplify the collection of parking attribute data related to each space. The test data were collected in June 2012. The collected data were downloaded from the PinPoint GPS and were transferred to a GIS for display and analysis. The mapped GPS points were overlaid onto a Bing digital orthophoto, and comparisons were made with regard to a point’s location and the underlying parking space displayed on the map. Clemson standard perpendicular and angled parking spaces are typically 8.5 feet wide, and location data were collected at an approximate midpoint between the parking space lines for each space. Thus, an error of +/- 4 feet, which is possible with a WAAS-enabled GPS, should theoretically fall within the parking space lines. Upon inspection, initially none of the GPS points fell within the parking spaces displayed in the orthophoto. However, the relative distance between each collected point was 8.5 feet on average. Thus, the relative accuracy of the PinPoint GPS was quite good. The absolute accuracy was actually better than first believed, because it was later found that the Bing map was shifted slightly from nearby ground control points. Moving the digital orthophoto so that one of the points fell within its corresponding parking space resulted in all of the points falling within their corresponding parking spaces.


Campus-wide Data Collection

After the successful test, the researchers proceeded to collect field data in each parking lot on campus. Because GPS satellites orbit the earth twice a day, satellite coverage varies continuously. A measure of satellite coverage is known as position dilution of precision (PDOP). A PDOP of 1 is the best value possible. Values under 3 are very good. Figure 1 shows an example of mission planning charts during one of the collection days. The top figure shows how PDOP varied throughout the day, deteriorating after 12:00 p.m. and recovering at about 1:30 p.m. The sky plot on the bottom left shows how all seven available satellites are clustered above an elevation angle of 30 degrees, resulting in a PDOP of 4.7. Only an hour later, there were 11 satellites at or above the 10-degree mask angle and the PDOP improved to 1.5. This illustrates the importance of conducting mission planning prior to doing a GPS survey to maximize accuracy.

Figure 1. GPS mission-planning PDOP and Skyplot charts.


Clemson University does not have parking garages but does have varied cover in the parking lots. Most lots have some spaces that are near or under trees, or nearby buildings that can degrade PDOP below the maximum value possible identified during mission planning. The buildings can also cause multipath error, further degrading accuracy.

Office Processing

After the parking space data were collected, they were transferred to a desktop computer using PinPoint GeoTech software. The data upload process was completed after every data collection session to minimize risk of data loss. Once a data set was moved from the device to the computer, it was exported into a format that could be opened in Microsoft Excel. The data were initially saved by lot name but were eventually merged into a single spreadsheet containing totals for each lot and for the entire campus. The data were also converted to AutoCAD and ArcGIS.

The next step was to check attribute information to make sure all of the fields were populated correctly. PDOP values were reviewed to make sure they were acceptable and matched the mission-planning values. Some values were higher, but this was expected due to the tree canopies and nearby buildings in some areas of the parking lots.

AutoCAD As-built Map Update

Once all data were collected and exported from the PinPoint software into Microsoft Excel, the data were used to update the CAD map of campus. The Clemson University Facilities Department maintains an AutoCAD as-built map of campus. This map includes buildings, roads, utility lines, and lighting. However, a glaring omission from this map was parking spaces. The outlines of parking lots, including curb area, are included in the Facilities map. Therefore, using the row and space number attribute information collected during the inventory, each parking lot row can be filled in with the appropriate number of parking spaces. The number of parking spaces in each row was obtained using the countif function in Microsoft Excel. Knowing the number of spaces in each row, an array of spaces was drawn in AutoCAD using common dimensions and accounting for the wider handicap spaces. Figure 2 shows an updated as-built map before and after the parking spaces were added.

Figure 2. AutoCAD as-built map before and after parking spaces were added.


The original as-built map from Campus Facilities was geo-referenced to South Carolina State Plane coordinates. This made it easy to import into ArcGIS and align with a background digital orthophoto from Bing Maps. Both the AutoCAD map and the digital orthophoto map had to be adjusted based on control points to ensure that the two maps matched and were spatially accurate. Figure 3 shows the corrected overlay for one of the parking lots. The figure shows that the CAD parking lot lines lined up reasonably well with the digital orthophoto lines. In some cases it was necessary to adjust some of the CAD parking lot lines because the parking space width was not always uniform.

Figure 3. Overlay of adjusted orthophoto and CAD maps.


Displaying and Adjusting GPS Points

The GPS point data were imported into the GIS using longitude and latitude coordinates (WGS84 Datum). When the data were first imported, it was found that the individual spaces did not match exactly with the AutoCAD map. While it was expected that there were going to be absolute positional errors, the relative positioning errors were greater than anticipated. Figure 4 shows a plot of the GPS points overlaid onto the AutoCAD drawing. Errors are clearly evident; however, one-to-one correspondence between GPS points and the parking lot spaces was not difficult for a number of reasons. First, the majority of the collected data were relatively close to the actual spaces. Second, because there was usually reasonable separation between parking lot rows, it was relatively easy to determine what spaces were associated with each row. In all cases, the number of spaces collected for a row matched the number of spaces in the CAD map. Note that the orthophoto map could not be used for this because some spaces were entirely covered by trees. Third, included in the collected attributes were the row numbers as well as parking space numbers. The researchers were careful to collect spaces in order along a row; thus, the space number could be used to check if the positional errors caused spaces to be out of order. In order to correct the positional errors, the editor tool in ArcGIS was used. Each space was moved on the display to match a particular parking space drawn in AutoCAD. This process was repeated for all of the lots on campus. Figures 4 and 5 show raw and corrected point data for one of the lots.

Figure 4. GPS point data before correction.


Figure 5. GPS point data after correction using the ArcGIS edit tool.

GPS ErrorsThe research team did an investigation on the cause of errors to see why the relative (and absolute) positioning of the point data was not as anticipated. PDOP was the first measure that was checked. Note the area circled in Figure 4. A few of the points at this location seem to have experienced greater relative error than other points. Initially, the researchers were confident that this was due to an instantaneous change in PDOP and satellite availability; however, this was found not to be the case. In fact, the PDOP values for the entire lot were all under 2.5 and were actually in the 1.5 to 1.7 range in the area circled in Figure 4. The satellite availability also did not change in that area. The next possible cause of relative error was thought to be GPS multipath error. While this parking lot has trees in the medians, there are no nearby buildings that would extend above the 10-degree mask angle. Further, there are relatively few trees in the vicinity of the circled area; thus, multipath problems did not seem logical for this location. One other possibility was that the WAAS differential correction was not applied to these points. Unfortunately, the PinPoint device did not include an attribute on whether or not a data point was corrected, so there was no way of telling if this was the case. Some other problem had to exist.

Data Collection Unit Truncated Decimals

Table 1 shows a sample of the GPS data collected for one of the parking lots. The attributes shown are those that are collected by the GPS unit (i.e., Point ID, Positional Dilution of Precision, Latitude, Longitude, and Number of Satellites Visible [SV]). Other attributes, such as those collected by the researcher in the field, are not shown. Upon exploring the data from the PinPoint device, it was evident that some of the coordinates for some of the points were truncated. It was not clear why this


happened—even after talking to the PinPoint representative—except that it likely had something to do with zeroes after the decimal point. In the worst cases, some coordinates were truncated to only four decimal places, while others were truncated after the fifth decimal and still others after the sixth decimal place. It was calculated that truncating to four decimal places could potentially lead to coordinate errors of +/- 40 feet or more depending on whether both the longitude and latitude values were truncated. Likewise, truncating to five decimal places could potentially lead to coordinate errors of +/- 4 to 6 feet. The circled points in Figure 4 had truncated coordinates, which explains the relative error problems with the points at that location.

Table 1. PinPoint device sample attributes.

Point PDOP Date and Time Latitude Longitude No of SV

25 2.0 6/19/2012 12:59 34.67293667 -82.83529833 7

26 2.0 6/19/2012 13:00 34.67291833 -82.835295 7

27 2.0 6/19/2012 13:01 34.67288 -82.835305 7

28 2.8 6/19/2012 13:01 34.67285667 -82.83532 7

29 2.8 6/19/2012 13:02 34.672845 -82.83533833 7

30 2.8 6/19/2012 13:02 34.67279333 -82.83529 7

31 2.8 6/19/2012 13:03 34.67283 -82.83526 7

32 2.8 6/19/2012 13:03 34.67287167 -82.83525833 7

33 2.9 6/19/2012 13:04 34.67289333 -82.83523833 6

34 2.9 6/19/2012 13:04 34.67290667 -82.83522333 6

35 2.9 6/19/2012 13:05 34.67293 -82.83522 6

Testing with a Different GPS Unit

Because of the systematic errors associated with the PinPoint device, the researchers decided to collect additional data with a different WAAS-enabled GPS unit. In this experiment, a slightly more expensive (~$5,000) Trimble GeoExplorer 6000 XT was used. The GeoExplorer is also capable of collecting a customizable set of attributes. The locations of the raw data overlaid onto a CAD as-built map are shown in Figure 6. For the most part, the points shown in Figure 6 line up nearly perfectly with the underlying parking spaces shown on the CAD layer. The PDOP values for all of the points were less than 3 except for the three points on the west side of the map that had PDOP values of more than 6. Higher PDOP values resulted in these points being farther from their correct locations (as shown by the arrows). These high PDOP values were due to the research team purposely blocking part of the receiver’s antenna so that some satellites would not be in view of the receiver.

Table 2 shows sample attribute data from the GeoExplorer. The coordinate data collected by the GeoExplorer were projected into South Carolina State Plane cordinates, with the units being feet. The data were collected to three decimal places. As can be seen, the values here are not rounded, resulting in more precise coordinates. Note that one of the attributes shown in the table is “Correction Type.” SBAS stands for satellite-based augmentation system. WAAS is classified as an SBAS. Not only did the erroneous points in Figure 6 have high PDOP values, but they were also classified as uncorrected, which would also explain their error relative to the other points. The test of the GeoExplorer demonstrates the effectiveness of a WAAS-enabled GPS device for collecting parking data assets. Data points with low PDOP are very accurate and can be used for inventory and asset management purposes.


Figure 6. GPS points collected with a Trimble GeoExplorer 6000 with associated PDOP values (note that higher PDOP values indicate greater positional error).

Table 2. Sample GeoExplorer attribute data.

Point PDOP Date Time Northing Easting Correction Type

45 1.9 7/1/2013 09:56:16 am 1040178.808 1449124.477 Real-time SBAS Corrected





50 3.5 7/1/2013 09:59:08 am 1040180.286 1449178.754 Uncorrected

51 3.5 7/1/2013 10:00:38 am 1040200.819 1449207.964 Uncorrected

52 6.9 7/1/2013 10:01:10 am 1040207.840 1449221.468 Uncorrected

53 6.6 7/1/2013 10:01:35 am 1040194.587 1449242.878 Uncorrected



Conclusions and RecommendationsThis paper investigated the use of multiple WAAS-enabled GPS units to inventory parking asset data on Clemson University’s campus. Using underlying CAD as-built and digital orthophoto maps, the GPS data were easily geocoded so that the points corresponded to the underlying parking spaces. The lower-cost unit experienced errors in the GPS data caused by coordinates being rounded or truncated by either the GPS unit or by the processing software. The second, more expensive WAAS-enabled GPS unit was tested, and its accuracy was found to be much better than that of the first GPS unit. Only a very minor correction was needed for a small percentage of points. The ability to collect parking space data using mapping-quality GPS receivers shows that much more expensive surveying-quality units are not necessary. Surveying-quality receivers are also more time-consuming


to use, because they require a base station to be set up for differential correction (unless continuously operating reference station [CORS]/online positional user service [OPUS] is used). While the second GPS unit proved to be much more accurate than the first, it still had issues in locations where the GPS and/or the WAAS signals were blocked. Even these erroneous points could be located accurately by examining the initial maps and the associated attributes.

The techniques used in this research should be easily transferable to other locations. While digital orthographic maps are available at a reasonable scale throughout the United States, the quality of these maps may not be as good outside the United States. One of the advantages of using Clemson University as a test bed for this application is the availability of CAD as-built maps. CAD map coverage is not widespread, even in the United States. However, the editing of collected GPS points could have been done almost as efficiently using just digital orthophotos as reference. It is noteworthy that using digital orthophotos to conduct the inventory without the use of GPS in the field would have resulted in a significant number of missed parking spots due to the tree canopy and the difficulty of clearly identifying parking lot lines.

With the completion of this parking inventory system, Clemson University now has an accurate count of spaces, designations, and accompanying utilization data. Spot utilization studies have been conducted on spaces that had lower than average utilization rates to determine if these would best serve the campus transportation system under a different permit designation. Several areas have been converted to provide additional spaces for commuter/residential and employee permit holders to bring these utilization rates down from 95 percent and 89 percent, respectively. While this is likely not a long-term solution given campus expansion plans, the short-term provision of redesignating spaces to meet demand does relieve the problems temporarily to buy time for longer-term solutions to be studied and funded. A key aspect of the inventory system was a fully documented change order system to identify when changes are made to add or remove spaces or redesignate spaces to ensure an accurate portrayal of Clemson’s parking supply. Several options are on the table for long-term planning, such as building a new surface lot, disallowing freshman from bringing cars to campus, and setting up off-campus park and ride lots, as well as a host of other transportation demand management strategies. However, it is clear that Clemson University will make every last space count along the way.

Works Cited1. “City of Bozeman, Montana Parking Study.” Bozeman Unified Development Ordinance. City

of Bozeman, Montana, January 30, 2010.

2. Denver Strategic Parking Plan. City of Denver, CO, October 2010.

3. Asset Management Primer. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, Office of Asset Management, 1999.

4. Asset Management Data Collection for Supporting Decision Processes. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, 2006.

5. Bonsall, P.W. “The Changing Face of Parking-Related Data Collection and Analysis: The Role of New Technologies.” Transportation, Vol. 18, No.1 (1991): 83–106.

6. Vereen, S., J. Hummer, and W. Rasdorf. A Sign Inventory to Assess and Control Liability and Cost. FHWA/NC/2002-017. Raleigh, NC: North Carolina Department of Transportation, 2002.

7. Wolshon, B. Louisiana Traffic Sign Inventory and Management System. Report No. 381. Baton Rouge, Louisiana: Louisiana Transportation Research Center, 2003.


8. Sarasua et al. “Design and Development of a Bus Stop Inventory to Support an Intelligent Transportation System: The Marta Experience.” Journal of Public Transportation, Vol. 1, No. 4 (1998).

9. Huang, B., T.F. Fwa, W.T. Chan, and S.B. Liu. “Implementation of GIS/GPS-Based Airport Pavement Management System: A Case Study.” 6th International Conference on Managing Pavements, Vol. 10, Queensland, Australia, October 2004.

10. Leigh, C., J. Peterson, and S. Chandra. “Campus Bicycle-Parking Facility Site Selection: Exemplifying Provision of an Interactive Facility Map.” Proceedings of Surveying & Spatial Sciences Institute Biennial International Conference, Vol. 28, September 2009.

11. Dunn, M., R. Shields, and W. Yue. “The Application of GIS in CBD Parking Management.” Journal of the Eastern Asia Society of Transportation Studies, Vol. 3, No. 3 (September 1999).

12. Lau, Wilfred W.T., et al. “The Hong Kong Second Parking Demand Study.” Proceedings of the ICE-Transport, Vol. 158, No.1 (2005): 53–59.

13. Rich and Associates. “City of Billings Downtown Parking Plan.” City of Billings, Montana, January 2010.

14. Carl Walker, Inc. and Crandall Arambula. “Task Report: Current Parking Program Assessment & Supply/Demand Summary.” City of Missoula, Montana, 2008.

15. Carl Walker, Inc. “City of Atascadero Parking Utilization and Management Study.” City of Atascadero, California, 2006.

16. Rich and Associates. “City of LaCrosse Parking Study.” City of LaCrosse, Wisconsin, May 2009.

17. Desman Associates. “City of Meadville Parking Study.” City of Meadville, Pennsylvania, December 2008.

18. Wilbur Smith Associates. “Menlo Park Downtown Parking Study.” Menlo Park, California, August 2010.

19. Rick Williams Consulting. “Downtown Oregon City Parking Study.” Oregon City, Oregon, April 2009.

20. Rick Williams Consulting. “Downtown Springfield Parking Study.” Springfield, Oregon, March 2010.

21. Rick Williams Consulting. “Downtown Redmond Parking Study.” Redmond, Washington, January 2008.

22. Walker Parking Consultants. “Parking Study, Downtown Valparaiso.” City of Valparaiso, Indiana, January 2010.

23. Rich and Associates. “Village of Westmont Parking Study.” Village of Westmont, Illinois, November 2008.

24. City of Bloomington. “GIS Parking Lot Spaces Layer Metadata.” (2001) [Online]. Available: http://bloomington.in.gov/documents/viewDocument.php?document_id=804.

25. Waten, G., D. Musiega, and C. Ndegwa. “A GIS Based Parking Management and Dissemination System.” International Journal of Science and Research, Vol. 2, No. 7 (July 2013).

26. Hofmann-Wellenhof, B. and H. Lichtenegger. Global Positioning System: Theory and Practice, 5th ed. Springer, 2008.


27. Federal Aviation Administration. Navigation Programs-Wide Area Augmentation System (WAAS). [Online]. Available: www.faa.gov/about/office_org/headquarters_offices/ato/service_units/techops/navservices/gnss/waas/. [Accessed July 2013.]

28. National Satellite Test Bed (NSTB). Wide-Area Augmentation System Performance Analysis Report, July 2006. FAA/William J. Hughes Technical Center, NSTB/WAAS T&E Team, Atlantic City International Airport, NJ [Online]. Available: www.nstb.tc.faa.gov/REPORTS/waaspan17.pdf. [Accessed July 2013.]

29. Bossler, J., R. McMaster, C. Rizos, and J. Campbell. Manual of Geospatial Science and Technology, 2nd ed. CRC Press, 2010.

Brian Maleck, E.I.T., received his B.S. in civil engineering from Rensselaer Polytechnic Institute in 2011 and his M.S. in civil engineering from Clemson University in 2013. He was an active member of ITE while at Clemson, and was the first Communications Chair of the organization. He is currently pursuing a career in the parking industry.

Wayne A. Sarasua, Ph.D., P.E., is an associate professor in civil engineering at Clemson University. He is a past president of the South Carolina Section of ITE and is currently a co-advisor of the Clemson ITE Student Chapter. He has chaired, co-chaired, or participated in numerous conferences and committees for the Southern District and the Georgia and South Carolina Sections of ITE. Wayne was the 2012 recipient of the

Southern District ITE Excellence in Transportation Engineering Education Award. His specialties are in planning and design of transportation systems, transportation asset management, and spatial data analysis. He is a member of ITE.

Jennifer H. Ogle, Ph.D., P.E., is an associate professor in the Glenn Department of Civil Engineering at Clemson University. She is currently a co-advisor of the Clemson ITE Student Chapter and former recipient of the ITE Dan Fambro Paper Award. Her research interests focus on transportation infrastructure management, highway design, transportation safety, and pedestrian accessibility.

Kweku Brown, E.I.T., received his civil engineering M.Sc. from the University of Connecticut. He is currently a doctoral candidate at Clemson University. He is a past president of the Clemson ITE Student Chapter and a member of the South Carolina Section. He was the recipient of a 2013 Dwight Eisenhower Graduate Fellowship. His research focuses on transportation safety utilizing geographic and spatial analysis

methods. He is a member of ITE.


IntroductionThe modern roundabout has become an increasingly popular form of intersection control in the United States due to its effectiveness in improving safety and reducing traffic congestion. Since the first modern roundabout was built in Nevada in 1990, the number has increased significantly, and as of December 2012, more than 2,000 have been constructed.1

The purpose of this paper is to provide insight as to how additional lane length on an approach and/or departure affects roundabouts. Currently, there is no United States guideline on how additional lane lengths affect roundabout operation, so most transportation professionals refer to studies conducted overseas that do not necessarily translate directly to domestic roundabout design and operation. As the number of modern roundabouts in the United States continues to increase, the desire is to provide effective information to professionals on roundabout operation and design for conditions suitable for their jurisdiction. Using delay as the measure of effectiveness, a hypothetical four-leg, double-lane roundabout with additional lane length design at both entry and exit was analyzed, and the additional lane lengths were varied at both the entry and exit. Similar length variations were applied to an existing roundabout with known data after calibration and validation. The findings from this study are intended to provide transportation professionals with the quantitative means of improving existing roundabout operational performance and also help design future roundabouts with appropriate additional lane lengths that yield better performance. While the design of an additional lane differs from a flared entry, findings from this study can also be applied to flare lengths if they are designed to operate in a similar fashion as additional lane entry.

The Effect of Additional Lane Length on Double-Lane Roundabout OperationBy Samuel Hammond, Christopher Hunter, Ph.D., and Kevin Chang, Ph.D., P.E.


As roundabouts have become increasingly popular in the United States, it is very important to establish some means of improving their performance in the near future when vehicle demand nears or exceeds capacity. At signalized intersections, U.S. transportation professionals regularly consider parameters such as green time, cycle length, and number of lanes in order to improve traffic operation. However, there has not been much research performed domestically that addresses how to vary different geometric parameters to improve operations for a roundabout when analysis shows that a nearby development will impact traffic operation. Hence, most transportation professionals refer to studies conducted overseas that do not necessarily translate directly to domestic roundabout design and operation.

One of the design requirements that needs further exploration is the entry approach. The entry can be designed to increase capacity either by adding a full lane upstream of the roundabout or by widening the approach gradually (flaring) through the entry geometry.2 Most of the studies on roundabout entry design have been looking at the widening effect of the width of the approach lane. However, little attention has been given to the length of the approach and its effect on roundabout operation.

The National Cooperative Highway Research Program (NCHRP) report on roundabout design neither provides recommended lengths nor gives information on how long the entry lane needs to be widened along the approach. The Federal Highway Administration (FHWA) roundabout guideline, which was superseded by the NCHRP report, suggests a minimum flare length of 80 feet in urban areas and 130 feet in rural areas.3 It is not clear whether the suggested flare length also applies to additional lane design, and the maximum flare length is not specified.

In general, the increasing popularity of roundabouts in the United States underscores the need for more research on roundabouts in the United States to address the issues that traffic engineers face in practice. The means of improving signalized intersections to meet specified demands have been well researched and documented, and methods for predicting their performance are well established. However, roundabouts lack such research on performance improvement. This research examines the effect of additional lane lengths on a double-lane roundabout operation using delay as the primary measure of effectiveness. The study is intended to provide transportation professionals with a means of improving existing roundabout operational performance and to aid the planning and design stages so that future roundabouts can be built with appropriate lane lengths to yield better performance.

Literature ReviewWhen compared to signalized intersections, the research available on improving operations at roundabouts due to increased traffic flow is comparably lacking. Signalized intersections usually implement several modifications to improve safety and performance over time due to new developments or increases in traffic flow. Since roundabouts handle traffic flow similarly to signalized intersections, it is possible for the volume-to-capacity (v/c) ratio to approach or exceed 1.00. Under such conditions, long queues form and delay increases at roundabouts. Such conditions require modifications to improve performance.

Earlier research on roundabout operation was conducted by the United Kingdom–based Transport and Road Research Laboratory (TRRL), where numerous experiments and observations were performed on existing roundabouts. Kimber incorporated findings from the TRRL studies and identified six geometric parameters as having a significant effect on capacity: entry width, approach half-width, effective flare length, flare sharpness, inscribed circle diameter, and entry radius.4,5 Out of the six parameters, entry width, approach width, and flare length were determined to be the most relevant with regard to capacity.

The approach width, typically 12 feet in the United States, is the width of the traveled way in advance of any entry flare; the entry width is the width of the traveled way at the point of entry. FHWA identifies the entry width as the “largest determinant of a roundabout’s capacity.”3 The entry can be designed to increase capacity by either adding a full lane upstream of the roundabout or by widening the approach


gradually (flaring) through the entry geometry.2 NCHRP recommends an entry width of 24 to 30 feet for two-lane entry and 36 to 45 feet for three-lane entry. It does not, however, specify how far back the additional lane or flaring should begin. In Europe, where flaring design is more common than an additional lane design, the U.K. Department of Transport Design Manual recommends flare lengths of about 82 feet (25 meters) for widening to effectively increase capacity.6 Flare lengths greater than about 328 feet (100 meters) result in higher speeds, which undermines the main purpose of the modern roundabout configuration. The configuration of a modern roundabout reduces driver approach speeds to improve safety and enhance traffic flow. Therefore, when increasingly long lane lengths are used, the safety benefit of roundabouts may be forfeited. The 82-foot recommendation by the U.K. Department of Transport Design Manual has not been tested in the United States, but some state agencies follow the overseas guidelines, since data on the additional lane or flare length have not been provided. Interim requirements and guidance on roundabouts by the New York Department of Transportation suggest a flare length of 41 feet (12.5 meters) to 328 feet (100 meters) for urban areas and 66 feet (20 meters) to 325 feet (100 meters) for rural areas.7

The FHWA used a model developed by Wu in determining the capacity of a roundabout, whereby short length widening at the approach is considered.3,8 Wu estimated the capacity of an unsignalized crossroad and T-junction intersection by taking into account the length of the turn lanes. Wu later analyzed this model at a roundabout intersection and introduced an enhancement/correction factor for determining the capacity of a double-lane entry at a roundabout.9 Wu was able to identify the effect of entry length, but the effect of the additional lane length at the exit was not mentioned, and it was assumed that the capacities of both lanes were identical and the traffic flow in both lanes at the entry was equally distributed. However, studies conducted on some double-lane roundabouts in the United States show that the right lane is utilized more frequently than the left lane and thus is usually considered to be the critical lane. For instance, one of the double-lane roundabouts in Brattleboro, Vermont, showed that the right lane had about 70 percent of the entry total flow, so capacity in the Wu model appears to have been overestimated.10 This research examines the effect of the flare/additional lane length on roundabout operation using typical U.S. driving behavior, where the right lane is considered the critical lane and is utilized more frequently than the left lane.

In order to model typical U.S. driving behavior, VISSIM was used for analysis purposes. VISSIM is a microsimulation software from Germany in which vehicles are modeled using parameters such as driver behavior, vehicle speeds, and vehicle type.11 VISSIM has the ability to control gaps and headways on a lane-by-lane basis to accurately replicate vehicle operations at roundabouts. Numerous studies have used VISSIM microsimulations to examine roundabout performance due to their unique ability to mimic real-world traffic operations. Trueblood and Dale considered VISSIM to be a very effective microsimulation software package for roundabout performance analysis and used VISSIM to model existing roundabouts in the state of Missouri.12 Bared and Afshar used VISSIM to model roundabouts for various ranges of circulating and entry traffic volumes.13 They found that simulation results from VISSIM matched field data more closely than those from the SIDRA analytical and RODEL empirical models.

MethodologyThis section details how the research effort was conducted with respect to modeling the impact of additional lane length on roundabout operation. In the flare design, a single lane is gradually widened into two lanes at the entry (see Figure 1). In the additional lane length design, a full lane is added at the entry and a taper of sufficient length is provided. Both design cases result in the widening of the entry and increases the rate that vehicles enter the roundabout at a given time. This means that in terms of operation, a single traffic stream separates into two streams for both the additional lane and flare designs. The additional lane design was used in this research to examine the effect on a roundabout; the findings applied to flared entry as well.


A hypothetical double-lane roundabout with four legs was modeled in VISSIM with varying additional lane lengths at the entry and exit. For comparison purposes, similar variations were then tested on an existing double-lane roundabout with data from NCHRP 572.10 In order to properly lay out the roundabout correctly in VISSIM, guidelines developed by Trueblood and Dale and Li, et al. were used to re-create the reduced-speed areas at the conflict points.12,14 These reduced-speed areas were kept at a length of 17 feet and placed 8 feet from the yield line on each lane of the approach. Reduced-speed areas were also placed in the circulatory roadway at a length of 17 feet before the entry areas. Travel speeds of 20 miles per hour were used in the reduced-speed zones, as recommended by Trueblood and Dale.12

Since VISSIM is a stochastic model whose results vary depending on the random seed number used, the model was run multiple times and the average results were used. For this study, 17 simulation runs with a cumulative running time of 1 hour were analyzed in VISSIM to achieve a 95 percent confidence interval.

Figure 1. Flaring and additional lane design.

Source: FHWA Roundabout Guide.

For the hypothetical model, VISSIM default values for headway were used. Data collection points used to capture delay data were placed at locations specified in NCHRP 572 to compare results.10

The travel time section in VISSIM was established starting at 250 feet from the yield line on the approach and ending at the exit where the vehicle left the circulatory roadway. The roundabout


used for comparison was set up in VISSIM with data collection points placed at similar locations to those used in NCHRP 57210; the model was then calibrated using field data from the same study. The calibration effort began with the VISSIM default values and was gradually adjusted until the reduced speed, driving behavior, yield bar placement, headway, and minimum gaps of the measured field travel time data closely matched the VISSIM data. The field travel time data were the same data used in NCHRP 572 and were obtained from Kittelson and Associates.

Hypothetical Double-Lane RoundaboutThe roundabout used in this study focused on the six important parameters given by TRRL.5 The design was based on the guidelines in NCHRP 672.2 As shown in Figure 2, the roundabout had two circulatory lanes and four legs with single lanes that diverged into two lanes at the entry and merged into one at the exit. An inscribed circle of 180 feet was used for this study, and the four approaches were aligned at 90 degrees. This AutoCAD-designed layout was subsequently uploaded into VISSIM.

Figure 2. Hypothetical roundabout design.

Figure 2. Hypothetical roundabout design.

For the purposes of this analysis, no specific volume was assigned on a lane basis. Vehicles were allowed to freely choose lanes, but the links and driving behavior assumptions were configured


such that the right lane would be used more frequently (about 70 percent usage was observed from the simulation). The roundabout model mirrored real-life driving behavior, as vehicles were free to change lanes when prevailing conditions were not favorable. One-quarter each of the traffic made a right or left turn, and the remaining one-half proceeded straight through the roundabout. A degree of saturation less than 0.80 was targeted based on the following assumptions:

■■ The major road was established in the north–south direction with a volume of 800 vehicles per hour in each direction;

■■ The minor road was established in the east–west direction with a volume of 350 vehicles per hour in each direction;

■■ The right lane was assumed to be the critical lane in all directions; and

■■ Fifteen percent of the volume consisted of heavy vehicles.

The v/c ratio for the northbound and southbound critical lane was 0.78, while the eastbound and westbound critical lane was 0.41 using the analytical method presented in the Highway Capacity Manual (HCM).

Starting with an additional lane length of zero (single lane entry and exit), the operational performance of the roundabout was analyzed in VISSIM for 17 simulation runs, and the delay data were averaged. Table 1 shows the different scenarios used for the hypothetical and existing model analysis:

■■ Scenario 1: Only the entry additional lane length was varied, while the exit additional lane length was kept at zero (single exit).

■■ Scenario 2: Both entry and exit additional lane lengths were varied.

Table 1. Model scenarios.

Hypothetical Model Scenario 1 Hypothetical Model Scenario 2

Additional lane East West South North East West South North

Location

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Variation 1 X X X X X X X X X X X X

Variation 2 X X X

Variation 3 X X X

Existing Model Scenario 1 Existing Model Scenario 2

Additional lane East West South North East West South North

Location

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Entr

y

Exit

Variation 1 X X X X X X X X X X X X

Variation 2 X X X

Variation 3 X X X


Under scenario 1, three variations were considered:

1. Additional lane lengths at the entry of all four legs were varied at the same time. This scenario was represented by HA in this study, where H represents the hypothetical model and A represents all legs.

2. An additional lane at the entry with the highest volume (south leg) was varied. This scenario was represented by HS, where H represents the hypothetical model and S represents the south leg.

3. An additional lane at the entry with the lowest volume (west leg) was varied. This scenario was represented by HW, where H represents the hypothetical model and W represents the west leg.

Under scenario 2, three variations were considered:

1. Additional lane lengths at the entry and exit of all four legs were varied at the same time. This scenario was represented by HAX in this study, where H represents the hypothetical model, A represents all legs, and X represents the exit.

2. An additional lane at the entry and exit with the highest volume (south leg) was varied at the same time. This scenario was represented by HSX, where H represents the hypothetical model, S represents the south leg, and X represents the exit.

3. Only one additional lane at the entry and exit with the lowest volume (west leg) was varied at the same time. This scenario was represented by HWX, where H represents the hypothetical model, W represents the west leg, and X represents the exit.

The additional lane lengths that were analyzed in VISSIM for both scenarios included 0 feet, 150 feet, 250 feet, 350 feet, 450 feet, and 550 feet. The VISSIM lane closure feature was utilized to make the zero foot length possible. While reducing the exit and entry lanes on a double-lane roundabout to a single lane is not practical, it was done in this study to illustrate the extent of the delay effect up to zero feet.

Existing Double-Lane RoundaboutFor analysis purposes, an existing roundabout was analyzed in a similar manner as the hypothetical model. The roundabout chosen for this analysis was the Brattleboro roundabout in Vermont, since it had a physical configuration comparable to the hypothetical model used in this study and was one of the roundabouts included in the NCHRP 572 study.10 The data from the NCHRP 572 study were used to calibrate and validate the hypothetical model in VISSIM. The Brattleboro roundabout was a double-lane roundabout with four legs aligned at 90 degrees. Its inscribed circle diameter was 176 feet, and all legs had additional lane lengths greater than 100 feet. Figure 3, obtained from the Vermont Transportation Agency, shows the different additional lane lengths on each approach. The south, east, and north legs have exceptionally long taper lengths, and these lengths were included in the model. The figure also shows new pavement markings where a three-lane alternative was being considered for the northbound entry. This research study used the exiting configuration of a two-lane entry, which represented the configuration when field data were collected for NCHRP 572.


Figure 3. Brattleboro roundabout design.

Source: Vermont Transportation Agency.

The field data collection determined that the volumes for the eastbound, westbound, southbound, and northbound legs were 832 vehicles/hour, 441 vehicles/hour, 515 vehicles/hour, and 1,051 vehicles/hour, respectively. Using the HCM analysis, the v/c ratio for each critical lane (right lane) was found to be 0.96, 0.58, 0.62, and 1.41, respectively. The field data used in NCHRP 572 did not include delay records for southbound traffic, so travel time data were used to validate this VISSIM model.10 Using a t-test, a two-tailed p-value of 0.764 was calculated, indicating no statistically significant difference between the VISSIM and field travel time data. For calibration, the headway, reduced-speed area, driving behavior, and link arrangement were adjusted until the VISSIM travel time mirrored the field data. Figure 4 compares the field data with the final VISSIM trial that gave an acceptable error. Individual movements are labeled showing the direction of travel; for example, W-S indicates a movement entering from the west leg and exiting at the south leg.


Figure 4. Field and VISSIM travel time comparison.

After the model was validated, various lane lengths were analyzed following the same procedure as described earlier for the hypothetical model (see Table 1). The additional lane length was varied for the same two scenarios as for the hypothetical model, and only the additional lane lengths were varied; all other parameters remained the same. For each scenario in the existing model, the letter E was used to differentiate these scenarios from the hypothetical model, which used H. (As an example, where additional lane lengths at the entry are varied at all four legs, this scenario was represented by EA in this study, where E represents the existing model and A represents all legs.) For variation 3 of the existing model, it should be noted that the volume from the north leg was used, as it represented the leg with the lowest entering volume.

Since the existing model had varying additional lane lengths of 150 to 180 feet, the following lengths were analyzed for both of the existing model scenarios: 0 feet, 50 feet, 100 feet, and the actual existing length (see Figure 3). In addition, lengths of 100 feet, 200 feet, 300 feet, and 400 feet were added to the existing additional lane lengths and analyzed in VISSIM to study the effect of longer lengths on roundabout operation.

Results And DiscussionThe delay and speed data for the hypothetical model are shown in Figure 5. The delay data represent the difference between the measured travel time and free-flow travel time from a location of 250 feet approaching the yield line on the entry leg to the exit point on the circulatory roadway. Figure 5 also shows the average speed between these two locations. Data from the hypothetical model show that the highest delay occurred when the model had a single lane (zero additional lane length) for each scenario. There was no significant difference between scenarios 1 and 2, though the delay data were slightly higher for scenario 2 (when the additional lane length at the entry and exit were varied at the same time). Increasing the length up to approximately 150 feet was effective in reducing delay, but beyond that point, there was no significant decrease. In general, an increase in lane length resulted in an increase in vehicle speed, and the most effective length was between 50 feet and 150 feet.


The analysis of the different variations showed that increasing the length on all four legs at the same time was more effective than just increasing the length on one leg. As the speed data show, increasing the lengths caused the speed to increase at the entries; this decreased the time required for vehicles to reach the circulatory roadway. When additional vehicles reach the circulatory roadway within a short period of time, the conflicting flow increases and reduces the likelihood of finding an acceptable gap. It is for this reason that the delay increases even though approach speeds are increasing. Increasing the length on just one leg reduced the delay for that entry but resulted in more vehicles in the circulatory roadway and increased the conflicting flow for other entries. Increasing the length on the entry with the lowest volume (minor road) increased the conflicting flow and caused delay on the major road. The delay on the minor road, which had minimum effect on the intersection, was decreased, but the delay on the major road increased, resulting in a slightly increased delay at the intersection overall. Increasing the length on the entry with the highest volume was more effective than increasing the length on the entry with the lowest volume, since the delay on the major road, which significantly contributes to the delay of the entire intersection, was reduced. Increasing the length of just one entry (either highest or lowest volume) was not as effective as increasing all four legs at the same time, as increasing the length on all four legs reduced the delay on each approach, and thus reduced the overall delay of the intersection.

Wu suggested balancing the exit and entry capacities so that the potential of widened entry could be achieved and bottleneck effects at the exit could be avoided.9 Based on this analysis, the double-lane exit did not affect delay at the intersection. The difference was more noticeable within short intervals between zero and 150 feet; beyond 150 feet, increasing the exit length did not result in any significant change in the delay. This could be due to the fact that low volumes (or v/c ratios) were considered. It is also possible that the conflict at the exit was minimal because the roundabout configuration was carefully laid out per NCHRP guidelines.2

The same variations were applied to the existing roundabout in Brattleboro, Vermont. Observations during site visits (spring 2013) to this roundabout determined that some adjustments to improve its operation during peak hours were needed. During off-peak hours, the roundabout operated exceptionally well on all approaches. During peak hours, the northbound approach exhibited long queues with delays up to about 23 seconds from approximately 250 feet upstream of the yield line. The high traffic in this direction was due to the increased development of restaurants, offices, and other businesses south of this roundabout. Under free-flow conditions, the travel time from approximately 250 feet upstream to the yield line was measured to be about 7 seconds, but during peak hours this short interval took about 30 seconds of travel time. During the peak hour, the east, west, and north legs that operated exceptionally well during off-peak hours also experienced an increase in delay.

In order to evaluate the operations at this roundabout, the length variation applied to the hypothetical model was also applied to the Brattleboro roundabout model in VISSIM after calibration and validation. The results in Figure 6 conform to the findings from the hypothetical model. On the east, west, north, and south legs of this roundabout, about 180, 160, 150, and 180 feet of respective lane length existed at both entry and exit (Figure 3). For the analysis the additional lane lengths were decreased so that all lengths were either zero (single lane), 50, or 100 feet using the above-stated scenarios. Shorter lengths within 100 feet on all legs resulted in the most significant decrease in delay. Increasing the existing length by 100-foot increments at all legs simultaneously resulted in less change in delay. Zero-foot lengths resulted in the highest delay, and delay decreased as increasing lengths approached the existing lengths. As noticed in the hypothetical model, adjusting just one leg was not as effective as adjusting all legs at the same time. Adjusting only the leg with the lowest volume was the least effective means of improving delay. There was no significant difference in varying the length of the exit lane; the


difference was more noticeable when the length was between zero and 150 feet, but beyond 150 feet, increasing the exit length did not result in any significant change in delay.

Figure 5. Delay and speed data for hypothetical model.


Figure 6. Delay and speed data for existing model.

Conclusion And RecommendationsThe findings from this study are based on double-lane roundabouts with varying approach geometries and additional lane length configurations. The delay values reported in this study were measured from a distance of 250 feet from the yield line on the approach to the point where the vehicle exited the circulatory roadway. Delays upstream of the 250 foot mark and beyond the exit


line were not recorded, though delays beyond these limits could add to the magnitude of the data reported in this study. Understanding delay variability within this short interval under the assumed conditions furthers the body of knowledge with regard to roundabout operations that was initially presented in NCHRP 572.10

The analyses of the hypothetical and existing roundabout models indicate that very long additional lane lengths were not effective in reducing delay at roundabouts. Shorter lengths of up to 150 feet were determined to be most effective. This finding corroborates the results from the U.K. Department of Transport Design Manual for Road and Bridges, which recommended shorter flare lengths of about 82 feet to effectively increase capacity and pointed out that longer flare lengths resulted in higher speeds.6 The findings from this study can also be applied to flare designs. Where flaring is used, additional analysis is needed if the flaring does not result in two entry lanes. At entries where two full lanes are used, longer lengths will result in increased speed and reduced delay.

From an operations standpoint, shorter additional lane lengths between 50 and 150 feet at both the entry and exit were most effective. While adjusting the lane length of all legs was determined to be more beneficial than only adjusting one leg, at a location where only one leg can be modified, the leg with the most volume should be adjusted and length variation should be within the 50- to 150-foot range. If lengths of 150 feet already exist, other modification techniques need to be applied, as longer lengths will be ineffective in reducing delay. Increasing the lane length allowed vehicles to reach the circulatory roadway faster, but with more vehicles entering the roundabout, the likelihood of conflicting flow increased as well. NCHRP identifies design procedures that balance entry, circulatory, and exit flow through lane numbers and arrangements.2 Shorter lengths help regulate the rate of entry at a slow but constant rate, but longer lengths can result in an instantaneous increase in circulatory roadway flow with less capacity to handle the flow.

The findings from this study will help transportation professionals with regard to roundabout design and operations. This study confirms that additional lane length can be varied in a manner that effectively reduces delay while avoiding unnecessary lane design and construction; this study can also be used during the planning and design stages of a new roundabout in order to determine the appropriate additional lane length. Additional analysis is needed to determine the effect of different lengths on safety, since this study has shown that increasing the lane length can increase the approach speed at a roundabout, and this trade-off could undermine the operational benefits of a modern roundabout facility.

Works Cited1. Kittelson & Associates. Modern Roundabouts Website, Inventory Activities [Online]. Available:

http://roundabouts.kittelson.com/. [Accessed December 2012.]

2. National Cooperative Highway Research Program. NCHRP Report 672: Roundabouts: An Informational Guide, 2nd ed. Washington, DC: Transportation Research Board, 2010.

3. Federal Highway Administration. Roundabouts: An Informational Guide. FHWA-RD-00-06. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, 2000.

4. Kimber, R.M. The Traffic Capacity of Roundabouts. TRRL Laboratory Report LR 942. Crowthorne, England: Transport and Road Research Laboratory, 1980.

5. U.K. Department of Transport. Design Manual for Roads and Bridges, Vol. 6, Section 2, Part 3. U.K. Department of Transport, 2007.


6. New York Department of Transportation. Roundabout: Interim Requirements and Guidance. Albany, NY: New York Department of Transportation, 2000.

7. Wu, N. “Capacity of Shared/Short Lanes at Unsignalized Intersections.” In Proceedings of the Third International Symposium on Intersections without Traffic Signals, M. Kyte, ed. Portland, OR, USA.: University of Idaho, 1997.

8. Wu, N. “Capacity Enhancement and Limitation at Roundabouts with Double-Lane or Flare Entries.” Proceedings of the 5th International Symposium on Highway Capacity and Quality of Service: Technical Papers, Vol. 2. Yokohama, Japan: Japan Society of Traffic Engineers, 2006.

9. National Cooperative Highway Research Program. NCHRP Report 572: Roundabouts in the United States. Washington, DC: National Research Council, Transportation Research Board, 2007.

10. PTV. VISSIM User’s Manual 5.30. Karlsruhe, Germany: Planung Transport Verkehr AG, 2010.

11. Trueblood, M. and J. Dale. “Simulating Roundabouts with VISSIM.” 2nd Urban Street Symposium. Anaheim, California: 2003.

12. Bared, J.G. and A.M. Afshar. Planning Capacity Models By Lane for 2- and 3-lane Roundabouts Using Simulation. Washington, DC: Transportation Research Board, National Research Council, 2009.

13. Li, Z., M. DeAmico, M. Chitturi, A. Bill, and D. Noyce. Calibration of VISSIM Roundabout Model: A Critical Gap and Follow-up Headway. Washington, DC: Transportation Research Board, National Research Council, 2013.

14. Transportation Research Board. Highway Capacity Manual. Washington, DC: Transportation Research Board, National Research Council, 2010.

Samuel Hammond is a graduate research assistant at the University of Rhode Island. His research interests include roundabout operation, sustainable transportation planning, traffic safety, traffic flow modeling, and simulation. Samuel has been the recipient of the Tom Joyner Scholarship, Academic Achievement Award (two consecutive years), and Dean’s Scholarship.

Christopher Hunter, Ph.D., is an associate professor in the Department of Civil & Environmental Engineering at the University of Rhode Island, where he primarily teaches courses in transportation and conducts research on traffic system operations and traffic safety. He is a member of the Institute of Transportation Engineers, American Society of Civil Engineers and the National Society of Black Engineers. He has been the

recipient of the Diversity Award for Faculty Excellence in Leadership and Service and the Edmund and Dorothy Marshall Faculty Excellence Award for Outstanding Service and Excellent Teaching in Civil Engineering.

Kevin Chang, Ph.D., P.E., is an assistant professor in the Department of Civil Engineering and Center for Secure and Dependable Systems at the University of Idaho. He is a graduate of the University of Washington and is registered as a Professional Engineer in Washington and California. Kevin is the current chair of the ITE Transportation Education Council and is a past president of the Washington

state section of ITE. He is a member of ITE.

Institute of Transportation Engineers1627 Eye Street, NW, Suite 600

Washington, DC 20006www.ite.org

© July 2014 Institute of Transportation Engineers. All rights reserved.

Documents

JOURNAL OF TRANSPORTATION