This PDF is available at http://nap.nationalacademies.

org/27816

Operational Performance and Safety

Effects of Arterial Weaving Sections
(2024)

DETAILS
78 pages | 8.5 x 11 | PAPERBACK
ISBN 978-0-309-70968-2 | DOI 10.17226/27816

CONTRIBUTORS
Lily Elefteriadou, Gustavo Zschaber, Ehsan Amini, Bob Aghdashi, Orestis
Karamouzis, Alain Carrer, Christopher Kinzel, Irfan Ahmed, Darren Torbic,
BUY THIS BOOK Boniphace (Boni) Kutela, Alexandra Kondyli; National Cooperative Highway
Research Program; Transportation Research Board; National Academies of
Sciences, Engineering, and Medicine

FIND RELATED TITLES SUGGESTED CITATION

National Academies of Sciences, Engineering, and Medicine. 2024. Operational
Performance and Safety Effects of Arterial Weaving Sections. Washington, DC:
The National Academies Press. https://doi.org/10.17226/27816.

Visit the National Academies Press at nap.edu and login or register to get:
– Access to free PDF downloads of thousands of publications
– 10% off the price of print publications
– Email or social media notifications of new titles related to your interests
– Special offers and discounts

All downloadable National Academies titles are free to be used for personal and/or non-commercial
academic use. Users may also freely post links to our titles on this website; non-commercial academic
users are encouraged to link to the version on this website rather than distribute a downloaded PDF
to ensure that all users are accessing the latest authoritative version of the work. All other uses require
written permission. (Request Permission)

This PDF is protected by copyright and owned by the National Academy of Sciences; unless otherwise
indicated, the National Academy of Sciences retains copyright to all materials in this PDF with all rights
reserved.
Operational Performance and Safety Effects of Arterial Weaving Sections

N AT I O N A L C O O P E R AT I V E H I G H W AY R E S E A R C H P R O G R A M

NCHRP RESEARCH REPORT 1094

Operational Performance and Safety
Effects of Arterial Weaving Sections

Lily Elefteriadou
Gustavo Zschaber
Ehsan Amini
Bob Aghdashi
Orestis Karamouzis
Alain Carrer
University of Florida Transportation Institute
Gainesville, FL

Christopher Kinzel
Irfan Ahmed
HDR Engineering, Inc.
Kansas City, MO

Darren Torbic
Boniphace (Boni) Kutela
Texas A&M Transportation Institute
College Station, TX

Alexandra Kondyli
University of Kansas
Lawrence, KS

Subscriber Categories
Design • Operations and Traffic Management • Safety and Human Factors

Research sponsored by the American Association of State Highway and Transportation Officials
in cooperation with the Federal Highway Administration

2024

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

NATIONAL COOPERATIVE HIGHWAY NCHRP RESEARCH REPORT 1094

RESEARCH PROGRAM
Systematic, well-designed, and implementable research is the most Project 15-66
effective way to solve many problems facing state departments of ISSN 2572-3766 (Print)
transportation (DOTs) administrators and engineers. Often, highway ISSN 2572-3774 (Online)
problems are of local or regional interest and can best be studied by ISBN 978-0-309-70968-2
state DOTs individually or in cooperation with their state universities Library of Congress Control Number 2024935155
and others. However, the accelerating growth of highway transporta-
© 2024 by the National Academy of Sciences. National Academies of
tion results in increasingly complex problems of wide interest to high-
Sciences, Engineering, and Medicine and the graphical logo are trade-
way authorities. These problems are best studied through a coordinated
marks of the National Academy of Sciences. All rights reserved.
program of cooperative research.
Recognizing this need, the leadership of the American Association
of State Highway and Transportation Officials (AASHTO) in 1962 ini-
tiated an objective national highway research program using modern COPYRIGHT INFORMATION
scientific techniques—the National Cooperative Highway Research Authors herein are responsible for the authenticity of their materials and for obtaining
Program (NCHRP). NCHRP is supported on a continuing basis by written permissions from publishers or persons who own the copyright to any previously
funds from participating member states of AASHTO and receives the published or copyrighted material used herein.
full cooperation and support of the Federal Highway Administration Cooperative Research Programs (CRP) grants permission to reproduce material in this
(FHWA), United States Department of Transportation, under Agree- publication for classroom and not-for-profit purposes. Permission is given with the
ment No. 693JJ31950003. understanding that none of the material will be used to imply TRB, AASHTO, APTA, FAA,
FHWA, FTA, GHSA, or NHTSA endorsement of a particular product, method, or practice.
The Transportation Research Board (TRB) of the National Academies
It is expected that those reproducing the material in this document for educational and
of Sciences, Engineering, and Medicine was requested by AASHTO to not-for-profit uses will give appropriate acknowledgment of the source of any reprinted or
administer the research program because of TRB’s recognized objectivity reproduced material. For other uses of the material, request permission from CRP.
and understanding of modern research practices. TRB is uniquely suited
for this purpose for many reasons: TRB maintains an extensive com-
mittee structure from which authorities on any highway transportation
subject may be drawn; TRB possesses avenues of communications and NOTICE
cooperation with federal, state, and local governmental agencies, univer-
The research report was reviewed by the technical panel and accepted for publication
sities, and industry; TRB’s relationship to the National Academies is an
according to procedures established and overseen by the Transportation Research Board
insurance of objectivity; and TRB maintains a full-time staff of special- and approved by the National Academies of Sciences, Engineering, and Medicine.
ists in highway transportation matters to bring the findings of research
The opinions and conclusions expressed or implied in this report are those of the
directly to those in a position to use them. researchers who performed the research and are not necessarily those of the Transportation
The program is developed on the basis of research needs iden- Research Board; the National Academies of Sciences, Engineering, and Medicine; the
tified by chief administrators and other staff of the highway and FHWA; or the program sponsors.
transportation departments, by committees of AASHTO, and by The Transportation Research Board does not develop, issue, or publish standards or spec-
the FHWA. Topics of the highest merit are selected by the AASHTO ifications. The Transportation Research Board manages applied research projects which
Special Committee on Research and Innovation (R&I), and each year provide the scientific foundation that may be used by Transportation Research Board
R&I’s recommendations are proposed to the AASHTO Board of Direc- sponsors, industry associations, or other organizations as the basis for revised practices,
procedures, or specifications.
tors and the National Academies. Research projects to address these
topics are defined by NCHRP, and qualified research agencies are The Transportation Research Board; the National Academies of Sciences, Engineering, and
Medicine; and the sponsors of the National Cooperative Highway Research Program do not
selected from submitted proposals. Administration and surveillance of
endorse products or manufacturers. Trade or manufacturers’ names or logos appear herein
research contracts are the responsibilities of the National Academies solely because they are considered essential to the object of the report.
and TRB.
The needs for highway research are many, and NCHRP can make
significant contributions to solving highway transportation problems
of mutual concern to many responsible groups. The program, however,
is intended to complement, rather than to substitute for or duplicate,
other highway research programs.

Published research reports of the

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

are available from

Transportation Research Board

Business Office
500 Fifth Street, NW
Washington, DC 20001

and can be ordered through the Internet by going to

https://www.mytrb.org/MyTRB/Store/default.aspx
Printed in the United States of America

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

The National Academy of Sciences was established in 1863 by an Act of Congress, signed by President Lincoln, as a private, non-
governmental institution to advise the nation on issues related to science and technology. Members are elected by their peers for
outstanding contributions to research. Dr. Marcia McNutt is president.

The National Academy of Engineering was established in 1964 under the charter of the National Academy of Sciences to bring the
practices of engineering to advising the nation. Members are elected by their peers for extraordinary contributions to engineering.
Dr. John L. Anderson is president.

The National Academy of Medicine (formerly the Institute of Medicine) was established in 1970 under the charter of the National
Academy of Sciences to advise the nation on medical and health issues. Members are elected by their peers for distinguished contributions
to medicine and health. Dr. Victor J. Dzau is president.

The three Academies work together as the National Academies of Sciences, Engineering, and Medicine to provide independent,
objective analysis and advice to the nation and conduct other activities to solve complex problems and inform public policy decisions.
The National Academies also encourage education and research, recognize outstanding contributions to knowledge, and increase
public understanding in matters of science, engineering, and medicine.

Learn more about the National Academies of Sciences, Engineering, and Medicine at www.nationalacademies.org.

The Transportation Research Board is one of seven major program divisions of the National Academies of Sciences, Engineering, and
Medicine. The mission of the Transportation Research Board is to mobilize expertise, experience, and knowledge to anticipate and solve
complex transportation-related challenges. The Board’s varied activities annually engage about 8,500 engineers, scientists, and other
transportation researchers and practitioners from the public and private sectors and academia, all of whom contribute their expertise in the
public interest. The program is supported by state transportation departments, federal agencies including the component administrations
of the U.S. Department of Transportation, and other organizations and individuals interested in the development of transportation.

Learn more about the Transportation Research Board at www.TRB.org.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

COOPERATIVE RESEARCH PROGRAMS

CRP STAFF FOR NCHRP RESEARCH REPORT 1094

Waseem Dekelbab, Deputy Director, Cooperative Research Programs, and Manager, National Cooperative
Highway Research Program
David M. Jared, Senior Program Officer
Mazen Alsharif, Senior Program Assistant
Natalie Barnes, Director of Publications
Heather DiAngelis, Associate Director of Publications
Kristin C. Sawyer, Editor

NCHRP PROJECT 15-66 PANEL

Field of Design—Area of General Design
Dirk B. Gross, Delaware, OH (Chair)
Tom Creasey, Alliance Transportation Group, Lexington, KY
Janice R. Daniel, New Jersey Institute of Technology, Newark, NJ
Ganesh J. Karkee, California Department of Transportation, Oakland, CA
Barbara Katherine Ostrom, WSP-USA, Vienna, VA
Howard R. Ressel, Popli Design Group, Rochester, NY
Andrew J. Warren, Arkansas Department of Transportation, Conway, AR
David Anthony Petrucci, Jr., FHWA Liaison
Richard A. Cunard, TRB Liaison

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

FOREWORD

By David M. Jared
Staff Officer
Transportation Research Board

NCHRP Research Report 1094: Operational Performance and Safety Effects of Arterial
Weaving Sections presents a deterministic methodology for evaluating the operational
performance of weaving sections on arterials. This methodology is based on field data
supplemented by microsimulation and driving simulation data and extends the urban
streets analysis of the seventh edition of the Highway Capacity Manual (HCM7). Accom-
panying the methodology are a computational engine for implementation and instruc-
tional materials for dissemination. Two case studies are also provided to illustrate the
methodology; these studies should be of interest to state departments of transportation
and other agencies seeking to reduce weaving-related crashes and conflicts on arterial
roadways.

Arterial roadways are crucial links in the national transportation system. Providing local
and regional mobility and access, they are critical to economic vitality and quality of life
in the United States. As drivers place increasing demands on transportation infrastruc-
ture and freeway congestion continues to increase, arterials have become an alternative
route for many drivers, resulting in the lower performance of these roadways. In many
areas, weaving maneuvers negatively affect arterial operations and safety. Furthermore,
weaving maneuvers can affect the operations and safety of many alternative intersection
configurations, including median U-turns and restricted crossing U-turns. Understand-
ing the operational and safety performance of arterials and the effect of weaving thereon is
important for transportation agencies tasked with maintaining these roadways. The HCM7
provides a methodological approach for estimating the performance of weaving sections
on freeway segments. This methodology, however, is not applicable to arterial weaving due
to differences between freeways and arterials in traffic control, traffic flow characteristics,
and driver behavior.
Under NCHRP Project 15-66, “Operational Performance and Safety Effects of Arterial
Weaving Sections,” the University of Florida was asked to (1) develop a deterministic meth-
odology for evaluating the operational performance of weaving sections on arterials and
(2) identify the safety effects of varying geometric, volume, and traffic control conditions
in weaving sections on arterials. The methodology and analysis are based on field data sup-
plemented by microsimulation data and driving simulation data. The operational analysis
methodology extends the HCM7’s urban streets analysis and is based on the calculation of
potential conflict opportunities and the turbulence index of the arterial weave. The proj-
ect included development of a computational engine that implements the new procedure,
along with instructional materials for disseminating the methodology and two case studies.
The safety analysis was based on crash data and conflict data obtained in the field and in the

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

driving simulator. The analysis concluded that as the weave length increases, the chance of
weaving-related crashes and conflicts decreases.
As a supplement to the report, Appendices A through I, none of which are printed
within this publication, are available on the National Academies Press website (nap.
nationalacademies.org) by searching for NCHRP Research Report 1094: Operational
Performance and Safety Effects of Arterial Weaving Sections.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

CONTENTS

1 Summary
4 Chapter 1 Background
7 Chapter 2 Research Approach
7 Literature Review
8 Performance Measures for Evaluating Arterial Weaving Segments
8 Data Collection
35 Data Analysis
49 Operational Analysis Methodology

62 Chapter 3 Findings and Applications

65 Chapter 4 Conclusions and Suggested Research
67 References
68 Appendices A through I
69 Bibliography

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

SUMMARY

Operational Performance and

Safety Effects of Arterial
Weaving Sections
The procedures detailed in the seventh edition of the Highway Capacity Manual (HCM7)
do not consider the impacts of arterial weaving, defined as the crossing of two or more traf-
fic streams in the same direction between two or more access points on a multilane urban
street with some type of traffic control. The primary objective of this project was to develop
a deterministic method for evaluating the operational performance of weaving sections on
arterials. The secondary objective was to identify the safety effects of varying geometric,
volume, and traffic conditions in weaving sections on arterials.
The research team first synthesized literature related to operational analysis methods and
performance measures for arterial weaving. Then the team synthesized previous research
related to safety along arterial weaving sections. It was concluded that there are very few
studies addressing operations along arterial weaving segments and no studies that evaluate
the safety of such sections. The project led to the development of a list of performance mea-
sures for evaluating arterial weaving sections and a recommendation for using travel time as
the service measure to ensure consistency with the HCM. It also led to the development of
two new intermediate measures, potential conflict opportunities (PCOs) and the turbulence
index (TI), to be used in the operational analysis. A PCO is defined as a potential encounter
between vehicles in conflicting traffic flows within a weaving segment. The TI is a function
of the PCO and site length, reflecting the density of these potential conflicts over the length
of the weaving segment. Both measures can be calculated by the analyst using the weaving
and nonweaving demand flows within the segment.
Based on what it found in the literature review, the research team acquired field data at
20 locations (18 of which were used in the analysis) in California, Nebraska, North Carolina,
Ohio, and Texas. A total of 1,118 min of video were collected by drones; approximately
6,000 weaving and 11,000 nonweaving trajectories were extracted. Simulations of five of
these sites were created to obtain additional data for analysis. A total of 1,080 additional sce-
narios were simulated; this produced over 413,000 simulated trajectories for use in model
development. In addition, a driving simulator experiment was conducted with 39 participants
to explore driver behavior around arterial weaving areas.
Data analyses were conducted to evaluate relationships between performance character-
istics and specific prevailing traffic, control, and geometric characteristics of arterial weav-
ing segments based on the field and microsimulation data. The analysis showed that speed
generally decreases as demand increases. Similarly, speed decreases as the PCO increases
and the length of the segment decreases. The analysis also found that the number of missed
maneuvers increases with demand and with the PCO.
The safety analysis evaluated the effects of varying geometric, volume, and traffic condi-
tions in weaving sections on arterials. The methods and analysis are based on field data

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

2   Operational Performance and Safety Effects of Arterial Weaving Sections

supplemented by microsimulation data and driving simulation data. Crash analyses were
conducted based on field data, while conflict analyses were conducted based on trajectories
obtained in the field and trajectories obtained through the driving simulator. The analyses
indicated that as the weave length increases, the chance of weaving-related crashes and
conflicts decreases. Regarding crash data, this impact is statistically significant at a 95%
confidence level for sideswipe crashes only. The statistically insignificant results for total
crashes may be attributed to the impact of other factors (such as signalization patterns), a
relatively small sample size, or both.
Based on the data obtained, the research team developed an operational analysis methodol-
ogy for evaluating arterial weaving segments, suitable for HCM-type analysis. The operational
analysis method extends the existing HCM’s urban streets analysis and is based on the cal-
culation of PCO and the TI of the arterial weave. The PCO for any movement is the product
of the respective conflicting demands. Since there may be several weaving maneuvers within
an urban segment, the total PCO for the segment must be estimated as the sum of all PCOs.
The TI is a measure of the overall impact of weaving movements on traffic operations
within the arterial weaving section. A higher TI value indicates a higher level of traffic tur-
bulence and potential inefficiencies in traffic flow. When the weaving flow is zero, the TI is
zero. As weaving demands increase, the PCOs increase, which results in an increase of the
TI. As the length over which weaving occurs increases, the constraints due to lane changing
decrease, and the TI decreases.
The new weaving analysis methodology is based on estimating the speed drop (or delay
and travel time increase) at an arterial segment due to weaving activity. The arterial weaving
segment performance is evaluated based on the same two service measures as that of urban
streets [i.e., the performance measures used to define level of service (LOS)]:
• Travel speed for through vehicles, which includes the impacts of the design and prevailing
traffic conditions within the link and the delay incurred to the through movement at the
downstream intersection, and
• Volume-to-capacity (v/c) ratio for the through movement at the downstream intersection.

The project developed a computational engine that implements the new procedure in
the Excel-based STREETVAL computational engine. The research team created two real-
world case studies based on the field data collected to assess the impact of weave maneuvers
in urban streets. To facilitate dissemination, the research team developed instructional
materials containing an overview of the HCM urban streets, the proposed method of
NCHRP Project 15-66, “Operational Performance and Safety Effects of Arterial Weaving
Sections,” and the two case studies.
The analysis showed that shorter weaving segments with higher demands could lead to
up to five missed weaving maneuvers in 15 min. The PCO and TI provide a good indication
of turbulence that can evaluate the impacts of access points.
To facilitate use of the new methodology, the research team recommends integrating the
method within the HCM. Additional limitations of the research are as follows:
• The data collection did not include sites with weaving that involve more than one seg-
ment. Additional data should be collected to evaluate operations at such sites.
• The data collection did not include sites with significant presence of pedestrians and bicy-
cles. Additional data should be collected to evaluate operations at such sites.
• The analysis of missed turns did not produce adequate data to develop a quantitative
model. Therefore, it is suggested to collect data at sites where such maneuvers are more
frequent and develop suitable methods to predict this measure.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Summary  3

The safety analysis of arterial weaving sections that investigated crash and conflict data
yielded expected results; although, with a few exceptions, the parameters in the statistical models
were not statistically significant. The analysis of time-to-collision (TTC) conflicts (defined as
the time measurement of how many seconds two vehicles would take to collide if they are
moving at a constant speed and direction) based on the driving simulator yielded the only
statistically significant results. These showed fewer TTC conflicts with an increase in site length,
and while the analyses of crash data and heavy-braking data from the study sites did not yield
statistically significant results in most cases, the signs of the parameters were as expected.
The analysis of crash data showed that the ratio of weave length to site length has a negative
association with crashes, suggesting that as the ratio of weave length to site length increases,
the number of crashes decreases. Similarly, the analysis of heavy-braking conflicts at the study
sites indicates that as the ratio of weave length to site length increases, the number of such
conflicts decreases.
Given the results of the safety analysis, designers should consider closing a midblock access,
which would result in increasing the length of the weaving area along the segment. Finally,
although the results of the safety analysis did not provide a definitive relationship between the
weave length of arterial weaving sections and crashes or conflicts, sufficient information was
found to suggest that additional research in this area would yield promising results. There-
fore, it is recommended that future research be conducted to develop methods that predict the
safety performance of arterial weaving sections, suitable for inclusion in the AASHTO Highway
Safety Manual.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

CHAPTER 1

Background

The procedures detailed in the seventh edition of the Highway Capacity Manual (HCM7)
do not consider the impacts of arterial weaving, defined as the crossing of two or more traffic
streams in the same direction between two or more access points on a multilane urban street
with some type of traffic control. The HCM7 provides methods for evaluating freeway weaving.
However, arterial weaving is conceptually different from freeway weaving in several ways:
• The operations of the weave (discharge patterns from the upstream intersection, queue formation)
depend on the signalization of the upstream and downstream intersections.
• In arterial weaving, the length of the weave is variable and depends on the length and lateral
location of the downstream queues.
• The lane-by-lane demands along urban streets are affected by the vehicles’ destinations, and
there is a complex interaction between weaving demands and mainline arterial demands.
• Weaving along urban streets is affected by midblock access, which may impact weaving opera-
tions in different ways, depending on the demands of each weaving, nonweaving, and turning
movement.
• Along urban streets and around alternative intersection configurations, two-sided weaving is
very common; such a maneuver is less common along freeways.
• Arterial weaving sections can have left and right exits and entrances at either end; thus, arterial
weave operations can be more complex because of the larger number of origin–destination
(O-D) pairs crossing each other.
Previous research on arterial operations and interchange ramp terminals (NCHRP Project 3-47,
“Capacity Analysis of Interchange Ramp Terminals,” NCHRP Project 3-60, “Capacity and Quality of
Service of Interchange Ramp Terminals,” and NCHRP Project 3-60A, “Validation and Enhance-
ment of the Highway Capacity Manual’s Interchange Ramp Terminal Methodology”) has indicated
that omission of arterial weaving considerations is a limitation of the existing methods. Figure 1
provides several examples of arterial weaving configurations. As shown, there are several possibili-
ties for weaving, each of which affects arterial operations and safety in a different way. Two-sided
weaves affect the entire segment, though one-sided weaves may affect only a subset of lanes. Short
weaving sections have the potential to affect operations and safety more extensively, particularly for
higher weaving demands. Configurations like those of urban arterial cloverleaves and diverging
diamond interchanges involve movements that are not controlled, and therefore their weaving
patterns will differ from those that are controlled.
Weaving along arterials may involve more than one segment (these are defined in HCM7
Chapter 18, Urban Street Segments). There is currently no clear definition of urban arterial weav-
ing in terms of spatial boundaries. HCM7 Chapter 16, Urban Street Facilities, evaluates urban
street facilities, which consist of combinations of segments. Consideration of weaving operations

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Background  5

Right-sided weave in dedicated lane Short separation WB weave from off-ramp RT

Short separation frontage road and on-/off-ramp

Urban arterial cloverleaf

DDI weave

Figure 1.   Arterial weaving configuration examples (WB = westbound, RT = right turn, DDI = diverging
diamond interchanges).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

6   Operational Performance and Safety Effects of Arterial Weaving Sections

along urban streets requires defining “arterial weaving sections” and including their operational
analysis within the context of the HCM7 methodologies of Chapter 16 and Chapter 18.
The primary objective of this project was to develop a deterministic method for evaluating
the operational performance of weaving sections on arterials. The secondary objective was to
identify the safety effects of varying geometric, volume, and traffic conditions in weaving sec-
tions on arterials.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

CHAPTER 2

Research Approach

The research team first synthesized the literature related to operational analysis methods and
performance measures for arterial weaving. The team also synthesized previous research related
to safety along arterial weaving sections. Next, researchers developed a list of performance mea-
sures for evaluating arterial weaving sections. The third stage of the research led to the devel-
opment and execution of a data collection plan, which involved field data collection and the
extraction of microsimulation data and driver simulator data. For the fourth stage, researchers
evaluated relationships between performance characteristics and specific prevailing traffic,
control, and geometric characteristics of arterial weaving segments based on the field and micro-
simulation data. The fifth stage involved the development of an operational analysis methodology
for evaluating arterial weaving segments, while for the sixth stage researchers analyzed crash and
conflict data at these types of segments. Each of these stages and their findings are provided in
the following subsections.

Literature Review
The research team conducted a thorough literature review related to arterial weaving and
confirmed the literature on this topic is very limited. Researchers examined and documented
various arterial weaving configuration types and previously reported weaving analysis methods
for arterials and freeways. Researchers also reviewed research on safety analyses conducted for
various types of maneuvers around urban intersections and arterial segments [including median
U-turn intersections and restricted crossing U-turn (RCUT) intersections] and along freeway
weaving segments. The team developed a summary of performance measures for operational
performance and safety performance of urban streets. The team also examined and summarized
driver simulator data collection methods for operational analysis and conflict analysis and other
data collection methods that are pertinent for weaving operations, including simulation software
for arterial weaving. The research team concluded the following based on the literature review:
• There is a wide variety and proliferation of arterial weaving configurations, particularly with
the increasing use of alternative intersections and interchanges.
• There is a limited amount of literature that focuses on the impact of weaving on arterial opera-
tions, and existing literature has focused on the prediction of weaving and nonweaving speeds.
• The literature on freeway weaving operations indicated that the HCM7 method, which is
based on speed prediction of weaving and nonweaving vehicles, has inherent limitations and
does not produce accurate capacity and speed estimates.
• Several articles have explored issues of access management around interchanges and the
impact of section length on operations and safety.
• No studies were identified that evaluate the safety of arterial weaving, nor the interaction of
pedestrians and other vulnerable travelers with weaving operations.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

8   Operational Performance and Safety Effects of Arterial Weaving Sections

• A limited number of studies have compared U-turn operations to left-turn operations and
have examined the safety of RCUT configurations.
• Driving simulators can be used to study driver gap-acceptance and lane-changing behaviors
that are prevalent in arterial weaving segments, while providing a rich data set of trajectories
for safety and operational purposes.
The detailed literature review is provided in Appendix A.

Performance Measures for Evaluating

Arterial Weaving Segments
This part of the project examined performance metrics suitable for analysis of arterial weaving
segments and developed a list of suitable performance measures to be used for operational analy-
sis. The measures selected are useful for practitioners and suitable for the evaluation of arterial
weaving sections. In summary, the performance measures used for the new operational analysis
methodology should satisfy the following requirements:
• Address possible applications and uses of the Highway Capacity Manual (HCM) and the High-
way Safety Manual (HSM).
• Include measures that can be practically obtained in the field and are readily available to
highway agencies.
• Consider the driver perspective as it relates to operational and safety analyses and provide
measures by O-D, segment/weaving section, and turning movement.
• Support analyses and evaluations beyond the scope of the HCM and HSM, and consider addi-
tional types of analysis, like access management, emissions, and fuel consumption.
• Allow for integration into a comprehensive performance measurement program, such that
agencies can consider the measures in assessing specific actions (e.g., signal control improve-
ments on arterials).
The research team conducted a thorough study of existing and potential performance met-
rics (see Appendix B) and concluded there are numerous performance measures that can be
employed depending on the type of application. The performance measures recommended for
this project are summarized in Table 1. The table provides the list of performance measures
by topic, categorizing them as primary (i.e., these should be obtained and can be collected or
derived directly) versus secondary (i.e., they are useful but not necessary in evaluating opera-
tional or safety performance). The last column provides the justification for using each measure
and additional discussion on its collection.

Data Collection
During this part of the research, researchers planned and executed a data collection plan that
included field data, microsimulation data, driving simulator data, and crash data. The number
of possible combinations of geometric design, traffic, and control conditions (design speed,
weaving demand levels and patterns, number of lanes, intersection configuration, queue length
by lane, etc.) is extremely high. In addition, field data collection and data reduction are time-
consuming and expensive. On the other hand, arterial weaving operations and driver behavior
around such sites have not been studied previously in detail, and thus, simulation models do
not necessarily consider their operations comprehensively. Therefore, data collection consisted
of a combination of field data collection, driving simulation, and microsimulation; for safety
analysis, crash data and conflict data were assembled at the sites where field data are collected for
operational purposes. Conflict data were also collected using a driving simulator. The remainder
of this section provides summaries of the traffic-related field data collected along with study

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   9

Table 1.   Summary of performance measures.

Primary (P),
Performance Measure Secondary (S) Comments
Operations
Travel speed P Essential for seamless analysis with HCM7 Chapter 18, Urban Street
Segments.
Capacity P Essential for seamless analysis with HCM7 Chapter 18, Urban Street
Segments.
Potential conflict P Defined as a potential encounter between vehicles in conflicting traffic
opportunities (PCOs) flows within a weaving segment. It can serve as an indicator of turbulence.
Turbulence index (TI) P Function of PCOs and site length, reflecting the density of these potential
conflicts over the length of the weaving segment.
Percent drivers unable S Cannot be measured easily in the field, but it provides an important
to complete weave perspective when designing and operating the segment.
Stop rate within the S Defined in the HCM7, cannot be easily measured in the field, but it can
weave provide an important perspective related to cycle failures. Furthermore, a
changing stop rate may be more strongly related to signalization patterns
rather than weaving activity, and it is not feasible to distinguish the cause
of each stop.
Safety
Crash frequency P Essential for consistency with the predictive methods in HSM Part C (e.g.,
Chapter 12, Predictive Method for Urban and Suburban Arterials).
Heavy braking (HB) S Can be used for detection of rapid deceleration and potential collisions.
Time-to-collision S Indirect measure of safety defined as the time measurement of how many
(TTC) seconds two vehicles would take to collide if they are moving at a constant
speed and direction.
Post-encroachment S Indirect measure of safety defined as the time measurement of the
time (PET) difference between the time a vehicle enters a conflict point until the time
another vehicle arrives to this point.
Other Applications
Lane-change frequency S Lane-change frequency (for mandatory lane changes) can be derived from the
and distribution site geometry and the origin–destination (O-D) pairs within the weave. The
longitudinally distribution of lane changes requires specialized data collection procedures.
However, it can be obtained through simulation and can help provide
guidance related to signing, pavement markings, and access management.
Volume-to-capacity per S Can be collected or derived from other measures and used for operational,
movement, lane, and O-D safety, and planning analyses.

site characteristics; summaries of the driving simulator data; summaries of the microsimulator
calibrations, scenarios, and simulated data; and summaries of the databases developed for the
safety analysis.

Operational Field Data

The research team employed drone-captured video as its primary means of operational data
collection. The areas where field drone data collection was conducted, supplemented by agency
data such as crash records, are shown in Figure 2. These areas were selected to minimize the cost
of data acquisition per site (the research team has experienced drone pilots with the necessary
equipment in each of the selected areas), while maximizing the number of suitable weave sites in
each area. The exception was Columbus, Ohio, for which drone data were collected and provided
to the team by the Ohio Department of Transportation.
Other considerations in selecting sites (except those from Ohio) included Federal Aviation
Administration drone flight ceilings and the ability to obtain approvals; availability of traffic vol-
umes and general assessment of weave operations; local input on available data, weave operations,
and permissions to collect data; diversity of weave types (including the desire to collect data at

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

10   Operational Performance and Safety Effects of Arterial Weaving Sections

Roseville, CA Omaha, NE
Columbus, OH
Sacramento, CA

Charlotte, NC

Georgetown, TX
Round Rock, TX
Austin, TX

Drone video data provided by ODOT Drone video data collected by study team

Figure 2.   Field data collection metro areas (ODOT = Ohio

Department of Transportation).

cloverleaf and frontage road weaves); and a desire for straight (not curved) segments if possible
(for ease of analysis). The detailed specifications of the drone camera settings are as follows:
• Video Size: 4K/25 frames per second (fps) (3,840 × 2,160 HQ),
• Video Format: MP4, and
• Encoding Format: H.264.

There is a notable drop in the accuracy of the vehicle localization if the angle with respect to the
ground is greater than 55 degrees, with zero degrees indicating that the drone is directly above
the intersection. Therefore, flights were planned to keep the recording angle below 55 degrees.
In addition, extra care was taken to avoid changing drone position when recording. Collecting
video against a strong source of light was avoided as well.
The research team collected field data at 15 locations spread around the country, most cover-
ing one weave, some covering two—one in each direction. In addition to the four geographic
areas in which the research team collected drone data, the Ohio Department of Transportation
(DOT) collected the same type of data for five sites within the Columbus area. Table 2 provides
a list of all 20 data collection sites with a summary of their characteristics. Appendix C discusses
the sites at which data were collected.
In each of the four geographic regions (Nebraska, Texas, California, North Carolina), the
research team initially identified three to four recommended weave sites where 15-min drone
flights would be conducted. For these, it was initially planned that four 15-min drone flights
would be conducted for each site, resulting in 1 h of video data for each site. However, there were
several sites at which the research team was not able to collect data during all four 15-min flights
because of either technical difficulty with the hardware or flight permission conflicts. Table 3
summarizes the field data collected by site. The orange and gray boxes in the table indicate short-
ened or canceled flights due to the issues already described. Despite the technical difficulties,
the research team was able to exceed the initial target of 900 min of data (15 sites × four 15-min
videos × 15 min). When combined with the video data received from the Ohio DOT, the team
was able to collect 1,118 min of video, well over the initial target.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   11

Table 2.   List of data collection sites and summary of their characteristics.

Down- Approx.
stream Site ADT in
Through LT Length Weaving
ID Type* Arterial Segment** Direction Lanes Lanes (ft) Area
RRamp to
CA4 RRamp (Clov) Harbor, I-80 Ramp Off to On NB 3 0 810 11,807
RRamp to
CA5 RRamp (Clov) Florin, Golden State Off to On EB 3 0 435 27,286

RRamp to
CA5 RRamp (Clov) Florin, Golden State Off to On WB 3 0 435 32,102
CA12 RRamp to Int Douglas, I-80 WB to Sunrise EB 3 2 735 28,610
CA13 Int to Int Galleria, Roseville to Antelope Cr NB 3 2 1,115 12,938
RRamp to
NC1 RRamp (Clov) Uni. City Blvd, I-485 Off to On WB 3 0 1,250 20,313

NC6 RRamp to Int Freedom, I-85 Off to Tuckaseegee SEB 3 1 710 24,740
W Woodlawn, I-77 Off to
NC7 RRamp to Int S Tyron WB 2 1 975 19,740
RRamp to
NC9 RRamp (Clov) Mathews, US-74 Off to On NEB 2 1 955 17,198
NE1 RRamp to Int S 84th St, I-80 Off to F St SB 2 1 815 22,340
NE4 RRamp to Int N 204th St, W Dodge NB 2 1 980 17,590
NE6 Int to RRamp L, S120 to I-80 On EB 3 0 1,300 26,475
RRamp to
NE6 RRamp L, I-80 Off to S120 WB 3 2 1,300 26,475
E Broad St, I-270 Off-Ramp to
OH1 RRamp to Int Taylor Station Rd EB 3 1 1,200 31,416
E Main St, I-270 Off-Ramp to
OH2 RRamp to Int Noe Bixby Rd WB 2 1 825 13,921
S State St, I-270 Off-Ramp to
OH3 RRamp to Int Huber Village Blvd NB 2 2 1,140 21,010
W Broad St, I-270 Off-Ramp to
OH4 RRamp to Int Grener Ave WB 3 1 750 20,581
Sawmill Rd, I-270 Off-Ramp to
OH5 RRamp to Int Snouffer Rd SB 3 1 1,115 25,462
I-35 FR, I-35 Ramp to Cesar
TX2 LRamp to Int Chavez NB 2 2 560 23,152
TX3 Int to Int SH29, I-35 to Wolf Lakes WB 2 2 1,000 16,763
Old Settlers (379), I-35 FR to
TX5 Int to Int May’s EB 2 1 535 35,134
Old Settlers (379), I-35 FR to
TX6 Int to Int Chisolm WB 2 1 540 12,801
Notes: ID = identification, LT = left turn, ADT = average daily traffic.
* LRamp = left-side ramp, RRamp = right-side ramp, Int = intersection, Clov = cloverleaf
** FR = frontage road

The research team used the online machine-vision processing tools of DataFromSky [(DFS),
datafromsky.com] to extract vehicle trajectory data from each video. These included speed, accel-
eration, and safety data like heavy-braking (HB) and time-to-collision (TTC) events. The videos
were manually georegistered to allow accurate extraction of data related to distance, speed, and
acceleration using the DFS Viewer (DFSV). Georegistration adds scale to videos, allowing data
to be extracted in real-world coordinates. The georegistration process involved setting at least
four registration points and inserting coordinates in the WGS-84 system. The coordinates were
extracted from Google Earth ProTM based on easily identifiable analogous points or objects in
the videos and Google Earth (these are typically traffic markings and road boundaries). DFS
conflates, or rubbersheets, the video image based on these user-input registration points and
computes a best-fit deviation for each point. Figure 3 provides an example of the georegistration

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

12   Operational Performance and Safety Effects of Arterial Weaving Sections

Table 3.   Drone video data collection summary.

15-Min. Time Slots Drone Video Start Times
Total
Sites 1 2 3 4
Mins
1 2 3 4

CA 4 30 11:00 a.m. 2:30 p.m.

CA 5
10.1 25.1 1:30 p.m. 12:45 p.m.
(EB)
CA 5
10.1 25.1 1:30 p.m. 12:45 p.m.
(WB)
CA 12 60 8:15 a.m. 8:45 a.m. 3:00 p.m. 3:30 p.m.
CA 13
60 8:15 a.m. 8:45 a.m. 3:00 p.m. 3:30 p.m.
(NB)

NC 1 60 7:45 a.m. 8:15 a.m. 8:45 a.m. 5:50 p.m.

NC 6 60 12:00 p.m. 4:45 p.m. 5:15 p.m. 5:45 p.m.

NC 7
60 10:30 a.m. 11:00 a.m. 10:45 a.m.* 4:00 p.m.
(EB)
NC 7
60 10:30 a.m. 11:00 a.m. 10:45 a.m.* 4:00 p.m.
(WB)
NC 9 60 4:15 p.m. 11:45 a.m.* 12:15 p.m. 12:45 p.m.

NE 1 60 4:00 p.m. 4:30 p.m. 5:00 p.m. 7:00 p.m.

NE 4 60 4:00 p.m. 4:30 p.m. 5:00 p.m. 5:30 p.m.

NE 6
60 7:00 a.m. 7:30 a.m. 5:30 p.m. 7:45 a.m.*
(EB)
NE 6
60 7:00 a.m. 7:30 a.m. 5:30 p.m. 7:45 a.m.*
(WB)

OH 1 13.53 8.78 9.13 10.12 4.98 8.5 55.1

OH 2 10.07 8.35 9.8 4 32.2
OH 3 13.88 8.35 7.67 10.07 7.72 47.7
OH 4 12.67 6.97 19.6
OH 5 11.25 13.72 9.97 10.75 45.7

TXO 2 7.5 52.5 7:30 a.m. 10:30 a.m. 1:30 p.m. 6:30 p.m.

TXO 3 7.5 22.5 4:45 p.m. 5:15 p.m.

TXO 5 60 7:45 a.m. 11:45 a.m. 3:30 p.m. 4:45 p.m.

TXO 6 12.1 42.1 12:15 p.m. 4:00 p.m. 5:15 p.m.

Total 1,118 mins * Next day

Complete 15 min.
Shortened drone flight
due to technical issues.
No flights due to
technical difficulties.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   13

Registration
points
(coordinates
obtained from
Google Earth)

Deviations of
conflated image

Figure 3.   Example of georegistration process (NC7) using DFSV.

process using the DFSV, and Figure 4 illustrates postprocessing speeds from site NC7. The
research team set a maximum tolerance of 6 in. to determine acceptable deviations.
DFS raw trajectories of each detected object were exported to a comma-separated values
(CSV) file, where information about position, speed, acceleration, and other useful information
for each frame of the video could be extracted for further analysis. After a cleanup process, the
data were stored in the format shown in Table 4.
The variables included in the raw trajectory files are defined as follows:
• Track ID and Type of Object—Unique identification (ID) and type of identified object (car,
bicycle, etc.).
• Entry Gate and Entry Time (s)—Exact time when the vehicle passed through a user-defined
entry gate (imaginary boundary).
• Exit Gate and Exit Time (s)—Exact time when the vehicle passed through a user-defined exit
gate.
• Traveled Dist. (ft) and Avg. Speed (mi/h)—Traveled distance and average speed between gates.

No gates were defined for exporting the raw trajectories. Hence, the gate fields are empty in
the raw trajectory files as seen in Table 5. The following columns are repeated for each frame of
the video:
• Trajectory—x (degrees), y (degrees)—Position of the object.
• Speed (mi/h)—Speed of the object at a specific moment.
• Tan. Acc. (ft/s2) and Lat. Acc. (ft/s2)—Tangential and lateral acceleration of the object at a
specific moment.
• Time (s)—Seconds from the beginning of the video, where the position, speed, and accelera-
tion of the object are measured.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

14   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 4.   Still from processed video of NC7 after georegistration with real-time speed annotations (mi/h).

Using a combination of the DFSV tools, manual observation of the videos, and external analysis
of the trajectories using secondary tools, the research team extracted operational measures, including
vehicle movements/positioning/O-D, speeds/acceleration/deceleration, headways, and vehicle
classification. The team also extracted conflict-based safety surrogate measures, including HB, TTC,
and post-encroachment time (PET) measures.
Vehicle trajectories are the single most powerful data item that can be collected because almost
every other data element can be derived from them. As mentioned previously, DFS trajectories
are available in a proprietary format for further analysis within the DFSV. The research team
exported all the trajectories from the DFSV in a 5-fps event rate to a CSV format with the following
information for each vehicle at each time step throughout the recording:
• Unique ID,
• Type of the vehicle,
• Entry gate (origin),
• Entry time,
• Exit gate (destination),
• Traveled distance,
• Average speed,
• Time stamp (exported in 5 fps),
• Latitude/longitude,
• Instantaneous speed,
• Instantaneous tangential acceleration, and
• Instantaneous lateral acceleration.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections
Table 4.   Example of raw trajectory file (NC 7; 4:45–5:00 p.m.).
Copyright National Academy of Sciences. All rights reserved.
Operational Performance and Safety Effects of Arterial Weaving Sections

16   Operational Performance and Safety Effects of Arterial Weaving Sections

Table 5.   Summary of trajectories obtained by site.

No. of No. of No. of

Site Length Mainline Weaving Nonweaving
Site Latitude Longitude (ft) Lanes Trajectories Trajectories
CA4 38°34'28.68"N 121°32'54.52"W 810 1 202 102
CA5 38°29'45.49"N 121°26'45.64"W 435 1 177 711
CA12 38°44'38.27"N 121°16'4.34"W 735 3 495 403
CA13 38°46'9.95"N 121°15'56.41"W 1,115 3 823 849
NC1 35°18'44.12"N 80°42'30.98"W 1,250 1 292 321
NC6 35°14'46.26"N 80°53'40.06"W 710 3 545 477
NC7 35°10'48.71"N 80°53'13.82"W 975 1 564 826
NC9 35°7'32.90"N 80°42'18.76"W 955 2 197 697
NE1 41°13'16.69"N 96°2'35.75"W 815 2 664 954
NE4 41°15'53.13"N 96°14'3.51"W 980 2 658 769
NE6 41°12'42.58"N 96°5'53.65"W 1,300 1 – –
OH1 39°58'44.34"N 82°50'48.50"W 1,200 2 57 1,023
OH2 39°57'17.87"N 82°50'59.87"W 825 1 144 235
OH3 40°6'10.86"N 82°55'29.25"W 1,140 2 – –
OH4 39°57'9.46"N 83°7'22.64"W 750 2 126 1,008
OH5 40°6'20.55"N 83°5'27.88"W 1,015 2 61 755
TX2 30°15'36.65"N 97°44'10.29"W 560 3 65 813
TX3 30°37'59.46"N 97°41'38.63"W 1,000 2 123 383
TX5 30°32'4.49"N 97°41'26.10"W 535 2 793 625
TX6 30°32'9.88"N 97°41'38.14"W 540 2 332 454
Total 6,318 11,405

Additionally, all trajectories were manually classified either as weaving or nonweaving trajecto-
ries according to the characteristics of each site (left- or right-hand arrivals and departures). Table 5
provides a summary of all collected trajectories from the DFSV. For sites OH3 and NE6, the research
team was not able to record the entire trajectories within the weaving sections because of the length
of the facility along with illumination conditions (either too bright or too dark). Therefore, the
analysis used the data from 18 sites out of the total of 20 where data were collected. Using the DFSV,
the research team obtained speeds through the arterial weaving segment and away from the influ-
ence of the upstream and downstream intersections when feasible. The average speed of the vehicles
within the entire weaving section was also gathered using the trajectories. The average speed of all
vehicles within the weaving section is presented in Appendix D, which is a summary of field data.
As shown in Table 5, more than 6,000 weaving trajectories and 11,000 nonweaving trajectories
were extracted from the DFSV and were used in this study. The remaining 18 data collection sites
provided a wide variety of possible configurations for urban weaving facilities. A sample of the
recorded trajectories is provided in Appendix E.
Regarding pedestrians, except for one site, the remaining ones had only a few pedestrians (under
10 pedestrians per 15-min video). TX6 is the only site where a higher number of pedestrians were
identified because of the presence of homeless people at the intersection, which did not seem to affect
the site’s operation. Because of this scarcity of data, researchers were not able to evaluate the impact
of the presence of pedestrians within a weave. However, it is unlikely that pedestrians would affect
the operation of arterial weaves in a different manner than those of urban segments without weaving.
For each study site discussed previously, researchers obtained available crash data from
the local agency or DOT. From the local agency or DOT or through review of online aerial

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   17

photography, the research team also obtained relevant characteristics such as posted speed limit,
striping patterns, lane geometry, signalization and phasing, area type and local context, presence
of driveways and roadways within the weave, and roadway geometry.
The following is a list of additional data the research team obtained at every site:
• Site length and weave length—Two lengths were defined and obtained for each site. The site
length was measured as the painted gore of the ramp or free right turn to the downstream stop
bar of the approach. For intersection-to-intersection weaves the site length is measured from the
upstream stop bar to the downstream stop bar. The weave length is shorter and was measured as
the distance between the painted gore of the ramp or free right turn and the diverge point (left
turn, right turn, or diverge ramp). The site length was used in the operational analysis, while
the weave length was used in the safety analysis. Both data points were collected through online
aerial photography and drone data.
• Other site geometric characteristics—Additional site characteristics include type of terminal
(i.e., intersection or ramp), number of through lanes, and number of downstream left-turn
lanes. These were also collected through aerial photography and drone data.
• Archived traffic volumes—These include major- and minor-road traffic volumes at the terminals
and traffic volumes along the arterial [i.e., annual average daily traffic (AADT) and peak-hour
volumes].
• Archived crash data—These include crash data in the vicinity of the weaving section and con-
sist of the following data elements: crash number, location, date, time of day, relationship to
junction, crash type and manner of collision, and severity level. Crash data were obtained for
all study sites either directly from the respective state DOT or from available online portals.
• Traffic control information—The research team obtained signal phasing and timing plans for
the upstream and downstream intersections of the weave (if signalized).

Driving Simulator Data

A driving simulator experiment was conducted at the University of Kansas (KU) to extract
driver behavior parameters in arterial weaving areas. The simulator was used to extract three
types of data: (1) subjective measures, (2) driving parameters (e.g., lane-change behaviors, accel-
erations, speeds, and vehicle trajectories), and (3) physiological measures.
The subjective data were collected using electronically administered questionnaires including
the prescreening questionnaire, post-completion questionnaire, NASA-Task Load Index (TLX),
Positive and Negative Affect Schedule (PANAS), Psychological Entitlement Scale (PES), Multi­
dimensional Driving Style Inventory (MDSI), and Reckless Driving Habit Scale (RDHS). Driving
parameters were extracted from the simulation tasks and include distance between the start
of the road segment and the initiation of the weaving maneuver, speeds during the weaving
maneuver, acceleration, gap sizes (measured in seconds from the back of the leading vehicle to
the front of the following vehicle) in existing and target lanes, number of failed weaves, number
of collisions, and TTC.
Physiological measures were collected using the eye tracker and include gaze duration, mental
workload, and others.
The driving simulation study took place at the KU driving simulator laboratory from April 18,
2022, to June 24, 2022. The driving simulator is a fixed-base simulator in an Acura MDX chassis
(half cab). The simulator provides a 170-degree horizontal field of view as shown in Figure 5, with
three forward screens and one rear screen. The rear screen renders the view of both side-view
mirrors and a rear-view mirror, providing an immersive driving experience.
All data obtained through the driving simulator are recorded on the MiniSim computer, while
the video of the participant’s drive is captured on the video capture computer. A separate machine

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

18   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 5.   Layout of the KU driving simulator.

is used for the FOVIO-FX3 eye-tracking equipment. The study was approved by the Human
Research Protection Program at KU (STUDY00147804). On approval, the study was advertised
in several public places (notice boards across the KU campus, public libraries, grocery stores,
etc.) around Lawrence, Kansas. Paper advertisements in the local newspapers, as well as social
media announcements were also used to recruit local participants. Thirty-nine participants were
invited to participate in the study (21 males and 18 females). The participants’ recruitment was
carried out in three age groups, 18–30, 31–50, and 51–65 years, based on the availability and
willingness to participate and their satisfactory completion of the prescreening questionnaire.
The selection criteria included age, possession of a valid U.S. driver’s license, having at least
3 years of driving experience with an annual mileage no less than 5,000 miles, and having good
health (free from heart conditions, seizures, motion sickness, hearing disability, pregnancy, etc.).
Table 6 summarizes the characteristics of the study participants.

Table 6.   Descriptive statistics of participants.

Age Male Female

Groups No. of No. of Mean SD Mean SD
(Years) Males Females (Age) (Age) (Age) (Age)
18–30 9 9 25.33 2.6 22.67 3.46
31–50 6 5 37.17 6.77 41.2 8.29
51–65 6 4 60.33 3.39 60 2.16
Note: SD = standard deviation.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   19

The research team developed driving scenarios using the driving simulator, designed to under-
stand operational and safety implications on different arterial weaves from a driver perspective.
These scenarios were tested during a pilot implementation. In the scenarios, the research team
replicated the following parameters to design the simulator network: lane configuration (two
through lanes per direction and three through lanes per direction), arterial length (short length
of 900 ft and long length of 1,500 ft), level of service (LOS) (good LOS A/B and poor LOS D/E),
and weaving maneuver type (left to right and right to left). The scenario configurations (main
scenario) are summarized in Table 7.
The combination of all these parameters resulted in 16 segments that were included in the
driving simulator route. A schematic of the roadway network, the weaving maneuvers, and their
sequence during the drive are shown in Figure 6.
In total, the main scenario ran for 35 min and included three 5-min breaks to reduce motion
sickness due to continuous weaving. After completing the main scenario, participants were asked
to fill out the post-completion survey, which elicited information regarding the overall experience.
In addition to the main scenario, participants also drove a second scenario with the same geometry
but without any traffic, to use as a baseline (base scenario). Half of the participants in each age
group drove the base scenario first, to eliminate any order bias. The base scenario ran for 20 min,
including three 2-min breaks. After the successful completion of the base scenario, participants
were asked to fill out the NASA-TLX, PANAS, PES, MDSI, and RDHS questionnaires. After the
successful completion of all these steps, participants were given a $50 gift card as compensation
for their time.

Subjective Measures
The research team collected several subjective measures. The prescreening survey solicited
primarily demographic information (e.g., age, gender, race), information on driving experience,
and medical conditions. The post-completion survey included questions related to the overall
experience during the data collection. Some questions were “Did you feel you had enough space

Table 7.   Main scenario configurations for the driving simulator.

Lane
Configuration Site Length (ft) Level of Service Weaving Maneuver No. in Figure 6
2×2a Short (900) Poor (LOS D) Right to left 1
2×2 Short (900) Poor (LOS D) Left to right 2
2×2 Short (900) Good (LOS A/B) Right to left 3
2×2 Short (900) Good (LOS A/B) Left to right 4
2×2 Long (1,500) Poor (LOS D) Right to left 9
2×2 Long (1,500) Poor (LOS D) Left to right 10
2×2 Long (1,500) Good (LOS A/B) Right to left 11
2×2 Long (1,500) Good (LOS A/B) Left to right 12
3×3b Short (900) Poor (LOS D) Right to left 5
3×3 Short (900) Poor (LOS D) Left to right 6
3×3 Short (900) Good (LOS A/B) Right to left 7
3×3 Short (900) Good (LOS A/B) Left to right 8
3×3 Long (1,500) Poor (LOS D) Right to left 13
3×3 Long (1,500) Poor (LOS D) Left to right 14
3×3 Long (1,500) Good (LOS A/B) Right to left 15
3×3 Long (1,500) Good (LOS A/B) Left to right 16
a2 lanes per direction.
b3 lanes per direction.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

20   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 6.   Schematic of weaving maneuvers modeled in driving simulator.

to change lanes to complete the weaving maneuvers?”; “How helpful were the scheduled breaks?”;
and “Do you think the same maneuvers would be more difficult at night?”
Several questionnaires available in the literature were used to categorize drivers based on their
self-reported driving style. The MDSI and the RDHS have been used in past driving simulator
studies to determine driving style scales (Farah et al. 2007, Rossi et al. 2012, Taubman-Ben-Ari
et al. 2017, Van Huysduynen et al. 2018, Contractor et al. 2020, Freuli et al. 2020, Sun et al. 2021).
The MDSI is a 44-item self-report questionnaire developed by Taubman-Ben-Ari et al. (2004). The
scale comprises eight coherent factors that are distributed over four global driving styles. These are
reckless and careless driving style (Factor 3: risky driving style and Factor 5: high-velocity driving
style), anxious driving style (Factor 1: dissociative driving style, Factor 2: anxious driving style, and
Factor 6: distress-reduction driving style), angry and hostile driving style (Factor 4: angry driving
style), and careful and patient driving style (Factor 7: patient driving style and Factor 8: careful
driving style). The reckless and careless driving style refers to speeding and intoxicated driving; the
anxious driving style reflects alertness and tension; the angry and hostile driving style refers to rage
and hostility while driving; and the patient and careful driving style refers to well-adjusted driving
behaviors (Taubman-Ben-Ari and Skvirsky 2016). For all eight factors and four driving styles, the
subscale rating can be obtained by taking the average of all the items for each subscale.
The RDHS is a 14-item self-report questionnaire developed by Taubman-Ben-Ari et al. (2004)
to assess reckless driving that might endanger surrounding road users (e.g., driving through a
red light, not using seatbelts, driving at a higher speed than allowed, violating a stop sign). The
scale is rated on a six-point scale (from 1, “never,” to 6, “always”). The rating then can be obtained
averaging all 14 responses.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   21

In addition, researchers used mood and personality as well as empathy and moral decision-
making questionnaires available in the literature. The PANAS is a 20-item self-report measure to
assess positive and negative affect (Watson et al. 1988). The negative affect indicates distress and
repulsive experience (e.g., upset or guilty) with the environment, and positive affect indicates
pleasurable experience (e.g., enthusiastic or determined) with the environment. Each of these
subscales has 10 items, constructed on a five-point Likert scale: 1 = “Very slightly or not at all” to
5 = “Extremely.” The subscale rating can be obtained by taking the average of all items for each sub-
scale. For empathy and moral decision-making, researchers used the PES, which refers to a stable
and pervasive sense that one deserves more and is entitled to more than others (Campbell et al.
2004). The scale is a nine-item self-report measure, where the participant rates each statement
(e.g., “Great things should come to me,” “I feel entitled to more of everything”) from 1 (strongly
disagree) to 7 (strongly agree). The mean PES score then can be obtained for each participant.
A high PES score indicates a more aggressive response to any situation than a low PES score.
Finally, the NASA-TLX was used to obtain the participants’ subjective workload. The NASA-TLX
calculates the workload experienced by each participant during the task, as a weighted average
of six subscales: mental demand, physical demand, temporal demand, performance, effort, and
frustration. Participants rated a 20-point scale ranging from “very low” to “very high” and then
assigned a weight, from 0 to 5, to a pair of subscales shown on flash cards (six subscales resulting
in 15 possible pairwise combinations) (Stojmenova and Sodnik 2015). The questionnaire is usually
administered after the completion of a task or several tasks. In this study, two tasks were performed
for the main scenario and base scenario. These were “right-to-left weaving” and “left-to-right weav-
ing.” The questionnaire was administered after the completion of each simulator run for both tasks.
A summary of the participant drivers’ demographics and the subjective measures of MDSI,
RDHS, PANAS, and PES are shown in Appendix F.

Driving Parameters
The driving simulator outputs an abundance of detailed data. Some of the data that were
extracted from each simulation run include the following:
• Weaving vehicle trajectories, including speeds and accelerations;
• Average vehicle speed and acceleration during the weaving maneuver execution (from first to
last lane change performed);
• Lead and lag gaps for each lane change performed by the weaving vehicle;
• Position of the lane-changing vehicle with respect to the start of the segment; and
• Average TTC and vehicle position when TTC is less than 5 s.

Physiological Measures
The KU driving simulator laboratory has a separate eye-tracking machine, which is used to
collect physiological measures. A FOVIO-FX3 eye tracker is mounted on the simulator chas-
sis (EyeTracking 2019). The machine collects workload data through task-evoked pupillary
response at 1 Hz. Gaze points are collected at 60 Hz. Several other measures, like pupil diameter,
blink rate, gaze fixations, and independent component analysis (ICA) data, were also collected.
The ICA is also a proxy for mental workload that has shown good correlation with the subjective
measures (NASA-TLX).

Microsimulation Data
Given the wide variety of potential arterial weaving configurations and demand scenarios, as well
as the cost of collecting the necessary field data, microsimulation experiments were necessary to fill
gaps in the field data collection. To accomplish this, the research team reviewed three commercially

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

22   Operational Performance and Safety Effects of Arterial Weaving Sections

available microsimulators (AIMSUN, TransModeler, and PTV VISSIM) to select the one most suit-
able for this project. All three packages provide the capability to replicate operations along arte-
rial segments and the functionality of various types of traffic control. The research team selected
TransModeler based on its flexibility in the calibration of driver behavior parameters and the
provision of a metric to record the number of failed weaving maneuvers.
A subset of five of the field data collection sites (Table 8) was selected for simulation and addi-
tional scenario generation. The sites were selected to ensure variability in geometry and demand
conditions.
These sites were modeled using TransModeler version 5.0. The field characteristics were mod-
eled based on drone video recorded during the field data collection and the most recent satellite
images available on Google Earth Pro. Every simulation scenario was run for 15 min with an
initialization time of 5 min. Appendix G provides an overview of the simulation assumptions,
calibration, and outputs from the simulation.
After calibration, a wide range of scenarios was implemented to assess the impact of various
independent variables on weaving operations along arterials. Five main attributes were selected
to establish the scenarios: weaving demand, nonweaving demand, number of lanes, presence of
turning bays, and signal timing. NC6 was initially selected as a pilot to evaluate the impact of
various parameters on operations, resulting in 26 scenarios. For the remaining four sites, the
research team developed 16 scenarios for each site. Each scenario was run 12 times to account for
stochasticity effects. Table 9 provides a summary of the total number of runs and corresponding
number of scenarios created for each site.
On completion of the simulations, over 413,000 trajectories were generated; these trajectories
were used for the explanatory analysis and the development of the proposed model.

Databases for Safety Analysis

Three types of safety data were collected to identify the safety effects of varying geometric,
volume, and traffic conditions in weaving sections on arterials:
• Crash data for the study sites,
• Conflict data for the study sites, and
• Conflict data from the driving simulator.

Crash Data for the Study Sites

To conduct the crash analysis for the study sites, crash data and traffic volume data were
obtained either directly from the respective state DOT or available online portals for study sites
in California, North Carolina, Nebraska, Ohio, and Texas. After reviewing the completeness of

Table 8.   Simulated sites.

Site Latitude Longitude Time of the Day

CA12 38°44'38.27"N 121°16'4.34"W 03:30 p.m. to 03:45 p.m.

NC1 35°18'44.12"N 80°42'30.98"W 11:45 a.m. to 12:00 p.m.

NC6 35°14'46.26"N 80°53'40.06"W 04:45 p.m. to 05:00 p.m.

NE1 41°13'16.69"N 96°2'35.75"W 04:00 p.m. to 04:15 p.m.

TX5 30°32'4.49"N 97°41'26.10"W 11:45 a.m. to 12:00 p.m.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   23

Table 9.   Total number of scenarios for microsimulation data.

Scenarios
Weaving Nonweaving No. of Turning Signal Simulations
Site Demand Demand Lanes Bay Timing Scenarios (Scenarios × 12)
CA12 4 4 3 3 3 16 192
NC1 4 4 3 3 3 16 192
NC6 5 5 3 3 9 26 312
NE1 4 4 3 3 3 16 192
TX5 4 4 3 3 3 16 192
Total Simulations 1,080

the crash data, sufficient information was available to assign crashes to study sites in California,
North Carolina, Ohio, and Texas for analysis purposes. Sufficient information was not available
to assign crashes to the Nebraska study sites. Therefore, the descriptive statistics provided below
for the crash database address study sites in California, North Carolina, Ohio, and Texas only;
similarly, the crash analysis focused on study sites in these four states.

Data Processing
The research team used ArcGIS tools to obtain crashes that fell within the area of interest.
The area of interest in this case was the direction of travel of the study locations where weav-
ing is likely to take place. This is between the O-D terminals of the arterial weaving section in
the direction of travel of interest. First, all crashes were mapped to their respective locations
in the vicinity of the study sites using their coordinates (see Figure 7). Then, only crashes that were
within the boundaries of the study sites and in the direction of travel of interest were identified
and assigned to the arterial weaving section for further processing. Furthermore, all intersection
and intersection-related crashes were removed from the database for purposes of the analysis.

Figure 7.   Mapped crashes at a study site in Texas.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

24   Operational Performance and Safety Effects of Arterial Weaving Sections

Descriptive Summary of Crashes

The crash database included up to 7 years of crash data for the study sites as follows:
• California study sites: 2016–2021,
• North Carolina study sites: 2015–2019,
• Ohio study sites: 2019–2021, and
• Texas study sites: 2015–2021.
The descriptive summary includes the temporal factors, collision type, and crash severity.
Table 10 presents the total number of crashes and the average number of crashes per year per site.
On average, study sites in North Carolina and Ohio had approximately eight crashes per year,
while sites in California and Texas had between one and two crashes per year.
Figure 8 presents the yearly distribution of crashes by state. It can be observed that North Carolina
has the highest number of crashes for all years where crash data are available, except for 2019.
Figure 9 presents the hourly distribution of crashes. As expected, two peaks, morning and
evening, are observed in the figure. The two peaks are normally the time when there are more
vehicles traversing the locations of interest; thus, there is an increased likelihood of crashes.
Figure 10 and Figure 11 present the collision type and injury severity, respectively, of the crashes.
Most of these crashes involved rear-end collisions followed by sideswipe collisions. This is likely due
to the nature of vehicle maneuvers in weaving sections. Some vehicles are changing lanes, while

Table 10.   Number of crashes by state and average number of crashes per year
per site.

Average Number of
Years of Total Number of Crashes per Year per Site
State No. of Study Sites Crash Data Crashes (Crashes/Year/Site)
CA 4 2016–2021 39 1.63
NC 4 2015–2019 166 8.30
OH 3 2019–2021 73 8.11
TX 4 2015–2021 53 1.89

Figure 8.   Yearly distribution of crashes.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   25

Figure 9.   Hourly distribution of crashes.

others are preparing to stop at the intersection. Regarding the injury severity, most crashes did not
result in any injuries. This may be attributed to lower speeds of vehicles within weaving sections.

Relationship Between Crash Rate and Weave Length

In this study, the site length is the length of the entire section from the gore of the initial access
point to the downstream stop bar of the subject approach. For the safety analysis, the analysis uses
the weave length, defined as the shortest distance from the gore of the initial access point to either
the beginning of the closest downstream left-turn or right-turn lane or the closest downstream
access point along the segment. Figure 12 illustrates the relationship between crash rate and weave
length. Crash rates, measured in million vehicle miles traveled (MVMT), were computed using
number of crashes, time, AADT, and site length. Table 11 presents the corresponding data used
to calculate the crash rates for the study sites. According to Figure 12, longer weave lengths are

Figure 10.   Collision types.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

26   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 11.   Crash severity.

Figure 12.   Crash rate (total crashes) and weave length.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   27

Table 11.   Crash rate by study site.

Site Weave Crash Frequency for the Crash Rate

Study Period (Crashes/MVMT)
Length Length AADT
State Site (ft) (ft) (veh/day) Total Rear-End Sideswipe Total Rear-End Sideswipe
CA4 800 600 21,807 5 3 1 0.69 0.41 0.14
CA5 435 60 36,015 13 5 6 2.00 0.77 0.92
CA12 735 445 6,000 10 6 2 5.47 3.28 1.09
CA CA13 1,115 400 11,000 11 8 2 2.16 1.57 0.39
NC1 1,250 1,130 8,500 29 21 3 7.90 5.72 0.82
NC6 710 275 24,500 29 13 10 4.82 2.16 1.66
NC7 975 585 9,000 71 34 19 23.41 11.21 6.26
NC NC9 955 665 7,500 37 17 11 14.95 6.87 4.44
OH1 1,200 350 31,522 30 14 0 3.82 1.78 0.00
OH2 825 525 19,884 23 9 0 6.76 2.65 0.00
OH OH3 1,140 590 21,597 20 6 1 3.92 1.18 0.20
TX2 560 300 8,000 4 2 0 1.85 0.92 0.00
TX3 1,000 680 9,500 9 3 4 1.96 0.65 0.87
TX5 535 300 10,500 25 17 5 9.20 6.25 1.84
TX TX6 540 410 7,500 15 4 4 7.65 2.04 2.04

associated with lower crash rates for total crashes (i.e., all crash types combined) in Texas, as indi-
cated by the slopes of the trend lines. On the other hand, longer weave lengths in Ohio are associ-
ated with an increase in crash rates, while in California and North Carolina, no significant change
of crash rate is observed as the weave length increases.
The crash rates may vary by crash type. Thus, combining all crashes together might not portray
the actual behaviors of individual crash types. Figure 13 and Figure 14 present relationships between
rear-end and sideswipe, same-direction crash rates and weave length. It is believed that these two
collision types are most likely directly related to the turbulence that occurs within a weaving area.
Comparing state by state, it can be observed that there are inconsistent patterns in terms of the
changes of crash rates for collision types of interest as the weave length increases. For instance,
in Texas an increase in weave length is associated with a decrease in rear-end crashes; in North
Carolina, an increase in weave length is associated with an increase in rear-end crashes. In California
and Ohio, there does not appear to be any significant change in rear-end crashes as the weave length
increases. Meanwhile, for study sites in California and North Carolina, an increase in weave length
is associated with a decrease of sideswipe crashes, while for study sites in Ohio and Texas, there
does not appear to be any significant change in sideswipe crashes as the weave length increases.

Conflict Data for the Study Sites

Motor vehicle conflicts are an indirect measurement of safety and provide an opportunity to
examine near-crashes as an indirect measure of crash potential at a site. Using a combination
of the DFSV, manual observations of the drone videos, and postprocessing of the vehicle trajecto-
ries, conflict data were obtained from the study sites from the drone video. The primary conflict-
based safety surrogate measures collected included HB, TTC, and PET, defined as follows:
• HB (heavy braking)—This is a detection of rapid deceleration, used to detect potential collisions.
• TTC (time-to-collision)—This is a time measurement of how many seconds two vehicles
would collide in if they were moving at a constant speed and direction.
• PET (post-encroachment time)—This is a time measurement of the difference between the
time a vehicle enters a conflict point until the time another vehicle arrives to this point.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

28   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 13.   Crash rate (rear-end) and weave length.

Figure 14.   Crash rate (sideswipe, same direction) and weave length.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   29

Data Processing
Information was gathered for these three types of conflicts for the study sites using the drone
video. The video data extraction yields two sets of data: traffic region data and individual event data.
The traffic region data contain the information related to the region name, track ID and type, entry
time and exit time, among others. In addition to the track ID and type, the HB event data had the
braking time, conflict image ID, x and y coordinates, and the tangential acceleration. The TTC on
the other hand had the first and second track ID and type, time exposed, time integrated, and TTC.
Finally, the PET events data constitute the PET in addition to the first and second track ID and type
and x and y coordinates that are available in the previous files.
The research team used the traffic region data to extract only the portion of data of interest in
the HB, TTC, and PET event data. For an observation in the event data to be considered for this
study, it should meet the following two criteria:
1. The track ID and type should match with its respective traffic region data.
2. The event time should be within the entry and exit time in the traffic region data.
For all event type data, the research team had a set of information to extract as described in
the following list.
1. HB: The research team considered HB as the one with deceleration greater than or equal to
8.80 m/s2. As a point of reference, studies show that the maximum deceleration of a vehicle during
an emergency braking maneuver or event is around 9.8 m/s2. To provide flexibility in HB calcu-
lations, a peak deceleration of 15 m/s2 was considered the maximum valid reported value. An HB
event with a deceleration value greater than 15 m/s2 was considered an invalid event and was not
included in the HB analysis. The events and location of interest for HB data are the beginning,
middle, and ending coordinates of HB and the associated decelerations. Additionally, the team
extracted the coordinates of the peak deceleration and their associated deceleration.
2. TTC: The research team considered 1.5 s as a critical TTC. Thus, any value greater than 1.5 s was
discarded. The team was interested in the coordinates for the beginning, middle, and end points
and their associated TTCs. Also, the minimum TTC for each observation and the associated
coordinates were extracted. Additionally, the team extracted the time exposed and time integrated.
3. PET: The team defined critical PET as 5 s. In addition to the PET, the research team extracted
the coordinates associated with the critical PET for each event.
Furthermore, for an event to be considered for further analysis, it should have a recorded distance
between the starting and ending points. This is to say, if the extracted starting point is the same as
the end point, then that observation was considered invalid and was discarded from further analysis.

Descriptive Summary of Conflict Data

This section presents a descriptive summary of the conflict data.

HB
Table 12 presents the overall distribution of peak decelerations observed at the study sites, by state,
and Table 13 presents the distribution of peak decelerations by study sites. Table 12 shows consistent
distribution of values across states, with the mean decelerations ranging from 10.88 to 11.00 m/s2
and the standard deviations ranging from 1.55 to 1.65 m/s2. Table 13 shows slightly more variation
in peak decelerations across sites, which could be expected. Table 14 shows the number of HB events
per total volume, weaving volume, and vehicle miles traveled (VMT) for each study site.
Figure 15 presents the relationship between weaving length and number of HB events across
the sites within each state. It can be observed that the relationship between weaving length and
number of HB events is inconsistent. In California, North Carolina, and Texas, the increased
weaving length is associated with an increase in HB rates. On the other hand, the increased weaving
length for sites in Ohio and Nebraska is associated with a decrease in HB rates.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Table 12.   Distribution of HB (peak deceleration) events by state.

Peak Deceleration (m/s2)

State Count (HB) Duration Mean Median Min. Max. St. Dev.
CA 2,267 175 10.88 10.44 8.00 14.99 1.60
NC 3,757 225 11.00 10.58 8.81 14.99 1.63
NE 986 180 10.95 10.54 8.80 14.98 1.60
OH 1,668 210 10.96 10.62 8.81 14.98 1.55

Table 13.   Distribution of HB (peak deceleration) events by study site.

Peak Deceleration (m/s2)

State Site Count (HB) Duration Mean Median Min. Max. St. Dev.
CA4 49 30 10.62 10.35 8.00 14.91 1.78
CA5 10 25 9.23 8.93 8.03 13.71 1.68
CA12 1,414 60 11.03 10.67 8.03 14.99 1.63
CA CA13 794 60 10.65 10.24 8.80 14.97 1.50
NC1 1,289 60 11.41 11.09 8.82 14.99 1.71
NC6 723 60 10.55 10.12 8.81 14.99 1.43
NC7 1,495 60 10.82 10.40 8.81 14.98 1.55
NC NC9 250 45 11.33 10.90 8.84 14.99 1.76
NE1 151 60 11.01 10.73 8.80 14.86 1.56
NE4 84 60 10.51 9.99 8.82 14.98 1.45
NE NE6 751 60 10.99 10.56 8.82 14.95 1.62
OH1 1,195 75 11.04 10.70 8.81 14.96 1.57
OH2 147 30 10.76 10.47 8.86 14.79 1.42
OH3 – – – – – – –
OH4 221 45 10.73 10.28 8.81 14.98 1.50
OH OH5 105 60 10.82 10.50 8.83 14.86 1.52
TX2 420 60 10.53 10.06 8.81 14.73 1.51
TX3 997 30 11.23 10.88 8.80 14.99 1.68
TX5 666 60 11.05 10.64 8.81 14.98 1.67
TX TX6 360 45 10.80 10.46 8.81 15.00 1.56

Table 14.   HB events per total volume, HB events per weaving volume, and HB
events per VMT by study sites.
HB Events HB Events
per per
Site Weave Total Weaving Total Weaving HB Events per
Length Length Count Volume Volume Volume Volume VMT
State Site (ft) (ft) (HB) (TV) (WV) (HB/TV) (HB/WV) (HB/VMT)
CA4 800 600 49 304 202 0.16 0.24 1.06
CA5 435 60 10 888 177 0.01 0.06 0.14
CA12 735 445 1,414 898 495 1.57 2.86 11.31
CA CA13 1,115 400 794 1,672 823 0.47 0.96 2.25
NC1 1,250 1,130 1,289 613 292 2.10 4.41 8.88
NC6 710 275 723 1,022 545 0.71 1.33 5.26
NC7 975 585 1,495 1,041 564 1.44 2.65 7.78
NC NC9 955 665 250 894 197 0.28 1.27 1.55
NE1 815 650 151 1,618 664 0.09 0.23 0.60
NE4 980 710 84 1,427 658 0.06 0.13 0.32
NE NE6 1,300 900 751 – – – – –
OH1 1,200 350 1,195 1,080 57 1.11 20.96 4.87
OH2 825 525 147 379 144 0.39 1.02 2.48
OH3 1,140 590 0 – – – – –
OH4 750 330 221 1,134 126 0.19 1.75 1.37
OH OH5 1,015 615 105 816 61 0.13 1.72 0.67
TX2 560 300 420 878 65 0.48 6.46 4.51
TX3 1,000 680 997 506 123 1.97 8.11 10.40
TX5 535 300 666 1,418 793 0.47 0.84 4.64
TX TX6 540 410 360 786 332 0.46 1.08 4.48

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   31

Figure 15.   Weave length and HB rates.

TTC
Like it did with the peak deceleration, the research team explored the minimum TTC at the
state and site levels. Table 15 and Table 16 present the distributions of minimum TTC by state and
study site. According to the results in Table 15, there exists variability across states whereby sites in
Ohio have the shortest average TTC, while those in North Carolina have the longest average TTC.
Considering the variability of the minimum TTC, the standard deviation varies from 0.431 s in
North Carolina to 0.596 s in Nebraska. According to the results in Table 16, there are observed varia-
tions of the minimum TTC across study sites within the state as indicated by the mean and median
values, as well as the standard deviation values.
Figure 16 presents the relationship between the TTC rate and weave length. The results in Fig-
ure 16 suggest that the increase in the weave length is associated with the increase in TTC rates for
California and North Carolina. On the other hand, the sites in Ohio, Nebraska, and North Carolina
did not show a significant change.

Table 15.   Overall distribution of minimum TTC.

Duration Mean Median Min. Max.
State Count (mins) (ms) (ms) (ms) (ms) St. Dev. (ms)
CA 1,681 175 643.76 640 40 1,497 516.9
NC 1,928 225 783.98 760 40 1,499 431.8
NE 56 180 689.82 706.5 33 1,468 595.9
OH 272 210 582.78 449.5 33 1,499 563.9
TX 428 195 759.93 880 40 1,480 556.9

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

32   Operational Performance and Safety Effects of Arterial Weaving Sections

Table 16.   Distribution of minimum TTC by study site.

Duration Mean Median Min. Max.
State Site Count (mins) (ms) (ms) (ms) (ms) St. Dev. (ms)
CA4 762 30 768.7 783.5 40 1,497 439.8
CA5 57 25 767.3 1,000.0 40 1,480 570.9
CA12 742 60 461.3 40.0 40 1,484 523.3
CA CA13 120 60 919.6 1,121.5 40 1,493 533.6
NC1 1,343 60 727.7 680.0 40 1,498 367.6
NC6 80 60 713.9 774.5 40 1,496 547.1
NC7 118 60 634.8 460.0 40 1,480 588.1
NC NC9 387 45 1,039.4 1,201.0 40 1,499 458.7
NE1 17 60 969.5 1,201.0 33 1,468 503.9
NE4 5 60 566.8 33.0 33 1,401 731.3
NE NE6 34 60 568.1 316.5 33 1,436 588.2
OH1 149 75 531.6 266.0 33 1,499 554.9
OH2 9 30 903.2 983.0 33 1,499 449.0
OH3 – – – – – – –
OH4 26 45 972.5 1,099.5 33 1,499 466.3
OH OH5 88 60 521.4 83.0 33 1,499 566.9
TX2 51 60 669.8 800.0 40 1,440 568.8
TX3 125 30 697.5 760.0 40 1,480 522.4
TX5 173 60 822.2 1,000.0 40 1,480 546.0
TX TX6 79 45 780.4 1,000.0 40 1,480 615.4
Notes: mins = minutes, ms = milliseconds.

Figure 16.   Weave length and TTC rates.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   33

PET
Table 17 and Table 18 present the distribution of PET by state and study site. Table 17 shows
that most of the PET data originated from California, followed by Texas, while other states have
a relatively small number of observations. On average, sites in North Carolina have the longest
PETs, while sites in Nebraska have the shortest. Further, sites in Ohio depicted more variability,
as indicated by the largest standard deviation value. Table 18 shows variation in PET across sites
within the state. For instance, the mean PET in California varies between 2043.2 milliseconds to
3387.1 milliseconds. A similar pattern is observed in other states.
Figure 17 presents the relationship between the weave length and PET rates. It can be observed
that for Texas, the rate of PET conflicts increases as the weave length increases, while for the
sites in California, the rate of PET conflicts decreases as the weave length increases. Further,
no significant change in the rate of PET conflicts relative to weave length is observed for
sites in the remaining states.

Conflict Data from Driving Simulator

This section focuses on safety data gathered from the driving simulator. Table 19 shows the
configurations of the driving simulator and the traffic operational conditions simulated, while also
summarizing the frequency of TTC conflicts for each condition and the average TTC duration.
Thirty-nine participants (21 males and 18 females) participated in the driving simulator study.

Table 17.   Overall distribution of minimum PET.

Duration Mean Median Min. Max.
State Count (mins) (ms) (ms) (ms) (ms) St. Dev. (ms)
CA 2,102 175 2,384.7 2,320 40 4,960 1,450.6
NC 44 225 3,050.9 3,400 480 4,960 1,296.7
NE 59 180 2,356.1 2,302 100 4,771 1,381.3
OH 11 210 2,399.7 2,033 33 4,766 1,572.3
TX 395 195 2,539.6 2,560 40 4,960 1,442.8

Table 18.   Distribution of minimum PET by study site.

Duration Mean Median Min. Max. St. Dev.
State Site Count (mins) (ms) (ms) (ms) (ms) (ms)
CA4 777 30 2,043.2 1,760 40 4,960 1,459.4
CA5 1,145 25 2,547.8 2,600 40 4,960 1,397.6
CA12 149 60 2,703.1 3,000 80 4,960 1,452.3
CA CA13 31 60 3,387.1 4,000 640 4,920 1,343.0
NC1 4 60 3,660.0 4,040 2,080 4,480 1,091.3
NC6 29 60 2,867.6 3,160 480 4,960 1,347.3
NC7 7 60 3,514.3 4,000 880 4,360 1,238.4
NC NC9 4 45 2,960.0 2,840 1,560 4,600 1,252.6
NE1 41 60 2,213.9 2,102 133 4,771 1,345.2
NE4 15 60 2,546.7 2,936 100 4,571 1,512.9
NE NE6 3 60 3,347.3 2,869 2,702 4,471 976.7
OH1 1 75 4,766.0 4,766 4,766 4,766 –
OH2 7 30 1,995.0 1,533 33 4,200 1,442.2
OH3 0 – – – – – –
OH4 3 45 2,555.3 2,033 1,233 4,400 1,646.8
OH OH5 1 60 2,280.0 2,280 2,280 2,280 –
TX2 1 60 2,280.0 2,280 2,280 2,280 –
TX3 118 30 2,554.6 2,820 80 4,960 1,452.1
TX5 15 60 3,610.7 4,120 1,640 4,920 1,184.6
TX TX6 261 45 2,472.3 2,480 40 4,960 1,434.9

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

34   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 17.   Weave length and PET rates.

Table 19.   Driving simulator—summary of TTC conflict data.

Frequency
Lane Weaving No. in of TTC Avg. Duration
Configuration Site Length (ft) LOS Maneuver Figure 6 Conflicts of TTC (ms)
2×2a Short (900) Poor (LOS D) Right to left 1 34 3,488
2×2 Short (900) Poor (LOS D) Left to right 2 35 3,128
2×2 Short (900) Good (LOS A/B) Right to left 3 20 3,826
2×2 Short (900) Good (LOS A/B) Left to right 4 9 4,112

2×2 Long (1,500) Poor (LOS D) Right to left 9 27 3,550

2×2 Long (1,500) Poor (LOS D) Left to right 10 2 3,686
2×2 Long (1,500) Good (LOS A/B) Right to left 11 8 3,933
2×2 Long (1,500) Good (LOS A/B) Left to right 12 4 3,776

3×3b Short (900) Poor (LOS D) Right to left 5 18 3,489

3×3 Short (900) Poor (LOS D) Left to right 6 37 3,039
3×3 Short (900) Good (LOS A/B) Right to left 7 11 4,136
3×3 Short (900) Good (LOS A/B) Left to right 8 15 3,842

3×3 Long (1,500) Poor (LOS D) Right to left 13 12 3,584

3×3 Long (1,500) Poor (LOS D) Left to right 14 7 3,193
3×3 Long (1,500) Good (LOS A/B) Right to left 15 4 3,000
3×3 Long (1,500) Good (LOS A/B) Left to right 16 3 3,653
a Two lanes per direction.
b Three lanes per direction.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   35

Data Analysis
This section presents the relationships between performance characteristics and specific prevail­
ing traffic, control, and geometric characteristics of arterial weaving segments based on the field and
microsimulation data. It also summarizes the findings from the driving simulator experiments. The
last section presents the safety analysis results.

Relationships Between Weaving Characteristics and Traffic

Performance Based on Field and Microsimulation Data
This section provides an overview of the relationships between performance metrics like speed
and number of stops versus various arterial weaving segment characteristics. To understand these
relationships the research team conducted an evaluation using data from five sites: CA12, NC1,
NC6, NE1, and TX5. As indicated earlier, NC6 includes a total of 26 simulation scenarios; the other
four sites have 16 scenarios each. These simulation scenarios evaluate changes in weaving demand,
nonweaving demand, number of lanes, turning-bay characteristics, and signal timing. To illustrate
those relationships, this section includes the graphs generated for site NC6; the exploratory analysis
graphs for the remaining sites are included in Appendix H.
Figure 18 illustrates the average speed of vehicles going through the downstream intersec-
tion of a weaving segment versus part a, the nonweaving demand; part b, the weaving demand;
and part c, the total demand arriving to the weaving segment. This figure includes only the
simulated data. The graphs show that with the increase in demand the speed drops and generally
seems to flatten for higher demands. The data set used for these graphs includes all simulation
data for NC6.
The variability shown in Figure 18 is due to the wide range of scenarios developed for NC6.
Never­theless, the graph shows that, as expected, by increasing the demand per lane (non-
weaving, weaving, or total), the average speed of vehicles going through in the downstream
intersection decreases.
Figure 19 plots the average speed of vehicles traveling through the downstream intersection
of the weaving segment versus part a, the nonweaving demand; part b, the weaving demand;
and part c, the total demand arriving to the weaving segment. This set of figures plots all the
simulation data (blue) and the field data obtained by drone and processed using the DFSV (red).
As shown, the trends for the simulation data set and the field data are similar, however, the
field data are significantly fewer (26 data points), resulting in significant variability. The simula-
tion data provide a clearer trend for a broader set of conditions.
In Figure 20 researchers plotted the average speed of vehicles going through the downstream
intersection versus the potential conflict opportunity (PCO) for the field and simulation data. A PCO
is defined as a potential vehicle encounter between a weaving vehicle and a conflicting through
vehicle. Graph a on the left plots the complete simulation data set, and graph b on the right presents
simulation (blue) and field data (red). As shown, these graphs show similar trends as those shown in
Figure 19. For higher PCOs, speeds begin to taper off to a relatively constant rate. Neither simulation
scenarios nor collected video data include oversaturated conditions.
Figure 21 presents the relationship between the average speed of vehicles traveling through
the downstream intersection of the segment versus total demand per lane in the segment, as
well as the relationship between the average speed of vehicles arriving during the green (green
dots and line) and vehicles arriving during the red (red dots and line). The graph on the left
(a) includes the simulation data for NC6, and the graph on the right (b) provides only the field
data for NC6.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

36   Operational Performance and Safety Effects of Arterial Weaving Sections

(a) (b)

(c)

Figure 18.   Relationships between speed and demand per lane for NC6 (simulated data only).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   37

(a) (b)

(c)

Figure 19.   Relationships between speed and demand for NC6 (field and simulated data).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

38   Operational Performance and Safety Effects of Arterial Weaving Sections

(a) (b)

Figure 20.   Relationships between speed and PCO for NC6 (field and simulated data).

(a) (b)

Figure 21.   Relationships between speed and total demand for vehicles arriving during the green and vehicles
arriving during the red for NC6.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   39

As shown, there is a clear change in pattern for vehicles arriving during the green and during
the red. For vehicles arriving during the green, when demand increases, speeds decreases. The
same trend can be seen for vehicles arriving during the red, with lower overall speeds. Again,
there are fewer datapoints for the field data.
Figure 22 presents the number of stops in the weaving section within one signal cycle versus
the weaving demand, nonweaving demand, and total demand in the facility. This figure shows
simulated data only. As shown, the increase in demand tends to increase the number of stops in the
weaving facility; however, the data show very high variability, as there are several other important
factors that affect the number of stops (most importantly, signalization).
Figure 23 provides the number of stops versus PCO for the simulation data set of NC6. Again,
there is a general increase in the number of stops as PCO increases.
In the microsimulation, another crucial parameter collected was the count of vehicles unable
to execute the weaving maneuver. This parameter is likely to be affected by traffic flows (weav-
ing and nonweaving) and the length of the weaving segment. This parameter did not vary
significantly for the NC6 site. Figure 24 presents the relationship of this parameter to demand
and PCO.
Because of the large number of scenarios created in simulation for all the sites, it is difficult to
see patterns when combining the data for all sites. For example, combining data from all the sites
produces graphs such as the ones shown in Figure 25. As shown, each demand level incorporates
a variety of design and control characteristics, which results in very high variability. Still, all three
graphs show that the speed is reduced with increasing demand, as expected.

(a) (b)

Figure 22.   Relationships between the number of stops and demand for NC6 (simulated data).
 (continued on next page)

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

40   Operational Performance and Safety Effects of Arterial Weaving Sections

(c)

Figure 22. (Continued).

Figure 23.   Relationships between number of stops and PCO for NC6 (simulated data and field data).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   41

Figure 24.   Relationships between missed turns versus demand/PCO for NC6 (simulated data).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

42 Operational Performance and Safety Effects of Arterial Weaving Sections

(a) (b)

(c) (d)

Figure 25. Relationships between speed and demand/PCO for all sites (simulated data).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   43

Relationships Between Weaving Characteristics and Traffic

Performance Based on Driver Simulator Experiments
As part of the driving simulator data analysis, the following data were considered and aggregated
to identify significant relationships between model parameters for operational and safety analyses:
• Weaving vehicle trajectories, including speeds and accelerations;
• Average vehicle speed and acceleration during the weaving maneuver execution (from first to
last lane change performed);
• Lead and lag gaps for each lane change performed by the weaving vehicle;
• Position of the lane-changing vehicle with respect to the start of the segment;
• Average TTC and vehicle position when TTC is less than 5 s;
• Number of through lanes;
• Maneuver orientation;
• Average density during weaving maneuver;
• Segment length; and
• Gender and age group of participants.
Since the data available in the driving simulator experiment relate to the behavior of the weav-
ing vehicle, the analysis was focused on identifying relationships between the weaving vehicle’s
maneuvers and the geometric or operational features of the weaving segments. Figure 26 illustrates
that the average speed of the weaving vehicle during the maneuver is decreasing as the segment
density increases, which is expected. In addition, the research evaluated the lead and lag gaps during
a lane change. A gap is defined as the time headway from the back of the front vehicle to the front
of the following vehicle. A lead gap is the gap measured from the subject vehicle to the front vehicle
in the target lane, while a lag gap is the gap to the trailing vehicle in the target lane. The research
indicated that lead and lag gaps accepted by the study participants decrease as the density increases,
as drivers are willing to accept shorter gaps. This finding is also consistent with literature and
field and simulation findings. Similarly, TTC is decreasing with increased density, which is
also expected.
Next, the research team performed analysis of variance and regression to test the effect of geo-
metric, operational, and demographic factors on the performance of the weaving vehicle. Table 20
summarizes the statistical analyses. Weaving speed was found to be dependent on the traffic den-
sity, as shown in Table 20. Increasing the number of through lanes also leads to reduced vehi-
cle speed because of the increased lane-changing activity that must be performed to execute the
maneuver. The direction of the weaving maneuver also negatively affects speed—it was found that
weaving from left to right led to lower speeds than right-to-left maneuvers. The segment length also
positively impacts average speed, with longer segments resulting in higher average weaving speeds.
Last, it was found that speed decreases with an increase in the average age. The adjusted R-squared
of this model is 34.6%.
The average vehicle acceleration during the weaving maneuver was found to be correlated with
the average density, the number of through lanes, the lead vehicle’s speed and gap, the subject vehi-
cle’s speed, the weaving maneuver orientation, and the gender and the age of the drivers. Increased
density leads to slightly reduced acceleration rate. Also, an increased number of lanes results in
lower acceleration rate. Higher speeds of the subject or the lead vehicles suggest higher accelerations,
while increased lead gap also results in higher accelerations. The weaving direction (left to right
versus right to left) was found to be statistically significant in this model as well. Vehicles perform-
ing left-to-right weaving maneuvers had higher acceleration rates compared to vehicles performing
right-to-left maneuvers. In terms of demographics, females or older adults had lower accelerations.
The coefficients of the regression analysis are presented in Table 21. The adjusted R-squared of this
model is 20.1%.

Copyright National Academy of Sciences. All rights reserved.

 Operational Performance and Safety Effects of Arterial Weaving Sections

(a) (b)

Copyright National Academy of Sciences. All rights reserved.

(c) (d)

Figure 26.   Relationship between traffic density and (a) average speed of weaving vehicle, (b) lead gap, (c) lag gap, and (d) TTC.
Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   45

Table 20.   Coefficients of regression analysis of average weaving

speed.
Term Coef SE Coef T-Value P-Value VIF
Constant 46.92 2.02 23.21 0.000 –
Density −0.3291 0.0257 −12.81 0.000 1.15
Through lanes −1.426 0.673 −2.12 0.035 1.05
Segment length (long) 4.278 0.332 12.90 0.000 1.02
Orientation (left to right) −0.979 0.340 −2.88 0.004 1.08
Age group (years)
> 50 −2.445 0.531 −4.61 0.000 1.45
18–30 2.167 0.447 4.85 0.000 1.44
Notes: Coef = coefficient, SE = standard error, VIF = variance inflation factor.

Table 21.   Coefficients of regression analysis of average weaving

acceleration.
Term Coef SE Coef T-Value P-Value VIF
Constant 4.328 0.457 9.47 0.000 –
Density −0.01827 0.00578 −3.16 0.002 1.13
Through lanes −1.269 0.153 −8.30 0.000 1.05
Orientation (left to right) 0.2106 0.0773 2.72 0.007 1.08
Age group
> 50 −0.265 0.121 −2.19 0.029 1.45
18–30 0.071 0.102 0.70 0.485 1.44
Gender (female) −0.1358 0.0748 −1.82 0.070 1.00

In terms of the gap sizes, the accepted lead gaps (ft) on the target lane were found to be a func-
tion of density (veh/mi/lane), the speed of the lead vehicle on the current lane [Vlead (mi/h)],
the difference between the average segment speed and the subject vehicle’s speed [Vavg − Vsubject
(mi/h)], and the proportion of segment traveled (l/D) (Table 22). Drivers were found to accept
longer lead gaps at lower density values. As the lead vehicle’s speed or the relative speed between
the subject vehicle and the surrounding vehicles decreased, vehicles accepted longer lead gaps
on the target lane. Also, small lead gaps on the current lane suggest small accepted lead gaps on
the target lane. The adjusted R-squared of this model is 36.9%.
The accepted lag gaps on the target lane (ft) are a function of density (veh/mi/lane), the speed
of the lead vehicle on the current lane [Vlead(mi/h)], the proportion of segment traveled (l/D), the
number of through lanes, and weaving maneuver direction (Table 23). Drivers were found to accept
longer lag gaps at lower density values. As the number of through lanes (and resulting number of
lane changes) increases, the size of accepted lag gaps decreases. In addition, weaving from left to
right results in longer lag gaps, probably because it is generally more difficult to see behind the right
side of the vehicle. Also, small lead gaps on the current lane suggest larger accepted lag gaps on the

Table 22.   Coefficients of regression analysis of accepted lead gap.

Term Coef SE Coef T-Value P-Value VIF
Constant 240.3 14.2 16.97 0.000 –
Density (veh/mi/lane) −1.420 0.236 −6.03 0.000 1.09
l/D −67.6 21.3 −3.17 0.002 1.11
Vlead (mi/h) −1.409 0.264 −5.33 0.000 1.16
ΔV = Vavg – Vsubject (mi/h) −5.535 0.623 −8.88 0.000 1.09

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

46   Operational Performance and Safety Effects of Arterial Weaving Sections

Table 23.   Coefficients of regression analysis of accepted lag gap.

Term Coef SE Coef T-Value P-Value VIF

Constant 183.3 29.7 6.17 < 0.001 –
Density (veh/mi/lane) −1.711 0.345 −4.97 < 0.001 1.39
l/D 74.1 24.7 3.00 0.003 1.28
Vlead (mi/h) 1.295 0.295 4.38 < 0.001 1.43
Lead gap on current lane (ft) 0.1175 0.0352 3.34 0.001 1.22
Through lanes −30.93 8.92 −3.47 0.001 1.34
Orientation (left to right) 24.29 4.44 5.47 < 0.001 1.35

target lane. Finally, lead vehicle speed on the current lane suggests higher accepted lag gaps. The
adjusted R-squared of this model is 22.6%.
Last, TTC is a function of density, number of through lanes, the distance of the vehicle from the
end of the weaving segment, and the weaving maneuver orientation (Table 24). A greater number
of lanes or left-to-right weaving maneuvers results in reduced values of TTC. Also, the further the
vehicle is from the end of the weaving segment, the higher the TTC. The adjusted R-squared of this
model is 14.1%.

Relationships Between Weaving Characteristics and Safety Data

Three types of safety analyses were conducted to identify the safety effects of varying geomet-
ric, volume, and traffic conditions in weaving sections on arterials:
• Analysis of crash data for the study sites,
• Analysis of conflict data for the study sites, and
• Analysis of conflict data from a driving simulator.

The goal of the safety analyses was to identify key design and safety issues that could poten-
tially help highway agencies reduce crashes along urban arterial weaving sections. Another goal
was to identify future research needs to develop methods that predict the safety performance of
arterial weaving sections, suitable for inclusion in the AASHTO HSM.

Statistical Analysis of Crash Data for Study Sites

The goal of the crash analysis was to investigate relationships between geometric design and traf-
fic control features and motor vehicle crashes along urban arterial weaving sections. To do so, the
research team developed mixed-effect negative binomial (MENB) regression models to explore the
influence of weave length and site length on the frequency of crashes on arterial weaving segments.
The application of the MENB to this study was because different sites had varying numbers of crash
data available. Thus, the possible analysis that would consider all sites and all available years would be
considering the number of crashes per year. This approach requires the repeated measure whereby
crashes are counted each year for the same location. Thus, the crash site would be a grouping vari-
able for the MENB.

Table 24.   Coefficients of regression analysis of TTC.

Term Coef SE Coef T-Value P-Value VIF

Constant 4.355 0.440 9.90 0.000 –

Density (veh/mi/lane) −0.01778 0.00407 −4.37 0.000 1.25

Through lanes −0.209 0.126 −1.66 0.097 1.27
Weaving length, vehicle
position (ft) 0.000497 0.000274 1.81 0.071 1.08
Orientation (left to right) −0.2185 0.0586 −3.73 0.000 1.11

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   47

The analysis was performed for total crashes, rear-end crashes, and sideswipe crashes. The
influence of weave length and site length is evaluated using the ratio of weave length to site
length (WL/SL ratio). Again, in this study, the site length is defined as the length of the entire
section from the gore of the initial access point to the downstream stop bar of the approach.
Also, for the safety analysis, the weave length is defined as the shortest distance from the gore of
the initial access point to either beginning of the closest downstream left-turn or right-turn lane
or the closest downstream access point along the segment. In addition to the WL/SL ratio, the
developed models controlled for the variations across states, traffic volume, number of through
lanes, presence of downstream left-turn lanes, and year of crash.
Table 25 and Table 26 present the final regression results for total crashes and sideswipe crashes,
respectively. For both models, the WL/SL ratio has a negative association with crashes. Therefore,
as the WL/SL ratio increases, the number of crashes decrease. This implies that as the weave length
increases along the entire site length from gore to the intersecting road, the chance of weaving-
related crashes decreases. According to the results, the impact is statistically significant at a 95%
confidence level for sideswipe crashes only. The statistically insignificant results for total crashes
may be attributed to a small sample size in terms of number of sites. They may also be due to the
impact of other factors (such as the signalization pattern), which may have a more pronounced
impact on various types of crashes.
In addition to the WL/SL ratio, other factors that indicated statistical significance in one or
more of the models included the state and the year of crash. Across states, sites in North Carolina
and Ohio are likely to have a higher number of total crashes compared to sites in California. On the
other hand, sites in North Carolina and Texas are likely to have a higher number of sideswipe crashes
compared to sites in California, while sites in Ohio are likely to have a lower number of sideswipe
crashes. The traffic volume factor was not statistically significant across crash types. Furthermore,
the crash year is only statistically significant for sideswipe crashes.

Table 25.   Mixed-effect negative binomial regression results for total crashes.

Estimate Std. Error Z-Value P-Value

Intercept 0.022 3.972 0.01 0.996
WL/SL ratio −0.822 0.715 −1.15 0.250
State
California – – – –
North Carolina 1.706 0.341 5.00 < 0.001
Ohio 1.713 0.315 5.45 < 0.001
Texas 0.277 0.357 0.78 0.439
Traffic volume
log(ADT) 0.069 0.384 0.18 0.858
Year
2015 – – – –
2016 0.266 0.249 1.07 0.286
2017 0.132 0.258 0.51 0.609
2018 0.163 0.255 0.64 0.523
2019 0.315 0.241 1.31 0.191
2020 −0.162 0.295 −0.55 0.582
2021 −0.094 0.292 −0.32 0.749
Mixed Effects
Grouping variable
Study site Variance = 0.099 Std. Dev. = 0.314
Number of sites 17
Number of observations 87

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

48   Operational Performance and Safety Effects of Arterial Weaving Sections

Table 26.   Mixed-effect negative binomial regression results for sideswipe

crashes.

Estimate Std. Error Z-Value P-Value

Intercept 1.111 5.955 0.19 0.852
WL/SL ratio −2.213 0.995 −2.22 0.026
State
California – – – –
North Carolina 2.161 0.472 4.58 < 0.001
Ohio −1.755 1.077 −1.63 0.103
Texas 1.035 0.498 2.08 0.038
Traffic volume
log(ADT) −0.181 0.579 −0.31 0.755
Year
2015 – – – –
2016 0.273 0.529 0.52 0.606
2017 0.848 0.481 1.77 0.078
2018 0.966 0.473 2.04 0.041
2019 0.762 0.486 1.57 0.117
2020 0.719 0.642 1.12 0.263
2021 0.902 0.616 1.47 0.143
Mixed Effects
Grouping variable
Study site Variance = 0.043 Std. Dev. = 0.207
Number of sites 17
Number of observations 87

Statistical Analysis of Conflict Data for Study Sites

The goal of the conflict analysis was to investigate relationships between geometric design and
traffic control features and motor vehicle conflicts along urban arterial weaving sections. To do so,
the research team developed regression models to explore the influence of weave length and site
length on the frequency of HB, TTC, and PET events along arterial weaving segments. Table 27
presents the final regression results for HB conflict as influenced by the weave length and site length,
represented by their ratio (WL/SL ratio). It can be observed that the WL/SL ratio has a negative
influence on HB. This means that as the weave length to the site length ratio increases, the number of
HB events decreases. Other site characteristics found to be associated with the increased HB events
included traffic volume, weave intensity, and maximum lane changes. Weave intensity was calculated
(i.e., measured) as the number of lane changes divided by total through volume. This final regression
model for HB is consistent with what one would expect to see; however, none of the parameters are

Table 27.   Influence of weave length on HB events.

Estimate Std. Error Z-Value P-Value

Intercept 2.914 1.492 1.95 0.051
WL/SL ratio −0.019 0.382 −0.05 0.961
log(volume) 0.231 0.175 1.32 0.188
Weave intensity 1.066 0.971 1.10 0.273
Maximum lane changes 0.168 0.354 0.48 0.635
Mixed Effects
Grouping variable
Study site Variance = 0.786 Std. Dev. = 0.886
Number of sites 18
Number of observations 56

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   49

Table 28.   Mixed-effect negative binomial regression results for simulator-

based conflicts.

Estimate Std. Error Z-Value P-Value

Intercept −1.3944 0.1849 −7.54 < 0.001
Site length (ft)
Long site (1,500) Base
Short site (900) 1.038 0.223 4.66 < 0.001
Traffic density
High density (LOS D) Base
Low density (LOS A/B) −0.832 0.259 −3.21 0.001
Mixed Effects
Grouping variable Variance Std. Dev.
PID 0.08098 0.2846
Logstreams 0.14369 0.3791
Number of PIDs 40
Number of logstreams 16
Number of observations 640
Note: PID = person ID (i.e., the number of participants in the simulator study).

statistically significant at a 90% confidence level. This might be attributed to a small sample size in
terms of number of sites.
No regression models were developed that provided reasonable results explaining the relation-
ship between weave length and the frequency of TTC or PET events.

Statistical Analysis of Conflict Data for Driving Simulator

The goal of the conflict analysis, using data from a driving simulator, was again to investigate
relationships between geometric design and traffic control features and the frequency of motor
vehicle conflicts along urban arterial weaving sections and provide a comparison to results from
the field data.
Table 28 presents the final regression results for the simulator-based TTC conflicts. The statis-
tical analysis indicates that the number of TTC conflicts increases with a decrease in site length.
Compared to long sites (1,500 ft), short sites are associated with an increase of about three conflicts
[exp (1.038)]. On the other hand, low traffic density is associated with fewer conflicts.

Operational Analysis Methodology

This section presents the suggested operational analysis methodology for arterial weaving sec-
tions. The methodology uses the data described previously to expand the existing HCM7 Chap-
ter 18, Urban Street Segments, procedure with additional steps and modifications to account for
the impact of weaving maneuvers on traffic flow. This approach allows for a seamless transition in
evaluating an urban street segment, starting with operations with zero weaving demands. The main
reason for doing so is that a particular urban street segment may have weaving demands for some of
the time; when the weaving demands are zero or small, the methodology should result in the same
performance as that of an urban street segment. The methodology is designed to evaluate arterial
weaving segments by using the travel speed and volume-to-capacity (v/c) ratio. The methodological
framework consists of several steps, including defining analysis objectives and data requirements,
determining arterial weaving section characteristics, and evaluating segment performance.
As in HCM7 Chapter 18, an arterial weaving segment is defined by and consists of its bound-
ary points and the link between them (Figure 27). A boundary point may consist of a signalized

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

50   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 27.   Schematic of an arterial weaving section.

intersection or a ramp terminal. The segment may contain unsignalized intersections along its link.
The arterial weaving segment performance is evaluated based on two service measures (i.e., the
performance measures used to define LOS):
• Travel speed for through vehicles, which includes the impacts of the design and prevailing traffic
conditions within the link and the delay incurred to the through movement at the downstream
intersection, and
• The v/c ratio for the through movement at the downstream intersection.

In addition to these, the research team evaluated the feasibility of using the percentage of
drivers who are unable to perform their weaving maneuver as a guide when designing weaving
sections.
The methodology is based on estimating the speed drop (or delay and travel time increase) at an
arterial segment due to weaving activity. Conceptually, with the same volumes, number of lanes, and
weaving length, an arterial segment with weaving will operate at a lower average space mean speed
than an arterial segment without weaving. The difference between speeds is caused by the turbulence
of the weaving flows. In addition to the weaving configuration (for example, the site length), the
turbulence level is a function of the weaving and nonweaving demands. One complication in the
analysis of arterial weaving is that there may be several weaving traffic streams with different origins
and destinations within an urban street segment. Therefore, the methodology must combine their
effect and anticipate a variety of weaving patterns.
Figure 28 illustrates the proposed methodological framework for operational analysis. The
required data for this analysis are similar to those in Exhibit 18-5 in HCM7, with the exception of
the distances from the gore of all access points to the downstream stop bar. As discussed earlier,
this figure resembles Exhibit 18-8 in HCM7, with an additional step (Step 1.5) to account for the
effects of weaving movements and a modified Step 2.

Step 1: Determine Traffic Demand Adjustments

In Step 1, the methodology estimates the traffic demand adjustments to ensure that the input
volumes accurately reflect segment traffic conditions. The adjustments include the following:
• Limiting entry to the segment because of capacity constraints. When the demand volume for
a traffic movement at an intersection surpasses its capacity, the discharge volume from the

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   51

Figure 28.   Motorized vehicle methodology for arterial weaving segments

(Steps highlighted in red represent additions or modifications to the HCM7
method. g/C 5 effective green time over cycle time ratio).

intersection becomes limited (or metered), which can potentially influence the volume entering
the segment.
• Balancing the volumes entering and exiting the segment by proportionally adjusting the exiting
volumes to match the entering volumes, maintaining equilibrium between both directions
of travel.
• Mapping entry to exit flow paths using an O-D distribution matrix, which represents the
combined volume from each upstream point of entry weighted by its percentage contribution

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

52   Operational Performance and Safety Effects of Arterial Weaving Sections

to the downstream movement, allowing for a comprehensive understanding of traffic flow

patterns within the segment. This mapping of O-Ds is also used to calculate the PCO within
the weaving segment, as described in the next step.

Step 1.5: Calculate TI

In this new step, the turbulence index (TI) is calculated for the weaving section to assess the
degree of traffic turbulence caused by weaving movements:
n
PCO i
TI = / (1)
i=1 SL i

where
TI = Turbulence index (veh/h)2/ft;
PCOi = The PCOs for the weaving movement i considering the conflicting volume the movement
must traverse to its destination, calculated as shown throughout the rest of Step 1.5;
SLi = Distance from the gore of the access point to the stop bar for weaving movement i; and
n = Total number of weaving movements within the weaving section (including weaving
to and from midblock access points).
A PCO is defined as a potential encounter between vehicles in conflicting traffic flows within a
weaving segment. For example, Figure 29 illustrates the estimation of the PCO for vehicles origi-
nating from the upstream free right turn and making a left turn at the downstream intersection.
The PCO for this movement is the product of the respective demands, since all vehicles origi-
nating from the ramp and making a left turn must cross all lanes and therefore have the potential
to conflict with all vehicles originating from the arterial:

PCO RL = VRL # VA (2)

where
VRL = Flow (veh/h) arriving from the ramp and making a left turn at the downstream intersection,
and
VA = Total flow (veh/h) arriving from the arterial.
Since there may be several weaving maneuvers within an urban segment, the total PCO for the
segment must be estimated as the sum of all PCOs. As an example, to calculate the PCO for site NC6,
Figure 30 provides a schematic of all the movements in the facility, with part a showing the move-
ments arriving from the arterial and part b showing the movements arriving from the upstream free

Figure 29.   Illustration of PCOs for movement VRL.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   53

(a) Movements Originating from the Arterial

(b) Movements Originating from the Upstream Free Right Turn

Figure 30.   Illustration of PCOs for movement.

right turn. The methodology calculates the PCOs from the perspective of the weaving and turn-
ing movements. In this example, the demand from the upstream free right turn conflicts with the
through traffic in three ways: (1) the demand from the free right turn moving to the left turn, (2) the
demand from the free right turn merging and going through at the downstream intersection, and
(3) the demand from the free right turn merging with other right-turning traffic arriving from the
upstream and turning right at the downstream intersection.
As shown in Figure 30, the weaving demand is composed of VAR, VRA, and VRL, and the non-
weaving demand is composed of VRR, VAA, and VAL. The potential conflicts in a weaving section
can occur between a weaving and a nonweaving movement or between two weaving movements.
Table 29 presents the PCO calculations for each of the movements originating from the free right
turn of Figure 30, when the total number of through lanes within the weaving segment is N.
In the second equation, VAA and VAL are divided by the number of through lanes originating from
the arterial (N), which assumes that the traffic arriving from the free right turn only conflicts with
a single lane (the average of the through and left-turning traffic) plus the entire right-turning

Table 29.   Calculation of PCO, for example,

of Figure 30.

Demand Potential Conflict Opportunity (PCO)

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

54   Operational Performance and Safety Effects of Arterial Weaving Sections

movement originating from the arterial. The equation for calculating PCOs within an urban
street segment is as follows:
n
PCO = / PCO i (3)
i=1

where n is the total number of movements in the upstream free right turn within the segment.
Depending on signalization patterns, two conflicting flows may experience such conflicts only
for part of the cycle. For example, when the upstream through movement is stopped upstream, the
free right turn may be entering the arterial weave with no conflicts or with minimal conflicts from
the opposing left turns. However, estimating the actual number of conflicting opportunities would
be unrealistic for an HCM-type analysis. Also, based on the data analysis shown in the previous
section, this new measure is very well correlated with the speed of the arterial weave.
After obtaining the PCO, the TI for the weaving section is obtained. The TI is a function of PCO
and site length, reflecting the density of these potential conflicts over the length of the weaving
segment. A higher TI value indicates a higher level of traffic turbulence and potential inefficiencies
in traffic flow. When the weaving flow is zero, the TI is zero. As weaving demands increase, the
PCOs also increase, which results in an increase of the TI. As the length over which weaving occurs
increases, the constraints due to lane changing decrease, and the TI decreases.
Both measures can be calculated by the analyst using the weaving and nonweaving demand
flows within the segment.
By adding Step 1.5 to the methodological framework, the analysis can better account for the
complex interactions between weaving and nonweaving traffic movements in arterial weaving
sections, ultimately leading to a more accurate assessment of the operational performance and
more effective traffic control measures.

Step 2: Determine Running Time

In this step, the procedure calculates the free-flow speed, the additional running time attribut-
able to the sources outlined in HCM7 Chapter 18 on urban street segments, and the additional
delay due to weaving.
As indicated previously, the analysis of arterial weaving segments is aligned with that of urban
street segments. There are three complications for using the same LOS criteria and methodological
approach to analyze arterial segments: First, weaving takes place across all lanes, including auxiliary
lanes and turning lanes. Restricting the operational performance evaluation only to the through
movement may not consider weaving impacts that affect these lanes. Therefore, the new method
must employ a series of assumptions to extract the running time and travel speed of the through
lanes only. Second, the methodology calculates the link travel time and the intersection delay sepa-
rately; however, obtaining these separately in the field or in simulation is not feasible. Therefore,
again, assumptions were made during the data extraction regarding the delay due to the traffic signal
versus the delay experienced midblock due to other causes. Third, one can make the argument that
the LOS thresholds should be different for arterial weaving segments, if driver expectation is differ-
ent, particularly for drivers of the weaving movements. A similar argument is made in the HCM
when providing LOS thresholds for signalized intersections versus unsignalized intersections. In this
document, researchers assume the same LOS thresholds as those for urban street segments.
To address the first two complications, the research team developed predictive models assum-
ing the link delay occurs during the green interval, while the intersection delay occurs during
the red interval. In this case, the difference in the speeds between the through movement and

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   55

the turning movements during the green is negligible. Regarding the delay due to the signal, the
methodology relies on the existing signalized intersection procedures because weaving should
not affect the through movement delay.
Therefore, the models developed in this project predict the delay due to weaving, dwv, which is
considered in the calculation of segment running time, tR (see HCM7 Equation 18-7), as follows:
N
6.0 - l 1 3, 600 L ap

tR = fx + fv + / d ap, i + d wv + d other (4)

0.0025 L 5, 280 Sf i=1

where
dwv is the new weaving turbulence term; all other parameters are explained in HCM7.
dwv is calculated as follows:

J TI N b J V - 150 Nc gJ
K
N
O

d wv = a ) KK OO ) KK OO ) d CK O
L P
(5)
N N P
L th P L

where
Nth = Number of through lanes at the downstream intersection;
N = Total number of lanes at the downstream intersection;
C = Average cycle length for the downstream intersection;
g = Average green time for the downstream intersection; and
α, β, γ, and δ = Model parameters.
All other parameters are as previously defined.
Therefore, the running time is now calculated as follows:

6.0 - l 1 3, 600 L N ap J TI N b J V - 150 Nc J GT N

fv + / d ap, i + a ) KK OO ) KK O )d
K O
tR = fx + + d other (6)
KCO

N O
L P

0.0025 L 5, 280 Sf N th
i=1
L P L P
As with arterial segments, the LOS criteria for arterial weaving segments are based on travel
speed and base free-flow speed (Table 30), while the v/c ratio is used to distinguish between under-
saturated and oversaturated (i.e., LOS F) conditions. Travel speed reflects the movement of through
vehicles, encompassing both the running speed within the link and the delay at the downstream inter-
section. Likewise, the v/c ratio refers to the through movement at the downstream intersection.
In this methodology, researchers rely on the HCM7 framework for the following steps: Step 3:
Determine the Proportion Arriving During Green, Step 4: Determine Signal Phase Duration,
Step 5: Determine Through Delay, Step 6: Determine Through Stop Rate, Step 7: Determine Travel

Table 30.   LOS criteria for urban street segments. (Source: HCM7 Exhibit 18-1)

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

56   Operational Performance and Safety Effects of Arterial Weaving Sections

Speed, Step 8: Determine Spatial Stop Rate, Step 9: Determine LOS, and Step 10: Determine Auto-
mobile Traveler Perception Score.

Model Calibration and Evaluation

Before diving into the data analysis, researchers first filtered out the data points where the
speed was less than the threshold for LOS F (as indicated in Table 30). This action was taken to
eliminate points representative of congested conditions.
Subsequently, the data set was split to first calibrate and then evaluate the model. A random
selection of 20% of the data was set aside for validation purposes, leaving the remaining 80% to
be used for model training.
To calibrate this model, researchers adopted the root mean square error (RMSE) as optimiza-
tion criterion. The RMSE was computed using the following equation:

1 n
` Yi - Yrı j
2
MSE = /
ni=1
(7)

where
Y_i is the predicted speed,
Y ı is the observed speed, and
n is the number of data points, each representing a 5-minute observation.
The training data set underwent a further data pruning process. Data points with a traffic flow
of less than 150 passenger cars per hour per lane (pc/h/lane) were eliminated. This decision was
informed by the optimization results, which indicated that volumes lower than 150 pc/h/lane
closely mirrored predictions from the HCM7 methodology.
Table 31 provides the estimated parameters of the proposed model, based on the predicted speed
for all data points in the evaluation data set.
The RMSE values indicate that the proposed model delivers a 28.8% improvement in speed
prediction compared to the HCM methodology. Therefore, with these calibrated parameters, the
final form of the proposed model follows:

6.0 - l 1 3, 600 L N ap J TI N 0.06 J V - 150 N0.86 J N

fv + / d ap, i + 0.05 ) KK OO ) KK
K GT O
tR = fx + O ) 0.1 KCO
+ d other (8)
N O
L P

0.0025 L 5, 280 Sf N th
i=1
L P L P

Quality of Model Fit

As presented in Table 31, the RMSE of the model, when considering the effect of weaving, can
be reduced from 9.83 mi/h to 7.00 mi/h. This substantial decrease in RMSE signifies an improve-
ment in the model’s predictive accuracy, and the overall model error is thus deemed acceptable
within the scope of this analysis.

Table 31.   Parameter estimation for proposed model.

Minimum Flow RMSE (mi/h)

α β γ δ
(pc/h/lane) Modeling Evaluation
Proposed
0.05 0.06 0.86 0.10 150 6.15 7.00
model
HCM – – – – – 8.07 9.83
methodology
Improvement % 28.80 23.84

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   57

For a more comprehensive evaluation, the fitted and observed speed data from the evaluation
database are plotted in Figure 31. This graphical representation provides a comparative analysis
of observed field speeds against the speeds predicted by both the HCM7 and the newly proposed
model as a function of flow.
As shown, the HCM7-predicted speeds are generally higher than the observed speeds, par-
ticularly for higher flows. The speeds predicted by the new model are generally well aligned with
field data, and significantly lower than those predicted by HCM7. The new model demonstrates
a more accurate representation of vehicle speeds across a variety of traffic conditions.
Figure 32 illustrates the comparative values of the RMSE based on the predictions made by
the HCM7 and the proposed model. As previously discussed, the HCM7 tends to overestimate
speeds, particularly under conditions of high traffic volume. This overestimation results in a
higher RMSE, indicating a greater discrepancy between the predicted and observed speeds.
The proposed model demonstrates a lower RMSE than HCM, particularly in high-demand
scenarios.
Figure 33 illustrates the relationship between speed and the TI, with the latter normalized by the
number of through lanes. The figure reveals a similar pattern in the trend lines for both the field data
and the HCM7 predictions. Consistent with previous observations, the HCM7 overestimates speeds
for a wide range of TI values. This finding underscores the necessity of considering the impact of
weaving maneuvers at arterial segments, a factor that could significantly enhance the accuracy of
speed predictions and the overall effectiveness of traffic management strategies.

Computational Engine
The HCM urban streets methodology is complex, and therefore performing a full set of analy-
ses is not feasible without using a computational engine. This project used the Excel STREETVAL

Figure 31.   Speed-flow diagram comparing field data, HCM7, and new model.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

58   Operational Performance and Safety Effects of Arterial Weaving Sections

Figure 32.   Comparing RMSE of HCM7 and proposed model.

computational engine to implement the proposed operational method. The computational pro-
cedure is implemented in cell-based formulas as well as Visual Basic for Applications (VBA) code.
The advantage of this approach is that the platform is available and is used as a research tool by the
Highway Capacity and Quality of Service (HCQS) Committee. This will make the results of this
project more easily accessible to the HCQS Committee, assure its usability in the future, and elimi-
nate the need for additional funding to maintain it beyond the period of performance of this project.

Figure 33.   Speed versus TI/ Nth for field, HCM 7, and proposed model.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   59

The implementation was carried over on Excel STREETVAL V8.30 (released in March 2022) and
developed version 8.30.1.
As part of this project, two real-world case studies were coded in the Excel STREETVAL
V8.30.1 to evaluate the impact of arterial weaving. These two tests confirmed a faithful opera-
tion of the method as well as the implementation in STREETVAL. Figure 34 shows the main
page of Excel STREETVAL computational engine, containing the generic advisory material
and disclaimer. The Excel STREETVAL V8.30.1 can be found at nap.nationalacademies.org

Figure 34.   The Excel STREETVAL V8.30.1 implements the NCHRP 15-66 operational analysis method for arterial
weaving segments.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

60   Operational Performance and Safety Effects of Arterial Weaving Sections

(search for NCHRP Research Report 1094: Operational Performance and Safety Effects of Arterial
Weaving Sections).
Figure 35 shows the “Set Up” sheet of the STREETVAL V8.30.1, where the user can edit global
information of the HCM urban streets analysis. As an example, Figure 35 shows the inputs for the
North Carolina case study, where a single urban streets segment (1,200 ft) is analyzed for assessing
the impact of a weaving maneuver from an access point.
Figure 36 shows the output page of STREETVAL V8.30.1 where the performance measures are
adjusted based on the proposed operational method in the NCHRP 15-66. Two interim variables,
TI and delay due to arterial weaving (d_wv), are also reported in this output sheet.

Figure 35.   The “Set Up” sheet of STREETVAL V8.30.1.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Research Approach   61

NCHRP 15-66 Interim Outputs

Figure 36.   Output sheet of STREETVAL V8.30.1 with NCHRP 15-66 weaving analysis method and interim
variables.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

CHAPTER 3

Findings and Applications

The procedures detailed in the HCM7 do not consider the impacts of arterial weaving, which is
defined as the crossing of two or more traffic streams in the same direction between two or more
access points on a multilane urban street with some type of traffic control. The primary objective of
this project was to develop a deterministic method for evaluating the operational performance of
weaving sections on arterials. The secondary objective was to identify the safety effects of varying
geometric, volume, and traffic conditions in weaving sections on arterials.
The research team first synthesized the literature related to operational analysis methods and
performance measures for arterial weaving. The team also synthesized previous research related to
safety along arterial weaving sections. It was concluded that there are very few studies addressing
operations along arterial weaving segments and no studies that evaluate the safety of such sections.
From the project, the team developed a list of performance measures for evaluating arterial weaving
sections and recommended using travel time as the service measure to ensure consistency with
the HCM. It also developed two new intermediate measures, the PCO and the TI, to be used in the
operational analysis.
Based on the literature review findings, the research team used drones to collect or obtain field
data at 20 locations (18 of which were used in further analysis) in California, Nebraska,
North Carolina, Ohio, and Texas. A total of 1,118 min of video, and approximately 6,000 weaving
and 11,000 nonweaving trajectories, were extracted from the videos. Five of these sites were also
simulated to obtain additional data for analysis. A total of 1,080 additional scenarios were simu-
lated, which produced more than 413,000 simulated trajectories for use in the model development.
In addition, a driving simulator experiment was conducted with 39 participants to explore driver
behavior around arterial weaving areas.
Data analyses were conducted to evaluate relationships between performance characteristics
and specific prevailing traffic, control, and geometric characteristics of arterial weaving segments
based on the field and microsimulation data. The analysis showed that speed generally decreases
as the demand increases. Similarly, speed decreases as the PCO increases and as the length of
the segment decreases. The analysis also found that the number of missed maneuvers increases
with demand and with PCO.
The safety analysis evaluated the effects of varying geometric, volume, and traffic conditions
in weaving sections on arterials. The methods and analysis are based on field data supplemented
by microsimulation data and driving simulation data. Crash analyses were conducted based on
field data, while conflict analyses were conducted based on trajectories obtained in the field and
trajectories obtained through the driving simulator. The analyses concluded that as the weave
length increases, the chance of weaving-related crashes and conflicts decreases. Regarding crash
data, this impact is statistically significant at a 95% confidence level for sideswipe crashes only.

62

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Findings and Applications   63

The statistically insignificant results for total crashes may be attributed to a relatively small sample
size in terms of the number of sites.
Based on the data obtained, the research team developed an operational analysis methodology
for evaluating arterial weaving segments suitable for HCM-type analysis. The operational analysis
method extends the existing HCM’s urban streets analysis and is based on the calculation of the PCO
and the TI of the arterial weave.
A PCO is defined as a potential encounter between vehicles in conflicting traffic flows within a
weaving segment. The PCO for any movement is the product of the respective conflicting demands.
Since there may be several weaving maneuvers within an urban segment, the total PCO for the seg-
ment must be estimated as the sum of all PCOs. The equation for calculating PCOs within an urban
street segment follows:
n
PCO = / PCO i (9)
i=1

where n is the total number of movements in the upstream free right turn within the segment.
After obtaining the PCO, the TI for the weaving section was obtained using the following
equation:
n
PCO i
TI = / (10)
i = 1 SL i

where
TI = Turbulence index [in units of (veh/h)2/ft],
PCOi = The potential conflict opportunities for the weaving movement i considering the con-
flicting volume the movement must traverse to its destination,
SLi = Distance from the gore of the access point to the stop bar for weaving movement i, and
n = Total number of weaving movements within the weaving section (including weaving
to and from midblock access points).
The TI serves as a measure of the overall impact of weaving movements on traffic operations
within the arterial weaving section. A higher TI value indicates a higher level of traffic turbulence
and potential inefficiencies in traffic flow. The TI is a measure that reflects the number of lane
changes combined with the density of lane changing, as expressed in the distance measure for each
weaving movement. When the weaving flow is zero, the TI is zero. As weaving demands increase, the
PCOs also increase, which results in an increase of the TI. As the length over which weaving occurs
increases, the constraints due to lane changing decrease, and the TI decreases.
The new weaving analysis methodology is based on estimating the speed drop (or delay and travel
time increase) at an arterial segment due to weaving activity. Conceptually, with the same volumes,
number of lanes and weaving length, an arterial segment with weaving will operate at a lower aver-
age space mean speed than an arterial segment without weaving. The difference between speeds is
caused by the turbulence (expressed quantitatively as a function of TI) of all weaving flows within
the segment including midblock access points.
The arterial weaving segment performance is evaluated based on the same two service mea-
sures as that of urban streets (i.e., the performance measures used to define LOS):
• Travel speed for through vehicles, which includes the impacts of the design and prevailing
traffic conditions within the link and the delay incurred to the through movement at the
downstream intersection, and
• The v/c ratio for the through movement at the downstream intersection.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

64   Operational Performance and Safety Effects of Arterial Weaving Sections

In addition to these, the research team evaluated the feasibility of using the percentage of
drivers who are unable to perform their weaving maneuver as a guide when designing weaving
sections. It was determined that shorter weaves and higher demands can lead from one to five
missed maneuvers per 15 min.
The project developed a computational engine that implements the new procedure in the Excel-
based STREETVAL computational engine. The computational procedure is implemented in cell-
based formulas and VBA code. The advantage of this approach is that the platform is available and
used as a research tool by the HCQS Committee. The new methodology was implemented in the
STREETVAL V8.30 (released in March 2022) and developed version 8.30.1.
The research team created two real-word case studies based on the field data collected to assess the
impact of weave maneuvers in urban streets. The first case study is based on the Round Rock, Texas,
study site, where in the westbound direction of travel there is an upstream unsignalized, channel-
ized right turn that creates weave maneuvers. The second case study is an urban street segment in
Charlotte, North Carolina, where an entry ramp from a freeway acts as an access point and creates
turbulence through weaving maneuvers. The data used for analysis are those obtained during the
data collection. However, the research team made necessary adjustments to the input data to better
demonstrate the impact of the weave maneuver within HCM urban streets analysis. The details
of the two case studies are documented in Appendix I.
To facilitate dissemination, the research team developed instructional materials in the form of
46 PowerPoint slides containing an overview of the HCM urban streets, the proposed method of
NCHRP 15-66, and the two case studies. These are available on the National Academies Press web-
site (nap.nationalacademies.org; search for NCHRP Research Report 1094: Operational Performance
and Safety Effects of Arterial Weaving Sections).

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

CHAPTER 4

Conclusions and Suggested

Research

In conclusion, the data analysis has confirmed the trends anticipated, and the driving simulation
analysis has provided some important insights for consideration in the operational analysis method.
The research team has developed an operational analysis method that is integrated with the existing
HCM7 urban streets segment analysis method. It has implemented this method in a computational
engine that is integrated with the existing STREETVAL engine used by the HCQS Committee.
The analysis showed that shorter weaving segments with higher demands could lead to up to
five missed weaving maneuvers in 15 min. The PCO and TI provide a good indication of turbu-
lence that can evaluate the impacts of access points.
To facilitate the use of the new methodology the research team recommends integrating the
method within the HCM. Additional limitations of the research are as follows:
• The data collection did not include sites with weaving that involves more than one segment.
Additional data should be collected to evaluate operations at such sites.
• The data collection did not include sites with significant presence of pedestrians and bicycles.
The presence of pedestrians may not affect the performance of arterial weaves differently than
that of any urban street segment. However, the presence of bicyclists along the segment may
affect operations. Additional data should be collected at sites with significant bicyclist pres-
ence to evaluate operations at such sites.
• The analysis of missed turns did not produce adequate data to develop a quantitative model.
Therefore, it is suggested to collect data at sites where such maneuvers are more frequent and
develop suitable methods to predict this measure.
The safety analysis of arterial weaving sections that investigated crash and conflict data yielded
expected results; although, with a few exceptions, the parameters in the statistical models were not
statistically significant. The analysis of TTC conflicts based on the driving simulator yielded the
only statistically significant results that showed fewer TTC conflicts with an increase in site length;
and while the analyses of crash data and HB data from the study sites did not yield statistically
significant results in most cases, the signs of the parameters were as expected. The analysis of crash
data showed that the ratio of weave length to site length has a negative association with crashes,
suggesting that as the ratio of weave length to site length increases, the number of crashes decreases.
Similarly, the analysis of HB events at the study sites indicates that as the weave length to the site
length ratio increases, the number of HB events decreases.
Given the results of the safety analysis, designers should consider the following option for
reducing crashes along arterial weaving sections:
• Determine if there is an opportunity to temporarily or permanently close midblock accesses
along the segment. Closing midblock access points, particularly when weaving demands are
high, will eliminate these weaving maneuvers and result in increasing the weave length. This
will likely reduce potential crashes that may be attributed to weaving maneuvers.

65

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

66   Operational Performance and Safety Effects of Arterial Weaving Sections

Finally, although the results of the safety analysis did not provide a definitive relationship
between the weave length of arterial weaving sections and crashes or conflicts, sufficient infor-
mation was found to suggest that additional research in this area would yield promising results.
Therefore, it is recommended that future research be conducted to develop methods that predict
the safety performance of arterial weaving sections, suitable for inclusion in the AASHTO HSM.
Such research should account for the following:
• Include enough study locations where minimal weaving takes place within the segment (i.e.,
control sites) and enough arterial weaving sections with a range of weave lengths and section
lengths. The analysis results also indicate that geographic diversity of study sites is necessary.
• Although crash data and conflict data were both analyzed in this project, the crash analysis
yielded sufficient results to suggest that future research for the HSM should focus on analysis
of crash data and not necessarily conflict data.
• Provide a recommendation as to whether intersection and intersection-related crashes should be
included in the analysis of the weaving section. In the analysis conducted as part of this research,
intersection and intersection-related crashes were excluded from the analysis, but additional con-
sideration should be given to determine whether this is the recommended approach.
• The fact that weaving intensity was a significant predictor of HB events suggests that an opera-
tional performance measure like weaving intensity should be incorporated into the analysis
and methodology.
• The impact of speed on crash frequency and severity should be investigated further. It has been
found that speeds decrease with increases in demand and weaving activity along arterial weaving
sections. It has also been found that as demand and weaving activity decrease, speed is expected
to increase. It is also known that as speed increases, the potential for a severe crash increases.
Thus, as demand and weaving activity increase during peak periods, higher crash frequencies may
be expected, but the severity of the crashes may decrease because of lower speeds. Similarly, as
demand and weaving activity decrease during nonpeak periods, fewer crashes may be expected,
but the crashes may be more severe. These relationships among speed, crash frequency, and crash
severity during peak and nonpeak periods should be further investigated along arterial weaving
sections.

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

References

Campbell, K. W., A. M. Bonacci, J. Shelton, J. J. Exline, and B. J. Bushman. 2004. Psychological Entitlement: Interper-
sonal Consequences and Validation of a Self-Report Measure. Journal of Personality Assessment, 83(1): 29–45.
Contractor, A. A., N. H. Weiss, N. T. Kearns, S. V. Caldas, and K. L. Dixon-Gordon. 2020. Assessment of Post-
traumatic Stress Disorder’s E2 Criterion: Development, Pilot Testing, and Validation of the Posttrauma
Risky Behaviors Questionnaire. International Journal of Stress Management, 27(3), p. 292.
EyeTracking. 2019. EyeWorks 3 User Manual. EyeTracking, Inc., Solana Beach, CA.
Farah, H., A. Polus, S. Bekhor, and T. Toledo. 2007. Study of Passing Gap Acceptance Behavior Using a Driving
Simulator. Advances in Transportation Studies: An International Journal, pp. 9–16.
Freuli, F., G. De Cet, M. Gastaldi, F. Orsini, M. Tagliabue, R. Rossi, and G. Vidotto. 2020. Cross-cultural Perspective
of Driving Style in Young Adults: Psychometric Evaluation Through the Analysis of the Multidimensional
Driving Style Inventory. Transportation Research Part F: Traffic Psychology and Behaviour, 73, pp. 425–432.
Highway Capacity Manual 7th Edition: A Guide for Multimodal Mobility Analysis. 2022. Transportation Research
Board, Washington, DC. https://doi.org/10.17226/26432.
Rossi, R., M. Gastaldi, G. Gecchele, and C. Meneguzzer. 2012. Comparative Analysis of Random Utility Models
and Fuzzy Logic Models for Representing Gap-Acceptance Behavior Using Data from Driving Simulator
Experiments. Procedia: Social and Behavioral Sciences, Vol. 54, pp. 834–844.
Stojmenova, K., and J. Sodnik. 2015. Methods for Assessment of Cognitive Workload in Driving Tasks. In ICIST
2015 5th International Conference on Information Society and Technology (pp. 229–234).
Sun, L., L. Cheng, and Q. Zhang. 2021. The Differences in Hazard Response Time and Driving Styles of Violation-
Involved and Violation-Free Taxi Drivers. Transportation Research Part F: Traffic Psychology and Behaviour,
82, pp. 178–186.
Taubman-Ben-Ari, O., M. Mikulincer, and O. Gillath. 2004. The Multidimensional Driving Style Inventory—Scale
Construct and Validation. Accident Analysis & Prevention, 36(3), pp. 323–332. https://doi.org/10.1016
/S0001-4575(03)00010-1.
Taubman-Ben-Ari, O., and V. Skvirsky. 2016. The Multidimensional Driving Style Inventory a Decade Later:
Review of the Literature and Re-evaluation of the Scale. Accident Analysis & Prevention, 93, pp. 179–188.
Taubman-Ben-Ari, O., T. Lotan, and C. G. Prato. 2017. Young Male Drivers’ Risky Driving 15 Months After
Licensure—The Role of Intervention, Attitudes Towards Accompanied Driving, and Parents’ Risk. Trans-
portation Research Part F: Traffic Psychology and Behaviour, 51, pp. 73–80.
Van Huysduynen, H. H., J. Terken, and B. Eggen. 2018. The Relation Between Self-Reported Driving Style and Driving
Behaviour. A Simulator Study. Transportation Research Part F: Traffic Psychology and Behaviour, 56, pp. 245–255.
Watson, D., L. A. Clark, and A. Tellegen. 1988. Development and Validation of Brief Measures of Positive and
Negative Affect: The PANAS Scales. Journal of Personality and Social Psychology, 47, pp. 1063–1070.

67

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Appendices A through I

Appendices A through I are available on the National Academies Press website (nap.
nationalacademies.org) by searching for NCHRP Research Report 1094: Operational Perfor-
mance and Safety Effects of Arterial Weaving Sections.
Appendix A: Literature Review
Appendix B: Review of Performance Measures
Appendix C: Description of Data Collection Sites
Appendix D: Field Data Summary
Appendix E: Trajectory Sample—TX05 Vehicle ID 161
Appendix F: Participants’ Demographics
Appendix G: Simulation Assumptions, Calibration, and Outputs
Appendix H: Explanatory Analysis Graphs
Appendix I: Real-World Case Studies

68

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Bibliography

The bibliography can be found on the National Academies Press website (nap.national
academies.org) by searching for NCHRP Research Report 1094: Operational Performance and
Safety Effects of Arterial Weaving Sections.

69

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Abbreviations and acronyms used without definitions in TRB publications:

A4A Airlines for America
AAAE American Association of Airport Executives
AASHO American Association of State Highway Officials
AASHTO American Association of State Highway and Transportation Officials
ACI–NA Airports Council International–North America
ACRP Airport Cooperative Research Program
ADA Americans with Disabilities Act
APTA American Public Transportation Association
ASCE American Society of Civil Engineers
ASME American Society of Mechanical Engineers
ASTM American Society for Testing and Materials
ATA American Trucking Associations
CTAA Community Transportation Association of America
CTBSSP Commercial Truck and Bus Safety Synthesis Program
DHS Department of Homeland Security
DOE Department of Energy
EPA Environmental Protection Agency
FAA Federal Aviation Administration
FAST Fixing America’s Surface Transportation Act (2015)
FHWA Federal Highway Administration
FMCSA Federal Motor Carrier Safety Administration
FRA Federal Railroad Administration
FTA Federal Transit Administration
GHSA Governors Highway Safety Association
HMCRP Hazardous Materials Cooperative Research Program
IEEE Institute of Electrical and Electronics Engineers
ISTEA Intermodal Surface Transportation Efficiency Act of 1991
ITE Institute of Transportation Engineers
MAP-21 Moving Ahead for Progress in the 21st Century Act (2012)
NASA National Aeronautics and Space Administration
NASAO National Association of State Aviation Officials
NCFRP National Cooperative Freight Research Program
NCHRP National Cooperative Highway Research Program
NHTSA National Highway Traffic Safety Administration
NTSB National Transportation Safety Board
PHMSA Pipeline and Hazardous Materials Safety Administration
RITA Research and Innovative Technology Administration
SAE Society of Automotive Engineers
SAFETEA-LU Safe, Accountable, Flexible, Efficient Transportation Equity Act:
A Legacy for Users (2005)
TCRP Transit Cooperative Research Program
TEA-21 Transportation Equity Act for the 21st Century (1998)
TRB Transportation Research Board
TSA Transportation Security Administration
U.S. DOT United States Department of Transportation

Copyright National Academy of Sciences. All rights reserved.

Operational Performance and Safety Effects of Arterial Weaving Sections

Transportation Research Board

500 Fifth Street, NW
Washington, DC 20001

ADDRESS SERVICE REQUESTED

ISBN 978-0-309-70968-2
90000

9 780309 709682

Copyright National Academy of Sciences. All rights reserved.