Citation
Development of a Rural Freeway Level of Service Model Based upon Traveler Perception

Material Information

Title:
Development of a Rural Freeway Level of Service Model Based upon Traveler Perception
Copyright Date:
2008

Subjects

Subjects / Keywords:
Data collection ( jstor )
Freeways ( jstor )
Motor vehicle traffic ( jstor )
Roads ( jstor )
Service quality assurance ( jstor )
Speed ( jstor )
Traffic surveys ( jstor )
Transportation ( jstor )
Travelers ( jstor )
Video clips ( jstor )
City of Gainesville ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
5/1/2005

Downloads

This item has the following downloads:


Full Text












DEVELOPMENT OF A RURAL FREEWAY LEVEL OF SERVICE MODEL BASED
UPON TRAVELER PERCEPTION















By

DAVID S. KIRSCHNER


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ENGINEERING

UNIVERSITY OF FLORIDA


2005
































Copyright 2005

By

David S. Kirschner














ACKNOWLEDGEMENTS

I would like to thank my committee chair, Dr. Scott Washburn, and my committee

members Dr. Lily Elefteriadou and Mr. Bill Sampson.















TABLE OF CONTENTS



A C K N O W L E D G E M E N T S ............................................................................................... iii

LIST OF TABLES .............. ............................................... ........ vi

LIST OF FIGURES .................................................... .............. ......... vii

A B S T R A C T .......................................... .................................................v iii

CHAPTER

1 INTRODUCTION ...... ............................... ........................... 1

B background ............................................................... ... ........ ............. 1
P problem Statem ent ................................................................... .... ...... ........ .. 2
R research O objective and T asks ...................................................................................... 4
Chapter O rganization......................... .... .. ................ 4

2 L ITER A TU R E R E V IEW .................................... ..................................................... 6

H CM Freew ay LO S M methodology .................................... ............................ ....... 6
Studies Investigating Traveler Perception of LOS ..................................................... 8

3 RESEARCH APPROACH ....... ................................................ .............. 16

A alternative Survey M ethods .......................................................... .............. 16
Video D ata Collection ......... .............................. .. ...... ........... 19
Survey Sessions ........................................ 30

4 ANALYSIS AND RESULTS...................................................... ......................... 35

A n aly sis M eth o d ................................................................... ................................ 3 5
Statistical A nalysis............................................ 39

5 CONCLUSIONS AND RECOMMENDATIONS ............................................... 45

Data Collection and Video Clip Creation............................................. ........... 45
Statistical A n aly sis................. .................................. ................. 4 6


iv









Study Limitations and Recommendations for Further Research............................ 47

REFERENCES .... ............... .... .............. .......... .. 49

APPENDIX

A LOCATIONS OF DATA COLLECTION SITES................................................ 51

B V ID E O CLIP SCR EEN SH O T S......................................................... .................... 55

C RURAL FREEWAY TRIP QUALITY SURVEY FORM.............................. .... 63

D SAM PLE LOOP DETECTOR DATA ........................................................................ 68

BIOGRAPHICAL SKETCH ................. ........... ................. 73














LIST OF TABLES



1. H C M L evel of Service Thresholds ....................................................... .................... 8

2. Data Collection Sites and Traffic Data ............................................. 21

3. Data Collection Times, Locations, and Directions ............................... .... ............... 24

4. Traffic D ata for 13 V ideo Clips ......... ................. ......... .................. .............. 29

5. C lip Sites, D ates, and T im es ........... .................................................... .............. 30

6. Dates and Locations of Survey Sessions ........................................................ 33

7. D ensity M odel E stim ation R esults...................................................... .... .. .............. 40

8. Comparison of Estimated and HCM LOS Thresholds ............. .............. 41

9. Traffic Characteristics Model Estimation Results.................................... ............... 42

10. Level of Service M odel Estim ation Results.......................................... .... .. .............. 43

11. Realism of Video Survey Responses .................................................................... 46














LIST OF FIGURES

1. Camera Setup-Front View, Side View, Speedometer................................................. 23

2. In-Vehicle Equipment Setup.......................................................... .............. 24

3. Setup of a Survey Session ............................................................................ .............. 26

4 Sam ple V ideo Screenshot .................................................................................... 26

5. Illustration of an Ordered Probability Model.............................................................. 37

6. Illustration of an Ordered Probability Model with an Increase in fP............................. 38















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Engineering

DEVELOPMENT OF A RURAL FREEWAY LEVEL OF SERVICE MODEL BASED
UPON TRAVELER PERCEPTION

By

David S. Kirschner

May 2005

Chair: Scott Washburn
Major Department: Civil and Coastal Engineering

The concept of Level of Service (LOS) is meant to reflect the trip quality a traveler will

experience on a roadway or other transportation facility. Despite this, there have been

relatively few studies that have tried to measure the association of prescribed level of service

assessment methods with traveler perceptions. The objective of this study is to provide

insight into how road users perceive their trip quality on rural freeways, and to examine how

the existing service measure (density) relates to these travelers' perceived trip quality.

Study participants were shown a series of video clips of rural freeway travel from a

driver's perspective, then filled out survey forms indicating their opinion of the trip quality

provided by the conditions in the video clip, and ranked these video clips on a scale from

'Excellent' to 'Very Poor'. In addition, the survey participants were asked to give

background information about themselves and their driving habits in case these factors also

turned out to be significant in influencing perceived trip quality. These video clips were

matched with inductance loop detector data that were collected simultaneously at the data









collection sites, in order to see how well the existing service measure (density) corresponded

to the participants' rankings.

The data from the surveys were analyzed using an ordered probability model to

determine which factors influenced the participants' decisions and how. Three models were

created. The first model used only density as a predictive factor. The second took into

account only roadway and traffic characteristics, and the third examined all the significant

factors that could be gathered from the survey. The 'density only' model showed that density

is indeed a strong indicator of travelers' perceptions of trip quality. A set of LOS thresholds

was also calculated using the survey participants' responses. While the survey thresholds and

the HCM thresholds had similar values for facility failure, the intermediate thresholds

estimated from the survey participants' responses were noticeably lower than the HCM

thresholds. This suggests that travelers' tolerance of congestion is lower on rural freeways

than the HCM indicates. The other models showed the significance of other factors in the

perception of trip quality in addition to density, such as socio-economic information and

personal driving habits.

This study provided some preliminary insight into travelers' perception of trip

quality, but further study is needed. It is suggested that more research be conducted regarding

the effects of different factors on the perception of trip quality, such as a more diverse

population sampling. Eventually, the results from this type of video-based study should also

be compared to results obtained from a comparable in-field driving experiment. This study

indicates the need for a further exploration into the differences between urban and rural

freeways, and possibly a different set of thresholds for rural freeways.















CHAPTER 1
INTRODUCTION

Background

Transportation engineers are responsible for targeting roadway infrastructure

improvements where they will have the most beneficial effect. Since the capital available

for these improvements is limited, engineers must carefully select the projects they

choose to fund so that investments will have the best cost-benefit ratio. In a large part

these decisions are guided by the procedures and methodologies found in the Highway

Capacity Manual (HCM) [1]. The HCM is considered to be the definitive reference guide

for traffic operations and analysis in the United States. The procedures in the HCM are

used to estimate the operational performance of a variety of transportation facilities (e.g.,

signalized intersections, two-lane highways) and the corresponding level of service

(LOS). The assignment of a LOS is based on designated performance measures and

corresponding threshold values for individual facilities. The HCM is published by the

Transportation Research Board (TRB) and its development and maintenance is the

responsibility of the Highway Capacity and Quality of Service (HCQS) committee of the

TRB. The current edition of the HCM was published in 2000.

The concept of LOS is a foundation of the HCM. The LOS of a facility is used in

the HCM as a qualitative indicator of the operating conditions being experienced by

travelers of that facility, under specific roadway, traffic, and control conditions. The

HCM describes LOS as "A qualitative measure describing operational conditions within









a traffic stream, based on service measures such as speed and travel time, freedom to

maneuver, traffic interruptions, comfort, and convenience." LOS is divided into six

categories, A through F in the 2000 HCM. LOS A indicates excellent service and LOS F

indicates extremely poor service. An analysis yielding LOS A would indicate that the

facility is performing extremely well, with low volumes and little congestion. If an

analysis shows a facility to be performing at LOS C, it is in the middle range of

congestion. If a facility is at LOS E, it is still permitting traffic flow but is experiencing

significant delays with conditions approaching capacity. At LOS F, a facility is

experiencing oversaturated conditions and the demand has exceeded the capacity of the

facility.


Problem Statement

The performance measures that are used to calculate LOS for a facility are

referred to as service measures. The currently designated service measures) for each

facility is (are) based on the collective experience and judgment of the members of the

HCQS committee. The same is true with the selection of the threshold values for the

various LOS designations. There is currently no quantitative procedure to define which

values are used as LOS thresholds. The LOS determination process, therefore, is based on

the perspective of transportation professionals. The selection of service measures by the

HCQS committee is, however, guided by two principles: 1) the service measure for each

facility should represent speed and travel time, freedom to maneuver, traffic

interruptions, and comfort and convenience in a manner most appropriate to

characterizing quality of service for the particular facility being analyzed, and 2) the

service measure chosen for a facility should be sensitive to traffic flow such that the









service measure accurately describes the degree of congestion experienced by users of the

facility [2].

The 1985 HCM described LOS as "A qualitative measure that characterizes

operational conditions within a traffic stream and their perception by motorists and

passengers. The descriptions of individual levels of service characterize these conditions

in terms of factors such as speed and travel time, freedom to maneuver, traffic

interruptions, and comfort and convenience" [3]. This statement indicates that the

selection of performance measures and thresholds for the determination of level of

service should be consistent with how operating conditions are perceived by the traveling

public. Until recently, road users' perceptions of quality of service were rarely compared

to the LOS assigned to a facility by the HCM, despite the above definition emphasizing

the importance of reflecting road users' perceived quality of service.

There have been suggestions from within the HCQS committee that a new

approach needs to be explored when selecting a service measure for a facility. Instead of

the measure and corresponding thresholds that transportation professionals (the HCQS

committee) believe represent the quality of service as perceived by travelers, the public's

opinion should be taken into account so as to determine what measure or measures they

associate with quality of service on a transportation facility. Under the current

methodology, the HCQS committee believed that the service measures were highly

correlated with public perception, but this was not known for sure [4]. Since billions of

dollars of transportation investment decisions are made every year based upon the

outcome of HCM level of service analyses, it is essential that the transportation









engineers' assessments of the impact of these investments be consistent with traveler

perception of the investment impacts.


Research Objective and Tasks

The objective of this study was to develop a model for assessing the LOS of a

roadway facility that takes into account the road user's perceived quality of service.

Specifically, this study was focused on rural freeways.

The following tasks were carried out in supporting the above research objective.

Determine appropriate rural freeway sites to perform field data collection

Collect video of roadway and traffic conditions from these sites

Collect traffic data from count stations at these sites

Produce video clips to be shown to survey participants

Develop a survey instrument

Recruit survey participants

Conduct survey sessions

Perform an analysis of survey responses

Develop a level of service model


Chapter Organization

Chapter 2 contains an overview of the current HCM freeway analysis

methodology as well as an overview of relevant literature. Chapter 3 describes the

research approach for this study, including the field data collection, survey instrument

development, survey response data collection, and the statistical analysis method used to

analyze the data. Chapter 4 contains the analysis results. Chapter 5 contains the






5


conclusions and recommendations. Additionally, several appendices with supporting data

and information are included.














CHAPTER 2
LITERATURE REVIEW

The Highway Capacity Manual [1] states that the level of service of a roadway

section should accurately reflect the perceptions of travelers, yet the current methodology

does not directly take these perceptions into account. There have been some recent

studies performed seeking travelers' opinions about what factors and qualities are

important to them in assessing the quality of their trip. A literature review was conducted

to identify these studies and note their findings with regard to the traveler's perception of

LOS.


HCM Freeway LOS Methodology

A freeway is a section of divided roadway with controlled access and two or more

lanes in one direction. Within this definition there are significant differences between

urban and rural freeways. Rural freeways have greater distances between interchanges

than urban freeways, higher speed limits than urban freeways, and a higher percentage of

social and recreational trips than urban freeways. Urban freeways have a higher

percentage of work and shopping trips than rural freeways. Despite these differences,

urban and rural freeways both use density as their service measure with the same

thresholds for LOS.

Traveler expectations and perceptions of quality of service are different for rural

and urban freeways. While urban freeways experience the full range of LOS conditions

from A to F, rural freeways rarely drop below LOS C. Rural freeway travelers have come









to expect these higher levels of service, therefore while urban freeway travelers are

concerned with their overall travel time and the reliability of this travel time, rural

freeway drivers take travel time for granted. Urban freeway drivers expect their ability to

change lanes to be restricted, while a restricted ability to change lanes negatively impacts

a rural freeway user's perceived quality of service [5].

The original HCM had a basic three-point scale to define level of capacity. In

1963 the Level of Service concept was introduced and replaced the previous scale. In

1965 the six-point LOS scale (from A to F) was introduced. In 1985 this six-point scale

was redefined to use traffic density (vehicles per unit length of roadway) as the service

measure for defining LOS on freeway sections. This is the method that is still used today.

Although the concept of LOS is meant to reflect the operational conditions as perceived

by motorists, no freeway LOS methodology in the history of the HCM has been based on

driver perception studies. Therefore, there can be no way to make sure that the LOS

thresholds freeways (as well as any other type of facility) accurately reflect users'

perception of the quality of service they receive.

Under the existing LOS methodology, rural and urban freeways have the same

service measure density, as well as the same thresholds for each rank on the LOS scale.

These thresholds for all freeway sections are shown in the table below.

Do these thresholds accurately reflect the quality of service perceived by travelers

on all freeways, urban and rural? In particular, the studies by Hostovsky [5] and

Washburn [4] indicate that rural freeway travelers may judge the quality of their trip

based on different qualities and criteria. A potential outcome of this study is a set of LOS

thresholds unique to rural freeways. This idea of differing service measures for different










categories of a specific facility type is not a new one. Currently there are two service

Table 1. HCM Level of Service Thresholds

Level of Service Density (pc/mi/ln)
A 0-11
B 11-18
C 18-26
D 26-35
E 35-45
F >45



measures used for assessing the LOS of a two-lane highway. These classes share a

common service measure, but the thresholds are different (one of the classes also uses an

additional service measure). In addition, the HCM procedure for analyzing arterial streets

uses the same service measure for all arterial streets but includes four sets of thresholds

for four different classifications of arterials [1].


Studies Investigating Traveler Perception of LOS


A study by Pecheux et al. [6] noted that the Transportation Research Board's

Committee on Highway Capacity and Quality of Service recognized a need to improve

the HCM methodology of assessing LOS. Specifically, concerns were raised that the LOS

of a roadway section did not correspond to road users' perceptions. The authors felt that

for LOS to accurately reflect travelers' perception of quality of service they would first

have to find out what performance measures were significant to travelers.

The study method involved test participants driving along a pre-selected 40-

minute route, encompassing mostly arterial streets, accompanied by an interviewer and a

traffic engineer. The participant would discuss what factors they personally found









important to the quality of their trip. Participants identified over 40 factors that were

important. These included such factors as intersection efficiency-if the intersection was

being utilized by opposing traffic while travelers were waiting, and the aesthetic qualities

of the intersection. Both of these topics are not covered by the HCM. The study

concluded that more research was needed to focus on traveler perception.

A study by Hostovsky et al. [5] used focus group participants to identify factors

important to trip quality on rural freeways and then compared those findings with those

from a focus group study using regular urban commuters and commercial truck drivers.

The participants in the rural freeway focus group identified three factors that were most

important to trip quality-low density, regular (predictable) travel time, and maintaining

a steady travel speed. Other topics discussed were the safety issues inherent to the

isolated locations of rural freeways, aesthetics, speed differential between cars, the

presence of heavy vehicles, and the need for better traveler information.

When compared to the results of a focus group study involving urban commuters,

it was found that urban commuters placed high importance on the overall speed of their

trip, where rural freeway travelers felt that the ability to choose their speed was a

positive. This reflects the fact that urban drivers rarely have the opportunity to choose

their speed in the traffic stream, so a faster speed is usually preferable over a slower one.

Urban commuters were also not as concerned with the ability to change lanes and move

about the facility at will. Most of the urban drivers were happy if they could stay in one

lane and maintain a desired speed for their trip. The rural drivers were pleased if the

density of the freeway section was low enough to allow movement between lanes and

passing at will. This study was significant due to the fact that it recognized the









differences in how travelers rate their trip quality on an urban versus a rural freeway. The

HCM uses the same methodology for freeways in both types of areas.

A study by Nakamura et al. [7] evaluated traffic flow conditions along an

expressway in Japan from a driver's viewpoint. The study intended to quantitatively

analyze the relationship between traffic flow conditions, drivers' perceptions, and

drivers' behaviors. The field data portion of this study was intended to collect data on

drivers' behavior and perception under various flow conditions. Drivers had a video

camera mounted in their own vehicle and were asked to drive a section along an

expressway. After each trip the subject was asked to complete a survey about the traffic

flow conditions. Twenty-two subject vehicles were used and 105 surveys were collected.

The behavioral data collected was number of lane changes, travel time by lane, and

percent time spent following.

This study found that the most important factor influencing drivers' satisfaction

with their trip was the traffic flow rate. Other factors affecting trip quality were found to

be number of lane changes, and the percent time spent following. Additionally, choosing

the LOS based on the driver's level of satisfaction was attempted and then compared to

the conventional LOS methodology. The results of this comparison suggested that the

traffic conditions on Japanese expressways are not satisfying drivers. The realistic

meaning of this result was that if facilities were designed to the driver's satisfaction level

rather than the conventional LOS it would require an enormous investment.

Several studies have been identified using road-user surveys and video selections

to evaluate LOS methodology. The first study, by Sutaria and Haynes [8], used a road

user survey to evaluate the LOS methodology for signalized intersections. Over 300









drivers were shown video clips taken both from a driver's perspective and from an

overhead camera at an intersection. The film segments were specifically chosen to

represent a specific LOS and were intended to be shown to drivers for one or two signal

cycles. The final compilation shown to drivers included both types of view and the clips

were put in a random order.

Their road user survey consisted of two parts-a group attitude survey and a film

survey. The group attitude survey used a questionnaire, to be answered before the film

portion of the survey. The questionnaire included demographic information such as

gender, age, and education, as well as questions about the participants' driving

experience and the type of roadways the participants usually drove on. They were then

asked to give the relative importance of factors including delay, number of stops, traffic

congestion, heavy vehicle density, and ability to change lanes as these factors applied to

the quality of service at an intersection. After the initial questionnaire the participants

were shown the video clips, consisting of a driver's view of a vehicle approaching,

waiting, and passing through an intersection. After each of these clips the participants

rated the quality of service they felt the intersection provided. At the end of the video

portion the participants were again asked to rate the factors important to quality of service

at an intersection to see if their initial opinion had changed. In all, 310 drivers

participated in this survey.

The results from the survey showed delay to be the most important factor both

before and after the film portion of the survey. This study provided the first results that

took into account the perceptions of travelers and changed the performance measure used

by the HCM to evaluate LOS in the 1985 edition. This study also recommended further









similar studies, and for further studies to simultaneously collect video and traffic flow

data to allow for accurate measurements of what is depicted on the films.

A study conducted by Pecheux et al. [9] addressed the issue of developing a study

method to assess the perceived LOS at signalized intersections. The first objective of all

the study methods was to determine how well the current LOS methodology reflects the

opinion of road users. The second objective was to determine the factors affecting users'

perceptions at signalized intersections.

The participants in this study represented a wide range of ages, education levels,

and incomes. The participants were first shown a series of approaches to signalized

intersections from a driver's perspective. After being shown a sequence of these clips, the

participants were asked to fill out a survey including their attitudes about certain driving

situations as well as their socio-economic information. After filling out these surveys the

participants were asked to discuss the factors that influenced their perception of quality of

service as a group. The study results showed that on average, the participants' delay

estimates were fairly accurate, however individual delay perceptions varied significantly.

Fifteen factors were identified that contributed significantly to quality of service. Finally,

the study found that participants tended to perceive service quality on three or four

distinct levels as opposed to the six HCM levels of service.

Another study using video clips and road user surveys was performed by

Choocharukul et al. [10] with the intention of evaluating the current HCM methodology

of assessing LOS. This study intended to provide a multivariate statistical analysis of the

factors that were important to road users' perception of trip quality and to compare those

factors to the current performance measures for LOS.









The data for this study were collected at various urban freeways. Cameras

mounted on overpasses were used and were focused on sections that included inductance

loop detectors. The cameras were focused so that only one direction of travel could be

observed. The data from the loop detectors were collected and synchronized with the time

of the video clips so the researchers would know the actual traffic flow conditions during

the time the video clips were recorded. Two sets of video clips were chosen, each

containing twelve clips. Two video clips were used for each HCM LOS designation, A-

F, with one clip on the high end of an LOS designation and the other clip at the low end.

These designations were determined by the loop detector information.

There were two groups of survey participants in this survey, one consisting of

students, transportation professionals, and environmental management professionals, and

the other consisting of commercial truck drivers and clerical and support staff. The

participants were provided with written descriptions of the six HCM LOS designations

(directly from the HCM). They were then asked to view the twelve video clips and rank

each of them with the LOS they thought was appropriate for the conditions. The

participants were also surveyed for demographic information such as age and education

levels, as well as information about their driving habits.

This study used an ordered probit statistical model to assess how users perceive

the LOS of the roadway sections. The results of the survey and analysis revealed that

perceived levels of service do not closely follow the HCM. Almost all the participants in

this study had a lower tolerance for LOS A than the HCM, with the average cut-off for

LOS A among the study participants shown to be 7 passenger cars per mile per lane

(pc/mi/ln) as opposed to the HCM cutoff of 11 pc/mi/ln. The HCM threshold for LOS F









also does not correspond with the findings of this study, with the participants selecting an

average of 82 pc/mi/ln as the upper bound of LOS F as opposed to the HCM LOS F of 45

pc/mi/ln. The study also found that factors other than density are likely to influence road

users' perception of quality of service. The results from both groups indicated that a

freeway with 4 lanes (instead of 6 or 8), an increase in traffic density, and an increase in

the standard deviation of vehicle speeds all contributed strongly to a worse perception of

LOS. It should be noted that the use of an overhead view of traffic could likely affect

survey participants' perceptions in a different way than that of an-vehicle view of traffic

and roadway conditions.


Background Study

A study was performed at the University of Florida by Washburn et al. [4] with

the objective of discovering what factors are important to drivers when evaluating the

quality of their trip on a rural freeway. Several methods were considered for this study

(focus groups, video/simulation viewing and review, interviews, etc.) with the final

choice being an in-field survey-based approach. Two hundred and thirty-three travelers

were surveyed at rest stops and service plazas along rural freeways in Florida. These

locations were chosen due to their access to travelers in the process of a rural freeway

trip. It was believed that this in-field survey approach would provide more reliable data,

than mail-back surveys for example, as the drivers' experiences would still be fresh in

their minds when filling out the surveys.

Drivers were asked to rank the factors that contributed to the quality of their trip

on a scale from 1 to 7. The most important factor, ranked in the top three 64.3% of the

time, was the ability to consistently maintain the desired travel speed. The factor with the









next highest ranking was the ability to change lanes freely and pass other vehicles. This

was ranked in the top three 33.3% of the time. The third most important factor was the

ability to maintain a speed no less than the posted speed limit. This factor was ranked in

the top three 33.0% of the time. This preliminary study showed that though density is

important to rural freeway travelers, it is not the most important factor in determining trip

quality. It also showed that drivers consider many other factors when determining trip

quality.


Conclusions

The studies detailed in this chapter have shown that, while some research has

been done on travelers' perception of quality of service, there is a need for more study.

The current HCM methodologies for evaluating LOS may be insufficient for determining

the perceived quality of service from the traveler's point of view. From the studies

summarized in this chapter, we can see that it is possible to understand and approximate a

traveler's perception of quality of service using the factors that are found to be important

to them. This type of research may ultimately assist decision makers when planning for

new roadways and roadway improvements.















CHAPTER 3
RESEARCH APPROACH

This chapter will describe the methods used in collecting the sample data for this

study as well as the methods used to refine the data for use in public surveys .Detailed

within the chapter are the choice of a survey method, the selection of data collection sites,

the creation of the survey form, and the process for conducting a road user survey.


Alternative Survey Methods

Common methods of data collection include the following: focus groups, field

surveys, in-vehicle surveys driven by a researcher, in-vehicle surveys driven by the

research participant, driving simulators, and video surveys.

Focus Groups This consists of recruiting test participants in order to arrange a

roundtable-type discussion about rural freeway travel. Participants would discuss

their rural freeway trip experiences and relate which aspects of rural freeway

travel are most important to them when evaluating the quality of their trip. The

advantage of a focus group is the relative ease of the survey, there is no video data

collection, field work, or liability on the part of the researchers. The disadvantage

is that participants may influence each other's responses and one particularly

vocal participant could swing the rest of the group towards his or her opinion.

Another disadvantage is the lack of a control element for the researchers there is

no one experience on which the participants are basing their opinions, so the

researchers can not look at the data or video record to interpret the responses.

16









Additionally, the potential lack of quantitative feedback upon which to build an

analytical model limits researchers in their ability to predict the responses of other

travelers faced with similar roadway and traffic conditions.

* Field Surveys Researchers distribute survey forms at locations frequented by

rural freeway travelers, such as rest stops or service plazas. The participants give

their opinions on rural freeway travel in a survey form, rating which factors are

most important when they judge their trip quality. One advantage to this method

is that participants surveyed have recently driven on a rural freeway and have this

experience fresh in their mind. Another advantage is that it is relatively easy to

recruit participants for this sort of survey; there is always a ready supply of people

in this type of location. The disadvantages are similar to the focus group.

* In-Vehicle Surveys (driven by research personnel) Participants are recruited and

driven along a section of rural freeway, then surveyed about their perception of

the trip quality. Advantages to this method include all participants would have

the same experience to draw upon for their responses, and there would be no need

to attempt to simulate the driving experience as participants would be

experiencing the conditions firsthand. The disadvantages to this method include

the liability to the researchers should the vehicle be involved in an accident, and

the time and effort involved in conducting a survey of this manner. The

controllability and repeatability of the conditions are also disadvantages because it

is not possible to ensure the same conditions will be experienced by multiple

survey participants.









* In-Vehicle Surveys (driven by research participants) Participants are recruited to

drive along a section of rural freeway and provide the researchers with feedback

on their trip once they return. Once again this method is advantageous in that it

would provide participants with a firsthand look at the conditions involved. The

disadvantages to this are similar to the previous method in that there is significant

liability attached to a method like this, and this method would be even more time-

consuming than the previous one. This method also suffers from the same lack of

control and repeatability as any in-car survey.

* Driving Simulator Participants are put behind the wheel of a real vehicle, but

the driving environment is simulated with the use of computer animation and

video display monitors. They would then participate in the virtual driving of a

rural freeway segment. This would give participants a closer likeness of actual

freeway travel without the liability of having them drive a real section themselves.

Disadvantages include cost (simulator time is expensive) and the well-

documented motion sickness problem for participants (which increases

recruitment time and costs).

* Video Surveys This method involves participants viewing pre-recorded video

scenes from actual rural freeway sites. The clips could be from one of two

perspectives:

o Overhead View A camera placed over the test section of rural freeway

records the traffic flow for survey participants to review at a later time.

While this method does not give a simulation of actually driving the









freeway section, it does give the participant a broader overview of the

traffic stream.

o Driver's Perspective A vehicle is equipped with a video camera to

record the rural freeway trip from the driver's perspective. This method

would better simulate actual rural freeway travel than an overhead view.



After considering all advantages and disadvantages, the method chosen was a

video survey from the driver's perspective. This method would allow larger groups of

people to be surveyed simultaneously while giving a reasonably accurate depiction of

rural freeway travel This method allows for complete control and repeatability of the

conditions experienced by the participants, as well as eliminating the liability issues

inherent in an in-vehicle survey.


Video Data Collection

The data collection method was developed after selecting the form the survey

would take. It included five specific tasks Site Selection, Equipment and Setup, Video

Data Collection, Video Clip Creation, and Loop Data Collection.


Site Selection

The sites at which the video clips were captured were all within Florida. Reasons

for this include the proximity to the University of Florida and the access to the Florida

Department of Transportation's (FDOT) network of more than 7,500 traffic monitoring

stations. The FDOT maintains a network of inductance loop detectors (ILD) along

Florida's Intrastate Highway System. There are two types of ILD stations permanent









and portable. The permanent stations were used in this study since they are continuously

recording data 24 hours a day, 365 days a year. The portable stations require a data

recorder to be hooked up at the location of the ILD station in a roadside cabinet. The

permanent stations are telemetered such that data can be downloaded and archived on a

daily basis. The data archived from the permanent stations are compiled every year by

the FDOT and published on the Florida Traffic Information (FTI) CD. Also included on

this CD for each state highway are the number of lanes in each direction, the ILD station

type (permanent or portable), a description of the site, the Average Annual Daily Traffic

(AADT), the percentage of trucks on the highway, and the peak hour in each direction,

Multiple sites were examined around the state. There were several factors leading

to the final site selections. South Florida was excluded due to the limited number of rural

freeway segments and the long driving times required to get there and back. Sites west of

Tallahassee were also not considered due to driving distance. These conditions hinged

upon the availability of suitable sites in north-central Florida. The final sites selected

represented a mix of four-lane and six-lane freeways, level and rolling terrain, truck

percentages, and a wide range of volume. This information was obtained from the Florida

Traffic Information CD published by the FDOT [11]. All sites selected were permanent

count stations instead of portable stations. Additional data were used from a similar study

conducted months before at the University of Florida. A list of the data collection sites

and their associated traffic data follows in Table 2. Maps of the locations of the data

collection sites can be found in Appendix A.















Table 2. Data Collection Sites and Traffic Data


Volume Two-way K D Truck
Site Site Type Description Direction Direction AADT Factor Factor %
1 2
189920 Telemetered SR-93/I-75, 3.5 mi south of Turnpike, Sumter Co. 20472 N 21250 S 41722 10.1 59.84 21.66
360317 Telemetered 1-75, 0.35 mi north of Williams Rd overpass, Marion Co. 37630 N 37844 S 75474 11.14 55.41 21.76
140190 Telemetered SR-93/I-75, 0.6 mi. south of SR 54, Pasco County 37443 N 37203 S 74646 8.76 53.67 11.71
730292 Telemetered SR-9/I-95, 1.4 mi south of Palm Coast Pkwy, Flagler Co. 29276 N 29980 S 59256 9.91 54.92 17.82
269904 Telemetered SR-93/I-75, 3 mi. north of Marion Co. line, Alachua Co. 31304 N 31023 S 62327 11.96 55.97 19.09
970428 Telemetered SR-91/F1. Turnpike, 797 ft. south of CR561, Lake Co. 17655 N 18088 S 35743 11.09 55.42 12.34










Equipment and Setup

The objective of video data collection was to depict travel along a section of rural

freeway from a driver's perspective. In order to give the survey participants a more

complete representation of the conditions on the freeway section, it was decided that two

more video images would be captured the vehicle's speedometer and the driver side

rear-view mirror. This brought the total number of cameras needed to three. The images

would be combined during the creation of the video clips.

The vehicle used for the video data collection was a minivan. As mentioned

above, three cameras were placed in the vehicle in order to capture different aspects of

the rural freeway trip. The first camera captured the view through the front windshield,

including a view of the interior rear-view mirror. It was placed on a mount secured to the

right side of the driver's seat. The second camera captured a view of the speedometer. It

was mounted on a suction mount affixed to the steering column. It was found during a

preliminary test that the instrument cluster needed to be shaded to reduce glare, so the

image would not appear washed-out. The final camera captured a view from the vehicle's

driver-side rear-view mirror. This camera was mounted on a pole secured between the

driver's seat and the door. Figure 1 shows these cameras as they were mounted in the

vehicle. The video images were captured by three portable VCRs placed inside the

vehicle. A microphone was also connected to one of the VCRs allowing the researcher to

announce when they crossed a detector loop and any other potentially important

information. This would allow the researcher to match the captured video clip to the loop

data collected in a later step. All these devices were powered by three 12-volt deep-cycle

batteries. A schematic of the equipment and connections is found in Figure 2.






























Figure 1. Camera Setup-Front View, Side View, Speedometer


Video Data Collection

The method used to capture the video clips remained constant throughout the

collection of data. The researcher would activate the three VCRs and start recording.

Then the researcher merged onto the freeway. The cameras captured conditions between

the exit ramps that came before and after the ILD station. The researcher would speak

into the microphone when the detector loop was crossed, giving the exact time and site

number so the clip could be matched with the relevant loop data. Up to four runs were

made at each location giving a number of clips to choose from when creating the clips for

the survey. The data collection for this project was performed during November of 2003

and March 6-9, 2004. A summary of each video data collection session is shown in

Table 3.











Camaer 2 Camera 3 Mirophone


Monitor


Figure 2. In-Vehicle Equipment Setup


Table 3. Data Collection Times, Locations, and Directions


Date
11/4/2003
11/5/2003
11/21/2003
11/21/2003
11/21/2003
11/21/2003
3/7/2004
3/7/2004
3/8/2004
3/8/2004
3/8/2004
3/8/2004
3/8/2004


Site
730292
269904
140190
140190
140190
140190
730292
730292
189920
360317
360317
360317
970428


Freeway
1-95
1-75
1-75
1-75
1-75
1-75
1-95
1-95
1-75
1-75
1-75
1-75
Turnike


Direction
NB
SB
NB
SB
SB
SB
SB
NB
SB
SB
NB
SB
SB


After reviewing the video data gathered on the first day of data collection, it was


deemed unusable. The mounting for the camera allowed too much vibration in the picture


and the video would not work for a public survey. Although the second round of video


Time
12:55
11:01
6:48
7:04
7:14
2:31
2:35
2:47
12:36
11:50
12:04
2:07
1:49


Camwa 1









data collection was scheduled for March 6-8, the runs made on the first day needed to be

redone due to problems with the camera placement, necessitating a fourth day of data

collection on March 9, 2004.


Video Clip Creation

The survey participants were shown a single video display that contained the

video scenes of the front windshield and interior rear-view mirror, the driver's side rear-

view mirror, and the speedometer. The display used was a video projector and a wall-

mounted screen, located between 5 and 20 feet away from the participants depending on

the specific survey location. The setup of one of the survey sessions is depicted below in

Figure 3. The majority of the screen was taken up by the view through the front

windshield. Since the front windshield view captured a portion of the dashboard as well

as the view from the front of the vehicle, the other two images could be overlaid on this

area. A screenshot from one of the video clips used in the survey is shown in Figure 4.

Screenshots from all 13 clips can be found in Appendix B.

The clips were assembled using a video-editing program (Adobe Premiere) [12].

They first had to be captured from the VHS tapes using an ADS digital encoder [13].

After they were stored on the computer hard drive they were combined using Adobe

Premiere into clips from 1.5 to 2.5 minutes in length. The length of an individual clip was

chosen based on events in the video that the researchers wanted to include or exclude, as

well as with a survey participant's attention span in mind. The clips shown to viewers

were chosen based on conditions they represented that were unique or different from

other clips. This selection process is explained further in the section entitled "Video Clip

Selection".























Figure 3. Setup of a Survey Session


Figure 4. Sample Video Screenshot


Inductance Loop Detector Data Collection

It was desired to calculate the LOS of these sites according the HCM

methodology in order to assess how strongly correlated it was with the responses









provided by the survey participants. In order to determine the HCM LOS of the rural

freeway segments, data were collected from the inductance loop detector (ILD) stations

at each test site. The data collected came in three files-speed, volume, and vehicle

classification. FDOT personnel programmed the detectors at the sites selected for the

study to record data in five-minute intervals (the hardware minimum interval) rather than

the usual one-hour interval. This shorter interval allowed for traffic data that more

accurately reflected the conditions depicted in the video clips. It should be noted that

even with a five-minute data collection interval the conditions shown in the video clips

could potentially vary from the average provided by the ILD data. These ILD data were

used to categorize the collected video data and provided a starting point for selecting a

range of conditions to be represented in the survey.

The ILD data were provided in the form seen in Appendix D. When available,

there were three data files for each site speed, count, and class. In the speed file, counts

are provided for each speed range. The midpoints of the speed ranges are shown at the

top of the table. In the class file, descriptions such as "CL01" are given to the columns.

These refer to the specific class of vehicle counted in that group and are explained by the

figure provided. From the data provided it was possible to calculate descriptive statistics

for the traffic flow at each site, such as the percentage of heavy vehicles in the traffic

stream, the total 5-minute volume, the average speed, and the density.


Video Clip Selection

There were thirteen video clips chosen for the final survey. The final number of

clips chosen was a result of five pilot test sessions, striking a balance between coverage

of alternatives and attention/focus span of participants. These preliminary tests had






28


shown that many participants lost interest after two minutes and had already started

writing their opinions down. The final video clips were chosen to represent a variety of

conditions in categories including lane configuration, traffic density, terrain, truck

percentage, the presence of a median or guardrail, and shoulder configuration. The

relevant data for each video clip included in the final survey is included below in Table 4

and Table 5.


















Table 4. Traffic Data for 13 Video Clips


ILD LD ILD
Clip Clip Truck Inside Middle Outside Inner Middle Outer Terran
# Road Dir Lanes Length Volume' Density Truck Speed Speed Speed Avg Lane Lane Lane 5min Terrain Speed
# Length %0 Speed Speed Speed Av Lane Lane Lane Volume
% Speed Volume
1 1-75 S 2 2:10 low none 8.00 0.13 77.1 72.2 74.30 42 57 99 flat 75
2 1-75 S 3 1:52 med-high high 63 75 66 204 flat 70-75
3 1-75 N 2 2:00 med-high med 13.79 0.20 76.4 69.4 72.20 67 99 166 flat 60-70
4 1-95 N 2 1:35 very high 26.45 56.5 55.4 56.00 102 145 247 flat 40-55
5 1-75 S 3 1:40 low-med med 6.30 0.17 77.6 74.0 66.4 72.30 26 51 37 114 rolling 70-75
6 1-95 S 2 1:59 med 10.21 76.9 74.7 75.80 63 66 129 flat 70
7 1-75 S 2 2:00 med-high high 26.31 0.34 71.6 65.5 68.90 167 135 302 flat 67-72
8 1-75 N 3 2:01 med-high low 61 98 80 239 flat 67-72
9 1-75 S 2 2:00 high high 26.11 0.32 71.3 66.1 69.20 179 122 301 flat 55-65
10 1-95 N 2 1:43 med 10.86 78.4 69.4 72.90 80 52 132 flat 75
11 1-75 S 3 1:26 med med 48 82 48 178 flat 70-75
12 1-75 S 2 1:27 med-high med 17.04 0.15 73.6 68.2 71.10 110 92 202 flat 60-65


13 Turnpike S 2 2:03 med high


rolling 75-80


1 These levels (low, med, high) indicate subjective judgments that were used to choose between clips.










Table 5. Clip Sites, Dates, and Times


Clip # Clip Site Time Date Closest City
1 189920 run 1 189920 12:36 3/8/2004 Wildwood
2 360317 run 1 360317 11:50 3/8/2004 Ocala
3 Tampa 0648 140190 6:48 11/21/2003 Tampa
4 730292 run 4 730292 14:47 3/7/2004 Daytona Beach
5 Micanopy 1101 269904 11:01 11/5/2003 Micanopy
6 730292 run 3 730292 14:35 3/7/2004 Daytona Beach
7 Tampa 0714 140190 7:14 11/21/2003 Tampa
8 360317 run 2 360317 12:04 3/8/2004 Ocala
9 Tampa 0704 140190 7:04 11/21/2003 Tampa
10 Daytona 1255 730292 12:55 11/4/2003 Daytona Beach
11 360317 run 3 360317 12:07 3/8/2004 Ocala
12 Tampa 1431 140190 14:31 11/21/2003 Tampa
13 970428 run 1 970428 13:49 3/8/2004 Winter Garden



Survey Sessions


Development of Survey Form and Participant Instructions


The survey form for this study had to serve two purposes-record the

participants' opinions about the rural freeway video clips and their reasons for these

opinions, and record characteristics about the participants that might influence their

ratings. Thus the form is divided into two sections.

The first section of the survey form is for personal information about the traveler

taking the survey. Examples of this information include education level, income, and

number of years possessing a driver's license. This section also records information about

the participant's rural freeway travel habits. It asks for information such as the amount of

rural freeway trips taken per month and the average length of the participant's rural

freeway trips. Finally it asks for some driving habits, such as any changes in the









participant's driving style when driving alone versus with a passenger. It also asks the

participant to rate their usual driving style, from Conservative to Aggressive.

The second section of the survey is for recording the participant's opinions and

rankings of the video clips. It is divided into two sections for each of the thirteen clips.

The first section asks the participant to rank the quality of the trip depicted in the video

clip on a scale from 'Very Poor' to 'Excellent' with 6 total ranking levels. A total of six

ranking levels was chosen so that there would be general correspondence with the six

levels of the HCM (A-F). Participants were asked to use the word ranking rather than a

numerical ranking (e.g., 1-6) to minimize the possibility that those familiar with the HCM

might try to equate the numerical rankings with the HCM LOS rankings. The second

section asks the participant to record why they ranked the video clip as they did, listing

all factors that significantly contributed to their ranking. The participants were to then

number these according to their relative significance to each other.

Finally the form includes questions about the survey itself. These include the

participant's opinion on the video clips as a representation of rural freeway travel and if

the participant would have changed their rankings based on the purpose of the trip (e.g.,

business, recreational, or social).

A one page written survey instruction sheet was developed because there was a

significant amount of information that needed to be communicated to the participants in

order for them to complete the survey form in a manner which would be useful as study

data. The participants could refer back to it if there were any questions about the survey

process. The instructions given to each survey participant are provided in Appendix C.









Conducting the Survey Sessions

Survey participants were recruited from various sources. They include the

following:

Undergraduate students in the University of Florida civil engineering program,

recruited from the introductory transportation engineering course,

Graduate students in the University of Florida civil engineering program,

recruited from the transportation degree program,

Employees of the University of Florida Technology Transfer Center,

Employees of the Florida Department of Transportation, and

Alachua county residents (Random participants recruited for a fee by the Florida

Survey Research Center)

The undergraduate students were recruited from the Principles of Highway

Engineering and Traffic Analysis course during the Fall 2004 semester. The graduate

students were those enrolled in a transportation engineering degree program during the

Fall 2004 semester. The University of Florida Technology Transfer Center is an

organization that provides training and technical assistance to Florida's transportation and

public works professionals. Their survey session was conducted at their off-site

headquarters in Gainesville, FL, with participants ranging from high-school educated

support staff to professionals with graduate degrees. The FDOT survey session was

conducted at the central office in Tallahassee, FL. This session also included participants

of varying backgrounds and demographics. The public sample was comprised of Alachua

county residents, recruited by the University of Florida Survey Research Center. The

survey center was instructed to recruit individuals with varying socio-demographic










characteristics and also make sure that the participants had experience driving on rural

freeways. Additionally, they did not recruit college students as there was already a

sufficient number in this group.

In total there were 126 surveys filled out for this study. The locations, dates, and groups

of participants taking the survey during each session are given in Table 6.


Table 6. Dates and Locations of Survey Sessions


Survey # of
Session Date City Location Participants Surveys
1 8/4/04 Gainesville UF Technology Transfer Center T2 employees 16
2 11/16/04 Tallahassee Florida DOT Central Office DOT employees 11
3 12/2/04 Gainesville University of Florida undergraduate students 14
4 12/2/04 Gainesville University of Florida undergraduate students 9
5 12/4/04 Gainesville UF Hilton Conference Center public' 13
6 12/4/04 Gainesville UF Hilton Conference Center public' 15
7 12/4/04 Gainesville UF Hilton Conference Center public' 11
8 12/9/04 Gainesville University of Florida undergraduate students 20
9 1/22/05 Gainesville University of Florida public' 9
10 1/27/05 Gainesville University of Florida graduate students 8
Total Number of Surveys 126
Participants were recruited through the University of Florida Survey Research Center




Because of the video format of the survey, multiple surveys could be filled out at

a time, the main limitations being the ability of the participants to comfortably view the

video screen and the length of time for which the participants could be expected to focus

on this task. The screen was placed as close as possible to eye level so participants

looking at the screen saw it as they would a car's windshield. Before viewing the clips

the participants were given the instruction sheet and time to read it. These written

instructions were also verbally reviewed by the session moderator, as well as some

supplemental information. The participants were also told that they could ask









interpretation questions in-between the viewing of the video clips. It was decided to

create two example clips, each 20 seconds long, to show the upper and lower ends of the

range of possible traffic flows. The first was a nearly empty four-lane freeway and the

second was stop-and-go traffic along a four-lane freeway. The participants were then

shown each of the 13 video clips and instructed to watch each clip entirely before writing

their responses. Since it was not intended for the order of the clips to have any effect on

the participants' rankings, the order was shifted for each survey session. After each clip

was finished, the participants were given time to record their rankings.















CHAPTER 4
ANALYSIS AND RESULTS

This chapter contains information about the methodology used to analyze the

survey data, as well as the results of these analyses.


Analysis Method

To determine how or if the participants' responses correspond to the six LOS

rankings, a statistical analysis was needed to predict the probability of selecting discrete

rankings (1-6 as included in the survey). While one of several multinomial discrete-

choice modeling methods would suffice to predict a discrete outcome, most do not take

into account the ordered nature of the responses in this survey (1 is better than 2, which is

better than 3, etc.). Using a standard multinomial discrete model, such as a multinomial

logit model, would still yield consistent parameter estimates, but with a loss of efficiency

[14]. In order to account for the discrete and ordered responses in this survey, an ordered

probability model was chosen as the statistical analysis approach.

An ordered probability model is derived by defining an unobserved variable, z,

that is the basis for modeling the ordinal ranking of data (in this case the six clip

rankings) [15]. This variable is specified as a linear function for each observation n such

that



Zn = Xn + En (1)









where X, is a vector of variables determining the discrete ordering for observation n, f is

a vector of estimable parameters, and En is a random disturbance. In this analysis, y is

defined as each participant's evaluation of each of the 13 video clips. Since there are 126

participants and 13 clips, there are a total of 1638 observations. Using this equation, the

observed clip ranking, yn for each observation is written as



Yn = 1 ifzn < u

Yn = 2 if/Ul z, < aU2

yn = 3 ifU2< z, < U3 (2)

Yn = 4 if U3< Zn < p4

Yn = 5 ifi4< Zn <_U/

Yn = 6 ifz, > us



where the p values are the thresholds that define Yn. The u values are estimated jointly

with the model parameters (P). The estimation problem then becomes one of determining

the probability that a participant will select a particular ranking for each clip. In using the

ordered probit model, it is assumed that the error term, en, is normally distributed with a

mean of 0 and a variance of 1. The resulting ordered probit model has the following

probabilities corresponding to each clip ranking:



P(yn = 1) = 0(-fX,)

P(yn = 2) = 1(u, PX,) 0(-fX,,)

P(~n = 3) = 0(U2 PX,) O(j1 fX,) (3)









PO2n

P(y'

P(y'


1(U flX.) 1(U2- flX.)

1(4 -- '-(U2 -PXn)

\-^4~-PXn)


It can be shown that threshold yi can be set equal to 0 without loss of generality [15]. In

the above equations, 0(.) represents the cumulative normal distribution:


1 ~1,
(u) = \ e 2 dw
v^-00


This model can be estimated using maximum likelihood procedures.

The thresholds u1 and u1 define the upper and lower thresholds for outcome i.

This is illustrated in Figure 5.


y -4


y=2


y= I


y 5


y=6


Figure 5. Illustration of an Ordered Probability Model


A positive increase in the f term implies that an increase in x will increase the probability

that the highest category response will be returned (in this case, y = 6). An increase in the


-f fly it) -/ fly n.,. P -PC









p term also implies that the probability of returning the lowest response (y

decreased. This is illustrated in Figure 6.


1) is


y=4


y=2


y=l


y=5


y=6


Figure 6. Illustration of an Ordered Probability Model with an Increase in f

A unique issue was present in this data set that complicated the analysis

procedure. Each of the 126 participants viewed 13 clips and thus generated 13

observations. The issue is that there are unobserved characteristics that are unique to each

participant that will be reflected in all 13 of their rankings. If this is not accounted for in

the model, the model will be estimated as though each of the 1638 observations came

from a unique participant. This approach would result in lower standard errors in the

model's estimated parameters, leading to inflated t-statistics and exaggerated degrees of

significance.

The solution to this problem is found in a standard random effects approach. The

first equation is rewritten as


- fly iA -tix -A 4-











zic = PXic + E,c + P, (5)



where i denotes each participant (i = 1,...,126), the c denotes each video clip (c =

1,...,13), p, is the individual random effect term and all other terms are as previously

defined. The random effect term p, is assumed to be normally distributed with mean 0

and variance oa. When this random effects model is estimated, an estimate of o is also

calculated, the significance of which determines the significance of the random effects

model relative to the standard ordered probit model [16].


Statistical Analysis

The results of the surveys were put into spreadsheet form, with unique cases for

each clip viewing. Each participant's rankings were kept together within the spreadsheet

for analysis purposes. The data were analyzed using LIMDEP [17] with a random effects

approach as detailed in the previous section.

The first analysis was performed to explore how the quality of service perceptions

of the participants in this survey correlated with the HCM LOS thresholds. The density

for each of the video clips was calculated from the loop detector data (and the video data,

in cases where the loop detector data was incomplete). A statistical analysis was

performed using density as the only independent variable to find out where the thresholds

of the survey participants fell relative to the six clip rankings. The results are given below

in Table 7.

The very high level of significance indicated by the t-statistic (coefficient divided

by standard error) calculated for density in the above model offers some evidence that










this performance measure correlates well with perceived LOS. The reference t-statistic

for these analyses is 1.282, representing a 90% confidence level in a one-tailed t-test. The

positive coefficient calculated for density indicates that, as density increases, the

likelihood of a traveler perceiving a worse LOS increases. The random effects term, a, is

also highly significant, meaning that the choice of a random effects model for this data

set was correct. Had this term not been significant, a normal ordered probability model

would have been sufficient.

One test for the goodness-of-fit of a model is calculating that model's p2 value.

The p2 value of a model is between 0 and 1. A p2 value of 1.0 indicates a perfect model

fit. The p2 value of a model is calculated as follows:

p2 LL()- K(6)
LL(0)

where K represents the number of variables in the model, LL(fl) represents the log

likelihood at convergence, and LL(O) represents the initial log likelihood [15].

Table 7. Density Model Estimation Results

Standard
Variable Coefficient Error t-statistic
Constant -0.138 0.076 -1.82

Traffic ( I i ,. i ....
Density (pc/mi/ln) 0.096 0.003 34.37

Threshold Values
/l 0.918 0.038 23.89
P2 1.922 0.048 39.92
/3 2.863 0.053 53.88
/P4 4.112 0.066 62.47

Standard Deviation of Random Effects
a 0.455 0.050 9.12

Initial Log Likelihood -2710.16
Log Likelihood at Convergence -2314.60
p2 0.15









Using the participants' responses it was possible to calculate a set of thresholds

for the participants' assigned LOS rankings. Using the calculated values in Table 7, the

threshold values can be calculated as (Uk o)/fll. In this equation, k designates the five

threshold values, uj = 0, and the other threshold values are given in Table 7. A

comparison between the calculated threshold values from this survey and the HCM LOS

thresholds is given in Table 8.


Table 8. Comparison of Estimated and HCM LOS Thresholds

Estimated Thresholds HCM thresholds
LOS
(pc/mi/ln) (pc/mi/ln)
A 0-2 0-11
B >2-11 >11-18
C >11-21 >18-26
D >21-31 >26-35
E >31-44 >35-45
F >44 >45


These thresholds are generally lower than the HCM thresholds for corresponding

rankings, indicating the participants in this survey had a lower tolerance for high-density

traffic conditions than could be inferred from the HCM LOS thresholds.

The second analysis that was performed was intended to take into account all the

traffic and roadway characteristics influencing the participants' perception of trip quality.

The results of this table are given below in Table 9.

The traffic characteristics examined produced effects according to expectations.

The calculated difference in speed between the inner lane and the outer lane was in the

model as "speed differential". As this value increased, participants were more likely to

assign a worse LOS to a given set of conditions. A higher average speed resulted in a

more favorable LOS ranking. Motorists in this survey found three lanes in one direction










to be a preferred configuration over two lanes and were more likely to assign a favorable

LOS ranking to those roadways with three lanes in one direction.


Table 9. Traffic Characteristics Model Estimation Results


Standard t-
Variable Coefficient Error statistic

Constant 6.296 0.597 10.55

Traffic ( Ii. i ....
Speed Differential (mi/h) 0.163 0.027 6.08
Average Speed (mi/h) -0.096 0.009 -10.97
3 Lanes (1 Yes, 0 No) -1.848 0.210 -8.82
Truck % 0.005 0.004 1.04
Density (pc/mi/ln) 0.061 0.006 10.59

Threshold Values
p/ 0.949 0.064 14.88
P2 2.192 0.077 28.48
P3 3.258 0.092 35.60
/P4 4.630 0.106 43.80

Standard Deviation of Random Effects
o 0.522 0.060 8.76

Initial Log Likelihood -2710.16
Log Likelihood at Convergence -1472.53
p2 0.45


An increase in the truck percentage resulted in a higher possibility of a worse LOS

ranking. While the t-statistic for the truck percentage was below 1.282, it was decided to

leave this variable in the model because it was felt that this was a very important variable

from a policy standpoint. As expected, the participants preferred not to have a high

percentage of trucks in the traffic stream. Finally, density was very significant in this

model as it was in the first. A higher density led to an increased possibility of a worse

LOS ranking. The random-effects term was again significant in this analysis, justifying

the use of a random-effects model.











The third analysis that was performed was aimed at discovering which factors are

important to travelers when judging their trip quality. This model was estimated

including demographic data as well as roadway and traffic flow characteristics. The

values given in Table 10 should be interpreted such that a positive parameter estimate

means that an increase in that variable will lead to a better perceived quality of service,

and a negative parameter estimate means that an increase in that variable will lead to a

worse perceived quality of service.


Table 10. Level of Service Model Estimation Results


Standard t-


Variable
Constant

Demographic and Background Information
Age > 35 (1 Yes, 0 No)
Income (thousands of $)
Average Number of Rural Freeway Trips per Month
Average One-Way Trip Distance > 100 miles? (1 Yes, 0 -
No)
Less Aggressive Driver with Passengers? (1 Yes, 0 No)

Traffic ( i ... .i,..i .
Speed Differential (mi/h)
Average Speed (mi/h)
3 Lanes (1 -Yes, 0 -No)
Truck %
Density (pc/mi/ln)

Threshold Values


Standard Deviation of Random Effects


Initial Log Likelihood
Log Likelihood at Convergence
p2
P


Coefficient Error statistic


6.156 0.622


-0.358
-0.003
0.025

0.395
0.267


0.162
-0.095
-1.836
0.005
0.062


0.939
2.181
3.247
4.613


0.435


0.121
0.002
0.017

0.127
0.186


0.028
0.009
0.217
0.005
0.006


0.065
0.078
0.093
0.107


0.059


9.90


-2.96
-1.89
1.49

3.11
1.43


5.85
-10.58
-8.47
1.03
10.56


14.56
27.93
34.90
43.21


7.42

-2710.16
-1447.34
0.46









In Table 10, a positive coefficient value indicates that as the variable increases,

there is an increased likelihood of a worse perception of LOS. Likewise, a negative

coefficient value indicates that as the variable increases, there is an increased likelihood

of a better perception of LOS.

The results indicate that, while density is important to travelers, it is not the only

factor influencing perceived quality of service. The survey results showed significant

effects of demographic and background information on drivers' LOS rankings. Table 10

indicates that participants with over 35 are more likely to assign a given set of conditions

a better LOS, as are those with higher incomes.

Travelers who drive on rural freeways more frequently are more likely to perceive a

worse LOS, as are those whose average rural freeway trip is over 100 miles in one-way

length. Those participants who indicated that they tend to drive less aggressively with

passengers in the car as opposed to driving alone were more likely to assign a worse LOS

to a given set of conditions. A possible explanation is that these drivers are more

aggressive than the average motorist. Participants were asked if they considered

themselves to be an aggressive driver, and the results of that model did not display

significance. Perhaps motorists were more reluctant to admit they drive aggressively, but

this tendency manifests itself in their responses to this question.

The results estimated using the traffic and roadway characteristic variables

showed similar significance and magnitude to the model estimated only using these

variables. The random effects term was once again significant.















CHAPTER 5
CONCLUSIONS AND RECOMMENDATIONS

Since 1963, the Level of Service concept has been integral to the Highway

Capacity Manual methodology for assessing the performance of transportation facilities.

There is, however, still relatively little known about how the HCM methodologies for

assigning LOS correspond to road users' perceptions of their quality of service. The

purpose of this study was to investigate what factors influenced road users' perceptions

of quality of service, and how that perception compares to HCM calculated LOS.


Data Collection and Video Clip Creation

The data collection process used for this study proved successful in gathering the

necessary video data. After deciding on the best camera positions and mounting

techniques, all cameras recorded clear, steady views of their intended targets. The

equipment in the vehicle performed exactly as intended, capturing the necessary

information while keeping all three VCR timers consistent so the video data could be

synchronized at a later time.

The sites chosen generally provided a good variety of traffic conditions, but some

clips from a pilot study were also used to provide additional roadway and traffic

conditions that were not captured in the data collection effort for this project. These clips

were re-edited using the same process as the clips filmed for this study so there would be

consistency in the screen views.









The loop detector data did not work out as well for some of the sites as was

initially hoped. Due to malfunctioning detectors or construction at the selected sites,

some of the desired data were not available.

The final form of the video clips and the presentation to survey participants

worked very well, exactly as intended. The last question on the survey form (as seen in

Appendix A) asked participants to rate how well the video clips simulated the driving

experience for the conditions depicted on the screen. The majority of participants found

the survey to be a "very good" representation of the actual driving experience, with 95%

of the participants rating the survey as a "good" or better representation of the actual

driving experience. The responses to this question are tabulated in Table 11. As shown in

this table, the average response from participants was approximately a 2 out of 6,

corresponding to "very good".


Table 11. Realism of Video Survey Responses


Ranking Excellent Very Good Good Fair Poor Very Poor
1 2 3 4 5 6
Frequency 21 64 36 5 1 0
Percent of Total 1
17 50 28 4 1 0
Responses (%)
Average Rank 2.2


Statistical Analysis


The analysis process chosen for this survey was an ordered probability model,

specifically the ordered probit model. The structure of the standard ordered probit model

formulation does not account for each participant providing 13 responses, so a random-

effects formulation was used. This modeling choice was justified, with the standard

deviation of random effects showing significance in all statistical analyses.









The first model developed was one incorporating only density as an independent

variable. This produced results that were as expected, that density is very significant to

travelers when they are judging the quality of service provided by a rural freeway.

A complimentary outcome of this analysis was that density thresholds for each

LOS were estimated according to the survey participants' responses. For LOS A-E, the

survey participants showed a lower tolerance for high-density traffic conditions, hence

their estimated thresholds were lower. The HCM thresholds and the estimated thresholds

showed similar values for LOS F.

The second model was estimated to include the influence of other roadway and

traffic characteristics. The results of this model showed that while density is significant to

user perception of LOS, there are other significant factors influencing this perception,

such as average speed of the traffic stream and the speed differential between lanes.

The final model included all factors from the survey that were found to be

significant, including demographic factors as well as roadway and traffic characteristics.

The results of this model indicated that the background and characteristics of the

individual road user can influence their perception of LOS. While this result was

expected, it is still significant due to the implications for a potential future modification

to the HCM LOS methodology.


Study Limitations and Recommendations for Further Research

Since the scope of this study was limited to North Central Florida, additional

testing with participants from a variety of other geographic regions would be needed to

adopt any findings on a national level. An expanded sample, both geographically and in

roadway conditions, would provide much more comprehensive coverage of the roadway









and traffic condition combinations. The video survey format has inherent limitations as

well. In a future study, it would be desirable to allow road users to drive in a traffic

stream with known characteristics (density, truck percentage, etc.), then express their

opinion regarding the LOS of the roadway section. This was not considered for this study

due to cost and liability. The results of this survey could be compared to the results of the

video survey to assess the accuracy of the video survey. If the video survey is shown to

be an accurate method of simulating traffic conditions, it can be used in future studies and

will be more effective than in-field surveys. Finally, although participants were told to

imagine the conditions in the video scenes as if they were occurring throughout the

duration of a trip, it is not known whether actually experiencing these conditions for an

equivalent time to an entire trip would change the outcome.

It is hoped that the findings of this study will lead to further developments in this

area. The study does show that density is significant in determining a road user's

perception of trip quality. It is also known that there are significant factors influencing

LOS other than density and these should be explored more completely. Ultimately, a

better understanding of travelers' perceptions of quality of service will lead to a better use

of the available resources to improve the roadway network where it is really needed, and

to more accurate planning and accommodating for future demands.















REFERENCES


1. Transportation Research Board (2000). Highway Capacity Manual. TRB,
National Research Council. Washington, D.C

2. Harwood, D., Flannery, A., McLeod, D., Vandehey, M. (July 2001). The Case for
Retaining the Level of Service Concept in the Highway Capacity Manual.
Presented at the 2001 Transportation Research Board Committee A3A10 -
Highway Capacity and Quality of Service Midyear Meeting, Truckee, California.

3. Transportation Research Board (1985). Special Report 209: Highway Capacity
Manual. TRB, National Research Council. Washington, D.C

4. Washburn, S., Ramlackhan, K., McLeod, D. (2004). Quality of Service
Perceptions by Rural Freeway Travelers: Exploratory Analysis. Transportation
Research Record: Journal of the Transportation Research Board, No. 1883.
Washington, D.C., pp. 132-139.

5. Hostovsky, C., Wakefield, S, Hall, F. (2004). Freeway users' Perception of
Quality of Service: A Comparison of Three Groups. In Transportation Research
Record: Journal of the Transportation Research Board, No. 1883. TRB, National
Research Council. Washington, D.C., pp. 150-157.

6. Pecheux, K., Flannery, A., Wochinger, K., Rephlo, J., Lappin, J. (2004).
Automobile Drivers' Perceptions of Service Quality on Urban Streets.
Transportation Research Record: Journal of the Transportation Research Board,
No. 1883. TRB, National Research Council. Washington D.C. pp. 167-175.

7. Nakamura, H., Suzuki, K., Ryu, S. (2000). Analysis of the Interrelationship
Among Traffic Flow Conditions, Driving Behavior, and Degree of Driver's
Satisfaction on Rural Motorways. Transportation Research Circular E-C018:
Proceedings of the Fourth International Symposium on Highway Capacity.
National Research Council. Washington, D.C., pp. 42-52


8. Sutaria, T.C., and Haynes, J.J. (1977). Level of Service at Signalized
Intersections. Transportation Research Record: Journal of the Transportation
Research Board, No. 644. TRB, National Research Council. Washington, D.C.,
pp. 107-113.









9. Pecheux, K., Pietrucha, M., Jovanis, P. (2000). User Perception of Level of
Service at Signalized Intersections: Methodological Issues. Transportation
Research Circular E-C018: Proceedings of the Fourth International Symposium
on Highway Capacity, National Research Council. Washington, D.C., pp. 322-
335.

10. Choocharukul, K., Sinha, K., Mannering, F. (2004). User Perceptions and
Engineering Definitions of Highway Level of Service: an Exploratory Statistical
Comparison. Transportation Research Part A, 38. pp. 677-689.

11. Florida Traffic Information 2003. (2003). Florida Department of Transportation,
Tallahassee, FL, CD-ROM.

12. Users Guide for Adobe Premiere Pro Software. (n.d.). Last Accessed November
17, 2003, from http://www.adobe.com/products/premiere

13. Users Guide for ADS Pyro A/V Link. Last Accessed March 15, 2005, from
http://www.adstech.com/products/API-555/intro/api555_intro.asp?pid=API-555

14. Amemiya, T. (1985). Advanced Econometrics. Harvard University Press.
Cambridge, MA.

15 Washington, S., Karlaftis, M., Mannering, F., 2003. Statistical and Econometric
Methods for Transportation Data Analysis. Chapman & Hall/CRC. Boca Raton,
FL.

16 Greene, W., 2003. Econometric Analysis. Prentice Hall. Upper Saddle River, NJ.

17. Users Guide for LIMDEP 8.0. 2004. http://www.limdep.com Econometric
Software, Inc. Last Accessed March 23, 2005.





















APPENDIX A
LOCATIONS OF DATA COLLECTION SITES







LEWISR

9189920

189920
-I I ',-- \






52








47 ,iop
L 7. .
-* PU- N :wl


M ant* .


6.
--- I_ I -
I %,- ,.- _Q


Jl -iK A,+.
"'


I, *.^A,
S73029








20
AGLER -- 1 ':' I




,jqi.
-, ...\, ..



200
K .






53








-i' / A/0190
,581








*--, HI BOROUGH /























415A
7 30317
328- -






54









/ ^26




234








,269904
'3 46 3 46. .3
S241 .. ,


d? Li





















APPENDIX B
VIDEO CLIP SCREENSHOTS









56





















. .. .-






57




58






I
mI






59






60








61
















































. .. .








62




































II





















APPENDIX C
RURAL FREEWAY TRIP QUALITY SURVEY FORM




















UNIVERSITY OF
FFLORIDA RC
Transportation Research Center

Rural Freeway Trip Quality Survey

In the exercise you are about to participate in, you will be watching a series of 13 short video
segments of various roadway and traffic conditions on rural frec'. a' s A rural freeway is a
freeway that travels through relatively unpopulated areas. Rural freeways are typically used for
longer trips, such as city-to-city trips. All freeway segments (whether in urban, rural, or other
types of areas) are characterized by opposing directions of traffic being separated by either a
physical barrier or open space. All freeways are also characterized by limited access, that is,
entry to and exit from a freeway can only be made at interchanges (on- and off-ramps). For rural
freeways, interchanges are spaced much further apart than along freeways in urban areas.

Each of the video clips is approximately 1.5 to 2 minutes in length. Each clip is intended to give
you a "snapshot" of the typical conditions experienced over the course of an extended trip on a
rural freeway. When watching each video clip, please imagine and/or keep the following points
in mind:
The conditions viewed on the video clip for about 2 minutes are intended to be
representative of what you would experience for a much longer trip (30 minutes or more).
Imagine how you would personally drive, or try to drive, in the given conditions. You
are not limited to the driving behavior of the vehicle from which the video is being.
viewed. The intent of the video vehicle is to provide you with a reasonable
representation of the typical conditions being experienced by ALL motorists on that
section of rural freeway. Therefore, your survey responses should not be specific to how
the video vehicle was being driven. If you feel like you would, and could, drive
differently under the given conditions, then base your survey responses on that. It is
important that your survey responses reflect how the given conditions affect your
perception of trip quality based upon your own desired driving behavior.

After watching each video clip, we ask that you do the following on the survey form:
Rank (from Very Poor to Excellent) the travel conditions
In the space provided, briefly list the reasons/factors for why you ranked the conditions in
that video clip as you did. Please be as specific as possible-for example, you might say
'opportunities to pass other vehicles in order to maintain my desired speed were limited',
as opposed to 'speed was too low'.

The video clips are intended to be weather neutral-that is, in developing the video clips it was
not our intent to have weather be a significant factor in your trip quality perceptions. Although
the lighting conditions may vary somewhat, please do not factor in the environmental conditions
unless you feel very strongly about a certain condition.

If you recognize the freeway section, disregard previous knowledge and experience and base
your ranking strictly upon the conditions observed in the video clip.


Thank you for ;, our ciotperaliori and participation.





















L UNIVERSITY OF ,
FLORIDA
-- Transportation Research Center

About Yourself

Gender: 0 Male a Female

Age: 0 16 to 25 years o 26 to 45 years D 46 to 65 years E Over 65 years

Marital Status: D Single L Married [ Separated/Divorced E Widowed

Highest level of education:
j Some or no high school 0 High school diploma or equivalent
P Technical college degree (A.A.) D College degree a Post-graduate degree

Approximate annual household income:
Dl No income u Under $25,000 0 $25,000- 49,999 0 $50,000 -74,999
j $75,000 99,999 o $100,000 149,999 O $150,000 or more

Number of years possessing a driver's license:

About Your Rural Freeway Driving

Typical number of rural freeway round trips made during a month?
O 1 to2 o 3 to4 E 5 to6 E 7 to 8 9 to 10 0 11 to 12 Over 12

Typical percentage of these trips made as a driver __ as a passenger__ (should sum to 100)

Typical one-way length of trip made on a rural freeway (in miles)?
C ess than16 miles n 16to30 0 31 to 45 E 46 to 60 E 61 to 75 o 76 to 100
0 101 to 125 [ 126 to 150 D 151 to 175 0 176to200 E Over200

Vehicle type most often used for rural freeway trips:
Sedan L Sports car E Pickup truck E SUV 0 Minivan
Full-size van L RV/Motorhome 0 Motorcycle D Other

Typical number of passengers in vehicle for rural freeway trips?
O 0-Driver only 0 1 E 2 o 3 o 4 or more

Typical driving style on rural freeways (on a scale from 1-5, with 1 being 'Very Conservative' and
5 being 'Very Aggressive'):

When driving alone, versus driving with passengers, does your driving style become:
E Less aggressive L Stay the same r More aggressive

























Your Opinions

Rank the overall quality of your trip (Excellent, Very Good, Good, Fair, Poor, Very Poor) for the given
roadway and traffic conditions observed in each video clip. In the space provided, list all the significant
factors/reasons that influenced your ranking of the trip quality for each video clip. After listing the factors,
please number them from most significant to least significant (with 1 being the most significant).

Video Clip Rank Comments







2



3



4



5



6



7



8



9


























10



11



12



13




In general, how would the purpose of your trip (such as business, recreational, social) affect the trip quality
rankings assigned above (e.g., higher, lower, not at all)?








If the conditions in the video clips were encountered in an urban setting, and the trip length was relatively
short, how would this affect the trip quality rankings assigned above (e.g., higher, lower, not at all)?









How would you rate this exercise in terms of its ability to give you a reasonable feel for the traffic and
roadway conditions you would experience if you were actually driving your vehicle along this roadway
under these traffic conditions?


Excellent E Very Good a Good F Fair L Poor 0 Very Poor





















APPENDIX D
SAMPLE LOOP DETECTOR DATA
























Tag County Site Lane Year Month Day Hour Min Int 15 23 28 33


SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03

SPD 18 9920 1 04 03

SPD 18 9920 2 04 03

SPD 18 9920 3 04 03

SPD 18 9920 4 04 03


0 0 0 0

0 0 0 0

0 0 0 1

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0

0 0 0 0
o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o

o o o


38 43 48 53 58 63 68 73

0 0 0 0 2 1 3 4

0 0 0 0 0 0 0 2

0 0 0 0 0 0 0 2

0 0 0 0 1 2 6 10

0 0 0 0 1 2 3 3

0 0 0 0 0 0 0 3

0 0 0 0 0 0 0 5

0 0 0 1 0 1 5 10

0 0 0 0 0 2 2 7

0 0 0 0 0 0 0 1

0 0 0 0 0 0 1 7

1 0 0 0 2 5 8 7

0 1 0 0 0 5 3 6

0 0 0 0 0 0 4 0

0 0 0 0 0 2 3 3

0 0 0 0 0 4 4 8

0 0 0 0 0 0 5 5

0 0 0 0 0 0 0 2

0 0 0 0 0 0 0 0

0 0 0 0 0 1 7 2

0 0 2 0 0 1 2 9

0 0 0 0 0 0 0 3

0 0 0 0 0 0 1 2

0 0 0 0 0 2 5 8

0 0 0 0 3 3 3 6

0 0 0 0 0 0 0 5

0 0 0 0 0 0 0 3

0 0 0 0 0 1 7 11

0 0 0 0 0 2 6 7

0 0 0 0 0 0 0 1

0 0 0 0 0 0 0 1

0 0 0 0 2 2 3 7


Total Avg. 5 min

78 83 91 Vol. Spdi vol1

3 1 1 15 722

5 1 0 8 774 23

4 1 1 9 73 9

5 1 0 25 71 8 34

7 2 0 18 733

6 1 2 12 793 30

7 2 1 15 77 9

4 1 0 22 71 9 37

3 0 2 16 743

2 2 3 8 835 24

5 1 0 14 75 1

4 1 0 28 685 42

4 0 1 20 702

3 0 1 8 746 28

6 1 2 17 754

12 0 0 28 73 0 45

3 2 1 16 748

3 2 0 7 780 23

2 2 0 4 80 5


0 0

0 0

1 0


4 2 1

3 5 0

1 0 0

2 4 0

8 0 0

2 1 0

2 1 0

4 1 2

6 1 1


19


veh/hr/ln' Density'



138 1 86



204 282



180 238



222 299



144 1 86



252 3 56



168 235



270 3 65



138 1 82



114 155


24 144 201



29 174 232



29 174 241



36 216 291



22 132 1 82



30 180 240


1These categories were calculated from the given loop detector data and added to the speed data spreadsheets.















Total Total Total

Tag County Site Yr. Mo. Day Hour Min Int Lane # Lane # Lane # Lane # NB SB Volume

CNT 18 9920 04 03 08 00 05 005 1 15 2 8 3 9 4 25 23 34 57

CNT 18 9920 04 03 08 00 10 005 1 18 2 12 3 15 4 22 30 37 67

CNT 18 9920 04 03 08 00 15 005 1 16 2 8 3 14 4 28 24 42 66

CNT 18 9920 04 03 08 00 20 005 1 20 2 8 3 17 4 28 28 45 73

CNT 18 9920 04 03 08 00 25 005 1 16 2 7 3 4 4 15 23 19 42

CNT 18 9920 04 03 08 00 30 005 1 18 2 6 3 7 4 22 24 29 53

CNT 18 9920 04 03 08 00 35 005 1 23 2 6 3 9 4 27 29 36 65

CNT 18 9920 04 03 08 00 40 005 1 18 2 4 3 8 4 22 22 30 52

CNT 18 9920 04 03 08 00 45 005 1 13 2 4 3 12 4 28 17 40 57

CNT 18 9920 04 03 08 00 50 005 1 8 2 4 3 14 4 29 12 43 55

CNT 18 9920 04 03 08 00 55 005 1 16 2 4 3 8 4 22 20 30 50

CNT 18 9920 04 03 08 01 00 005 1 12 2 3 3 4 4 24 15 28 43

CNT 18 9920 04 03 08 01 05 005 1 19 2 3 3 8 4 19 22 27 49

CNT 18 9920 04 03 08 01 10 005 1 6 2 2 3 8 4 25 8 33 41



























CL CL CL CL CL CL CL CL CL CL CL CL CL CL CL Total
Tag County Site Lane Year Month Day Hour Min Int
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 Vol.


18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920

18 9920


Trucks
Buses'


%HV Total %
HV1
HV1


1 04

2 04

3 04

4 04

1 04

2 04

3 04

4 04

1 04

2 04

3 04

4 04

1 04

2 04

3 04

4 04

1 04

2 04

3 04

4 04

1 04

2 04

3 04

4 04

1 04

2 04

3 04

4 04


1These categories were calculated from the given loop detector data and added to the class data spreadsheets.







72



CLASSIFICATION SCHEME "F"


DESCRIPTION


NO. OF
AXLES


1 MOTORCYCLES 2
ALL CARS 2
2 CARS W/ 1-AXLE TRLR 3
CARS W/2-AXLE TRLR 4


PICK-UPS &VANS
I & 2 AXLE TRLRS


2,3. & 4


BUSES 2 & 3


2-AXLE, SINGLE UNIT 2


p 3-AXLE, SINGLE UNIT 3


B 4-AXLE, SINGLE UNIT 4


2-AXLE TRACTOR.
S-AXLE TRLR(2S1) 3

2-AXLE TRACTOR. 4
2-AXLE TRLR(2S2)

H 3-AXLE TRACTOR, 4
1 -AXLE TRLR(3S 1)

13-AXLE TRACTOR,
2-AXLE TRLR(3S2) 5


JI 3-AXLE TRUCK.
W/2-AXLE TRLR 5

TRACTOR W/ SINGLE
TRLR 6 & 7


5-AXLE MULTI- 5
TRLR

6-AXLE MULTI-
TRLR 6


7 or more


System Usage Data 1/9/90


CLASS.
GROUP


3

4

5


6


7


9




10



11



12

13


ANY 7 OR MORE AXLE















BIOGRAPHICAL SKETCH

David S. Kirschner is a 23-year old graduate student at the University of Florida.

He is studying towards his Master of Engineering degree, specializing in transportation

engineering. He received a Bachelor of Science in Civil Engineering degree from the

University of Florida in December of 2004.




Full Text

PAGE 1

DEVELOPMENT OF A RURAL FREEWAY LEVEL OF SERVICE MODEL BASED UPON TRAVELER PERCEPTION By DAVID S. KIRSCHNER A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 By David S. Kirschner

PAGE 3

ACKNOWLEDGEMENTS I would like to thank my committee chair, Dr. Scott Washburn, and my committee members Dr. Lily Elefteriadou and Mr. Bill Sampson. iii

PAGE 4

TABLE OF CONTENTS ACKNOWLEDGEMENTS ...............................................................................................iii LIST OF TABLES .............................................................................................................vi LIST OF FIGURES ..........................................................................................................vii ABSTRACT .....................................................................................................................viii CHAPTER 1 INTRODUCTION ..........................................................................................................1 Background .....................................................................................................................1 Problem Statement ..........................................................................................................2 Research Objective and Tasks ........................................................................................4 Chapter Organization ......................................................................................................4 2 LITERATURE REVIEW ...............................................................................................6 HCM Freeway LOS Methodology .................................................................................6 Studies Investigating Traveler Perception of LOS .........................................................8 3 RESEARCH APPROACH ...........................................................................................16 Alternative Survey Methods .........................................................................................16 Video Data Collection ..................................................................................................19 Survey Sessions ............................................................................................................30 4 ANALYSIS AND RESULTS .......................................................................................35 Analysis Method ...........................................................................................................35 Statistical Analysis ........................................................................................................39 5 CONCLUSIONS AND RECOMMENDATIONS .......................................................45 Data Collection and Video Clip Creation .....................................................................45 Statistical Analysis ........................................................................................................46 iv

PAGE 5

Study Limitations and Recommendations for Further Research ..................................47 REFERENCES .................................................................................................................49 APPENDIX A LOCATIONS OF DATA COLLECTION SITES .......................................................51 B VIDEO CLIP SCREENSHOTS ...................................................................................55 C RURAL FREEWAY TRIP QUALITY SURVEY FORM ..........................................63 D SAMPLE LOOP DETECTOR DATA ........................................................................68 BIOGRAPHICAL SKETCH............................................................................................73 v

PAGE 6

LIST OF TABLES 1. HCM Level of Service Thresholds .................................................................................8 2. Data Collection Sites and Traffic Data .........................................................................21 3. Data Collection Times, Locations, and Directions .......................................................24 4. Traffic Data for 13 Video Clips ....................................................................................29 5. Clip Sites, Dates, and Times .........................................................................................30 6. Dates and Locations of Survey Sessions ......................................................................33 7. Density Model Estimation Results ................................................................................40 8. Comparison of Estimated and HCM LOS Thresholds .................................................41 9. Traffic Characteristics Model Estimation Results ........................................................42 10. Level of Service Model Estimation Results ................................................................43 11. Realism of Video Survey Responses ..........................................................................46 vi

PAGE 7

LIST OF FIGURES 1. Camera Setup-Front View, Side View, Speedometer ...................................................23 2. In-Vehicle Equipment Setup .........................................................................................24 3. Setup of a Survey Session .............................................................................................26 4. Sample Video Screenshot .............................................................................................26 5. Illustration of an Ordered Probability Model ................................................................37 6. Illustration of an Ordered Probability Model with an Increase in .............................38 vii

PAGE 8

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Engineering DEVELOPMENT OF A RURAL FREEWAY LEVEL OF SERVICE MODEL BASED UPON TRAVELER PERCEPTION By David S. Kirschner May 2005 Chair: Scott Washburn Major Department: Civil and Coastal Engineering The concept of Level of Service (LOS) is meant to reflect the trip quality a traveler will experience on a roadway or other transportation facility. Despite this, there have been relatively few studies that have tried to measure the association of prescribed level of service assessment methods with traveler perceptions. The objective of this study is to provide insight into how road users perceive their trip quality on rural freeways, and to examine how the existing service measure (density) relates to these travelers perceived trip quality. Study participants were shown a series of video clips of rural freeway travel from a drivers perspective, then filled out survey forms indicating their opinion of the trip quality provided by the conditions in the video clip, and ranked these video clips on a scale from Excellent to Very Poor. In addition, the survey participants were asked to give background information about themselves and their driving habits in case these factors also turned out to be significant in influencing perceived trip quality. These video clips were matched with inductance loop detector data that were collected simultaneously at the data viii

PAGE 9

collection sites, in order to see how well the existing service measure (density) corresponded to the participants rankings. The data from the surveys were analyzed using an ordered probability model to determine which factors influenced the participants decisions and how. Three models were created. The first model used only density as a predictive factor. The second took into account only roadway and traffic characteristics, and the third examined all the significant factors that could be gathered from the survey. The density only model showed that density is indeed a strong indicator of travelers perceptions of trip quality. A set of LOS thresholds was also calculated using the survey participants responses. While the survey thresholds and the HCM thresholds had similar values for facility failure, the intermediate thresholds estimated from the survey participants responses were noticeably lower than the HCM thresholds. This suggests that travelers tolerance of congestion is lower on rural freeways than the HCM indicates. The other models showed the significance of other factors in the perception of trip quality in addition to density, such as socio-economic information and personal driving habits. This study provided some preliminary insight into travelers perception of trip quality, but further study is needed. It is suggested that more research be conducted regarding the effects of different factors on the perception of trip quality, such as a more diverse population sampling. Eventually, the results from this type of video-based study should also be compared to results obtained from a comparable in-field driving experiment. This study indicates the need for a further exploration into the differences between urban and rural freeways, and possibly a different set of thresholds for rural freeways. ix

PAGE 10

CHAPTER 1 INTRODUCTION Background Transportation engineers are responsible for targeting roadway infrastructure improvements where they will have the most beneficial effect. Since the capital available for these improvements is limited, engineers must carefully select the projects they choose to fund so that investments will have the best cost-benefit ratio. In a large part these decisions are guided by the procedures and methodologies found in the Highway Capacity Manual (HCM) [1]. The HCM is considered to be the definitive reference guide for traffic operations and analysis in the United States. The procedures in the HCM are used to estimate the operational performance of a variety of transportation facilities (e.g., signalized intersections, two-lane highways) and the corresponding level of service (LOS). The assignment of a LOS is based on designated performance measures and corresponding threshold values for individual facilities. The HCM is published by the Transportation Research Board (TRB) and its development and maintenance is the responsibility of the Highway Capacity and Quality of Service (HCQS) committee of the TRB. The current edition of the HCM was published in 2000. The concept of LOS is a foundation of the HCM. The LOS of a facility is used in the HCM as a qualitative indicator of the operating conditions being experienced by travelers of that facility, under specific roadway, traffic, and control conditions. The HCM describes LOS as A qualitative measure describing operational conditions within 1

PAGE 11

2 a traffic stream, based on service measures such as speed and travel time, freedom to maneuver, traffic interruptions, comfort, and convenience. LOS is divided into six categories, A through F in the 2000 HCM. LOS A indicates excellent service and LOS F indicates extremely poor service. An analysis yielding LOS A would indicate that the facility is performing extremely well, with low volumes and little congestion. If an analysis shows a facility to be performing at LOS C, it is in the middle range of congestion. If a facility is at LOS E, it is still permitting traffic flow but is experiencing significant delays with conditions approaching capacity. At LOS F, a facility is experiencing oversaturated conditions and the demand has exceeded the capacity of the facility. Problem Statement The performance measures that are used to calculate LOS for a facility are referred to as service measures. The currently designated service measure(s) for each facility is (are) based on the collective experience and judgment of the members of the HCQS committee. The same is true with the selection of the threshold values for the various LOS designations. There is currently no quantitative procedure to define which values are used as LOS thresholds. The LOS determination process, therefore, is based on the perspective of transportation professionals. The selection of service measures by the HCQS committee is, however, guided by two principles: 1) the service measure for each facility should represent speed and travel time, freedom to maneuver, traffic interruptions, and comfort and convenience in a manner most appropriate to characterizing quality of service for the particular facility being analyzed, and 2) the service measure chosen for a facility should be sensitive to traffic flow such that the

PAGE 12

3 service measure accurately describes the degree of congestion experienced by users of the facility [2]. The 1985 HCM described LOS as A qualitative measure that characterizes operational conditions within a traffic stream and their perception by motorists and passengers. The descriptions of individual levels of service characterize these conditions in terms of factors such as speed and travel time, freedom to maneuver, traffic interruptions, and comfort and convenience [3]. This statement indicates that the selection of performance measures and thresholds for the determination of level of service should be consistent with how operating conditions are perceived by the traveling public. Until recently, road users perceptions of quality of service were rarely compared to the LOS assigned to a facility by the HCM, despite the above definition emphasizing the importance of reflecting road users perceived quality of service. There have been suggestions from within the HCQS committee that a new approach needs to be explored when selecting a service measure for a facility. Instead of the measure and corresponding thresholds that transportation professionals (the HCQS committee) believe represent the quality of service as perceived by travelers, the publics opinion should be taken into account so as to determine what measure or measures they associate with quality of service on a transportation facility. Under the current methodology, the HCQS committee believed that the service measures were highly correlated with public perception, but this was not known for sure [4]. Since billions of dollars of transportation investment decisions are made every year based upon the outcome of HCM level of service analyses, it is essential that the transportation

PAGE 13

4 engineers assessments of the impact of these investments be consistent with traveler perception of the investment impacts. Research Objective and Tasks The objective of this study was to develop a model for assessing the LOS of a roadway facility that takes into account the road users perceived quality of service. Specifically, this study was focused on rural freeways. The following tasks were carried out in supporting the above research objective. Determine appropriate rural freeway sites to perform field data collection Collect video of roadway and traffic conditions from these sites Collect traffic data from count stations at these sites Produce video clips to be shown to survey participants Develop a survey instrument Recruit survey participants Conduct survey sessions Perform an analysis of survey responses Develop a level of service model Chapter Organization Chapter 2 contains an overview of the current HCM freeway analysis methodology as well as an overview of relevant literature. Chapter 3 describes the research approach for this study, including the field data collection, survey instrument development, survey response data collection, and the statistical analysis method used to analyze the data. Chapter 4 contains the analysis results. Chapter 5 contains the

PAGE 14

5 conclusions and recommendations. Additionally, several appendices with supporting data and information are included.

PAGE 15

CHAPTER 2 LITERATURE REVIEW The Highway Capacity Manual [1] states that the level of service of a roadway section should accurately reflect the perceptions of travelers, yet the current methodology does not directly take these perceptions into account. There have been some recent studies performed seeking travelers opinions about what factors and qualities are important to them in assessing the quality of their trip. A literature review was conducted to identify these studies and note their findings with regard to the travelers perception of LOS. HCM Freeway LOS Methodology A freeway is a section of divided roadway with controlled access and two or more lanes in one direction. Within this definition there are significant differences between urban and rural freeways. Rural freeways have greater distances between interchanges than urban freeways, higher speed limits than urban freeways, and a higher percentage of social and recreational trips than urban freeways. Urban freeways have a higher percentage of work and shopping trips than rural freeways. Despite these differences, urban and rural freeways both use density as their service measure with the same thresholds for LOS. Traveler expectations and perceptions of quality of service are different for rural and urban freeways. While urban freeways experience the full range of LOS conditions from A to F, rural freeways rarely drop below LOS C. Rural freeway travelers have come 6

PAGE 16

7 to expect these higher levels of service, therefore while urban freeway travelers are concerned with their overall travel time and the reliability of this travel time, rural freeway drivers take travel time for granted. Urban freeway drivers expect their ability to change lanes to be restricted, while a restricted ability to change lanes negatively impacts a rural freeway users perceived quality of service [5]. The original HCM had a basic three-point scale to define level of capacity. In 1963 the Level of Service concept was introduced and replaced the previous scale. In 1965 the six-point LOS scale (from A to F) was introduced. In 1985 this six-point scale was redefined to use traffic density (vehicles per unit length of roadway) as the service measure for defining LOS on freeway sections. This is the method that is still used today. Although the concept of LOS is meant to reflect the operational conditions as perceived by motorists, no freeway LOS methodology in the history of the HCM has been based on driver perception studies. Therefore, there can be no way to make sure that the LOS thresholds freeways (as well as any other type of facility) accurately reflect users perception of the quality of service they receive. Under the existing LOS methodology, rural and urban freeways have the same service measure density, as well as the same thresholds for each rank on the LOS scale. These thresholds for all freeway sections are shown in the table below. Do these thresholds accurately reflect the quality of service perceived by travelers on all freeways, urban and rural? In particular, the studies by Hostovsky [5] and Washburn [4] indicate that rural freeway travelers may judge the quality of their trip based on different qualities and criteria. A potential outcome of this study is a set of LOS thresholds unique to rural freeways. This idea of differing service measures for different

PAGE 17

8 categories of a specific facility type is not a new one. Currently there are two service Table 1. HCM Level of Service Thresholds Level of Service Density (pc/mi/ln) A 0-11 B 11-18 C 18-26 D 26-35 E 35-45 F > 45 measures used for assessing the LOS of a two-lane highway. These classes share a common service measure, but the thresholds are different (one of the classes also uses an additional service measure). In addition, the HCM procedure for analyzing arterial streets uses the same service measure for all arterial streets but includes four sets of thresholds for four different classifications of arterials [1]. Studies Investigating Traveler Perception of LOS A study by Pcheux et al. [6] noted that the Transportation Research Boards Committee on Highway Capacity and Quality of Service recognized a need to improve the HCM methodology of assessing LOS. Specifically, concerns were raised that the LOS of a roadway section did not correspond to road users perceptions. The authors felt that for LOS to accurately reflect travelers perception of quality of service they would first have to find out what performance measures were significant to travelers. The study method involved test participants driving along a pre-selected 40-minute route, encompassing mostly arterial streets, accompanied by an interviewer and a traffic engineer. The participant would discuss what factors they personally found

PAGE 18

9 important to the quality of their trip. Participants identified over 40 factors that were important. These included such factors as intersection efficiencyif the intersection was being utilized by opposing traffic while travelers were waiting, and the aesthetic qualities of the intersection. Both of these topics are not covered by the HCM. The study concluded that more research was needed to focus on traveler perception. A study by Hostovsky et al. [5] used focus group participants to identify factors important to trip quality on rural freeways and then compared those findings with those from a focus group study using regular urban commuters and commercial truck drivers. The participants in the rural freeway focus group identified three factors that were most important to trip qualitylow density, regular (predictable) travel time, and maintaining a steady travel speed. Other topics discussed were the safety issues inherent to the isolated locations of rural freeways, aesthetics, speed differential between cars, the presence of heavy vehicles, and the need for better traveler information. When compared to the results of a focus group study involving urban commuters, it was found that urban commuters placed high importance on the overall speed of their trip, where rural freeway travelers felt that the ability to choose their speed was a positive. This reflects the fact that urban drivers rarely have the opportunity to choose their speed in the traffic stream, so a faster speed is usually preferable over a slower one. Urban commuters were also not as concerned with the ability to change lanes and move about the facility at will. Most of the urban drivers were happy if they could stay in one lane and maintain a desired speed for their trip. The rural drivers were pleased if the density of the freeway section was low enough to allow movement between lanes and passing at will. This study was significant due to the fact that it recognized the

PAGE 19

10 differences in how travelers rate their trip quality on an urban versus a rural freeway. The HCM uses the same methodology for freeways in both types of areas. A study by Nakamura et al. [7] evaluated traffic flow conditions along an expressway in Japan from a drivers viewpoint. The study intended to quantitatively analyze the relationship between traffic flow conditions, drivers perceptions, and drivers behaviors. The field data portion of this study was intended to collect data on drivers behavior and perception under various flow conditions. Drivers had a video camera mounted in their own vehicle and were asked to drive a section along an expressway. After each trip the subject was asked to complete a survey about the traffic flow conditions. Twenty-two subject vehicles were used and 105 surveys were collected. The behavioral data collected was number of lane changes, travel time by lane, and percent time spent following. This study found that the most important factor influencing drivers satisfaction with their trip was the traffic flow rate. Other factors affecting trip quality were found to be number of lane changes, and the percent time spent following. Additionally, choosing the LOS based on the drivers level of satisfaction was attempted and then compared to the conventional LOS methodology. The results of this comparison suggested that the traffic conditions on Japanese expressways are not satisfying drivers. The realistic meaning of this result was that if facilities were designed to the drivers satisfaction level rather than the conventional LOS it would require an enormous investment. Several studies have been identified using road-user surveys and video selections to evaluate LOS methodology. The first study, by Sutaria and Haynes [8], used a road user survey to evaluate the LOS methodology for signalized intersections. Over 300

PAGE 20

11 drivers were shown video clips taken both from a drivers perspective and from an overhead camera at an intersection. The film segments were specifically chosen to represent a specific LOS and were intended to be shown to drivers for one or two signal cycles. The final compilation shown to drivers included both types of view and the clips were put in a random order. Their road user survey consisted of two partsa group attitude survey and a film survey. The group attitude survey used a questionnaire, to be answered before the film portion of the survey. The questionnaire included demographic information such as gender, age, and education, as well as questions about the participants driving experience and the type of roadways the participants usually drove on. They were then asked to give the relative importance of factors including delay, number of stops, traffic congestion, heavy vehicle density, and ability to change lanes as these factors applied to the quality of service at an intersection. After the initial questionnaire the participants were shown the video clips, consisting of a drivers view of a vehicle approaching, waiting, and passing through an intersection. After each of these clips the participants rated the quality of service they felt the intersection provided. At the end of the video portion the participants were again asked to rate the factors important to quality of service at an intersection to see if their initial opinion had changed. In all, 310 drivers participated in this survey. The results from the survey showed delay to be the most important factor both before and after the film portion of the survey. This study provided the first results that took into account the perceptions of travelers and changed the performance measure used by the HCM to evaluate LOS in the 1985 edition. This study also recommended further

PAGE 21

12 similar studies, and for further studies to simultaneously collect video and traffic flow data to allow for accurate measurements of what is depicted on the films. A study conducted by Pcheux et al. [9] addressed the issue of developing a study method to assess the perceived LOS at signalized intersections. The first objective of all the study methods was to determine how well the current LOS methodology reflects the opinion of road users. The second objective was to determine the factors affecting users perceptions at signalized intersections. The participants in this study represented a wide range of ages, education levels, and incomes. The participants were first shown a series of approaches to signalized intersections from a drivers perspective. After being shown a sequence of these clips, the participants were asked to fill out a survey including their attitudes about certain driving situations as well as their socio-economic information. After filling out these surveys the participants were asked to discuss the factors that influenced their perception of quality of service as a group. The study results showed that on average, the participants delay estimates were fairly accurate, however individual delay perceptions varied significantly. Fifteen factors were identified that contributed significantly to quality of service. Finally, the study found that participants tended to perceive service quality on three or four distinct levels as opposed to the six HCM levels of service. Another study using video clips and road user surveys was performed by Choocharukul et al. [10] with the intention of evaluating the current HCM methodology of assessing LOS. This study intended to provide a multivariate statistical analysis of the factors that were important to road users perception of trip quality and to compare those factors to the current performance measures for LOS.

PAGE 22

13 The data for this study were collected at various urban freeways. Cameras mounted on overpasses were used and were focused on sections that included inductance loop detectors. The cameras were focused so that only one direction of travel could be observed. The data from the loop detectors were collected and synchronized with the time of the video clips so the researchers would know the actual traffic flow conditions during the time the video clips were recorded. Two sets of video clips were chosen, each containing twelve clips. Two video clips were used for each HCM LOS designation, AF, with one clip on the high end of an LOS designation and the other clip at the low end. These designations were determined by the loop detector information. There were two groups of survey participants in this survey, one consisting of students, transportation professionals, and environmental management professionals, and the other consisting of commercial truck drivers and clerical and support staff. The participants were provided with written descriptions of the six HCM LOS designations (directly from the HCM). They were then asked to view the twelve video clips and rank each of them with the LOS they thought was appropriate for the conditions. The participants were also surveyed for demographic information such as age and education levels, as well as information about their driving habits. This study used an ordered probit statistical model to assess how users perceive the LOS of the roadway sections. The results of the survey and analysis revealed that perceived levels of service do not closely follow the HCM. Almost all the participants in this study had a lower tolerance for LOS A than the HCM, with the average cut-off for LOS A among the study participants shown to be 7 passenger cars per mile per lane (pc/mi/ln) as opposed to the HCM cutoff of 11 pc/mi/ln. The HCM threshold for LOS F

PAGE 23

14 also does not correspond with the findings of this study, with the participants selecting an average of 82 pc/mi/ln as the upper bound of LOS F as opposed to the HCM LOS F of 45 pc/mi/ln. The study also found that factors other than density are likely to influence road users perception of quality of service. The results from both groups indicated that a freeway with 4 lanes (instead of 6 or 8), an increase in traffic density, and an increase in the standard deviation of vehicle speeds all contributed strongly to a worse perception of LOS. It should be noted that the use of an overhead view of traffic could likely affect survey participants perceptions in a different way than that of an-vehicle view of traffic and roadway conditions. Background Study A study was performed at the University of Florida by Washburn et al. [4] with the objective of discovering what factors are important to drivers when evaluating the quality of their trip on a rural freeway. Several methods were considered for this study (focus groups, video/simulation viewing and review, interviews, etc.) with the final choice being an in-field survey-based approach. Two hundred and thirty-three travelers were surveyed at rest stops and service plazas along rural freeways in Florida. These locations were chosen due to their access to travelers in the process of a rural freeway trip. It was believed that this in-field survey approach would provide more reliable data, than mail-back surveys for example, as the drivers experiences would still be fresh in their minds when filling out the surveys. Drivers were asked to rank the factors that contributed to the quality of their trip on a scale from 1 to 7. The most important factor, ranked in the top three 64.3% of the time, was the ability to consistently maintain the desired travel speed. The factor with the

PAGE 24

15 next highest ranking was the ability to change lanes freely and pass other vehicles. This was ranked in the top three 33.3% of the time. The third most important factor was the ability to maintain a speed no less than the posted speed limit. This factor was ranked in the top three 33.0% of the time. This preliminary study showed that though density is important to rural freeway travelers, it is not the most important factor in determining trip quality. It also showed that drivers consider many other factors when determining trip quality. Conclusions The studies detailed in this chapter have shown that, while some research has been done on travelers perception of quality of service, there is a need for more study. The current HCM methodologies for evaluating LOS may be insufficient for determining the perceived quality of service from the travelers point of view. From the studies summarized in this chapter, we can see that it is possible to understand and approximate a travelers perception of quality of service using the factors that are found to be important to them. This type of research may ultimately assist decision makers when planning for new roadways and roadway improvements.

PAGE 25

CHAPTER 3 RESEARCH APPROACH This chapter will describe the methods used in collecting the sample data for this study as well as the methods used to refine the data for use in public surveys .Detailed within the chapter are the choice of a survey method, the selection of data collection sites, the creation of the survey form, and the process for conducting a road user survey. Alternative Survey Methods Common methods of data collection include the following: focus groups, field surveys, in-vehicle surveys driven by a researcher, in-vehicle surveys driven by the research participant, driving simulators, and video surveys. Focus Groups This consists of recruiting test participants in order to arrange a roundtable-type discussion about rural freeway travel. Participants would discuss their rural freeway trip experiences and relate which aspects of rural freeway travel are most important to them when evaluating the quality of their trip. The advantage of a focus group is the relative ease of the survey, there is no video data collection, field work, or liability on the part of the researchers. The disadvantage is that participants may influence each others responses and one particularly vocal participant could swing the rest of the group towards his or her opinion. Another disadvantage is the lack of a control element for the researchers there is no one experience on which the participants are basing their opinions, so the researchers can not look at the data or video record to interpret the responses. 16

PAGE 26

17 Additionally, the potential lack of quantitative feedback upon which to build an analytical model limits researchers in their ability to predict the responses of other travelers faced with similar roadway and traffic conditions. Field Surveys Researchers distribute survey forms at locations frequented by rural freeway travelers, such as rest stops or service plazas. The participants give their opinions on rural freeway travel in a survey form, rating which factors are most important when they judge their trip quality. One advantage to this method is that participants surveyed have recently driven on a rural freeway and have this experience fresh in their mind. Another advantage is that it is relatively easy to recruit participants for this sort of survey; there is always a ready supply of people in this type of location. The disadvantages are similar to the focus group. In-Vehicle Surveys (driven by research personnel) Participants are recruited and driven along a section of rural freeway, then surveyed about their perception of the trip quality. Advantages to this method include all participants would have the same experience to draw upon for their responses, and there would be no need to attempt to simulate the driving experience as participants would be experiencing the conditions firsthand. The disadvantages to this method include the liability to the researchers should the vehicle be involved in an accident, and the time and effort involved in conducting a survey of this manner. The controllability and repeatability of the conditions are also disadvantages because it is not possible to ensure the same conditions will be experienced by multiple survey participants.

PAGE 27

18 In-Vehicle Surveys (driven by research participants) Participants are recruited to drive along a section of rural freeway and provide the researchers with feedback on their trip once they return. Once again this method is advantageous in that it would provide participants with a firsthand look at the conditions involved. The disadvantages to this are similar to the previous method in that there is significant liability attached to a method like this, and this method would be even more time-consuming than the previous one. This method also suffers from the same lack of control and repeatability as any in-car survey. Driving Simulator Participants are put behind the wheel of a real vehicle, but the driving environment is simulated with the use of computer animation and video display monitors. They would then participate in the virtual driving of a rural freeway segment. This would give participants a closer likeness of actual freeway travel without the liability of having them drive a real section themselves. Disadvantages include cost (simulator time is expensive) and the well-documented motion sickness problem for participants (which increases recruitment time and costs). Video Surveys This method involves participants viewing pre-recorded video scenes from actual rural freeway sites. The clips could be from one of two perspectives: o Overhead View A camera placed over the test section of rural freeway records the traffic flow for survey participants to review at a later time. While this method does not give a simulation of actually driving the

PAGE 28

19 freeway section, it does give the participant a broader overview of the traffic stream. o Drivers Perspective A vehicle is equipped with a video camera to record the rural freeway trip from the drivers perspective. This method would better simulate actual rural freeway travel than an overhead view. After considering all advantages and disadvantages, the method chosen was a video survey from the drivers perspective. This method would allow larger groups of people to be surveyed simultaneously while giving a reasonably accurate depiction of rural freeway travel This method allows for complete control and repeatability of the conditions experienced by the participants, as well as eliminating the liability issues inherent in an in-vehicle survey. Video Data Collection The data collection method was developed after selecting the form the survey would take. It included five specific tasks Site Selection, Equipment and Setup, Video Data Collection, Video Clip Creation, and Loop Data Collection. Site Selection The sites at which the video clips were captured were all within Florida. Reasons for this include the proximity to the University of Florida and the access to the Florida Department of Transportations (FDOT) network of more than 7,500 traffic monitoring stations. The FDOT maintains a network of inductance loop detectors (ILD) along Floridas Intrastate Highway System. There are two types of ILD stations permanent

PAGE 29

20 and portable. The permanent stations were used in this study since they are continuously recording data 24 hours a day, 365 days a year. The portable stations require a data recorder to be hooked up at the location of the ILD station in a roadside cabinet. The permanent stations are telemetered such that data can be downloaded and archived on a daily basis. The data archived from the permanent stations are compiled every year by the FDOT and published on the Florida Traffic Information (FTI) CD. Also included on this CD for each state highway are the number of lanes in each direction, the ILD station type (permanent or portable), a description of the site, the Average Annual Daily Traffic (AADT), the percentage of trucks on the highway, and the peak hour in each direction, Multiple sites were examined around the state. There were several factors leading to the final site selections. South Florida was excluded due to the limited number of rural freeway segments and the long driving times required to get there and back. Sites west of Tallahassee were also not considered due to driving distance. These conditions hinged upon the availability of suitable sites in north-central Florida. The final sites selected represented a mix of four-lane and six-lane freeways, level and rolling terrain, truck percentages, and a wide range of volume. This information was obtained from the Florida Traffic Information CD published by the FDOT [11]. All sites selected were permanent count stations instead of portable stations. Additional data were used from a similar study conducted months before at the University of Florida. A list of the data collection sites and their associated traffic data follows in Table 2. Maps of the locations of the data collection sites can be found in Appendix A.

PAGE 30

Table 2. Data Collection Sites and Traffic Data Volume Site Site Type Description Direction 1 Direction 2 Two-way AADT K Factor D Factor Truck % 189920 Telemetered SR-93/I-75, 3.5 mi south of Turnpike, Sumter Co. 20472 N 21250 S 41722 10.1 59.84 21.66 360317 Telemetered I-75, 0.35 mi north of Williams Rd overpass, Marion Co. 37630 N 37844 S 75474 11.14 55.41 21.76 140190 Telemetered SR-93/I-75, 0.6 mi. south of SR 54, Pasco County 37443 N 37203 S 74646 8.76 53.67 11.71 730292 Telemetered SR-9/I-95, 1.4 mi south of Palm Coast Pkwy, Flagler Co. 29276 N 29980 S 59256 9.91 54.92 17.82 269904 Telemetered SR-93/I-75, 3 mi. north of Marion Co. line, Alachua Co. 31304 N 31023 S 62327 11.96 55.97 19.09 970428 Telemetered SR-91/Fl. Turnpike, 797 ft. south of CR561, Lake Co. 17655 N 18088 S 35743 11.09 55.42 12.34 21

PAGE 31

22 Equipment and Setup The objective of video data collection was to depict travel along a section of rural freeway from a drivers perspective. In order to give the survey participants a more complete representation of the conditions on the freeway section, it was decided that two more video images would be captured the vehicles speedometer and the driver side rear-view mirror. This brought the total number of cameras needed to three. The images would be combined during the creation of the video clips. The vehicle used for the video data collection was a minivan. As mentioned above, three cameras were placed in the vehicle in order to capture different aspects of the rural freeway trip. The first camera captured the view through the front windshield, including a view of the interior rear-view mirror. It was placed on a mount secured to the right side of the drivers seat. The second camera captured a view of the speedometer. It was mounted on a suction mount affixed to the steering column. It was found during a preliminary test that the instrument cluster needed to be shaded to reduce glare, so the image would not appear washed-out. The final camera captured a view from the vehicles driver-side rear-view mirror. This camera was mounted on a pole secured between the drivers seat and the door. Figure 1 shows these cameras as they were mounted in the vehicle. The video images were captured by three portable VCRs placed inside the vehicle. A microphone was also connected to one of the VCRs allowing the researcher to announce when they crossed a detector loop and any other potentially important information. This would allow the researcher to match the captured video clip to the loop data collected in a later step. All these devices were powered by three 12-volt deep-cycle batteries. A schematic of the equipment and connections is found in Figure 2.

PAGE 32

23 Figure 1. Camera Setup-Front View, Side View, Speedometer Video Data Collection The method used to capture the video clips remained constant throughout the collection of data. The researcher would activate the three VCRs and start recording. Then the researcher merged onto the freeway. The cameras captured conditions between the exit ramps that came before and after the ILD station. The researcher would speak into the microphone when the detector loop was crossed, giving the exact time and site number so the clip could be matched with the relevant loop data. Up to four runs were made at each location giving a number of clips to choose from when creating the clips for the survey. The data collection for this project was performed during November of 2003 and March 6-9, 2004. A summary of each video data collection session is shown in Table 3.

PAGE 33

24 Figure 2. In-Vehicle Equipment Setup Table 3. Data Collection Times, Locations, and Directions Date Site Freeway Direction Time 11/4/2003 730292 I-95 NB 12:55 11/5/2003 269904 I-75 SB 11:01 11/21/2003 140190 I-75 NB 6:48 11/21/2003 140190 I-75 SB 7:04 11/21/2003 140190 I-75 SB 7:14 11/21/2003 140190 I-75 SB 2:31 3/7/2004 730292 I-95 SB 2:35 3/7/2004 730292 I-95 NB 2:47 3/8/2004 189920 I-75 SB 12:36 3/8/2004 360317 I-75 SB 11:50 3/8/2004 360317 I-75 NB 12:04 3/8/2004 360317 I-75 SB 2:07 3/8/2004 970428 Turnpike SB 1:49 After reviewing the video data gathered on the first day of data collection, it was deemed unusable. The mounting for the camera allowed too much vibration in the picture and the video would not work for a public survey. Although the second round of video

PAGE 34

25 data collection was scheduled for March 6-8, the runs made on the first day needed to be redone due to problems with the camera placement, necessitating a fourth day of data collection on March 9, 2004. Video Clip Creation The survey participants were shown a single video display that contained the video scenes of the front windshield and interior rear-view mirror, the drivers side rear-view mirror, and the speedometer. The display used was a video projector and a wall-mounted screen, located between 5 and 20 feet away from the participants depending on the specific survey location. The setup of one of the survey sessions is depicted below in Figure 3. The majority of the screen was taken up by the view through the front windshield. Since the front windshield view captured a portion of the dashboard as well as the view from the front of the vehicle, the other two images could be overlaid on this area. A screenshot from one of the video clips used in the survey is shown in Figure 4. Screenshots from all 13 clips can be found in Appendix B. The clips were assembled using a video-editing program (Adobe Premiere) [12]. They first had to be captured from the VHS tapes using an ADS digital encoder [13]. After they were stored on the computer hard drive they were combined using Adobe Premiere into clips from 1.5 to 2.5 minutes in length. The length of an individual clip was chosen based on events in the video that the researchers wanted to include or exclude, as well as with a survey participants attention span in mind. The clips shown to viewers were chosen based on conditions they represented that were unique or different from other clips. This selection process is explained further in the section entitled Video Clip Selection.

PAGE 35

26 Figure 3. Setup of a Survey Session Figure 4. Sample Video Screenshot Inductance Loop Detector Data Collection It was desired to calculate the LOS of these sites according the HCM methodology in order to assess how strongly correlated it was with the responses

PAGE 36

27 provided by the survey participants. In order to determine the HCM LOS of the rural freeway segments, data were collected from the inductance loop detector (ILD) stations at each test site. The data collected came in three filesspeed, volume, and vehicle classification. FDOT personnel programmed the detectors at the sites selected for the study to record data in five-minute intervals (the hardware minimum interval) rather than the usual one-hour interval. This shorter interval allowed for traffic data that more accurately reflected the conditions depicted in the video clips. It should be noted that even with a five-minute data collection interval the conditions shown in the video clips could potentially vary from the average provided by the ILD data. These ILD data were used to categorize the collected video data and provided a starting point for selecting a range of conditions to be represented in the survey. The ILD data were provided in the form seen in Appendix D. When available, there were three data files for each site speed, count, and class. In the speed file, counts are provided for each speed range. The midpoints of the speed ranges are shown at the top of the table. In the class file, descriptions such as CL01 are given to the columns. These refer to the specific class of vehicle counted in that group and are explained by the figure provided. From the data provided it was possible to calculate descriptive statistics for the traffic flow at each site, such as the percentage of heavy vehicles in the traffic stream, the total 5-minute volume, the average speed, and the density. Video Clip Selection There were thirteen video clips chosen for the final survey. The final number of clips chosen was a result of five pilot test sessions, striking a balance between coverage of alternatives and attention/focus span of participants. These preliminary tests had

PAGE 37

28 shown that many participants lost interest after two minutes and had already started writing their opinions down. The final video clips were chosen to represent a variety of conditions in categories including lane configuration, traffic density, terrain, truck percentage, the presence of a median or guardrail, and shoulder configuration. The relevant data for each video clip included in the final survey is included below in Table 4 and Table 5.

PAGE 38

Table 4. Traffic Data for 13 Video Clips 29 Clip # Road Dir Lanes Clip Length Volume 1 Truck % 1 Density ILD Truck % Inside Speed Middle Speed Outside Speed LD Avg Speed Inner Lane Middle Lane Outer Lane ILD 5min Volume Terrain Speed 1 I-75 S 2 2:10 low none 8.00 0.13 77.1 72.2 74.30 42 57 99 flat 75 2 I-75 S 3 1:52 med-high high 63 75 66 204 flat 70-75 3 I-75 N 2 2:00 med-high med 13.79 0.20 76.4 69.4 72.20 67 99 166 flat 60-70 4 I-95 N 2 1:35 very high 26.45 56.5 55.4 56.00 102 145 247 flat 40-55 5 I-75 S 3 1:40 low-med med 6.30 0.17 77.6 74.0 66.4 72.30 26 51 37 114 rolling 70-75 6 I-95 S 2 1:59 med 10.21 76.9 74.7 75.80 63 66 129 flat 70 7 I-75 S 2 2:00 med-high high 26.31 0.34 71.6 65.5 68.90 167 135 302 flat 67-72 8 I-75 N 3 2:01 med-high low 61 98 80 239 flat 67-72 9 I-75 S 2 2:00 high high 26.11 0.32 71.3 66.1 69.20 179 122 301 flat 55-65 10 I-95 N 2 1:43 med 10.86 78.4 69.4 72.90 80 52 132 flat 75 11 I-75 S 3 1:26 med med 48 82 48 178 flat 70-75 12 I-75 S 2 1:27 med-high med 17.04 0.15 73.6 68.2 71.10 110 92 202 flat 60-65 13 Turnpike S 2 2:03 med high rolling 75-80 1 These levels (low, med, high) indicate subjective judgments that were used to choose between clips.

PAGE 39

30 Table 5. Clip Sites, Dates, and Times Clip # Clip Site Time Date Closest City 1 189920 run 1 189920 12:36 3/8/2004 Wildwood 2 360317 run 1 360317 11:50 3/8/2004 Ocala 3 Tampa 0648 140190 6:48 11/21/2003 Tampa 4 730292 run 4 730292 14:47 3/7/2004 Daytona Beach 5 Micanopy 1101 269904 11:01 11/5/2003 Micanopy 6 730292 run 3 730292 14:35 3/7/2004 Daytona Beach 7 Tampa 0714 140190 7:14 11/21/2003 Tampa 8 360317 run 2 360317 12:04 3/8/2004 Ocala 9 Tampa 0704 140190 7:04 11/21/2003 Tampa 10 Daytona 1255 730292 12:55 11/4/2003 Daytona Beach 11 360317 run 3 360317 12:07 3/8/2004 Ocala 12 Tampa 1431 140190 14:31 11/21/2003 Tampa 13 970428 run 1 970428 13:49 3/8/2004 Winter Garden Survey Sessions Development of Survey Form and Participant Instructions The survey form for this study had to serve two purposesrecord the participants opinions about the rural freeway video clips and their reasons for these opinions, and record characteristics about the participants that might influence their ratings. Thus the form is divided into two sections. The first section of the survey form is for personal information about the traveler taking the survey. Examples of this information include education level, income, and number of years possessing a drivers license. This section also records information about the participants rural freeway travel habits. It asks for information such as the amount of rural freeway trips taken per month and the average length of the participants rural freeway trips. Finally it asks for some driving habits, such as any changes in the

PAGE 40

31 participants driving style when driving alone versus with a passenger. It also asks the participant to rate their usual driving style, from Conservative to Aggressive. The second section of the survey is for recording the participants opinions and rankings of the video clips. It is divided into two sections for each of the thirteen clips. The first section asks the participant to rank the quality of the trip depicted in the video clip on a scale from Very Poor to Excellent with 6 total ranking levels. A total of six ranking levels was chosen so that there would be general correspondence with the six levels of the HCM (A-F). Participants were asked to use the word ranking rather than a numerical ranking (e.g., 1-6) to minimize the possibility that those familiar with the HCM might try to equate the numerical rankings with the HCM LOS rankings. The second section asks the participant to record why they ranked the video clip as they did, listing all factors that significantly contributed to their ranking. The participants were to then number these according to their relative significance to each other. Finally the form includes questions about the survey itself. These include the participants opinion on the video clips as a representation of rural freeway travel and if the participant would have changed their rankings based on the purpose of the trip (e.g., business, recreational, or social). A one page written survey instruction sheet was developed because there was a significant amount of information that needed to be communicated to the participants in order for them to complete the survey form in a manner which would be useful as study data. The participants could refer back to it if there were any questions about the survey process. The instructions given to each survey participant are provided in Appendix C.

PAGE 41

32 Conducting the Survey Sessions Survey participants were recruited from various sources. They include the following: Undergraduate students in the University of Florida civil engineering program, recruited from the introductory transportation engineering course, Graduate students in the University of Florida civil engineering program, recruited from the transportation degree program, Employees of the University of Florida Technology Transfer Center, Employees of the Florida Department of Transportation, and Alachua county residents (Random participants recruited for a fee by the Florida Survey Research Center) The undergraduate students were recruited from the Principles of Highway Engineering and Traffic Analysis course during the Fall 2004 semester. The graduate students were those enrolled in a transportation engineering degree program during the Fall 2004 semester. The University of Florida Technology Transfer Center is an organization that provides training and technical assistance to Floridas transportation and public works professionals. Their survey session was conducted at their off-site headquarters in Gainesville, FL, with participants ranging from high-school educated support staff to professionals with graduate degrees. The FDOT survey session was conducted at the central office in Tallahassee, FL. This session also included participants of varying backgrounds and demographics. The public sample was comprised of Alachua county residents, recruited by the University of Florida Survey Research Center. The survey center was instructed to recruit individuals with varying socio-demographic

PAGE 42

33 characteristics and also make sure that the participants had experience driving on rural freeways. Additionally, they did not recruit college students as there was already a sufficient number in this group. In total there were 126 surveys filled out for this study. The locations, dates, and groups of participants taking the survey during each session are given in Table 6. Table 6. Dates and Locations of Survey Sessions Survey Session Date City Location Participants # of Surveys 1 8/4/04 Gainesville UF Technology Transfer Center T 2 employees 16 2 11/16/04 Tallahassee Florida DOT Central Office DOT employees 11 3 12/2/04 Gainesville University of Florida undergraduate students 14 4 12/2/04 Gainesville University of Florida undergraduate students 9 5 12/4/04 Gainesville UF Hilton Conference Center public 1 13 6 12/4/04 Gainesville UF Hilton Conference Center public 1 15 7 12/4/04 Gainesville UF Hilton Conference Center public 1 11 8 12/9/04 Gainesville University of Florida undergraduate students 20 9 1/22/05 Gainesville University of Florida public 1 9 10 1/27/05 Gainesville University of Florida graduate students 8 Total Number of Surveys 126 1 Participants were recruited through the University of Florida Survey Research Center Because of the video format of the survey, multiple surveys could be filled out at a time, the main limitations being the ability of the participants to comfortably view the video screen and the length of time for which the participants could be expected to focus on this task. The screen was placed as close as possible to eye level so participants looking at the screen saw it as they would a cars windshield. Before viewing the clips the participants were given the instruction sheet and time to read it. These written instructions were also verbally reviewed by the session moderator, as well as some supplemental information. The participants were also told that they could ask

PAGE 43

34 interpretation questions in-between the viewing of the video clips. It was decided to create two example clips, each 20 seconds long, to show the upper and lower ends of the range of possible traffic flows. The first was a nearly empty four-lane freeway and the second was stop-and-go traffic along a four-lane freeway. The participants were then shown each of the 13 video clips and instructed to watch each clip entirely before writing their responses. Since it was not intended for the order of the clips to have any effect on the participants rankings, the order was shifted for each survey session. After each clip was finished, the participants were given time to record their rankings.

PAGE 44

CHAPTER 4 ANALYSIS AND RESULTS This chapter contains information about the methodology used to analyze the survey data, as well as the results of these analyses. Analysis Method To determine how or if the participants responses correspond to the six LOS rankings, a statistical analysis was needed to predict the probability of selecting discrete rankings (1-6 as included in the survey). While one of several multinomial discrete-choice modeling methods would suffice to predict a discrete outcome, most do not take into account the ordered nature of the responses in this survey (1 is better than 2, which is better than 3, etc.). Using a standard multinomial discrete model, such as a multinomial logit model, would still yield consistent parameter estimates, but with a loss of efficiency [14]. In order to account for the discrete and ordered responses in this survey, an ordered probability model was chosen as the statistical analysis approach. An ordered probability model is derived by defining an unobserved variable, z, that is the basis for modeling the ordinal ranking of data (in this case the six clip rankings) [15]. This variable is specified as a linear function for each observation n such that z n = X n + n (1) 35

PAGE 45

36 where X n is a vector of variables determining the discrete ordering for observation n, is a vector of estimable parameters, and n is a random disturbance. In this analysis, y is defined as each participants evaluation of each of the 13 video clips. Since there are 126 participants and 13 clips, there are a total of 1638 observations. Using this equation, the observed clip ranking, y n for each observation is written as y n = 1 if z n 1 y n = 2 if 1 < z n 2 y n = 3 if 2 < z n 3 (2) y n = 4 if 3 < z n 4 y n = 5 if 4 < z n 5 y n = 6 if z n 5 where the values are the thresholds that define y n The values are estimated jointly with the model parameters (). The estimation problem then becomes one of determining the probability that a participant will select a particular ranking for each clip. In using the ordered probit model, it is assumed that the error term, n is normally distributed with a mean of 0 and a variance of 1. The resulting ordered probit model has the following probabilities corresponding to each clip ranking: P(y n = 1) = (-X n ) P(y n = 2) = ( 1 X n ) (-X n ) P(y n = 3) = ( 2 X n ) ( 1 X n ) (3)

PAGE 46

37 P(y n = 4) = ( 3 X n ) ( 2 X n ) P(y n = 5) = ( 4 X n ) ( 3 X n ) P(y n = 6) = 1 ( 4 X n ) It can be shown that threshold 1 can be set equal to 0 without loss of generality [15]. In the above equations, (.) represents the cumulative normal distribution: uwdweu22121)( (4) This model can be estimated using maximum likelihood procedures. The thresholds 1 and 1 define the upper and lower thresholds for outcome i. This is illustrated in Figure 5 Figure 5. Illustration of an Ordered Probability Model A positive increase in the term implies that an increase in x will increase the probability that the highest category response will be returned (in this case, y = 6). An increase in the

PAGE 47

38 term also implies that the probability of returning the lowest response (y = 1) is decreased. This is illustrated in Figure 6. Figure 6. Illustration of an Ordered Probability Model with an Increase in A unique issue was present in this data set that complicated the analysis procedure. Each of the 126 participants viewed 13 clips and thus generated 13 observations. The issue is that there are unobserved characteristics that are unique to each participant that will be reflected in all 13 of their rankings. If this is not accounted for in the model, the model will be estimated as though each of the 1638 observations came from a unique participant. This approach would result in lower standard errors in the models estimated parameters, leading to inflated t-statistics and exaggerated degrees of significance. The solution to this problem is found in a standard random effects approach. The first equation is rewritten as

PAGE 48

39 z ic = X ic + ic + i (5) where i denotes each participant (i = 1,,126), the c denotes each video clip (c = 1,,13), i is the individual random effect term and all other terms are as previously defined. The random effect term i is assumed to be normally distributed with mean 0 and variance 2 When this random effects model is estimated, an estimate of is also calculated, the significance of which determines the significance of the random effects model relative to the standard ordered probit model [16]. Statistical Analysis The results of the surveys were put into spreadsheet form, with unique cases for each clip viewing. Each participants rankings were kept together within the spreadsheet for analysis purposes. The data were analyzed using LIMDEP [17] with a random effects approach as detailed in the previous section. The first analysis was performed to explore how the quality of service perceptions of the participants in this survey correlated with the HCM LOS thresholds. The density for each of the video clips was calculated from the loop detector data (and the video data, in cases where the loop detector data was incomplete). A statistical analysis was performed using density as the only independent variable to find out where the thresholds of the survey participants fell relative to the six clip rankings. The results are given below in Table 7. The very high level of significance indicated by the t-statistic (coefficient divided by standard error) calculated for density in the above model offers some evidence that

PAGE 49

40 this performance measure correlates well with perceived LOS. The reference t-statistic for these analyses is 1.282, representing a 90% confidence level in a one-tailed t-test. The positive coefficient calculated for density indicates that, as density increases, the likelihood of a traveler perceiving a worse LOS increases. The random effects term, is also highly significant, meaning that the choice of a random effects model for this data set was correct. Had this term not been significant, a normal ordered probability model would have been sufficient. One test for the goodness-of-fit of a model is calculating that models 2 value. The 2 value of a model is between 0 and 1. A 2 value of 1.0 indicates a perfect model fit. The 2 value of a model is calculated as follows: )0()(12LLKLL (6) where K represents the number of variables in the model, LL() represents the log likelihood at convergence, and LL(0) represents the initial log likelihood [15]. Table 7. Density Model Estimation Results Variable Coefficient Standard Error t-statistic Constant -0.138 0.076 -1.82 Traffic Characteristics Density (pc/mi/ln) 0.096 0.003 34.37 Threshold Values 1 0.918 0.038 23.89 2 1.922 0.048 39.92 3 2.863 0.053 53.88 4 4.112 0.066 62.47 Standard Deviation of Random Effects 0.455 0.050 9.12 Initial Log Likelihood -2710.16 Log Likelihood at Convergence -2314.60 2 0.15

PAGE 50

41 Using the participants responses it was possible to calculate a set of thresholds for the participants assigned LOS rankings. Using the calculated values in Table 7, the threshold values can be calculated as ( k 0 )/ 1 In this equation, k designates the five threshold values, 1 = 0, and the other threshold values are given in A comparison between the calculated threshold values from this survey and the HCM LOS thresholds is given in Table 8. Table 7 Table 8. Comparison of Estimated and HCM LOS Thresholds LOS Estimated Thresholds (pc/mi/ln) HCM thresholds (pc/mi/ln) A 0-2 0-11 B >2-11 >11-18 C >11-21 >18-26 D >21-31 >26-35 E >31-44 >35-45 F >44 >45 These thresholds are generally lower than the HCM thresholds for corresponding rankings, indicating the participants in this survey had a lower tolerance for high-density traffic conditions than could be inferred from the HCM LOS thresholds. The second analysis that was performed was intended to take into account all the traffic and roadway characteristics influencing the participants perception of trip quality. The results of this table are given below in Table 9. The traffic characteristics examined produced effects according to expectations. The calculated difference in speed between the inner lane and the outer lane was in the model as speed differential. As this value increased, participants were more likely to assign a worse LOS to a given set of conditions. A higher average speed resulted in a more favorable LOS ranking. Motorists in this survey found three lanes in one direction

PAGE 51

42 to be a preferred configuration over two lanes and were more likely to assign a favorable LOS ranking to those roadways with three lanes in one direction. Table 9. Traffic Characteristics Model Estimation Results Variable Coefficient Standard Error t-statistic Constant 6.296 0.597 10.55 Traffic Characteristics Speed Differential (mi/h) 0.163 0.027 6.08 Average Speed (mi/h) -0.096 0.009 -10.97 3 Lanes (1 Yes, 0 No) -1.848 0.210 -8.82 Truck % 0.005 0.004 1.04 Density (pc/mi/ln) 0.061 0.006 10.59 Threshold Values 1 0.949 0.064 14.88 2 2.192 0.077 28.48 3 3.258 0.092 35.60 4 4.630 0.106 43.80 Standard Deviation of Random Effects 0.522 0.060 8.76 Initial Log Likelihood -2710.16 Log Likelihood at Convergence -1472.53 2 0.45 An increase in the truck percentage resulted in a higher possibility of a worse LOS ranking. While the t-statistic for the truck percentage was below 1.282, it was decided to leave this variable in the model because it was felt that this was a very important variable from a policy standpoint. As expected, the participants preferred not to have a high percentage of trucks in the traffic stream. Finally, density was very significant in this model as it was in the first. A higher density led to an increased possibility of a worse LOS ranking. The random-effects term was again significant in this analysis, justifying the use of a random-effects model.

PAGE 52

43 The third analysis that was performed was aimed at discovering which factors are important to travelers when judging their trip quality. This model was estimated including demographic data as well as roadway and traffic flow characteristics. The values given in Table 10 should be interpreted such that a positive parameter estimate means that an increase in that variable will lead to a better perceived quality of service, and a negative parameter estimate means that an increase in that variable will lead to a worse perceived quality of service. Table 10. Level of Service Model Estimation Results Variable Coefficient Standard Error t-statistic Constant 6.156 0.622 9.90 Demographic and Background Information Age > 35 (1 Yes, 0 No) -0.358 0.121 -2.96 Income (thousands of $) -0.003 0.002 -1.89 Average Number of Rural Freeway Trips per Month 0.025 0.017 1.49 Average One-Way Trip Distance > 100 miles? (1 Yes, 0 No) 0.395 0.127 3.11 Less Aggressive Driver with Passengers? (1 Yes, 0 No) 0.267 0.186 1.43 Traffic Characteristics Speed Differential (mi/h) 0.162 0.028 5.85 Average Speed (mi/h) -0.095 0.009 -10.58 3 Lanes (1 Yes, 0 No) -1.836 0.217 -8.47 Truck % 0.005 0.005 1.03 Density (pc/mi/ln) 0.062 0.006 10.56 Threshold Values 1 0.939 0.065 14.56 2 2.181 0.078 27.93 3 3.247 0.093 34.90 4 4.613 0.107 43.21 Standard Deviation of Random Effects 0.435 0.059 7.42 Initial Log Likelihood -2710.16 Log Likelihood at Convergence -1447.34 2 0.46

PAGE 53

44 In Table 10, a positive coefficient value indicates that as the variable increases, there is an increased likelihood of a worse perception of LOS. Likewise, a negative coefficient value indicates that as the variable increases, there is an increased likelihood of a better perception of LOS. The results indicate that, while density is important to travelers, it is not the only factor influencing perceived quality of service. The survey results showed significant effects of demographic and background information on drivers LOS rankings. Table 10 indicates that participants with over 35 are more likely to assign a given set of conditions a better LOS, as are those with higher incomes. Travelers who drive on rural freeways more frequently are more likely to perceive a worse LOS, as are those whose average rural freeway trip is over 100 miles in one-way length. Those participants who indicated that they tend to drive less aggressively with passengers in the car as opposed to driving alone were more likely to assign a worse LOS to a given set of conditions. A possible explanation is that these drivers are more aggressive than the average motorist. Participants were asked if they considered themselves to be an aggressive driver, and the results of that model did not display significance. Perhaps motorists were more reluctant to admit they drive aggressively, but this tendency manifests itself in their responses to this question. The results estimated using the traffic and roadway characteristic variables showed similar significance and magnitude to the model estimated only using these variables. The random effects term was once again significant.

PAGE 54

CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS Since 1963, the Level of Service concept has been integral to the Highway Capacity Manual methodology for assessing the performance of transportation facilities. There is, however, still relatively little known about how the HCM methodologies for assigning LOS correspond to road users perceptions of their quality of service. The purpose of this study was to investigate what factors influenced road users perceptions of quality of service, and how that perception compares to HCM calculated LOS. Data Collection and Video Clip Creation The data collection process used for this study proved successful in gathering the necessary video data. After deciding on the best camera positions and mounting techniques, all cameras recorded clear, steady views of their intended targets. The equipment in the vehicle performed exactly as intended, capturing the necessary information while keeping all three VCR timers consistent so the video data could be synchronized at a later time. The sites chosen generally provided a good variety of traffic conditions, but some clips from a pilot study were also used to provide additional roadway and traffic conditions that were not captured in the data collection effort for this project. These clips were re-edited using the same process as the clips filmed for this study so there would be consistency in the screen views. 45

PAGE 55

46 The loop detector data did not work out as well for some of the sites as was initially hoped. Due to malfunctioning detectors or construction at the selected sites, some of the desired data were not available. The final form of the video clips and the presentation to survey participants worked very well, exactly as intended. The last question on the survey form (as seen in Appendix A) asked participants to rate how well the video clips simulated the driving experience for the conditions depicted on the screen. The majority of participants found the survey to be a very good representation of the actual driving experience, with 95% of the participants rating the survey as a good or better representation of the actual driving experience. The responses to this question are tabulated in As shown in this table, the average response from participants was approximately a 2 out of 6, corresponding to very good. Table 11Table 11. Realism of Video Survey Responses Ranking Excellent Very Good Good Fair Poor Very Poor 1 2 3 4 5 6 Frequency 21 64 36 5 1 0 Percent of Total Responses (%) 17 50 28 4 1 0 Average Rank 2.2 Statistical Analysis The analysis process chosen for this survey was an ordered probability model, specifically the ordered probit model. The structure of the standard ordered probit model formulation does not account for each participant providing 13 responses, so a random-effects formulation was used. This modeling choice was justified, with the standard deviation of random effects showing significance in all statistical analyses.

PAGE 56

47 The first model developed was one incorporating only density as an independent variable. This produced results that were as expected, that density is very significant to travelers when they are judging the quality of service provided by a rural freeway. A complimentary outcome of this analysis was that density thresholds for each LOS were estimated according to the survey participants responses. For LOS A-E, the survey participants showed a lower tolerance for high-density traffic conditions, hence their estimated thresholds were lower. The HCM thresholds and the estimated thresholds showed similar values for LOS F. The second model was estimated to include the influence of other roadway and traffic characteristics. The results of this model showed that while density is significant to user perception of LOS, there are other significant factors influencing this perception, such as average speed of the traffic stream and the speed differential between lanes. The final model included all factors from the survey that were found to be significant, including demographic factors as well as roadway and traffic characteristics. The results of this model indicated that the background and characteristics of the individual road user can influence their perception of LOS. While this result was expected, it is still significant due to the implications for a potential future modification to the HCM LOS methodology. Study Limitations and Recommendations for Further Research Since the scope of this study was limited to North Central Florida, additional testing with participants from a variety of other geographic regions would be needed to adopt any findings on a national level. An expanded sample, both geographically and in roadway conditions, would provide much more comprehensive coverage of the roadway

PAGE 57

48 and traffic condition combinations. The video survey format has inherent limitations as well. In a future study, it would be desirable to allow road users to drive in a traffic stream with known characteristics (density, truck percentage, etc.), then express their opinion regarding the LOS of the roadway section. This was not considered for this study due to cost and liability. The results of this survey could be compared to the results of the video survey to assess the accuracy of the video survey. If the video survey is shown to be an accurate method of simulating traffic conditions, it can be used in future studies and will be more effective than in-field surveys. Finally, although participants were told to imagine the conditions in the video scenes as if they were occurring throughout the duration of a trip, it is not known whether actually experiencing these conditions for an equivalent time to an entire trip would change the outcome. It is hoped that the findings of this study will lead to further developments in this area. The study does show that density is significant in determining a road users perception of trip quality. It is also known that there are significant factors influencing LOS other than density and these should be explored more completely. Ultimately, a better understanding of travelers perceptions of quality of service will lead to a better use of the available resources to improve the roadway network where it is really needed, and to more accurate planning and accommodating for future demands.

PAGE 58

REFERENCES 1. Transportation Research Board (2000). Highway Capacity Manual. TRB, National Research Council. Washington, D.C 2. Harwood, D., Flannery, A., McLeod, D., Vandehey, M. (July 2001). The Case for Retaining the Level of Service Concept in the Highway Capacity Manual. Presented at the 2001 Transportation Research Board Committee A3A10 Highway Capacity and Quality of Service Midyear Meeting, Truckee, California. 3. Transportation Research Board (1985). Special Report 209: Highway Capacity Manual. TRB, National Research Council. Washington, D.C 4. Washburn, S., Ramlackhan, K., McLeod, D. (2004). Quality of Service Perceptions by Rural Freeway Travelers: Exploratory Analysis. Transportation Research Record: Journal of the Transportation Research Board, No. 1883. Washington, D.C., pp. 132-139. 5. Hostovsky, C., Wakefield, S, Hall, F. (2004). Freeway users Perception of Quality of Service: A Comparison of Three Groups. In Transportation Research Record: Journal of the Transportation Research Board, No.1883. TRB, National Research Council. Washington, D.C., pp. 150-157. 6. Pcheux, K., Flannery, A., Wochinger, K., Rephlo, J., Lappin, J. (2004). Automobile Drivers Perceptions of Service Quality on Urban Streets. Transportation Research Record: Journal of the Transportation Research Board, No. 1883. TRB, National Research Council. Washington D.C. pp. 167-175. 7. Nakamura, H., Suzuki, K., Ryu, S. (2000). Analysis of the Interrelationship Among Traffic Flow Conditions, Driving Behavior, and Degree of Drivers Satisfaction on Rural Motorways. Transportation Research Circular E-C018: Proceedings of the Fourth International Symposium on Highway Capacity. National Research Council. Washington, D.C., pp. 42-52 8. Sutaria, T.C., and Haynes, J.J. (1977). Level of Service at Signalized Intersections. Transportation Research Record: Journal of the Transportation Research Board, No. 644. TRB, National Research Council. Washington, D.C., pp. 107-113. 49

PAGE 59

50 9. Pcheux, K., Pietrucha, M., Jovanis, P. (2000). User Perception of Level of Service at Signalized Intersections: Methodological Issues. Transportation Research Circular E-C018: Proceedings of the Fourth International Symposium on Highway Capacity, National Research Council. Washington, D.C., pp. 322-335. 10. Choocharukul, K., Sinha, K., Mannering, F. (2004). User Perceptions and Engineering Definitions of Highway Level of Service: an Exploratory Statistical Comparison. Transportation Research Part A, 38. pp. 677. 11. Florida Traffic Information 2003. (2003). Florida Department of Transportation, Tallahassee, FL, CD-ROM. 12. Users Guide for Adobe Premiere Pro Software. (n.d.). Last Accessed November 17, 2003, from http://www.adobe.com/products/premiere 13. Users Guide for ADS Pyro A/V Link. Last Accessed March 15, 2005, from http://www.adstech.com/products/API-555/intro/api555_intro.asp?pid=API-555 14. Amemiya, T. (1985). Advanced Econometrics. Harvard University Press. Cambridge, MA. 15 Washington, S., Karlaftis, M., Mannering, F., 2003. Statistical and Econometric Methods for Transportation Data Analysis. Chapman & Hall/CRC. Boca Raton, FL. 16 Greene, W., 2003. Econometric Analysis. Prentice Hall. Upper Saddle River, NJ. 17. Users Guide for LIMDEP 8.0. 2004. http://www.limdep.com Econometric Software, Inc. Last Accessed March 23, 2005.

PAGE 60

APPENDIX A LOCATIONS OF DATA COLLECTION SITES 51

PAGE 61

52

PAGE 62

53

PAGE 63

54

PAGE 64

APPENDIX B VIDEO CLIP SCREENSHOTS

PAGE 65

56

PAGE 66

57

PAGE 67

58

PAGE 68

59

PAGE 69

60

PAGE 70

61

PAGE 71

62

PAGE 72

APPENDIX C RURAL FREEWAY TRIP QUALITY SURVEY FORM

PAGE 73

64

PAGE 74

65

PAGE 75

66

PAGE 76

67

PAGE 77

APPENDIX D SAMPLE LOOP DETECTOR DATA

PAGE 78

69 0 Tag County Site Lane Year Month Day Hour Min Int 15 23 28 33 38 43 48 53 58 63 68 73 78 83 91 Total Vol. Avg. Spd 1 5 min vol1 veh/hr/ln1 Density 1 SPD 18 9920 1 04 03 08 00 05 005 0 0 0 0 0 0 0 0 2 1 3 4 3 1 1 15 72.2 SPD 18 9920 2 04 03 08 00 05 005 0 0 0 0 0 0 0 0 0 0 0 2 5 1 0 8 77.4 23 138 1.86 SPD 18 9920 3 04 03 08 00 05 005 0 0 0 1 0 0 0 0 0 0 0 2 4 1 1 9 73.9 SPD 18 9920 4 04 03 08 00 05 005 0 0 0 0 0 0 0 0 1 2 6 10 5 1 0 25 71.8 34 204 2.82 SPD 18 9920 1 04 03 08 00 10 005 0 0 0 0 0 0 0 0 1 2 3 3 7 2 0 18 73.3 SPD 18 9920 2 04 03 08 00 10 005 0 0 0 0 0 0 0 0 0 0 0 3 6 1 2 12 79.3 30 180 2.38 SPD 18 9920 3 04 03 08 00 10 005 0 0 0 0 0 0 0 0 0 0 0 5 7 2 1 15 77.9 SPD 18 9920 4 04 03 08 00 10 005 0 0 0 0 0 0 0 1 0 1 5 10 4 1 0 22 71.9 37 222 2.99 SPD 18 9920 1 04 03 08 00 15 005 0 0 0 0 0 0 0 0 0 2 2 7 3 0 2 16 74.3 SPD 18 9920 2 04 03 08 00 15 005 0 0 0 0 0 0 0 0 0 0 0 1 2 2 3 8 83.5 24 144 1.86 SPD 18 9920 3 04 03 08 00 15 005 0 0 0 0 0 0 0 0 0 0 1 7 5 1 0 14 75.1 SPD 18 9920 4 04 03 08 00 15 005 0 0 0 0 1 0 0 0 2 5 8 7 4 1 0 28 68.5 42 252 3.56 SPD 18 9920 1 04 03 08 00 20 005 0 0 0 0 0 1 0 0 0 5 3 6 4 0 1 20 70.2 SPD 18 9920 2 04 03 08 00 20 005 0 0 0 0 0 0 0 0 0 0 4 0 3 0 1 8 74.6 28 168 2.35 SPD 18 9920 3 04 03 08 00 20 005 0 0 0 0 0 0 0 0 0 2 3 3 6 1 2 17 75.4 SPD 18 9920 4 04 03 08 00 20 005 0 0 0 0 0 0 0 0 0 4 4 8 12 0 0 28 73.0 45 270 3.65 SPD 18 9920 1 04 03 08 00 25 005 0 0 0 0 0 0 0 0 0 0 5 5 3 2 1 16 74.8 SPD 18 9920 2 04 03 08 00 25 005 0 0 0 0 0 0 0 0 0 0 0 2 3 2 0 7 78.0 23 138 1.82 SPD 18 9920 3 04 03 08 00 25 005 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 4 80.5 SPD 18 9920 4 04 03 08 00 25 005 0 0 0 0 0 0 0 0 0 1 7 2 5 0 0 15 71.7 19 114 1.55 SPD 18 9920 1 04 03 08 00 30 005 0 0 0 0 0 0 2 0 0 1 2 9 4 0 0 18 70.2 SPD 18 9920 2 04 03 08 00 30 005 0 0 0 0 0 0 0 0 0 0 0 3 2 1 0 6 76.3 24 144 2.01 SPD 18 9920 3 04 03 08 00 30 005 0 0 0 0 0 0 0 0 0 0 1 2 1 1 2 7 79.6 SPD 18 9920 4 04 03 08 00 30 005 0 0 0 0 0 0 0 0 0 2 5 8 4 2 1 22 73.6 29 174 2.32 SPD 18 9920 1 04 03 08 00 35 005 0 0 0 0 0 0 0 0 3 3 3 6 3 5 0 23 71.9 SPD 18 9920 2 04 03 08 00 35 005 0 0 0 0 0 0 0 0 0 0 5 1 0 0 6 73.8 29 174 2.41 SPD 18 9920 3 04 03 08 00 35 005 0 0 0 0 0 0 0 0 0 0 0 3 2 4 0 9 78.6 SPD 18 9920 4 04 03 08 00 35 005 0 0 0 0 0 0 0 0 0 1 7 11 8 0 0 27 72.8 36 216 2.91 SPD 18 9920 1 04 03 08 00 40 005 0 0 0 0 0 0 0 0 0 2 6 7 2 1 0 18 71.3 SPD 18 9920 2 04 03 08 00 40 005 0 0 0 0 0 0 0 0 0 0 0 1 2 1 0 4 78.0 22 132 1.82 SPD 18 9920 3 04 03 08 00 40 005 0 0 0 0 0 0 0 0 0 0 0 1 4 1 2 8 81.3 SPD 18 9920 4 04 03 08 00 40 005 0 0 0 0 0 0 0 0 2 2 3 7 6 1 1 22 72.7 30 180 2.40 1 These categories were calculated from the given loop detector data and added to the speed data spreadsheets.

PAGE 79

70 Tag County Site Yr. Mo. Day Hour Min Int Lane # Lane # Lane # Lane # Total NB Total SB Total Volume CNT 18 9920 04 03 08 00 05 005 1 15 2 8 3 9 4 25 23 34 57 CNT 18 9920 04 03 08 00 10 005 1 18 2 12 3 15 4 22 30 37 67 CNT 18 9920 04 03 08 00 15 005 1 16 2 8 3 14 4 28 24 42 66 CNT 18 9920 04 03 08 00 20 005 1 20 2 8 3 17 4 28 28 45 73 CNT 18 9920 04 03 08 00 25 005 1 16 2 7 3 4 4 15 23 19 42 CNT 18 9920 04 03 08 00 30 005 1 18 2 6 3 7 4 22 24 29 53 CNT 18 9920 04 03 08 00 35 005 1 23 2 6 3 9 4 27 29 36 65 CNT 18 9920 04 03 08 00 40 005 1 18 2 4 3 8 4 22 22 30 52 CNT 18 9920 04 03 08 00 45 005 1 13 2 4 3 12 4 28 17 40 57 CNT 18 9920 04 03 08 00 50 005 1 8 2 4 3 14 4 29 12 43 55 CNT 18 9920 04 03 08 00 55 005 1 16 2 4 3 8 4 22 20 30 50 CNT 18 9920 04 03 08 01 00 005 1 12 2 3 3 4 4 24 15 28 43 CNT 18 9920 04 03 08 01 05 005 1 19 2 3 3 8 4 19 22 27 49 CNT 18 9920 04 03 08 01 10 005 1 6 2 2 3 8 4 25 8 33 41

PAGE 80

71 Tag County Site Lane Year Month Day Hour Min Int CL01 CL02 CL03 CL04 CL05 CL06 CL07 CL08 CL09 CL10 CL11 CL12 CL13 CL14 CL15 Total Vol. Buses 1 Trucks 1 HV 1 %HV 1 Total % HV 1 CLS 18 9920 1 04 03 08 00 05 005 0 4 5 0 0 0 0 0 5 0 1 0 0 0 0 15 0 6 6 0.4 CLS 18 9920 2 04 03 08 00 05 005 0 6 2 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0.26 CLS 18 9920 3 04 03 08 00 05 005 0 6 2 0 0 0 0 0 0 0 0 0 0 0 1 9 0 1 1 0.11 CLS 18 9920 4 04 03 08 00 05 005 0 9 1 0 0 0 0 0 13 0 0 0 0 0 2 25 0 15 15 0.6 0.47 CLS 18 9920 1 04 03 08 00 10 005 0 9 5 1 0 0 0 0 2 0 0 0 0 0 1 18 1 3 4 0.22 CLS 18 9920 2 04 03 08 00 10 005 0 9 1 0 0 0 0 0 2 0 0 0 0 0 0 12 0 2 2 0.17 0.2 CLS 18 9920 3 04 03 08 00 10 005 0 11 0 0 0 0 0 0 3 1 0 0 0 0 0 15 0 4 4 0.27 CLS 18 9920 4 04 03 08 00 10 005 0 5 4 0 0 0 0 1 10 0 1 0 0 0 1 22 0 13 13 0.59 0.46 CLS 18 9920 1 04 03 08 00 15 005 0 8 3 0 1 1 0 0 3 0 0 0 0 0 0 16 0 3 3 0.19 CLS 18 9920 2 04 03 08 00 15 005 0 7 1 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0.13 CLS 18 9920 3 04 03 08 00 15 005 0 13 0 0 0 0 0 0 1 0 0 0 0 0 0 14 0 1 1 0.07 CLS 18 9920 4 04 03 08 00 15 005 0 11 7 0 0 0 0 0 7 0 1 0 0 0 2 28 0 10 10 0.36 0.26 CLS 18 9920 1 04 03 08 00 20 005 0 9 3 0 1 1 0 0 6 0 0 0 0 0 0 20 0 6 6 0.3 CLS 18 9920 2 04 03 08 00 20 005 0 6 0 0 0 0 0 0 1 0 1 0 0 0 0 8 0 2 2 0.25 0.29 CLS 18 9920 3 04 03 08 00 20 005 0 8 3 0 1 0 0 0 5 0 0 0 0 0 0 17 0 5 5 0.29 CLS 18 9920 4 04 03 08 00 20 005 0 7 3 0 1 0 0 1 16 0 0 0 0 0 0 28 0 17 17 0.61 0.49 CLS 18 9920 1 04 03 08 00 25 005 0 8 0 0 1 0 0 0 6 0 0 0 0 0 1 16 0 7 7 0.44 CLS 18 9920 2 04 03 08 00 25 005 0 5 2 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0.30 CLS 18 9920 3 04 03 08 00 25 005 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 CLS 18 9920 4 04 03 08 00 25 005 0 5 2 0 0 0 0 1 7 0 0 0 0 0 0 15 0 8 8 0.53 0.42 CLS 18 9920 1 04 03 08 00 30 005 0 9 2 0 0 1 0 0 5 0 0 0 0 0 1 18 0 6 6 0.33 CLS 18 9920 2 04 03 08 00 30 005 0 5 1 0 0 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0.25 CLS 18 9920 3 04 03 08 00 30 005 0 2 4 0 1 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 CLS 18 9920 4 04 03 08 00 30 005 0 9 2 0 1 0 0 0 10 0 0 0 0 0 0 22 0 10 10 0.45 0.34 CLS 18 9920 1 04 03 08 00 35 005 0 12 1 0 1 0 0 0 9 0 0 0 0 0 0 23 0 9 9 0.39 CLS 18 9920 2 04 03 08 00 35 005 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0.31 CLS 18 9920 3 04 03 08 00 35 005 0 5 3 0 0 0 0 0 1 0 0 0 0 0 0 9 0 1 1 0.11 CLS 18 9920 4 04 03 08 00 35 005 0 9 2 0 1 0 0 1 13 0 0 0 0 0 1 27 0 15 15 0.56 0.44 1 These categories were calculated from the given loop detector data and added to the class data spreadsheets.

PAGE 81

72

PAGE 82

BIOGRAPHICAL SKETCH David S. Kirschner is a 23-year old graduate student at the University of Florida. He is studying towards his Master of Engineering degree, specializing in transportation engineering. He received a Bachelor of Science in Civil Engineering degree from the University of Florida in December of 2004. 73