HCM 2020 PDF
HCM 2020 PDF
HCM 2020 PDF
DETAILS
CONTRIBUTORS
GET THIS BOOK University of Florida Transportation Institute, Cambridge Systematics, and Alex
Skabardonis; National Cooperative Highway Research Program; Transportation
Research Board; National Academies of Sciences, Engineering, and Medicine
FIND RELATED TITLES
SUGGESTED CITATION
Visit the National Academies Press at NAP.edu and login or register to get:
Distribution, posting, or copying of this PDF is strictly prohibited without written permission of the National Academies Press.
(Request Permission) Unless otherwise indicated, all materials in this PDF are copyrighted by the National Academy of Sciences.
NCHRP
Web-Only Document 290:
Cambridge Systematics
Gainesville, FL
Alex Skabardonis
Consultant
Berkeley, CA
COPYRIGHT INFORMATION
Authors herein are responsible for the authenticity of their materials and for obtaining written permissions from publishers or persons who
own the copyright to any previously published or copyrighted material used herein.
Cooperative Research Programs (CRP) grants permission to reproduce material in this publication for classroom and not-for-profit
purposes. Permission is given with the understanding that none of the material will be used to imply TRB, AASHTO, FAA, FHWA, FTA,
GHSA, NHTSA, or TDC endorsement of a particular product, method, or practice. It is expected that those reproducing the material in this
document for educational and not-for-profit uses will give appropriate acknowledgment of the source of any reprinted or reproduced
material. For other uses of the material, request permission from CRP.
DISCLAIMER
The opinions and conclusions expressed or implied in this report are those of the researchers who performed the research. They are not
necessarily those of the Transportation Research Board; the National Academies of Sciences, Engineering, and Medicine; the FHWA; or
the program sponsors.
The information contained in this document was taken directly from the submission of the author(s). This material has not been edited
by TRB.
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.national-academies.org.
The Transportation Research Board is one of seven major programs of the National Academies of Sciences, Engineering, and Medicine.
The mission of the Transportation Research Board is to provide leadership in transportation improvements and innovation through
trusted, timely, impartial, and evidence-based information exchange, research, and advice regarding all modes of transportation. The
Board’s varied activities annually engage about 8,000 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.
ACKNOWLEDGMENT
The research reported herein was performed under NCHRP Project 15-57 by the University of Florida
Transportation Institute (UFTI). UFTI was the contractor for this study, with Cambridge Systematics
and Dr. Alexander Skabardonis as subcontractors. Dr. Lily Elefteriadou, Director at UFTI, was the
Principal Investigator. The other authors of this report are Fabio Sasahara, Graduate Research Assistant
at UFTI; Luan Carvalho, Graduate Research Assistant at UFTI; Tanay Datta Chowdhury, Graduate
Research Assistant at UFTI; Marilo Martin Gasulla, Graduate Research Assistant at UFTI; Richard
Margiotta, PhD, Principal at Cambridge Systematics and Alexander Skabardonis, PhD, Professor at
University of California Berkeley. The authors are thankful to the valuable contributions from the
members from the NCHRP Panel and from the Highway Capacity and Quality of Service Committee
for providing valuable feedback and facilitating contacts with public agencies for the data collection
task. The authors also thank the following public agency staff members who assisted with the site
identification and data collection: Daniel Buidens, Florida DOT; Rodney Carrero-Villa, Florida DOT;
Vicki Haskell, Wisconsin DOT; Brian Kary, Minnesota DOT; Sanhita Lahiri, Virginia DOT; David
Miranda, Georgia DOT; Rina Patolilic, Louisiana DOT; and Peter Vega, Florida DOT. Finally, the
authors acknowledge the valuable contribution of the following student assistants who contributed to
data analysis and reduction: Valerie Bonar, Christian Breau, Michael Hunter, Corbin Kramer, Brandon
Lahore, Sophia Semensky, and Hope Sotolongo-Miranda.
TABLE OF CONTENTS
SUMMARY ............................................................................................................................................... 1
1. INTRODUCTION ............................................................................................................................. 3
2. LITERATURE REVIEW ................................................................................................................ 4
3. CRITICAL REVIEW OF THE HCM .......................................................................................... 8
4. PERFORMANCE MEASURES ..................................................................................................12
5. QUEUE SPILLBACK INTO FREEWAYS ..............................................................................14
6. QUEUE SPILLBACK INTO URBAN STREETS ...................................................................27
7. LANE-BY-LANE SPEED AND FLOW ESTIMATION
METHODS FOR FREEWAYS ....................................................................................................35
8. SYSTEM ANALYSIS .....................................................................................................................41
9. SOFTWARE IMPLEMENTATION ..........................................................................................43
10. CONCLUSIONS AND RECOMMENDATIONS ....................................................................43
REFERENCES .......................................................................................................................................45
- iv -
LIST OF FIGURES
-v-
- vi -
Summary
The procedures detailed in the 6th Edition of the Highway Capacity Manual (HCM) estimate capacity
and several operational measures, including those determining Level of Service (LOS), for freeway
facilities as well as surface streets. However, the HCM lacks a methodology for evaluating the operational
performance of networks involving freeway-urban street interactions. The current HCM methods focus on
operational analysis of individual segments or sections, which may not correspond to the traveler’s
experience of an entire trip which traverses multiple different types of facilities. These methods do not
consider cases in which spillback occurs from one type of facility to another, or across an entire origin-
destination route, which is necessary for evaluation of network performance.
The objective of this research project is to develop materials for the HCM in order to modify the
freeway analysis methods and the urban street methods so that the effects of operations from one facility to
the other can be evaluated. This report summarizes the entire research effort and provides the proposed new
HCM Chapter (Chapter 38) in Appendix A- Chapter 38 - System Analysis. The new methods produced can
be used to evaluate operations along networks that include both freeways and urban streets. The methods
can also evaluate the impact of spillback into freeways and into urban streets from downstream facilities.
In summary, the following were accomplished:
Selection of appropriate performance measures: Travel time was selected as the performance measure
to evaluate highway systems. Travel time measures are already used in the HCM to evaluate urban streets
facilities. For freeway facilities, we developed additional models and methods to evaluate freeway
performance by lane, as spillback affects each lane of the freeway differently. In addition, a trip-based
performance measurement framework was developed to provide travel time estimates for given O-Ds
within a highway network. O-D measures reflect traveler experience and are well aligned with recently
available data collection methods which track individual trips. These new measures are intended to
complement segment-based measures provided in other HCM chapters.
Evaluation of queue spillback into freeways: Queue spillback into the freeway occurs due to insufficient
capacity in at least one element of the off-ramp: either the ramp proper, or the downstream ramp terminal.
The blockage of one or more freeway lanes adversely affects performance, and the extent of the blockage
effects depend on various factors including the design of the facility, the cause of the blockage, and the
length of the queue. Video and detector data from several locations were obtained, and used along with
microsimulation to develop the methodological framework. The methodology developed is based on the
calculation of demand and capacity at the downstream ramp terminal using the respective Interrupted Flow
methods. It expands the Oversaturated Segment Evaluation for freeway facilities (HCM Chapter 25) and
accounts for spillback and its effects by lane along the freeway mainline.
Evaluation of queue spillback into urban streets: Queue spillback into urban streets occurs due to
insufficient discharge capacity into the freeway merge. It may occur due to oversaturated conditions at the
merge segment or the presence of ramp metering. Video and detector data from several locations were
collected to understand how intersections are affected by on-ramp queue spillback. Microsimulation was
used to complement the evaluation of signalized ramp terminals and to analyze systems with unsignalized
intersections. The proposed methodology integrates the Interrupted Flow methodologies with the Freeway
Facilities procedure to account for constraints of the on-ramp capacity. Several adjustments were developed
to estimate the impacts of queue spillback from an on-ramp into upstream signalized and unsignalized
intersections, including roundabouts.
-1-
Lane by lane performance measures for freeways: Freeway speeds can vary widely depending on the
lane used, and each O-D is likely to use a specific lane or set of lanes along each segment. Spillback affects
each freeway lane differently, and its effects depend on the site design and the length of the queue. This
project developed models that estimate speeds and flows by lane. Detector data were collected from a
variety of locations around the US, and analytical models were developed to predict the lane flow
distribution and lane speed. These models considered a variety of factors including v/c ratio, presence of
nearby ramps, heavy vehicle percent, and grade. Regression models built from the field data demonstrated
that FFS and capacity values for each lane can be obtained as a percentage of the segment average with
satisfactory results.
Development of an O-D analysis framework: A new methodology was developed to estimate travel
times by O-D within a highway system. This methodology combines the tools of several HCM chapters
within the Uninterrupted Flow and Interrupted Flow volumes. It also builds on the research previously
described to evaluate queue spillback and system effects.
System Analysis
• HCM Travel time reliability methodologies for Freeways (Chapter 11) and Urban Streets
(Chapter 17) have significant differences in their procedures. Therefore this project made no
attempt to incorporate reliability analysis in the new procedures.
In the future, it would be useful to develop a LOS framework for the travel time performance measures
yielded by this methodology to evaluate highway systems. Similar to other performance measures used in
the HCM, communicating the values of performance measures to a lay audience may be challenging, while
LOS may be a more useful construct.
-2-
1. Introduction
Background
The procedures detailed in the 6th Edition of the Highway Capacity Manual (HCM) estimate capacity
and several operational measures, including those determining Level of Service (LOS), for freeway
facilities as well as surface streets. However, the HCM lacks a methodology for evaluating the operational
performance of networks involving freeway-urban street interactions. The current HCM methods focus on
operational analysis of individual segments or sections, which may not correspond to the traveler’s
experience of an entire trip which traverses multiple different types of facilities. These methods do not
consider cases in which spillback occurs from one type of facility to another, or across an entire origin-
destination route, which is necessary for evaluation of network performance. Currently, the procedure for
evaluating Signalized Intersections (HCM 2016 Chapters 19 and 31) predicts the average expected queue
length at an approach given various combinations of geometric- or traffic-related inputs. Similarly, the
Freeway Facilities procedure (HCM 2016 Chapters 10 and 25) estimates the maximum expected queue
length at an on-ramp in the case of oversaturated conditions on the freeway mainline. However, the effects
of these queues as they propagate upstream – onto a freeway mainline or a surface street intersection – are
not considered.
Appendix A - Chapter 38 - System Analysis (This appendix provides the new HCM Chapter 38, which is
recommended for inclusion in the HCM 6th Edition.)
Appendix B - Off-ramp Queue Spillback Check
Appendix C - Off-ramp Queue Spillback Analysis
Appendix D - On-ramp Queue Spillback Check
Appendix E - On-Ramp Queue Spillback Analysis
Appendix F – Freeway Facilities Lane-by-lane Analysis
Appendix G - Glossary
-3-
2. Literature Review
A review of the literature regarding operational effects of freeway-urban street interactions yielded a
limited number of publications. Although much of the published research on this topic does not directly
address performance measures in a manner consistent with HCM procedures, the concepts and framework
presented in the research are useful in developing the necessary procedural adjustments. For example,
previous research has quantified the delay incurred to freeway mainline vehicles as a result of off-ramp
queue spillback, but the HCM Freeway Facilities procedure does not estimate delay explicitly. Rather,
average travel speed is estimated, and the difference between that and the freeway’s free-flow speed can be
used to estimate delay.
This section first presents an overview of the literature found related to performance measures for
networks, followed by an overview related to the effects of oversaturation on freeway performance. The
last subsection discusses literature related to the effects of oversaturation on surface street performance.
Performance Measurement
A literature review on recent developments regarding performance measures yielded a range of different
approaches for the subject. Dong et al. (2016) propose a travel-time reliability prediction framework for
evaluating service quality on urban streets. According to the authors, “from a driver’s perspective, travel
time and its reliability are considered more intuitive measures of service quality than the levels of service
defined in the Transportation Research Board’s 2010 Highway Capacity Manual”. The proposed
framework, which takes into consideration the weather conditions on travel-time classification, was tested
on a freeway network in Iowa using probe vehicle data (it was also compared to VISSIM simulations with
similar results) with good results in representing field conditions for small networks. The work did not
consider lane-changing behavior and the percent of heavy vehicles in the analysis.
Friedrich (2016) presented an approach for evaluating entire multimodal journeys based on O-D points,
based on the German Guideline for Integrated Network Planning (RIN, 2008). In this framework typical
performance indicators include direct speed (direct distance over time, where direct distance is defined as
the straight-line distance) trip time ratio of public transport/car, and number of transfers. For each one of
the performance indicators, the evaluation outputs are presented as LOS ranges from A to F, similar to the
HCM, as the author states that results in this format are very easy to understand for decision makers.
Arun et al. (2016) present an operational analysis approach focused on the variability of driver
characteristics. The authors indicate that they had difficulties in implementing HCM methods in India due
to the diverse driver behavior and characteristics. According to the authors, the simple use of adjustment
factors is not adequate for representing the heterogeneity in Indian traffic conditions. Therefore, they
explored the possibility of adopting different sets of performance measures to better suit different locations,
and highlighted the strengths and limitations of each one.
In summary, the research team did not find any research directly pertaining to the HCM and systems
analysis. Most research on highway system performance prior to the start of this project had addressed
multimodal trips (for which performance measures are needed to address specific questions such as transit
to private automobile trip times), and performance measurement under varying driver populations.
-4-
spillback, but the HCM Freeway Facilities procedure does not account for this delay caused by queue
spillback into the mainline. Rather, average travel speed is estimated and the difference between that and
the freeway’s free-flow speed can be used to estimate delay.
In a study performed to compare field observations against this theory, Daganzo and Munoz (2000)
confirmed that off-ramp queues spilling over onto the freeway mainline indeed cause a significant reduction
in discharge flows downstream of the exit. During the queued steady state, an average discharge rate of
4,520 vehicles per hour on a three-lane freeway mainline (a 37.2% reduction from the HCM “base” capacity
of 7,200 vehicles per hour, according to the Basic Freeway Segments procedure) was observed. Note that
this rate should be compared to the “discharge capacity” of a mainline section, and the current version of
the HCM uses a generic capacity value, which does not consider the two-capacity phenomenon.
The proportion of exiting vehicles on the freeway also proved to have a significant effect on capacity: on
average, discharge rates increased from 4,520 to 5,720 vehicles per hour (26.5% increase) when the
proportion of exiting vehicles decreased from 29% to 24%, even though the actual number of exiting
vehicles remained constant. In a separate observation, vehicles were found to transition from free-flow
speeds to “queueing speeds” approximately 1 kilometer, or 3,280 feet, upstream of the queue. It was also
observed that drivers tend to adopt larger headway spacings over time “en masse”, and the authors indicated
that this happened likely because once the length and severity of the queue was collectively realized, driver
aggression as a whole subsided.
In a separate paper based on the data obtained in the previously described study, Munoz & Daganzo
(2002) focused instead on the “behavior” of the queue. It was found that the variation in speeds across
mainline lanes is greatest closest to the diverge point, whereas occupancy detectors positioned further
upstream indicated less variation. Specifically, lanes closer to the off-ramp queue were more affected in
terms of speed reduction, whereas the leftmost lane(s) showed very little difference between the presence
of a queue and free-flow conditions. Additionally, non-exiting vehicles in the vicinity of the off-ramp queue
“traveled more cautiously, with slightly wider but predictable spacings … and [that] more lane changes can
be expected.” In terms of capacity reduction, an average discharge flow of 1,500 veh/h/ln was recorded
immediately beyond the diverge point – 25% lower than that which this particular freeway’s geometric
conditions (as estimated by the authors) could potentially accommodate.
A similar study by Cassidy et al (2002) found that, in general, longer exit queues from the over-saturated
off-ramp were accompanied by lower discharge rates for the non-exiting vehicles, although no exact
measure of correlation between the two was established. The authors also note that exiting drivers
sometimes obstructed non-exiting vehicles by attempting to force their way into the queue rather than wait
for their turn. The authors do not discuss the possible correlation between queue length and forced queue
entries. The presence of a queue also affected the non-exiting vehicles’ average speed: upon the onset of
queueing at an off-ramp, non-exiting vehicles reduced their speed across all lanes, reaching speeds as low
as 25 kilometers per hour (15.5 miles per hour) before returning to free-flow speed downstream of the
diverge point.
Traffic operations during an incident may be similar to operations when a queue is present and thus we
briefly review here studies related to the capacity and traffic operational quality during incidents. In a
comprehensive comparison between various incident-related studies in the literature, Lu & Elefteriadou
(2013) present sets of capacity “adjustment factors” based on various conditions. In a regression analysis
developed based on past studies, capacity additions (denoted by a “plus” sign) and reductions (denoted by
a “minus” sign) were found to be dependent on congestion occurrence (+320 veh/h), number of lanes
(+1,213 veh/h/ln) and the number of lanes/shoulder affected (–1,948 veh/h, –1,116 veh/h and –182 veh/h
for shoulder, 1 lane and 2 lanes blocked, respectively). A similar regression structure could conceivably be
developed for queueing-related capacity reduction, albeit one that incorporates the probability of further
lane blockage in consideration of the queue length variability as well as driver variability (and particularly
the probability of forced lane changes into the queue.)
-5-
Based on the review of the literature, we can conclude that estimating operational measures in the case
of spillback from an off-ramp analytically is challenging, as it is very difficult to anticipate the wide variety
of driver actions. The following trends and observations documented in previous research are used in this
research to develop the framework of the proposed methodologies:
• Discharge flows along the mainline are affected by the presence of an off-ramp queue, with one
study observing 4,520 vehicles per hour on a three-lane freeway mainline (≈ 1,500 veh/h/ln.)
• Discharge rates along the mainline increase with decreasing off-ramp flows; they were also found
to increase with decreasing queue lengths, which are correlated with off-ramp demands.
• Rightmost lanes are more affected in terms of speed reduction, whereas the leftmost lane(s) show
very little difference between the presence of a queue and free-flow conditions.
• Exiting drivers sometimes obstruct through lanes by attempting to force their way into the queue.
However, research has not established any quantitative measures for the probability of such
blockage and its potential association with the off-ramp queue length.
• The presence of a queue at the off-ramp reduces the mainline vehicles’ speed with values
observed as low as 25 kilometers per hour (15.5 mi/h)
If the off-ramp queue blocks the right-most lane and is relatively short, non-exiting mainline drivers may
be willing to remain in the queued lane – albeit at a significantly reduced travel speed – and accept a small
amount of delay.
An FHWA-sponsored study (Saxton DTFH61-12-D-00020, Task Order 15: Highway Capacity Manual
(HCM) Systems Analysis Methodology) conducted by UF (PI: L. Elefteriadou) developed preliminary
procedures for conducting network analysis. However, the scope of that project did not include data
collection. The report proposed a series of modifications to the HCM in order to address spillback
conditions, and the research team used some of the findings and recommendations from that report in
crafting the data collection effort for this project.
-6-
In conclusion, there are tools available to estimate the lost time incurred at the upstream intersections as
a function of a downstream queue length and storage availability. Still, extensive field observations are
necessary to document discharge rates at congested on-ramps and based on these estimate the resultant
queue length along the on-ramp.
-7-
This section presents a critical review of the methods in the Highway Capacity Manual 6th Edition,
highlighting strengths and identifying weaknesses related to the evaluation of trips using freeways and
surface streets.
A detailed discussion on performance measures and the proposed trip-based evaluation framework is
presented in Section 4 – Discussion on Performance Measures.
Freeway Facilities
This subsection presents and discusses specific issues related to HCM freeway facilities analysis that
pertain to systems analysis.
Performance measurement: Freeway analysis is based on tools that estimate the operational
performance of each type of segment. The freeway systems analysis method evaluates operations in time
and space considering, to some degree, interactions (queue effects) between consecutive segments.
Operational performance is determined based on density and speed along each segment. In the case of
-8-
ramps, performance measures are estimated separately for the influence area (consisting of the two
rightmost lanes, and for a length of 1,500 ft.) and the remaining leftmost lanes. Based on these methods,
one can derive travel time for each segment and stitch together the travel time for the entire facility.
However, the quality of a particular trip from an origin to a destination cannot be obtained, because the
method does not provide operational performance by lane. In the field, speed can vary widely depending
on the lane used, and each origin-destination (O-D) is likely to use a specific lane or set of lanes along each
segment. For example, travelers exiting at a congested off-ramp will experience a much different travel
time than those using the left most lanes of the same segment. Currently, LOS for the facility is estimated
as a function of the LOS for each segment, and does not consider the quality of the trip from a traveler’s
perspective. However, travelers experience travel time over the entire system, as well as based on their O-
D. A performance measurement scheme for the entire system should also consider the origins and
destinations of the travelers within each segment.
Furthermore, five of the six LOS ranges exist where speeds are relatively high (above approximately 50
mi/h) and only one LOS range is used to define the congested regime, which is of the highest interest in
large urbanized areas. Many congestion management techniques will improve congestion (e.g., delay) but
the facility will still be classified as LOS F under the current framework.
Spillback from a downstream facility: The HCM addresses queue spillback only when it is fully
contained within the boundaries of a freeway facility. Spillback onto the freeway may occur either due to
inadequate capacity of the ramp proper, or due to inadequate capacity at the ramp terminal (typically the
signal at the downstream interchange). The capacity of the ramp proper is defined as the off-ramp’s
maximum allowable hourly flow rate based on its geometric characteristics (number of lanes, free-flow
speed, etc.). The capacity of the ramp terminal is defined as the capacity of the signalized or unsignalized
approach to the surface street. The current procedures in HCM provide guidance on estimating queue
storage ratio on ramp terminals, but do not directly address situations when ramps have queue-to-storage
ratio > 1.0 (LOS F at diverge ramps).
Influence area for merge and diverge areas: Currently, the HCM methodology for merge and diverge
areas focuses on predicting performance within a 1,500 ft. influence area and for the two rightmost lanes.
Figure 2 provides a schematic of the diverge influence area, according to the HCM. When spillback occurs,
it is likely that queue length extends upstream beyond 1,500 ft. Thus, the influence area of the junction may
be significantly longer. Also, the influence area may vary by time period depending on the demand-to-
capacity ratio and its variability. Lastly, the effects of spillback may affect additional lanes, as through
vehicles attempt to avoid the spillback and they increasingly use the leftmost lanes to maintain their speed.
The current freeway systems analysis framework, which uses a constant area (longitudinal distance and
cross section) to define this influence area, must be revised to consider spillback effects.
-9-
Impacts of spillback on weaving segments: Similar considerations to those described above are also
relevant for weaving segments. Along weaves, the on-ramp that is part of the weave may be affected in
different ways based on whether the spillback reaches its gore point or not. The area available to the two
weaving traffic streams would be reduced as a function of spillback. When longer queues prevail, the
upstream on-ramp may be entirely blocked, and it is possible that in this case the on-ramp and off-ramp
function independently, as there is no space available for weaving maneuvers.
Capacity constraints and speed reduction due to spillback: Spillback along the mainline results in
blockage of one or more lanes. Thus, capacity is reduced during such conditions, and the capacity reduction
depends on the frequency and duration of lane blockage. Also, lanes adjacent to those blocked are likely to
have reduced operating speeds. Those effects and their impact on the overall capacity and quality of service
must be addressed.
Ramp throughput during congested conditions: In order to consider the effects of an on-ramp queue to
upstream surface streets, it is necessary to estimate the queue length as a function of demand and capacity.
However, when the freeway mainline is congested, it is not clear what the capacity of the ramp is. In other
words, the HCM does not provide the actual relative contributions of the ramp demand vs. the mainline
demand at the merge during congested conditions; it assumes that the merge operates as a “zipper”, with
equal contributions from the shoulder lane and the on-ramp. However, the resulting ramp flow may be a
function of geometry, the extent and duration of congestion on the mainline, or prevailing driver behavior
in the region. The ramp capacity may be higher for weaving segments with lower freeway to off-ramp
demands and a longer weaving length. Estimating this discharge rate is necessary in order to estimate the
resulting queue length along the on-ramp. In cases where ramp metering is present, the discharge rate can
be determined and used to obtain the mainline input flow as defined by the HCM for oversaturated freeway
segments in Chapter 25.
Spillback effects along the mainline: Revised analyses should also consider the effects of spillback to
upstream segments, including basic freeway segments. For example, current procedures determine whether
the subject diverge ramp operates as an isolated ramp, considering the respective freeway and ramp
demands. These procedures should be revised to consider that the area of influence may vary as a function
of queue length and other factors related to spillback (for example the number of lanes blocked and their
respective frequency). Additionally, the lanes of the upstream segments may be differently affected,
depending on the spillback queue length.
Urban Streets
This subsection presents and discusses specific issues related to HCM urban streets analysis that pertain
to systems analysis.
Performance measurement. The urban streets methodology considers a variety of different facilities,
including street segments, signalized intersections, interchange ramps terminals (IRT), two-way stop-
controlled intersections (TWSC), all-way stop-controlled intersections (AWSC) and roundabouts. Each one
of these facilities have an associated MOE; intersections use control delay, while segments use travel speed.
Thus, development of a performance measurement framework for this group of facilities is easier, as these
existing measures can be converted to travel time.
Spillback from a downstream facility onto a signalized IRT. The connection between freeways and
urban streets is typically at an IRT. The HCM does not currently address spillback from freeways into
IRTs, however, the IRT methodology does address spillback from one signalized intersection to the other.
- 10 -
The procedure includes an adjustment to consider spillback from the downstream intersection to the
upstream in the form of additional lost time. This lost time is estimated for each upstream movement as a
function of the downstream queue length and storage availability. Similarly, the HCM Urban Streets
methodology provides an adjustment to the saturation flow rate to account for spillback effects. It is not
clear whether one approach may be preferable than the other. However, we recommend that the HCM
provides a consistent approach for addressing spillback onto signalized intersections and interchanges,
whether the queue is originating from a downstream intersection or from a freeway on-ramp.
Spillback from a downstream facility onto a roundabout IRT. Roundabouts are especially sensitive to
queue spillback, since it can result in complete gridlock for all movements. The current roundabouts
procedure does not evaluate roundabouts considering spillback.
Evaluation of actuated control for an intersection/interchange with spillback. The current signalized
methodology analysis framework considers both pre-timed and actuated control. In the case of pre-timed
control, signal control is an input, and the effective green can be adjusted to account for spillback. However,
in the case of actuated control, the signal phase duration is variable. This creates the following issues when
spillback occurs: a) the phase duration estimation would be impacted and thus the methodology needs to
be adjusted accordingly; and b) actuated control results in variable effective greens and thus variable
arrivals to the downstream ramp. These variable arrivals would result in variable queues at the ramp, which
will in turn affect the phase duration at the signal. This creates an iterative process which an HCM-type
analysis cannot address. This research project uses suitable assumptions and simplifications to provide a
reasonable framework for systems analysis, and uses equivalent pre-timed control to estimate phase
durations.
Queue length at the ramp receiving traffic from an intersection/interchange. Any type of queueing
analysis depends on the arrival patterns and the service patterns. In the case of a ramp junction receiving
traffic from an upstream intersection, the queue is calculated as a function of: a) the arrival patterns
upstream of the ramp - these are a function of the type of control, as well as the arrivals from each incoming
traffic stream; b) the departure rate at the ramp into the mainline - this is equal to the arrivals when
conditions are undersaturated. However, for oversaturated conditions, it is not clear what the discharge rate
is, and how it varies within an analysis period.
- 11 -
4. Performance Measures
This project develops trip-based (or O-D based) travel time measures in addition to the traditional facility-
based measures used in the HCM. Performance measures that relate to the trips taken by travelers are
considered the best approach, for the following reasons:
• Trips relate more directly to the traveler experience – they are how travelers actually experience
the system. Facility- and segment-based measures are still useful in this context because they
help practitioners identify and treat specific problems, such as bottlenecks.
• Each O-D within a freeway and urban street system uses a specific lane or set of lanes. Under
congested conditions, each lane across the facility may be operating very differently, and thus
evaluating the performance of the O-D would take into account the specific lane usage for those
trips.
• Measuring performance at the trip level is much more revealing for advanced operations
strategies, such as Active Traffic and Demand Management (ATDM) and Integrated Corridor
Management (ICM). Many components of these strategies seek to influence demand and to
balance demand with available capacity. Understanding what effect strategies have on trips is
important for selecting and modifying these strategies.
• Moving to trip-based measures is a logical step in the evolution of the HCM. The discipline of
performance measurement is moving in the direction of trip-based measures. As additional data
become available over the next several years to track the movement of individuals (hand-held
devices) and vehicles (connected vehicles, and eventually, automated vehicles), development of
trip-based measures from direct measurements will be enabled. Currently, it is possible to derive
trip-based measures synthetically, but when direct measurements become available, we expect a
rapid increase in the interest for trip-based measures.
• Trip-based measures are the basis for measuring accessibility. A logical extension of trip-based
measures, which are based on travel between specific origins and destinations, is the
measurement of accessibility. Accessibility is defined by two components: 1) the presence of
opportunities; and 2) the ease with which those opportunities can be obtained or “reached.”
Accessibility can be improved by improving the movement between opportunities or by moving
opportunities closer to travelers. Because it considers not only the performance of trips on the
system but also how trips interact with the entire built environment, accessibility is a key
indicator of the quality of life for a transportation agency’s customers.
Figure 3 shows the strategy behind selecting performance measures for the project. Both facility and trip
measures are recommended. Trip-based travel time measures are the most direct indicators of the traveler
experience but facility-based travel time measures also are useful for analysts to understand since many
improvement strategies are facility oriented. Other facility-based measures not defined in terms of travel
time are also recommended. These measures describe how the system is performing and are useful for
diagnosing problems. All three types of measures are needed to provide a complete picture of performance.
- 12 -
• Traveler Focus
Trip-Based Measures
• Multiple Facilities
• Multiple Modes
Travel Time • Precursor to Accessibility
Outcome Measures
• Individual Facilities
Facility-Based Measures
• Bottleneck/Problem Identification
• Individual Facilities
Facility-Based Output Measures
• Vehicle-Oriented
• Non-travel time measures
• Help to Explain Outcomes
Segment d/c
VMT/PMT
Queue Statistics
- 13 -
Queue spillback into freeway facilities occurs due to insufficient capacity in at least one element of the
off-ramp, namely the ramp proper or the downstream ramp terminal. The impact of the spillback on the
freeway mainline can be restricted to the exit area or can extend for miles, depending on a series of factors.
This section summarizes the data collection and analysis process for developing the methodology to
evaluate queue spillback into freeways. Field data and simulation were used to develop the procedure,
which is presented in Appendix B - Off-ramp Queue Spillback Check and Appendix C - Off-ramp – Queue
Spillback Analysis.
Data Collection
Data collection (Task 6) is a critical part in the development of analytical procedures to evaluate the
interactions between freeway and urban street facilities. The research team used a combination of field data
and simulation modeling in order to comprehensively evaluate a variety of designs and traffic conditions
that can be found throughout the U.S. The use of simulated data to complement field data was made
necessary for two important reasons:
1. The number of feasible traffic, control, and highway design characteristics available is extremely
high, making a pure field data collection approach unrealistic.
2. Simulation allows us to isolate the effects of a specific variable (for example, the length of the
acceleration lane) on traffic operations. Field data are subject to day-to-day noise in traffic
attributes (because of fluctuations in demand, weather, incidents, special events, etc.) that make it
difficult to isolate the effects of a single variable.
Even though microsimulation cannot and should not replace field data collection, it can very effectively
supplement field data. Properly calibrated simulation models can replicate driver behavior and
aggressiveness based on simple observations of gap acceptance, headway distributions, number of lane
changes per mile, etc.
This section describes the data collection for each of the three major components, including: data
requirements, dataset description, data reduction efforts, agencies contacted and application of
microsimulation.
Data Requirements
The process of identifying potential study locations started by identifying sites that regularly experience
spillback into freeways, and have instrumentation and data available that can be used by the research
team. Another key requirement for suitable study locations was that the off-ramp is the primary and only
freeway bottleneck to allow the observation of queue spillback effects in isolation. Any locations
experiencing overlapping bottlenecks at the diverge region were discarded from the process.
The data required for studying spillback into freeways is shown in Figure 4, and each of the components
is briefly discussed next.
- 14 -
Figure 4 – Data collection framework for queue spillback into freeway facilities
1. Video Cameras: video observations are the most critical data for the study of queue spillback, as
they provide information that cannot be accurately captured by loop detectors. Video recordings
can provide views of the off-ramp area so that the queues forming longitudinally and laterally due
to spillback along the freeway can be observed and measured. It also provides important insight
on driver behavior, lane changing activity, and forced merging.
2. Upstream Detectors: detectors are a critical source for speed and flow data. It is essential that
detectors are able to provide data on individual lanes and in raw format (minimal aggregation).
The available sources used for this project provide raw data aggregated in intervals between 20s
to 60s. Locations with multiple detectors upstream of the off-ramp were also preferred, as they
provide more comprehensive data to evaluate the changes in performance measures along the
freeway longitudinally.
3. Downstream Detectors: downstream detectors are not strictly required for the analysis, but are
useful to ensure that conditions downstream are undersaturated. Any sites where downstream
detectors showed oversaturated conditions were discarded.
4. Ramp Detectors: ramp detectors along the off-ramp can provide data to estimate the off-ramp
demand and capacity.
5. Geometry: geometric features such as number of lanes in the freeway and ramp, deceleration lane
length, lane width and others can be obtained using satellite tools such as Google Maps.
Agencies Contacted
The strict requirements for data collection to observe off-ramp queue spillback, as previously mentioned,
created a significant challenge. The research team contacted state Department of Transportation (DOT)
staff in each of the 50 states, obtaining a positive response from 21 states (Figure 5). From this sample, we
were able to identify study sites from five states that met all the required criteria: California, Georgia,
Virginia, Florida and Minnesota. The remaining 16 states that responded were able to identify locations
with queue spillback, but they did not meet the full data requirements previously stated, and therefore these
locations could not be used.
- 15 -
Figure 5 – Survey for off-ramp queue spillback study sites – outcome by state
Dataset Description
Several potential locations were selected for data collection, and only those that met all the criteria
previously described were kept in the final dataset. For the selected locations, detector data were extracted
from online repositories for each State DOT. Video recordings were extracted from online repositories or
recordings provided by DOT staff. Table 1 summarizes the data sources by state.
Table 2 summarizes the final list of ten study locations, with a description of their key characteristics and
number of video observations. A video observation consists of one peak period recording, where the
development and discharge of queues can be observed, which typically lasts between 2-4 hours.
- 16 -
Segment Type
Observations
Downstream
Peak Period
Directional
Off-Ramp
Terminal
Spillback
Regime
Ramp
Lanes
Video
lanes
LOCATION
Signalized
Miami/FL - I-95 SW 25th Rd. 3 Diverge 2 13 PM 4
Intersection
Freeway
Tampa/FL - I-275 NB to W Kennedy Blvd 4 Diverge 2 6 PM 4
Merge
Signalized
Norfolk/VA - I-64 WB to Northampton Blvd 4 Diverge 2 10 PM 3
Intersection
Centreville/VA - I-66 WB to SR-28 (Sully Signalized
4 Diverge 1 11 AM 3
Rd) Intersection
Signalized
Centreville/VA - I-66 EB to SR-28 (Sully Rd) 4 Diverge 1 11 AM 3
Intersection
Signalized
Minneapolis/MN - I-35W SB to 35th St. 4 Weaving 1 5 PM 3
Intersection
Freeway
Atlanta/GA - I-285 NB to I-20 5 Diverge 1 7 PM 3
Merge
Freeway
Miami/FL - I-75 SB to SR 826 5 Diverge 2 13 AM 4
Merge
Freeway
Atlanta/GA - I-285 NB to GA-141 5 Diverge 1 12 PM 3
Merge
Freeway
Sacramento/CA - SR-99 NB to SR-50 5 Weaving 2 5 AM 4
Merge
TOTAL 93
Schematics and sample screenshots of the study locations are presented in Figure 6 through Figure 15.
Source: Photo provided by the Florida Department of Source: Photo provided by the Florida Department of
Transportation Transportation
Figure 6 – Miami, FL – I-95 SB to SW 25th Rd. Figure 7 – Tampa, FL – I-275 NB to SR-60
- 17 -
Source: Photo provided by CATT Lab (Ritis.org) Source: Photo provided by CATT Lab (Ritis.org)
Figure 8 – Norfolk, VA – I-64 WB to Northampton Figure 9 – Centreville, VA – I-66 WB to SR-28
Blvd (Sully Rd)
Source: Photo provided by CATT Lab (Ritis.org) Source: Photo provided by the Minnesota Department of
Figure 10 – Centreville, VA – I-66 EB to SR-28 Transportation
(Sully Rd) Figure 11 – Minneapolis, MN – I-35W SB to 35th
St.
- 18 -
Source: Photo provided by the Georgia Department of Source: Photo provided by the Florida Department of
Transportation Transportation
Figure 12 – Atlanta, GA – I-285 NB to I-20 Figure 13 – Miami, FL – I-75 SB to SR 826
Source: Photo provided by the Georgia Department of Source: Photo provided by the California Department of
Transportation Transportation
Figure 14 – Atlanta, GA – I-285 NB to GA-141 Figure 15 – Sacramento, CA – SR-99 NB to SR-50
- 19 -
Data Reduction
Video observations were the most challenging part of the data reduction, as the process of obtaining the
recordings was very time-consuming. After the recordings were obtained from the respective agencies, the
research team performed a first screening at the videos for an initial validation, which included the
following conditions:
• Camera is able to show the queues in the diverge area as required for project purposes;
• No occurrence of incidents, disabled vehicles, lane closures or other factors that may affect the
daily dynamics of vehicular traffic;
• No overlapping bottlenecks – the off-ramp is the only source of congestion.
After this first screening, the next step was downloading detector raw data (speed and flow) from online
repositories for the periods captured in the video files. Then, any observations with the following issues
were also excluded from the dataset:
The valid observations remaining after the initial screenings make up the study dataset, previously
presented in Table 2. The next step in the data reduction was to measure the queue lengths from video
observations. Queue lengths were measured visually from video starting from the exit point by lane, as
illustrated in Figure 16. Vehicles queued on the adjacent lane were counted as part of the deceleration lane
queue, even though they are in a mainline lane. We chose this approach because the downstream end of
this queue is not always at the same location. Also, this approach makes the L2 queue measurements
consistent with the L1 queue, which starts from the end of the deceleration. Measurements were taken in
time intervals consistent with the detector aggregation times so queue lengths and flow/speed data can be
matched for further analyses. For example, if raw detector data from a given location are aggregated in 20-
second intervals, then queue lengths were measured every 20 seconds.
After the data reduction, queue lengths were matched with detector data to provide insights on the effects
of queue spillback in the freeway performance. Figure 17 provides a sample of data illustrating queue
spillback, where (a) shows the development of queues over time, while (b) shows the speed drops consistent
with the development of mainline queues. At this site, queues along lane 2 were longer than those in lane 1
due to demands at the downstream intersection. Speeds on lane 1 (S1) are lower than lane 4 (S4), consistent
with the occurrence of queue spillback on the right side of the freeway.
- 20 -
Figure 17 – Sample of data illustrating queue spillback: (a) queue length and (b) speeds by lane
Qualitative Observations
The videos obtained were used to observe operations during off-ramp queue spillback, in order to inform
the development of analytical models. These observations are discussed in the following paragraphs.
For off-ramps with two lanes, it was observed that queues may not distribute evenly along the two lanes.
Drivers typically choose a position at the off-ramp based on their next movement at the downstream ramp
terminal. Therefore, if the downstream ramp terminal has one particular movement with excessive demand-
to-capacity ratio, queues are likely to develop along the lane connected to that specific movement, while
the other off-ramp lane remains underutilized. Figure 18 shows examples of unbalanced demand at the off-
ramp on (a) the right lane and (b) the left lane of a two-lane off-ramp.
- 21 -
Source: Photos provided by (a)CATT Lab (Ritis.org) and (b)Florida Department of Transportation
Figure 18 – Unbalanced lane usage at the off-ramp: (a) right lane and (b) left lane
When multiple detector stations are present upstream of the off-ramp, it is possible to compare the effects
of queue spillback along multiple locations during the exact same time. Field data show that in the area
close to the ramp exit, the effects of queue spillback are restricted to the blocked lanes, while the through
vehicles in the leftmost lanes are not affected. However, for locations further upstream, speed drops are
more homogenous and evenly spread across all lanes of the freeway. Figure 19 illustrates the impacts of
queue spillback at three locations upstream of an off-ramp bottleneck: (a) at the ramp influence area, (b)
1,800 ft. upstream of the exit and (c) 5,500 ft. upstream of the exit.
When there is queue spillback from an off-ramp, queues may block one or two lanes in the freeway
mainline (Figure 20). Video observations of queues show that at each site experiencing recurrent queue
spillback that extends beyond the deceleration lane, the number of blocked lanes does not change. In other
words, queue length does not affect whether the site experiences the conditions of Figure 20(a) or (b). The
condition shown in Figure 20(b) occurs more frequently in locations with more aggressive driver behavior.
- 22 -
When the segment is a lane-drop (rather than a diverge) the exiting traffic can access the off-ramp with
a single lane change. Therefore, drivers are more likely to wait until they are closer to the exit to change
lanes, blocking the adjacent through lane.
Figure 20 – Blockage of one (a) or two (b) mainline lanes due to off-ramp queue spillback
Simulated Sites
Given the complexity of obtaining field data to evaluate queue spillback, simulation was used to
complement the data collection. The selection of locations for simulation took into consideration the use of
pre-calibrated locations that were available to the research team. The simulated sites were used with
increasing input demands so that queue spillback into freeways would occur. Table 3 presents the list of
simulated locations for this study.
Number of Number of
Downstream
Location freeway Segment type off-ramp
ramp terminal
lanes lanes
I-105 SB to Bellflower Blvd. 3 Diverge Signalized int. 1
I-105 NB to Bellflower Blvd. 4 Diverge Signalized int. 1
I-710 SB to Martin Luther King Blvd. 4 Diverge Signalized int. 2
I-105 WB to Garfield Ave. 4 Major Diverge Signalized int. 2
I-710 SB to I-105 5 Diverge Freeway merge 2
I-710 SB to Martin Luther King Blvd. 5 Weaving Signalized int. 2
Simulation models developed and calibrated with field data in AIMSUN were used to obtain simulated
data. The following assumptions were made for the simulated data:
Queue measurement: Queued vehicles in a freeway off-ramp are not completely stationary. During
queue spillback, exiting vehicles are typically moving at slow speeds. Additionally, vehicles are closely
spaced but not as close as vehicles stopped in an intersection. Queue lengths were measured visually from
video observations. However, quantitative criteria must be set to define freeway queue lengths in
microsimulation. Vehicles were considered in a queue if they meet the following criteria:
• Vehicle speeds are not greater than 5 mi/h
• Distance between vehicles is shorter than 40 ft.
Detectors: The greatest benefit of microsimulation is the ability to obtain a variety of performance
measures at any desired location of the simulated network. For field data, speed and flow readings were
- 23 -
available according to the position of detectors and were different for each study location. For the simulated
sites, detector stations were placed at three locations upstream of the exit: 1,500 ft, 4,000 ft and 8,000 ft.
Speeds and flows on a lane-by-lane basis were extracted in 20-second intervals, similarly to the lowest
resolution obtained from field data.
Forced merging: As queues extend further upstream of the ramp exit, drivers wishing to exit are more
likely to change lanes aggressively to join the back of the queue. Some vehicles may fail to join the back
of the queue and then attempt to find a gap to merge into the queue (forced merging). A microsimulation
model where drivers can perfectly anticipate a lane blockage ahead and always join the back of queue would
fail to replicate the field conditions of an off-ramp queue spillback. Therefore, for the simulation models, a
vehicle with an O-D that includes the off-ramp as a destination adjusts its position stochastically, which
allows a possibility that it may miss the back of the queue. If that happens, it attempts to find gaps in the
queue to perform a lane change, as illustrated in Figure 21.
Figure 21 – Example of a forced merging maneuver for a vehicle attempting to join the off-ramp
queue
Figure 22 provides an example of a simulated location where queue spillback occurs (I-105@Bellflower
Interchange). As shown, at the vicinity of the off-ramp the blockage occurs in the rightmost lane, with
negligible effects on the adjacent lanes. At sections further upstream, additional turbulence starts affecting
the performance of the other lanes in the freeway mainline. As shown, the simulation replicates the field
observations as expected.
- 24 -
The first step to evaluate queue spillback into freeways is to assess whether it is expected to occur. The
detailed methodology for off-ramp spillback check is provided in Appendix B - Off-ramp Queue Spillback
Check.
The methodology evaluates two potential capacity bottlenecks in the off-ramp (Figure 23):
• Ramp proper: If the off-ramp demand is greater than the capacity of the ramp proper, spillback is
expected to occur.
• Downstream ramp terminal (intersection or a merge segment, in the case of a freeway-to-freeway
connector): if there is insufficient capacity at the ramp terminal, queues will develop along the off-
ramp. The procedure then compares the expected queue length and the available queue storage. If
the expected queue length is greater than the queue storage, spillback is expected to occur.
If queue spillback is not expected to occur, no adjustments are required for the current HCM methods. If
queue spillback is expected to occur, the procedure for evaluating queue spillback must be applied.
- 25 -
- 26 -
Queue spillback into an urban street occurs due to insufficient merging capacity in a freeway on-ramp,
and may have the following causes:
• Insufficient capacity at the freeway merge segment, per guidance provided on HCM Chapter 14
– Merge and Diverge Segments;
• Insufficient capacity of the on-ramp (HCM Exhibit 14-12);
• Active ramp metering at the on-ramp.
If insufficient capacity at the on-ramp occurs and its queue storage is also insufficient, queue spillback is
expected to occur at the upstream intersection. The project has developed procedures for evaluating
spillback into signalized intersections as well as unsignalized intersections (TWSC, AWSC and
roundabouts). This section presents the data collection process for queue spillback into urban streets. It
discusses field data and simulation, qualitative analysis, and the developed methodology to evaluate the
occurrence and impacts of queue spillback from an on-ramp.
Data Collection
The data collection process for queue spillback into urban streets is very similar to the method used for
off-ramp queue spillback. This section describes the details of data collection for spillback into urban
streets.
Data Requirements
The data collection needs for studying spillback into urban streets are shown in Figure 25. The figure
refers specifically to a signalized intersection as this type of facility has the most complex data collection
requirements. Unsignalized intersections have a similar set of data collection requirements, excluding the
signal controller configuration.
Figure 25 – Data collection framework for queue spillback into urban streets (signalized
intersection)
The data collection specifics for this portion of the project are as follows:
- 27 -
1. Mainline detectors: Data were obtained from loop detectors located upstream and downstream of
the merge area, providing lane-by-lane speed/volume/occupancy measurements, with resolution
of 1 minute or less.
2. Queue/ramp detectors: Data from detectors along the ramp were obtained to provide the queue
length and available queue storage along the ramp. If the ramp is metered, the metering rates were
also provided.
3. Signal controller configuration: For signalized intersections, the following signal timing
parameters were collected: pre-timed/actuated, phasing, cycle length, green/yellow/all red
intervals, and actuation parameters (min green, max green, passage time, etc.)
4. Video cameras: It was essential to have video at the interchange in order to view spillback
occurrence and the available queue storage along the ramp together with the approaching
movements’ discharge rates.
5. Automated traffic volume counts: If available, traffic volume counts were obtained to save time
and effort in data reduction.
Agencies Contacted
The same agencies that were contacted for off-ramp queue spillback data (Figure 5) were also contacted
for potential locations to observe on-ramp queue spillback. The data collection for urban street queue
spillback was significantly more challenging when compared to off-ramp queue spillback, for the following
reasons:
• A very limited number of DOTs have access to cameras located in urban streets. In most cases,
these cameras are operated by municipalities or counties, which frequently lacked the resources to
capture and record videos for our research purposes;
• Identifying locations with queue spillback into unsignalized intersections was extremely difficult.
The research team was able to identify a few locations where roundabout ramp terminals experience
on-ramp queue spillback, but in all cases these locations did not have cameras able to capture video.
For this reason, unsignalized intersections were analyzed using microsimulation.
Dataset Description
Table 4 summarizes the list of the four study locations where field data were collected, with a description
of their key characteristics and number of video observations. Each video observation typically lasts
between 1-3 hours, and it corresponds to one peak period recording, where the development and discharge
of queues can be observed.
- 28 -
Table 4 – Summary of study locations – field data collection – on-ramp queue spillback
Number of
Interchange Intersection # video
Location on-ramp
type type observations
lanes
Atlanta/GA - I-285W NB @
Diamond Signalized 2 9
South Cobb Dr NB
Atlanta/GA - I-285W SB to
Diamond Signalized 1 10
Cobb Pkwy Dr.
Atlanta/GA - I-285W NB to
Diamond Signalized 1 10
Cobb Pkwy Dr.
San Diego/CA - Friars Rd. @ I-
Parclo Signalized 2 10
15 NB
Schematics and sample screenshots of the study locations are presented next.
This diamond interchange connects the I-285W NB freeway with the South Cobb Dr. arterial corridor.
During the AM peak period, the freeway facility experiences heavy congestion, while ramp metering at the
on-ramp limits the number of vehicles merging into the freeway. This causes queues to develop along the
on-ramp, which has insufficient storage capacity and causes spillback into the upstream signalized
intersection.
Two movements at the intersections are affected by the queue spillback. First, the SB-L movement Figure
26a) discharges directly into the on-ramp, which reaches its capacity before the end of green. When this
occurs, SB-L vehicles are held behind the stop bar as they are unable to proceed to the on-ramp. The second
affected movement is the NB-Th/R (Figure 26b), which receives green immediately after the SB-L. The
on-ramp is completely occupied by the queue from the previous movement, restricting the capacity of the
NB-R movement. The NB-Th movement is indirectly affected, as the last vehicles discharging from the
SB-L movement get trapped inside the box and partially block the movement of NB-Th vehicles at the
beginning of their green.
- 29 -
Source: Photo provided by the Georgia Department of Transportation and Google Maps
Figure 26 – I-285 NB @ South Cobb Dr. and affected movements by on-ramp queue spillback: (a)
NB-Th/R and (b) SB-L
This diamond interchange connects the I-285W SB freeway with the Cobb Parkway Dr. arterial corridor.
During the PM peak period, the freeway facility experiences heavy congestion, and the number of vehicles
merging into the freeway is constrained by the capacity of the merge. The spillback pattern observed at this
intersection is similar to the previously described location (I-285W NB@ South Cobb Dr). The heavy
demand of the left-turn movement from the major street causes queue spillback at the on-ramp.
Consequently, the opposing through movement is also impacted by the residual left-turn vehicles which
block the intersection.
- 30 -
Source: Photos provided by the Georgia Department of Transportation and Google Maps
Figure 27 – I-285W SB @ Cobb Pwky Dr. and affected movements by on-ramp queue spillback: (a)
NB-L and (b) SB-Th/R
This intersection is also part of the diamond interchange connecting the I-285W freeway with the Cobb
Parkway Dr. arterial corridor, previously described. The northbound direction of the freeway also
experiences heavy congestion, causing queue spillback into the signalized intersection. The intersection
does not have a high demand for the SB-L movement, therefore it does not cause queue spillback. However,
the NB-R movement’s demand is often higher than the on-ramp capacity and causes spillback (Figure 28).
Source: Photos provided by the Georgia Department of Transportation and Google Maps
Figure 28 – I-285W NB @Cobb Pkwy Dr. with NB-R movement affected by on-ramp queue spillback
- 31 -
This partial cloverleaf interchange connects the I-15 NB freeway facility to the Friars Rd. arterial
corridor. The freeway experiences heavy congestion during the AM peak period. Active ramp metering
also limits the on-ramp throughput into the freeway. As a consequence, queue spillback occurs frequently
into this intersection.
The only movement affected at this intersection is the WB-R (Figure 29). This is a protected-only right-
turn movement with a high demand. At the onset of green, the on-ramp has considerable storage space, but
it is occupied quickly as the WBR discharges into the on-ramp. After several cycles of high demand,
vehicles are unable to discharge into the on-ramp during the green interval, causing queues to grow further
upstream.
Source: Photos provided by the City of San Diego and Google Maps
Figure 29 – Friars Rd. @ I-15 NB with WB-R movement affected by on-ramp queue spillback
Data Reduction
Similar to the data collection for off-ramps, the process of obtaining video observations was very
challenging. After the recordings were obtained from the respective agencies, the research team performed
a first screening of the videos for an initial evaluation, which considered the following:
• Camera shows the approaches at the intersection and at least part of the on-ramp;
- 32 -
• No occurrence of incidents, disabled vehicles, lane closures or other factors that may affect the
natural dynamics of vehicular traffic;
Qualitative Observations
The videos obtained were used to observe operations during queue spillback, in order to inform the
development of analytical models. These observations are discussed in the following paragraphs.
When queue spillback from an on-ramp reaches the upstream intersection, the movements that discharge
into the on-ramp are directly affected. However, field observations show that other movements that do not
discharge into the on-ramp can be affected by vehicles trapped inside the intersection box. The most
common case occurs at intersections with leading left turn operation, as illustrated in the example of Figure
30. The SB-L movement discharges into the on-ramp that has insufficient storage to accommodate the
demand. At the end of the SB-L movement, some vehicles are trapped inside the intersection box.
Therefore, at the onset of green for the NB-Th movement, only the leftmost lane (L1) is clear to discharge
at the saturation flow rate. The other lanes (L2, L3 and L4) cannot discharge immediately at the start of
green and have to wait until the on-ramp queue clears, resulting in additional lost time for these lanes.
Figure 30 – Partial blockage of northbound through movement by vehicles trapped inside the
intersection box
- 33 -
Simulated Sites
The use of microsimulation is especially important for on-ramp queue spillback, due to the difficulties
collecting data for unsignalized intersections. The same interchanges used for simulation of off-ramp queue
spillback were used to simulate on-ramp queue spillback. These models were calibrated with a signalized
intersection operation, and these were replaced with stop-controlled and roundabout intersections, as
illustrated in Table 5. The intersection volumes were also changed to ensure on-ramp queue spillback
occurs.
- 34 -
If spillback is detected, the queue length along the on-ramp is estimated and compared to the available
storage. If the queue length is greater than the available storage length, then spillback is expected to occur.
The analyst must then refer to the methodology described in Appendix E – On-Ramp Queue Spillback
Analysis to evaluate the impacts of spillback on the performance of the arterial intersection.
Spillback into freeways results in uneven operations of mainline lanes, with some lanes blocked, while
some operating in undersaturated conditions. Also, the development of trip-based travel time measures
requires the evaluation of performance measures on individual lanes, as each O-D within a freeway system
uses a specific lane or set of lanes. Therefore, estimating travel time measures in a freeway facility requires
two key components: identifying the set of lanes selected for the trip, and estimating the operating speeds
on each of these lanes.
- 35 -
This section summarizes the framework developed for lane-by-lane analysis of freeway facilities. The
full methodology for evaluation of lane-by-lane performance measures in freeway facilities is described in
detail in Appendix F: Freeway Facilities – Lane-by-Lane Analysis.
- 36 -
Figure 32 – Comparison of lane flow distribution on two different 2-lane segments: (a) Minneapolis,
MN and (b) Salt Lake City, UT
Next, two different 3-lane freeway segments are compared (Figure 33). The lane distributions for these
are different than those for 2-lane segments. At low demand most of the flow of 3-lane segments is
concentrated in the center lane (lane 2), followed by lanes 1 and lane 3. As demand increases, lane flow
distribution increases in lane 3 and decreases in lanes 1 and 2.
- 37 -
Figure 33 – Comparison of lane flow distribution on two different 3-lane segments: (a) Tampa, FL
and (b) St. Paul, MN
As for 2-lane freeways, the values for boundary conditions at 3-lane freeways differ from one location
to another. For example, Figure 33(a) shows that during near-capacity conditions lane 3 carries the majority
of flow but it is similar to that of lane 2. However, in the location represented by Figure 33(b), the proportion
of flow allocated to lane 3 is higher when compared to the other lanes.
Finally, two 4-lane segments were examined. As shown in Figure 34, lane 4 is typically underused during
free-flow conditions, but when demand approaches capacity it carries the majority of flow. However, in
Figure 34(a), flow is highly concentrated on lanes 1 and 2 during free-flow, while Figure 34(b) shows that
flow is more concentrated in lanes 2 and 3 for similar demand levels.
- 38 -
Figure 34 – Comparison of lane flow distribution on two different 4-lane segments: (a) Salt Lake
City, UT and (b) Tampa, FL
𝐿𝐹𝑅 𝑎 𝑙𝑛 𝑏 (Equation 1)
Where:
LFRi = share of the total flow on lane i, where i ranges from 1 to n-1 (n = total number of segment
lanes)
LFRn = share of the total flow on the leftmost lane (lane n);
a= multiplicative calibration parameter
v/c = volume/capacity ratio
b = additive calibration parameter
- 39 -
The adjustment factors fa and fc are calculated as a function of a series of parameters, as follows:
where:
G = grade (%)
a0 = empirical constant
aG = empirical coefficient due to impact of grade
cG = empirical coefficient due to impact of grade
t = truck percentage (%)
at = empirical coefficient due to impact of trucks
bt = empirical coefficient due to impact of trucks
n = access point density – number of ramps half a mile upstream and half mile downstream
an = empirical coefficient due to impact of access point density
b = empirical constant
bn = empirical coefficient due to impact of access point density
vR = ramp flow (vph)
avR = empirical coefficient due to impact of ramp flow
bvR = empirical coefficient due to impact of ramp flow Evaluation of speeds on individual lanes
After the flows for individual lanes are obtained, the next step in the methodology calculates the speeds
on individual lanes. HCM Chapter 12 (Basic segments) proposes the following equation to describe the
speed-flow relationship of a basic segment:
𝑆 = 𝐹𝐹𝑆 − (Equation 5)
Where
𝑆 = segment speed (mi/h)
𝐵𝑃 = breakpoint value (pc/h/ln)
𝑐 = segment capacity (pc/h/ln)
𝐹𝐹𝑆 = segment free-flow speed
𝑣 = demand flow rate for the segment (pc/h/ln)
The same speed-flow relationship is used to predict speeds on individual lanes, as long as the free-flow
speed (FFS) and capacity (c) inputs can be provided on an individual lane basis. Appendix F provides a
procedure to estimate these values based on the segment-wise average values of FFS and capacity. Figure
35 illustrates the field data and the resulting speed-flow relationship of each lane for a 2-lane basic freeway
segment.
- 40 -
Figure 35 – Field data and predicted speed-flow curve for (a) lane 1 and (b) lane 2 (CA-1 NB – Santa
Cruz, CA)
8. System Analysis
This section describes briefly the methodology developed for estimating travel time performance
measures for specific O-D trips considering both urban streets and freeways. The details of the methodology
are provided in Appendix A: Chapter 38 – Systems Analysis. The methodology provides the detailed steps
to evaluate networks with freeway and urban street movements, as illustrated in the example of Figure 36.
- 41 -
Freeway Facilities:
• Flow, free-flow speed, operating speed, and capacity for individual lanes
• Expected travel speed along each segment and each lane
Urban Streets Facilities:
• Travel time along each segment
• Expected travel speed along each segment
System Analysis:
• Total and free-flow travel times
• Travel Time Index
• Average travel speed
- 42 -
9. Software Implementation
A computational engine with the proposed methodology has been developed and is available to HCQSC
and Panel Members on request for beta-testing. The computational engine has been evaluated using a series
of examples and use cases to ensure it replicates the methods accurately. The computational engine is not
a commercial product expected to address all use cases that may arise; however, the chapter provides
guidance for addressing a wide variety of use cases that may be encountered. Appendix A: Chapter 38 –
Systems Analysis and the other report appendices contain a series of examples illustrating the application
of the developed procedures.
This project conducted research on the analysis of highway systems and produced a new chapter (Chapter
38-System Analysis) for the HCM 6th Edition. These new methods can be used to evaluate operations along
networks that include both freeways and urban streets. The methods can also evaluate the impact of
spillback into freeways and into urban streets from downstream facilities. In summary, the following were
accomplished:
Selection of appropriate performance measures: Travel time was selected as the performance measure
to evaluate highway systems. Travel time measures are already used in the HCM to evaluate urban streets
facilities. For freeway facilities, we developed additional models and methods to evaluate freeway
performance by lane, as spillback affects each lane of the freeway differently. In addition, a trip-based
performance measurement framework was developed to provide travel time estimates for given O-Ds
within a highway network. O-D measures reflect traveler experience and are well aligned with recently
available data collection methods which track individual trips. These new measures are intended to
complement segment-based measures provided in other HCM chapters.
Evaluation of queue spillback into freeways: Queue spillback into the freeway occurs due to insufficient
capacity in at least one element of the off-ramp: either the ramp proper, or the downstream ramp terminal.
The blockage of one or more freeway lanes adversely affects performance, and the extent of the blockage
effects depend on various factors including the design of the facility, the cause of the blockage, and the
length of the queue. Video and detector data from several locations were obtained, and used along with
microsimulation to develop the methodological framework. The methodology developed is based on the
calculation of demand and capacity at the downstream ramp terminal using the respective Interrupted Flow
methods. It expands the Oversaturated Segment Evaluation for freeway facilities (HCM Chapter 25) and
accounts for spillback and its effects by lane along the freeway mainline.
Evaluation of queue spillback into urban streets: Queue spillback into urban streets occurs due to
insufficient discharge capacity into the freeway merge. It may occur due to oversaturated conditions at the
merge segment or the presence of ramp metering. Video and detector data from several locations were
collected to understand how intersections are affected by on-ramp queue spillback. Microsimulation was
used to complement the evaluation of signalized ramp terminals and to analyze systems with unsignalized
intersections. The proposed methodology integrates the Interrupted Flow methodologies with the Freeway
Facilities procedure to account for constraints of the on-ramp capacity. Several adjustments were developed
to estimate the impacts of queue spillback from an on-ramp into upstream signalized and unsignalized
intersections, including roundabouts.
Lane by lane performance measures for freeways: Freeway speeds can vary widely depending on the
lane used, and each O-D is likely to use a specific lane or set of lanes along each segment. Spillback affects
- 43 -
each freeway lane differently, and its effects depend on the site design and the length of the queue. This
project developed models that estimate speeds and flows by lane. Detector data were collected from a
variety of locations around the US, and analytical models were developed to predict the lane flow
distribution and lane speed. These models considered a variety of factors including v/c ratio, presence of
nearby ramps, heavy vehicle percent, and grade. Regression models built from the field data demonstrated
that FFS and capacity values for each lane can be obtained as a percentage of the segment average with
satisfactory results.
Development of an O-D analysis framework: A new methodology was developed to estimate travel
times by O-D within a highway system. This methodology combines the tools of several HCM chapters
within the Uninterrupted Flow and Interrupted Flow volumes. It also builds on the research previously
described to evaluate queue spillback and system effects
The limitations of the methods include the following:
System analysis:
• HCM Travel time reliability methodologies for Freeways (Chapter 11) and Urban Streets
(Chapter 17) have significant differences in their procedures. Therefore this project made no
attempt to incorporate reliability analysis in the new procedures.
In the future, it would be useful to develop a LOS framework for the travel time performance measures
yielded by this methodology to evaluate highway systems. Similar to other performance measures used in
the HCM, communicating the values of performance measures to a lay audience may be challenging, while
LOS may be a more useful construct.
- 44 -
References
- 45 -
- 46 -
CHAPTER 38
SYSTEM ANALYSES (DRAFT)
CONTENTS
1. INTRODUCTION..................................................................................................... 56
Overview ................................................................................................................. 56
Chapter Organization ............................................................................................ 56
Related HCM Content............................................................................................ 56
2. CONCEPTS ................................................................................................................ 58
Overview ................................................................................................................. 58
Spillback Impact on Freeways .............................................................................. 58
Spillback Impact on Urban streets........................................................................ 62
Lane-by-Lane Analysis .......................................................................................... 63
Performance Measurement for Systems and O-D.............................................. 63
3. METHODOLOGY .................................................................................................... 64
Scope of the Methodology ..................................................................................... 64
Required Data and Sources ................................................................................... 66
Computational Steps .............................................................................................. 67
4. EXAMPLE PROBLEMS............................................................................................ 84
Example Problem 1: O-D Based Travel Time Estimation for I-75 NB Freeway
in Gainesville, Florida ..................................................................................... 84
Example Problem 2: I-10 On-Ramp Spillback analysis in Baton Rouge,
Louisiana .......................................................................................................... 95
Example Problem 2, Part 1: Signalized Intersection Ramp Terminal .............. 97
Example Problem 2, Part 2: TWSC Ramp Terminal......................................... 118
Example Problem 2, Part 3: AWSC Intersection Ramp Terminal .................. 123
Example Problem 3: Off-Ramp Queue Spillback Analysis for a Freeway-to-
Freeway Ramp in Miami, Florida. .............................................................. 127
Example Problem 4: On-Ramp Queue Spillback Analysis into a Single-Lane
Roundabout in Los Angeles, California ..................................................... 137
LIST OF EXHIBITS
Exhibit 38-54 Freeway Facility (I-10 EB) – Demand Inputs ................................... 103
Exhibit 38-55 Performance Measures for the Freeway Facility (I-10 EB) ............. 104
Exhibit 38-56 Spillback Check – I-10 EB on-Ramp .................................................. 104
Exhibit 38-57 Freeway Facility, Segment 5 (merge) Performance: a)
Merge Capacities and b) Queue Lengths .......................................................... 106
Exhibit 38-58 Freeway Performance During Time Period 4 – with and
without the Queue Storage Constraint .............................................................. 106
Exhibit 38-59 Estimated Queue Lengths And Merge Capacities – Time
Period 2 .................................................................................................................. 107
Exhibit 38-60 Discharge Flow Rates into the On-Ramp for each Phase
Throughout the Cycle – Time Period 2 .............................................................. 110
Exhibit 38-61 Estimated Queue Lengths and Merge Capacities – Time
Period 3 .................................................................................................................. 111
Exhibit 38-62 Discharge Flow Rates Into the On-Ramp for Each Phase
Throughout the Cycle – Time Period 3 .............................................................. 113
Exhibit 38-63 Calculation of Spillback Capacity Reduction Factor for the
SBL Movement for Time Period 3 ...................................................................... 114
Exhibit 38-64 Estimated Queue Lengths and Merge Capacities – Time
Period 4 .................................................................................................................. 115
Exhibit 38-65 Calculation of Spillback Capacity Reduction Factor for the
SBL Movement During Time Period 4............................................................... 116
Exhibit 38-66 Comparison of Performance Measures – with and without
Consideration of Spillback Effects...................................................................... 117
Exhibit 38-67 TWSC Intersection Geometry – Acadian Thruway @ I-10
EB. ........................................................................................................................... 118
Exhibit 38-68 Calculation of the On-Ramp Demand (vR) Based on the
TWSC Intersection Operation. ............................................................................ 119
Exhibit 38-69 Queue Accumulation Plot Calculations for On-Ramp –
TWSC Intersection ................................................................................................ 120
Exhibit 38-70 Queue Accumulation Polygon for the On-Ramp – TWSC
Intersection ............................................................................................................ 121
Exhibit 38-71 Comparison of Performance Measures in a TWSC
Intersection – Time Period 3 - with and without Spillback Effects ................ 122
Exhibit 38-72 AWSC Intersection Geometry – Acadian Thruway @ I-10
EB ............................................................................................................................ 123
Exhibit 38-73 Calculation of the On-Ramp Demand (vR) Based on the
AWSC Intersection Operation ............................................................................ 124
Exhibit 38-74 Check for Spillback Occurrence – AWSC Intersection ................... 124
Exhibit 38-75 Queue Accumulation Plot Calculations for On-Ramp –
AWSC Intersection ............................................................................................... 125
1. INTRODUCTION
2. CONCEPTS
OVERVIEW
This section discusses concepts related to spillback on the freeway, spillback
on the urban street, lane-by-lane analysis, and performance measurement for
systems and O-Ds. Concepts related to freeway analysis and urban street
analysis are described in the respective chapters of this manual.
Off-ramp elements
A freeway off-ramp typically consists of three components (Exhibit 38-1):
• Deceleration lane(s): its distance is measured from the beginning of the
taper of the auxiliary lane to the gore
• Ramp roadway: the road section connecting the deceleration lane and the
downstream ramp terminal; its distance is measured from the gore to the
taper of the ramp terminal;
• Ramp terminal: ramp terminals connecting to urban street facilities can be
uncontrolled, stop-controlled or signalized intersections; its distance is
measured from the point where additional lanes are added to the
intersection approaches to the stop bar of the approach. The length of this
section should be at least as long as the turn bay lengths of the approach.
When the ramp connects two freeway facilities, the downstream ramp
terminal is replaced by the merge section of the on-ramp, with no storage length.
Exhibit 38-1
Off-Ramp Components
a) Regime 1
Under this regime, the queue ends within the deceleration lane and does not
spill back into the mainline freeway (Exhibit 38-2 (a)) Deceleration lanes typically
serve as a transition zone between speeds on the mainline (typically 55 – 75 mi/h)
and advisory speeds posted along the off-ramp roadway (typically 20 – 50 mi/h).
When queues begin to form on the deceleration lane, the available deceleration
distance is reduced, and speeds begin to be affected in the rightmost lane.
b) Regime 2
Under this regime, the queue of vehicles extends upstream beyond the
deceleration lane, but sufficient lateral clearance on the right-hand shoulder
allows for additional queue storage. In this case the deceleration lane does not
serve as a transition zone and drivers decelerate and join the back of the queue
more abruptly, resulting in turbulence and reduced speeds in the rightmost lane
(Exhibit 38-2 (b)). If no lateral clearance exists immediately upstream of the
deceleration lane, Regime 2 conditions are not possible. In some cases, this
regime does not occur even if storage is available; this depends on local driver
behavior and is site-specific.
c) Regime 3
Under this regime, the queue extends to the rightmost lane of the freeway
mainline (Exhibit 38-2 (c)). This may occur either when there is no shoulder
available for additional queue storage, or when drivers choose to queue in the
rightmost lane once the deceleration lane is entirely occupied. Non-exiting
vehicles on the rightmost lane are delayed or change lanes, which causes
increased turbulence and reduced speeds in the two rightmost lanes.
d) Regime 4
Under this regime, the queue blocks the rightmost lane, and drivers
occasionally or often use the adjacent freeway mainline lane next to the
rightmost freeway mainline lane to force their way into the queue, blocking thus
an additional lane (Exhibit 38-2 (d)). During this regime, mainline speed and
capacity are significantly reduced.
The effects of spillback vary from site to site and from time period to time
period due to driver behavior and site geometry. Data collection has shown that
at some sites, drivers block the adjacent lane, while at other sites they do not,
regardless of the queue spillback length at the site.
Exhibit 38-2
Definition of Spillback
Regimes
(a) Regime 1 – Queue within the (b) Regime 2 – Queue along the
deceleration lane shoulder
Exhibit 38-3
Capacity Adjustment Factors
(CAFBL) for Through Lanes
Adjacent to Blocked Lanes
during Queue Spillback
2 0.70 N/A
3 0.74 0.51
4 0.77 0.50
5 0.81 0.67
6 0.85 0.75
7 0.88 0.8
8 0.89 0.84
Exhibit 38-4
Queue Influence Area with
Increased Turbulence
Segment Free-Flow Speed (mi/h) Queue Influence Area (ft) Exhibit 38-5
50 810 Length of Queue Influence
55 900 Area as a Function of the
60 980 Segment Free-Flow Speed
65 1060 (FFS)
70 1140
75 1220
Exhibit 38-6
Queue Spillback from an On-
Ramp into Urban Street
Intersections
The queue length along the on-ramp also depends on the upstream
demands. In the example shown in Exhibit 38-6(a), there are three possible
movements contributing to this demand: NB right, SB left, and EB through. If the
NB right movement is very heavy and/or has the right-of-way for a significant
amount of time, the SB left movement may not have as much of an opportunity
to contribute to the demand and may spill back upstream, affecting the adjacent
SB through movement, as well as the upstream intersection. Thus, in the case of
signalized intersections, the relative contribution of demands to the queue length
will depend on the relative demands of these movements and the respective
signal timings and right-of-way allocation. The discharge rate of these upstream
intersection movements will depend on the storage availability on the on-ramp
during the respective phase. The analysis estimates the additional lost time due
to the presence of the downstream queue and adjusts the effective green of these
movements.
In the roundabout example shown in Exhibit 38-6(b), the same three
movements contribute to the on-ramp demand. However, in this case the
movements have priority in the following order: (1) SB left, (2) EB through and
(3) NB right. A high-priority movement with a heavy demand may constrain the
entry capacity of lower priority movements, resulting in total throughput that is
lower than the sum of the three contributing movement demands.
LANE-BY-LANE ANALYSIS
Spillback affects each lane of a facility differently. For example, when
spillback occurs at a freeway off-ramp, the right-most lanes of the freeway may
be blocked, while the left-most lanes operate in free-flow conditions. Therefore,
the methodology estimates operating conditions by lane as well as by segment.
The lane-by-lane performance metrics are also used to obtain O-D based travel
times.
The demand flow rates by lane The lane-by-lane analysis provides lane flow ratios
are estimated as a percentage
of the segment demand. (LFR) which represent the percentage of the entering
demand by lane. LFR is a function of the segment-wide
v/c ratio and values are provided for each segment type (basic, merge, diverge
and weaving). In addition, FFS, speeds, and capacities are estimated by lane.
When the facility reaches oversaturated conditions, the speeds are estimated
based on the Chapter 10, Freeway Systems method, which is based on
interactions between successive segments.
3. METHODOLOGY
Limitations
The methodology has the following limitations:
1. Multiple overlapping breakdowns or bottlenecks cannot be fully
evaluated by this methodology. Consult Chapter 6, HCM and Alternative
Analysis Tools for a discussion of simulation and other models.
2. Demand is an input into the process, and the methodology does not
address any changes in demand that are due to traffic operational
conditions.
3. Managed lanes can be analyzed as part of the freeway system. However,
the interaction of managed lanes operations with spillback conditions are
not addressed.
4. The methodology does not explicitly consider alternative intersection and
interchange designs, such as DDI and SPUI. However, it can be extended
to consider these, assuming turning movements, demands, and queues
can be accurately estimated for the movements of interest.
5. The methodology does not consider two-lane roundabouts and their
interaction with freeway on-ramps.
6. The reliability method cannot be applied for systems analysis because the
process for developing reliability scenarios is different for freeways and
arterials.
ramp junctions with a lane drop. At these locations, the exiting traffic can
access the off-ramp with a single lane change. Therefore, drivers are more likely
to wait until they are closer to the exit to change lanes, blocking the adjacent
through lane. However, not all lane drop exits experience a Regime 4 queue
spillback. Generally, Regime 4 occurs more frequently in locations with more
aggressive driver behavior. Local information and driver behavior should be
taken into consideration in determining the prevailing regime at a given site.
For operational analyses of existing locations, it is recommended that the
analyst provides the expected spillback regime based on observed field
conditions. For planning level purposes where no field data is available, Exhibit
38- 8 provides the expected queue spillback regime as a function of the number
of exiting lanes and driver aggressiveness.
Driver Aggressiveness
Ramp Geometry Exhibit 38- 8
Low Medium High
Diverge with Default spillback regimes as a
Regime 3 Regime 3 Regime 3 function of ramp geometry
deceleration lane
and driver aggressiveness
Diverge with lane drop Regime 3 Regime 4 Regime 4
COMPUTATIONAL STEPS
This section describes the methodology’s computational steps. Exhibit 38-9
illustrates the process used to evaluate systems operations.
Exhibit 38-9
Systems Analysis Methodology
Flowchart
Exhibit 38-10
Sample Study Network, with
Multiple Origins and
Destinations
Exhibit 38-11
Potential Bottlenecks
Constraining the Ramp
Terminal Demand
However, if the demand at the off-ramp exceeds its capacity, the flow that
will reach the ramp terminal will be lower than the off-ramp demand vR. In this
case, the following adjustments are performed:
𝑇
Equation 38- 1 𝑣 , 𝑂𝐹𝑅𝐹 𝑖, 𝑡, 𝑝
𝑆
where
𝑣 , = adjusted demand at the subject off-ramp (pc/h);
The parameter OFRF(i, t, p) 𝑂𝐹𝑅𝐹 𝑖, 𝑡, 𝑝 = actual flow that can exit at off-ramp 𝑖 during time step 𝑡 in time
is defined as the “actual
flow that can exit at off- interval 𝑝;
ramp i during time step t in
time interval p” (Chapter
𝑇 = number of time steps in 1 h; and
25). It can account for the 𝑆 = number of computational time steps in an analysis period (typically
effects of bottlenecks
upstream of the off-ramp S=240 for time steps of 15s)
that can meter the traffic
that arrives to the ramp. If the freeway facility operates at undersaturated conditions, the value of
𝑣 , is equal to the off-ramp demand 𝑣 .
If the subject freeway facility operates at oversaturated conditions, the total
demand of the off-ramp may be metered at an upstream bottleneck segment. The
Oversaturated Segment Evaluation methodology (HCM Chapter 25) provides
equations to estimate the off-ramp flow parameter OFRF (Equations 25-23
through 25-25) at every 15-second time step.
where
𝑣, = adjusted demand for movement 𝑖 at the downstream intersection
(pc/h);
𝑣 = demand for movement 𝑖 at the downstream intersection (pc/h);
𝑣 = off-ramp demand (pc/h);
Exhibit 38-12
Potential Bottlenecks
Constraining the On-Ramp
Demand
However, if capacity is exceeded at any of those locations, the flow that will
reach the freeway merge will be lower than the on-ramp demand vR and
adjustments should be made to the respective volumes.
If any of the ramp terminal movements that discharge into the on-ramp
operates over capacity, the total throughput to the on-ramp will be:
Equation 38-3
𝑣 , 𝑚𝑖𝑛 𝑣 , 𝑐
where
𝑣 , = adjusted on-ramp demand (veh/h);
𝑣 = demand for movement 𝑖 at the intersection (veh/h);
𝑐 = demand for movement 𝑖 at the intersection (veh/h);
𝑁 = number of intersection movements that discharge into the on-ramp
If the total on-ramp demand vR is greater than the ramp roadway capacity cR,
the adjusted on-ramp demand is:
Equation 38-4 𝑣 , 𝑚𝑖𝑛 𝑣 , 𝑐
where
𝑣 , = adjusted on-ramp demand (veh/h);
Exhibit 38-13
Spillback Check Procedure for
Off-Ramps
The process evaluates whether the spillback originates from the demand to the
ramp roadway, or from the demand to the ramp junction at the surface street, or
from the downstream freeway on-ramp. Based on this determination, the
procedure uses the demand and the capacity for the analysis interval, as well as
the previous queue length, to calculate the anticipated queue length for this
interval. The detailed calculations for off-ramp spillback check are presented in
Appendix A.
Exhibit 38-14
Spillback Check Procedure for
On-Ramps
Step 5A: Compute Operating Speeds for Individual Lanes Along the
Freeway Facility
Along freeway facilities, operational performance is determined based on the
density and speed at each segment along the network. The average travel time
for each segment can be derived based on the respective average speeds.
For a system analysis, the speed along a segment is function of:
• Estimated speeds for individual lanes;
• Probability that a lane will be selected by the subject O-D.
To estimate the speeds and capacities for individual lanes, a set of models
have been developed for each type of freeway segment considering the total
number of mainline freeway lanes. These models are valid only for
undersaturated conditions, and they predict the Lane Flow Ratio (LFR) for each
lane. They are of the form:
𝐿𝐹𝑅 = 𝑎 × 𝑙𝑛 𝑣/𝑐 + 𝑏
Equation 38-5
where
𝑎 = multiplicative calibration parameter (Equation 38-C3, Equation 38-C5, The full methodology to predict lane by
and Equation 38-C7); lane speeds on freeway facilities is
discussed in Appendix C.
𝑏 = additive calibration parameter (Equation 38-C4, Equation 38-C6, and
Equation 38-C8);
𝐿𝐹𝑅 = share of the total flow on lane 𝑖, where 𝑖 ranges from 1 to n-1 (n = total
number of segment lanes);
𝐿𝐹𝑅 = share of the total flow on the leftmost lane (lane n); and
𝑣/𝑐 = volume/capacity ratio 0 < 𝑣/𝑐 ≤ 1 .
Using these LFR values, the methodology next estimates the lane-by-lane
free-flow speeds and capacities. These are used to obtain the speeds of each lane
using the speed-flow models defined in HCM Equation 12-1.
The LFR models and their coefficients, along with the procedures for
estimating lane-by-lane free-flow speeds, capacities, and speeds are provided in
Appendix C. These models can be used to analyze basic, merge, diverge and
weaving segments and mainline freeways with two to four lanes.
Freeway segments with 5 or more lanes were not modeled due to insufficient
data. Limited field observations for these facilities indicate that flow distributions
become more homogenous at wider segments. Therefore, the flow distribution
for these segments can be estimated as:
𝑣
𝐿𝐹𝑅 = Equation 38-7
𝑛
where
𝑣 = segment entering demand (pc/h) and
𝐿𝐹𝑅 = share of the total flow on lane 𝑖, where 𝑖 ranges from 1 to n (n = total
number of segment lanes).
For all segment types the share of flow is estimated on the mainline
upstream of the segment. The oversaturated portion of the speed-flow curve
(when density is greater than density at capacity) cannot be addressed by the
speed flow models, as this is a limitation of the existing methods. The lane-by-
lane flows for oversaturated conditions are estimated using the procedures of
Chapter 25, adjusted to determine the incoming and outgoing flow on a lane-by-
lane basis. If off-ramp queue spillback occurs in the freeway facility, then the
methodology in Appendix A provides a procedure to determine the lane-by-lane
flow distribution.
The probability that a given lane is selected depends on the type of O-D. For
segments where a driver enters (merge segment) or leaves a freeway facility
(diverge segment), the probability of lane selection is shown in Exhibit 38-15
(assuming right-side ramps).
Lane choice probability for lane i Number of lanes in the segment
Exhibit 38-15
2 3 4+
Probability of Lane Choice for
p1 0.90 0.90 0.90
Entry/Exit Segments on
p2 0.10 0.05 0.05
Freeway Facilities
p3 - 0.05 0.05
p4+ - - 0
For other segments within a freeway facility the probability of choice for a
given lane i is equal to the Lane Flow Ratio of lane i (LFRi), defined as the
percentage of the total flow assigned to lane i:
Exhibit 38-16
Illustration of Lane Choice
Probabilities Along a Freeway
Facility
The speed for each segment is then computed as the sum of products of
speeds for each lane and the corresponding probability of lane choice:
Equation 38-9 𝑆 = 𝑝 ×𝑠
where
𝑆 = expected speed for the segment (mi/h)
𝑝 = probability that lane i is selected
𝑁 = number of lanes in the segment
𝑆 = speed at lane i (mi/h) (Equation 38-C14)
The speed-flow relationship for ramps is linear and speed decreases with
higher ramp flows, as shown in Exhibit 38-17. The maximum allowed values of
vR are bounded by ramp capacity, consistent with guidance provided by Chapter
14 – Merge and Diverge segments (Exhibit 14-12).
Exhibit 38-17
Speed-flow Curves for
Freeway Ramps
ramp roadway will increase until the limit value of jam density. The NV
parameter for the ramp roadway is computed every time step (15 seconds) and
then aggregated to the 15-minute period to compute the average density at the
ramp roadway. Similarly, the flow through the ramp roadway is aggregated to
15-minutes and then the speed at the off-ramp is obtained through Equation 12-1,
which is repeated here for convenience: below:
𝑆 = 𝐹𝐹𝑆 , 𝑖𝑓 𝑣 ≤ 𝐵𝑃
where
𝑆 = mean speed of a basic segment (mi/h)
𝐹𝐹𝑆 = adjusted free-flow speed (mi/h)
𝑐 = adjusted segment capacity (pc/h/ln)
𝐵𝑃 = breakpoint (pc/h/ln)
𝐷 = density at capacity, typically 45 pc/mi/ln
𝑣 = adjusted 15-min demand flow rate (pc/h/ln)
Step 8: Compute Travel Times for the Network and Each O-D
This step computes the total travel time TTO-D for the network as the sum of
travel times over all segments along the route. For multi-period analysis, it is
important that the travel time for the correct time period at each segment is
selected, as a long O-D may encompass several time periods. Exhibit 38-18
presents a sample calculation for a facility with two time periods (15 minutes
each). The first segment in the O-D is traversed during Time Period 1, and the
Cumulative Travel Time column is updated with the respective value.
Subsequent segments follow the same procedure until the cumulative travel time
exceeds the length of the first time period (900 seconds). For the next segment in
the network, travel times from Time Period 2 are added to the Cumulative Travel
Time. This procedure is then repeated until the final segment is reached. The
total travel time is obtained as the last value of the Cumulative Travel Time
column
Equation 38-14 𝑇𝑇
𝑇𝑇𝐼 , =
𝑇𝑇 ,
where
𝑇𝑇 = total travel time for a specific O-D (s)
𝑇𝑇 , = free-flow travel time for a specific O-D during (s)
Freeway Facilities
Lane flow ratio (LFR) v/c 0.1
Speed by lane Free-flow speed by lane (FFSi) Equation 38-C9
Urban Street Segments
Travel speed Running time Equation 18-7
Urban Street Intersections Exhibit 38-19
Reference Input Values
Control delay - Signalized Intersections Demand-to-capacity ratio (X) 0
for O-D Analysis at
Control delay - TWSC Intersections Movement demand (vx) 0 Free-Flow Conditions
Control delay - AWSC Intersections Demand-to-capacity ratio (x) 0
Control delay - Roundabouts Demand-to-capacity ratio (x) 0
Freeway ramps
Ramp speed Ramp free-flow speed Analyst input
Freeway Facilities
At free-flow, the speed at freeway segments is computed as equal to their
free-flow speed. When a lane-by-lane analysis is applied, the methodology
computes the free-flow speed for each lane (Equation 38-C9).
Next, the probabilities of lane choice on each segment are calculated for each
segment. If the subject segment is a entry/exit segment (segments where the
driver enters or leaves the freeway facility, as illustrated in Exhibit 38-16), the
lane choice probabilities are obtained from Exhibit 38-15. For other segments, the
lane choice probability is equal to its LFR (Equation 38-5). For the calculations of
LFR under free-flow conditions, a value of v/c = 0.1 is recommended to provide
results consistent with field data for free-flow conditions. Due to the log form of
the LFR equation (Equation 38-5), using v/c = 0 is mathematically unfeasible, and
very low values of v/c would yield unrealistic results
6.0 − 𝑙 3,600𝐿
𝑡 = 𝑓 + 𝑓 + 𝑑 , +𝑑
0.0025𝐿 5,280 𝑆
where
𝑆 , = travel speed of through vehicles for the segment (mi/h) (s)
𝑡 = segment running time (s)
𝑙 = start-up lost time (2s if signalized, 2.5s if stop or yield-controlled)
𝐿 = segment length (ft)
TWSC Intersections
The control delay d for TWSC intersections (Rank 2 through Rank 4
movements) is computed through Equation 20-64:
⎡ 3,600 𝑣 ⎤
3,600 ⎢𝑣 𝑣 𝑐 , 𝑐 , ⎥
𝑑= + 900𝑇 ⎢ −1+ −1 + ⎥+5
𝑐 , 𝑐 , 𝑐 , 450𝑇
⎢ ⎥
⎣ ⎦
where
𝑐 , = capacity of movement x (veh/h)
𝑣 = flow rate for movement x (veh/h)
𝑇 = analysis time period (0.25 h for a 15-min period) (h)
AWSC Intersections
The control delay d for AWSC intersections is computed through Equation
21-30:
ℎ 𝑥
𝑑 = 𝑡 + 900𝑇 𝑥 − 1 + 𝑥−1 + +5
450𝑇
where
𝑡 = service time (s)
𝑥 = volume-to-capacity ratio of the subject lane
ℎ = departure headway (s)
𝑇 = analysis time period (0.25 h for a 15-min period) (h)
At free-flow conditions, the demand-to-capacity ratio 𝑥 is set at zero, which
allows Equation 21-30 to be reduced to the following form:
Equation 38-16
𝑑 =𝑡 +5
The estimation of service time 𝑡 requires an iterative and computationally
intensive procedure described in the AWSC Intersections methodology (Chapter
21). It must be performed setting x = 0.
Roundabouts
The control delay d for roundabouts is computed through Equation 22-17:
3,600
3,600 𝑥
𝑑= + 900𝑇 𝑥 − 1 + 𝑥−1 + 𝑐 + 5 × 𝑚𝑖𝑛 𝑥, 1
𝑐 450𝑇
where
𝑥 = volume-to-capacity ratio of the subject lane
𝑐 = capacity of the subject lane (veh/h)
𝑇 = analysis time period (0.25 h for a 15-min period) (h)
Freeway Ramps
Freeway ramp speeds at free-flow are equal to the ramp free-flow speed 𝑆 ,
as provided by the analyst, and do not require additional adjustments.
4. EXAMPLE PROBLEMS
Exhibit 38-21
Example Problem 1
Network Interchanges,
with indication of origins
and destinations:
(a) Williston Rd.(b)
Archer Rd.(c) Newberry
Rd.(d) NW 39th Ave.
Exhibit 38-22
Freeway Origins and
Destinations for Example
Problem 1
The analysis steps for evaluating this network are discussed below.
As shown in Exhibit 38-21, the system has 9 nodes and 72 O-Ds, shown in
Exhibit 38-23. This case study estimates the travel time for the O-D from Archer
Rd. East (D) to NW 39th Ave (H), as depicted in red in Exhibit 38-21.
The total average travel time for each O-D can be obtained by adding the
travel times on each segment plus any delay experienced at all intersections
traversed. Travel time for each ramp traversed is also computed.
The O-D from node D to node H will traverse two urban street facilities, as
shown in Exhibit 38-24:
• Archer Rd. Westbound, comprised of two urban street segments and two
signalized intersections (SW 40th Blvd. and I-75 NB on-ramp);
• NW 39th Ave. Eastbound, comprised of one urban street segment and
two signalized intersections (I-75 NB off-ramp and NW 95th Blvd).
Exhibit 38-24
Urban Street Facilities
Evaluated for Example
Problem 1
The temporal scope of the analysis must also be defined. Given the short
length of the subject network, a single-period analysis will be performed. When
the final travel time is obtained, its value will be checked and if it exceeds the 15-
minute study period the temporal scope of the study will be reevaluated.
Additional input parameters for the urban street facility are as follows:
• Urban area, 3 lanes in each direction;
• Base FFS: 75.4 mi/h;
• Ramp FFS: 35 mi/h;
• Ramp side: Right;
• Lane width: 12 ft;
• Right side clearance: 10 ft;
• Traffic composition: 2% trucks on both freeway and ramps; and
• Familiar facility users.
Archer Rd. @ Demand (veh/h) 120 2348 88 36 864 548 60 208 96 36 480 304
SW 40th Blvd. Phase Split (s) 20 50 - 20 50 - 20 30 - 20 30 -
Exhibit 38-28 Input Parameter SW 40th Blvd - I-75 WB SW 37th Blvd - SW 40th Blvd
Input Data for Segment Segment length (ft) 530 1288
Analysis – Archer Rd. WB Speed limit 45 45
Through lanes 3 3
Restrictive median length (ft) 0 0
Upstream Intersection width (ft) 50 50
Curb proportion (%) 70 70
Base FFS (mi/h) 46.42 46.42
Running Speed (mi/h) 32.24 41.37
Running time (s) 11.21 21.23
Percent of base FFS 50.84 52.04
Parameter
Intersection Movement Demand c Merge Exhibit 38-31
v/c min(v, c)
(v), veh/h (veh/h) demand vR Demands at the On-Ramps
Along the Freeway Facility for
Williston Rd. EBL 160 0.15 1055 160
580 Example Problem 1
@ I-75 NB WBR 420 0.43 985 420
Archer Rd. @ EBL 320 0.34 935 320
868
I-75 NB WBR 548 0.53 1037 548
Newberry Rd. EBL 216 0.25 862 216
380
@ I-75 NB WBR 164 0.14 1163 164
NW 39th Ave. EBL 72 0.14 501 72
148
@ I-75 NB WBR 76 0.075 1012 76
Next, the off-ramp volumes are checked against the intersection turning
movement demands. The first check determines whether there are bottlenecks
along the freeway facility that may meter off-ramp demands. Exhibit 38-32
shows the estimated LOS for all 19 segments in the freeway facility. Since no
segment is oversaturated, the off-ramp demand is not metered, and no
adjustments are necessary.
The second check compares the off-ramp demands to the respective ramp
roadway capacity, as shown in Exhibit 38-33. The demand does not exceed
capacity for any of the ramps, therefore no adjustments to the intersection
volumes are performed.
2 6 11 16
(Williston (Archer (Newberry (NW 39th
Rd.) Rd.) Rd.) Ave.)
Parameter LT RT LT RT LT RT LT RT
Ramp length (ft) 900 1650 660 2380
Number of Ramp Lanes 1 1 1 2 Exhibit 38-34
Upstream Ramp Lane L1 L1 L1 L2 L2 L1 Queue Length Estimation and
Turn Bay Length (ft) 250 210 480 480 800 800 1260 1200 Queue Storage Checks for
Back of Queue Length Q95 (ft/ln) 689 21 120 363 223 363 193 482 Off-Ramps
Exceeding Turn Bay Queue Length (ft) 439 - - - - - -
Queue Storage Ratio (RQ) 0.49 - - - - - - -
Step 5A: Obtain Speeds for Individual Lanes in the Freeway Facility
First, the flow distribution among freeway lanes must be determined for the
segments in the freeway facility. Using the estimated flow rates, lane speeds are
computed as shown in Exhibit 38-35. The highlighted rows (8 through 16)
represent the segments included in the O-D and used to compute the overall
travel time. The rightmost lane is labeled Lane 1.
The expected speed for each segment is then computed as the sum of
products of speeds for each lane and the corresponding probability of lane The step by step calculations
choice, as provided in Equation 38-9. (Exhibit 38-36). to determine lane-by-lane
flows and speeds on segment
16 (diverge) are presented in
Even though the travel times of the remaining segments are not directly used an example problem under
in calculating the O-D travel time, the entire facility must be analyzed, as any Appendix C
Facility 2 I-75 NB - NW
510 46.42 31.53 11.03 Through 26.2 9.34
(NW 39th 95th Blvd
St. EB)
Segment ID Expected Speed (mi/h) Segment Length (ft) Travel Time (s) Exhibit 38-39
8 68.4 1500 15.0 Travel Times for Freeway
9 66.5 6300 64.6 Segments
10 67.2 5385 54.6
11 64.4 1500 15.9
12 72.0 2014 19.1
13 73.8 1500 13.9
14 71.2 6494 62.2
15 71.2 2480 23.7
16 53.1 1500 19.3
Step 8: Compute Travel Times for the Network and Each O-D
All segments within the subject O-D (E1-H1) are sorted according to the
travel sequence and their respective travel times are listed, as shown in Exhibit
38-41. The cumulative travel time for the O-D must also be computed to evaluate
if the network analysis is correctly contained within the temporal scope defined
in Step 1. For this example, a single 15-min analysis period was considered, for a
total time of 900s. Since the cumulative travel time does not exceed this boundary
value, all travel times obtained from time period 1 are valid for the analysis.
The subject network has three freeway lanes throughout its entire length.
One interchange connects the freeway to an urban street network (Acadian
Thruway), as illustrated in Exhibit 38-42. The selected origin and destination
points for analysis are H and F, respectively, with the traveled segments
highlighted in red.
Exhibit 38-42
Example Problem 2 Network
Intersections:
(a) Perkins Rd.; (b) Acadian
Center; (c) I-10 EB;(d) I-10
WB
The freeway facility (I-10 EB) is modeled according to the Freeway Facilities
methodology (Chapter 10), while the ramp terminal is modeled according to its
respective intersection methodology. First a check is performed to confirm the
occurrence of queue spillback. Next, the respective spillback analysis is applied
to evaluate the impacts of queue spillback in the capacity of each movement at
the intersection. With the estimated reduced capacities at the intersection, the
control delay values considering queue spillback are computed and compared to
the delay values without consideration of queue spillback.
Exhibit 38-43 illustrates the schematic representation of the freeway network
in the eastbound direction. Segments 3 (merge) and 5 (diverge) connect the
freeway to the urban street facility (Acadian Thruway).
Exhibit 38-43
Origins and Destinations for
the freeway facility (I-10 EB)
in Baton Rouge, LA
The analyzed urban street facility comprises four signalized intersections and
three segments, as shown in Exhibit 38-44. The on-ramp terminal subject to
analysis is the I-10 EB intersection.
Exhibit 38-44
Acadian Thruway Urban
Street Facility
Input data
Signalized Intersection
The geometry of the intersection connected to the I-10 EB on-ramp (I-10 EB)
is shown in Exhibit 38-45. There are three movements leading into the on-ramp:
• NBR: One channelized, unsignalized right-turn lane;
• SBL: One exclusive left turn lane with a protected phase; and
• EBT: One through lane.
Exhibit 38-45
Signalized Intersection
Geometry – Acadian Thruway
@ I-10 EB
Exhibit 38-46
Phasing Sequence – I-10 EB
Intersection
The demand volumes for each time period are presented in Exhibit 38-47.
Additional input data are summarized in Exhibit 38-48.
Phase Information
Maximum Green (Gmax), s 20 20 - 53 - 47 100
Yellow Change Interval (Y), s 4.7 4.7 - 4.7 - 4.7 4.7
Red Clearance Interval (Rc), s 1 1 - 1 - 1 1
Minimum Green (Gmin), s 5 5 - 15 - 5 15
Start-Up Lost Time (lt), s 2 2 - 2 - 2 2
Green Extension (e), s 2 2 - 2 - 2 2
Passage (PT), s 2 2 - 2 - 2 2
Recall Mode Off Off - Off - Off Off
Dual Entry No No - Yes - No Yes
Exhibit 38-49
Freeway Facility
Segmentation– I-10 EB
The geometric features of the freeway facility are summarized in Exhibit 38-
50.
Acceleration /
Segment Length Grade Ramp
Type deceleration
Exhibit 38-50 ID (ft) (%) length (ft)
lane length (ft)
Freeway facility (I-10 EB) -
Geometric Features 1 Basic 5280 0 - -
2 Diverge 1500 0 800 1139
3 Diverge 720 0 0 965
4 Basic 732 0 - -
5 Merge 1000 0 1000 924
6 Basic 1200 0 - -
7 Basic 900 0 - -
where
PHV = percentage heavy vehicles in the corresponding movement group (5%)
PHV = approach grade for the corresponding movement group (0%)
𝑣 𝑒 ,
𝑠 =
1−𝑒 ,
where
sp = saturation flow rate of a permitted movement (veh/h/ln)
v0 = opposing demand flow rate (veh/h);
tcg = critical headway = 4.5 (s); and
tfh = follow-up headway = 2.5 (s);
The computation of the permitted saturation flow rates must take into
consideration that the conflicting phase may have two distinct flow rates on
signalized intersection operation, as discussed in Chapter 31 (Signalized
Intersections Supplemental):
• During the queue service time (gs) portion of the conflicting phase green,
the opposing movement flow rate is equal to its saturation flow rate;
• During the green extension time (ge), the opposing movement flow rate is
equal to its arrival flow rate during the effective green (qg);
Exhibit 38-51 illustrates the calculation of the NBR capacity for a single cycle
during time period 1. For each active phase, the procedure identifies the
respective conflicting flow to the on-ramp along with its duration and flow rate.
The NBR saturation flow rate is then computed using HCM Equation 31-100. The
last column computes the maximum number of vehicles that can be discharged
during each phase as the product of the NBR saturation flow rate and the phase
duration. Clearance times between consecutive phases are also taken into
consideration assuming that they have no conflicting flow rate to the on-ramp.
NBR
NBR
Exhibit 38-51 Conflicting saturation
Conflicting Duration discharge
Calculation of NBR Capacity Active phase flow rate flow rate
flow (s) volume
for a Single Cycle – Time (veh/h) sNBR
(veh)
Period 2 (veh/h)
gs: queue service time; ge: green extension time; qg: arrival flow rate during effective green; s: saturation flow rate
As shown, for a 120s cycle the capacity of the unsignalized NBR movement is
34.8 vehicles. Aggregated to an hourly flow rate:
3600
𝑐 = 34.8 × = 1045 𝑣𝑒ℎ/ℎ
120
Because of the actuated control operation, the discharging rates to the on-
ramp are different during each time cycle, and during each period. Therefore,
this procedure must be repeated for every time period to compute the capacity of
the NBR unsignalized movement cNBR (Exhibit 38-52).
Exhibit 38-52
NBR Capacity, Computed for Time Period NBR capacity (veh/h)
Each Time Period
1 1213
2 1045
3 978
4 1182
Exhibit 38-53 summarizes the calculations for this step. During time period 3,
the SBL movement operates at demand over capacity (v/c = 1.56), therefore its
throughput to the ramp is constrained by its capacity value (685 veh/h). For all
other movements and time periods the throughput to the on-ramp is equal to its
demand because v/c < 1.
Time Movements
Parameter Exhibit 38-53
Period EBT NBR SBL Calculation of the On-Ramp
Demand (veh/h) 8 315 652 Demand (vR) Based on the
v/c 0.064 - 0.96 Intersection Operation.
1 c (veh/h) 125 1213 677
min (v, c) 8 315 652
Merge demand vR (veh/h) 975
Demand (veh/h) 96 521 586
v/c 0.768 - 0.93
2 c (veh/h) 125 1045 630
min (v, c) 96 521 586
Merge demand vR (veh/h) 1203
Demand (veh/h) 96 630 1071
v/c 0.77 - 1.56
3 c (veh/h) 125 978 685
min (v, c) 96 630 685
Merge demand vR (veh/h) 1411
Demand (veh/h) 24 80 463
v/c 0.39 - 0.62
4 c (veh/h) 62 1182 746
min (v, c) 24 80 463
Merge demand vR (veh/h) 567
The calculated on-ramp demand is then provided as input into the freeway
facility analysis (Exhibit 38-54). As shown, the ramp flow rates for the merge
segment (segment 5) are obtained from Exhibit 38-53 and highlighted in bold.
Time Period 1 Time Period 2 Time Period 3 Time Period 4 Exhibit 38-54
ML Ramp ML Ramp Ramp Ramp Freeway Facility (I-10
Segment ML flow ML flow
flow flow flow flow flow flow EB) – Demand Inputs
ID rate rate
rate rate rate rate rate rate
(veh/h) (veh/h)
(veh/h) (veh/h) (veh/h) (veh/h) (veh/h) (veh/h)
1 5209 - 6300 - 5300 - 5000 -
2 5209 348 6300 450 5300 1200 5000 50
3 4861 135 5850 116 4100 1000 4950 96
4 4726 - 5734 - 3100 - 4854 -
5 4726 975 5734 1203 3100 1411 4854 567
6 5701 - 6937 - 4511 - 5421 -
7 5701 - 6937 - 4511 - 5421 -
The results of the freeway facility analysis are provided in Exhibit 38-55.
Oversaturated conditions occur during time periods 2 and 3, therefore queueing
may occur along the on-ramp.
The next step will estimate the on-ramp queue length compared to the
available queue storage length to determine whether spillback is expected to
occur. Exhibit 38-56 shows the expected on-ramp queues from the freeway
facility analysis. For each time period, the ramp storage ratio (RQ) is computed by
dividing the ramp queue by the available storage length (924 ft). During time
period 2, a queue is expected on the ramp, but it is not long enough to cause
queue spillback (RQ < 1). During time period 3, however, the on-ramp is expected
to have RQ = 2.31, which indicates that spillback will occur at the intersection
during this time period.
Exhibit 38-56 Time vR Ramp queue Ramp queue Ramp storage Spillback
Spillback Check – I-10 EB on- period (veh/h) (veh) (ft) ratio (RQ) expected?
Ramp
1 975 0.0 0.0 0.00 No
2 1,203 15.0 388.6 0.42 No
3 1,411 82.1 2,133.6 2.31 Yes
4 567 0.0 0.0 0.00 No
Since spillback will occur for at least one time period, the impacts on the
operation of the signalized intersection must be evaluated. The next section
illustrates the application of the methodology to evaluate spillback effects at a
signalized intersection.
Time Period 2
The procedure to evaluate queue spillback into intersections is applied for
time period 2, even though spillback is not expected to occur during this time
period. The application of the methodology is presented for this time period to
facilitate the understanding of the calculations.
Exhibit 38-57
Freeway Facility, Segment 5
(merge) Performance: a)
Merge Capacities and b)
Queue Lengths
Step 7C – Plot queue accumulation polygon for the on-ramp and unsignalized
movements
In this step, a queue accumulation polygon is plotted for the on-ramp as a
function of all protected and permitted movements entering the on-ramp, on a
cycle-by-cycle basis. Since an unsignalized movement (NBR) also discharges into
the on-ramp, a queue accumulation polygon must be developed for this
movement as well. This is required to: (a) determine the discharge pattern of the
unsignalized movement throughout the cycle and (b) allow the estimation of
control delay for this movement.
Exhibit 38-59 presents the queue accumulation profiles for (a) the on-ramp
and (b) for the NBR movement.
Exhibit 38-59
Estimated Queue Lengths And
Merge Capacities – Time
Period 2
The cycle starts with a permitted left-turn movement (Φ1: SBL) discharging
into the on-ramp with a green time g1 = 43.9s, divided in a queue service time gs1
= 40.2s and a queue extension time ge1 = 3.7s (as defined in Chapter 31 –
Signalized Intersections Supplemental). During the green interval for SBL, the
capacity of the NBR movement is constrained since drivers must yield to the
protected left-turn vehicles. The estimated saturation flow rate for the NBR
movement with a conflicting flow vSBL can be obtained by the following equation,
based on HCM equation 31-100:
𝜆 𝑒 ,
𝑠 , =
1−𝑒 ,
where
sNBR,perm = saturation flow rate of the NBR movement (veh/h/ln)
λSBL = throughput of the opposing SBL movement(veh/h)
tcg = critical headway = 4.5 (s)
tfh = follow-up headway = 2.5 (s)
The saturation flow rates of the NBR movement during Φ1 are determined
next. During the SBL queue service time is:
𝜆 = 𝑠 = 1,739 veh/h/ln → 𝑠 , = 282 veh/h/ln
where
sSBL = saturation flow rate of the SBL movement (veh/h/ln)
sNBR,perm1 = saturation flow rate of the NBR movement during the SBL queue
service time (veh/h/ln)
The throughput for the NBR movement is obtained as the minimum of the
demand and saturation flow rate. Since the demand flow rate is greater than the
saturation flow rate, a queue will develop for the NBR movement:
𝜆 , = 𝑚𝑖𝑛 𝑠 , ,𝑣 = 𝑚𝑖𝑛 282, 521
𝜆 , = 282 𝑣𝑒ℎ/ℎ
where
λNBR,1 = throughput for the NBR movement during the SBL queue service time
(veh/h/ln)
vNBR = demand flow rate of the NBR movement (veh/h)
During the SBL green extension time ge, the SBL throughput λSBL is equal to
the arrival flow rate during the effective green (qg,SBL, from Equation 19-32):
𝑣 𝐶
𝜆 =𝑞 , =𝑃 × ×
3600 𝑔
586 120
𝜆 = 0.08 × × = 0.0356 𝑣𝑒ℎ/𝑠/𝑙𝑛 = 128 𝑣𝑒ℎ/ℎ/𝑙𝑛
3600 43.9
where
P = proportion of vehicles arriving during the green indication (decimal);
VSBL = SBL demand flow rate (veh/h);
C = cycle time (s); and
gSBL = SBL effective green time (s)
For this conflicting flow, therefore, the NBR saturation flow rate sNBR,perm2 is
obtained using Equation 31-100:
𝜆 𝑒 ,
𝑠 , =
1−𝑒 ,
586 120
𝜆 = 0.08 × × = 0.0356 𝑣𝑒ℎ/𝑠/𝑙𝑛 = 128 𝑣𝑒ℎ/ℎ/𝑙𝑛
3600 43.9
where
λNBR,2 = throughput for the NBR movement during the SBL green
extension(veh/h/ln)
sNBR,perm2 = saturation flow rate of the NBR movement during the SBL green
extension time (veh/h/ln)
With the discharge patterns for the NBR determined, the queue profile in the
on-ramp during Φ1 can be determined. During the SBL queue service time (cycle
time t = 0 to t = 40.2s), the throughput to the on-ramp is given by:
𝜆 = 𝜆 + 𝜆 , = 1,739 + 282 = 2,021𝑣𝑒ℎ/ℎ 𝑜𝑟 0.561𝑣𝑒ℎ/𝑠
Given the merge capacity cmerge = 1,142 veh/h for the current time period, the
on-ramp queue will grow at the following rate during the SBL queue service
time:
𝜆 −𝑐 = 2,021 − 1,142
Therefore, at the end of the SBL queue service time (t = 40.2s), the queue at
the on-ramp will be 0.244 x 40.2 = 9.8 vehicles (Exhibit 38-59a).
This process is then repeated for all phases throughout the cycle. The results
for a single cycle (120 sec) are presented in Exhibit 38-60, where the maximum
on-ramp queue occurs at t = 50.48s, with 10.82 vehicles (t = 50.48s). The expected
on-ramp queue at the end of the cycle is 2.02 vehicles. The remaining cycles
within time period 2 show the same pattern, where the on-ramp queue at the end
of each cycle becomes the initial queue at the start of the next cycle.
Each row in Exhibit 38-60 describes a portion of the cycle, as follows:
• gs1: queue service time for SBL (Φ1), as previously discussed
• ge1: green extension time for SBL (Φ1). The NBR movement discharges at
the permitted saturation flow rate due to the queue that has developed
during gs1, and the on-ramp queue grows at a rate of 0.07 veh/s
• r1: effective red time for SBL (Φ1). There is no throughput from protected
movements and the NBR movement discharges freely at the saturation
flow rate. The on-ramp queue grows at a rate of 0.11 veh/s
• g2*: effective green for NBT (Φ2), with no throughput from protected
movements. The duration of 0.88s is calculated based on the queue
service time of the NBR approach. The on-ramp queue grows at a rate of
0.11 veh/s
• g2**: remaining effective green for NBT (Φ2). For this portion, no queue
remains on the NBR approach, therefore the NBR throughput is equal to
its demand flow rate (vNBR). The on-ramp queue discharges at a rate of
0.17 veh/s
• r2: effective red time for NBT (Φ2). There is no throughput from protected
movements and the NBR throughput is equal to its demand flow rate
(vNBR). The on-ramp queue discharges at a rate of 0.17 veh/s
• gs7: queue service time for EBT (Φ7). The EBT discharges into the on-ramp
at the saturation flow rate. The throughput of the NBR movement is
At the end of the time period, a residual queue of 23.32 vehicles is expected
along the on-ramp, and this value is carried to the start of the next time period.
The time period length of 900s does not correspond to an exact number of signal
cycles, and the last cycle is interrupted at t = 60s. Therefore, the next time period
will start the analysis from the same timestamp to maintain consistency.
Time Period 3
The same steps performed for the analysis of time period 2 are applied again
for the analysis of time period 3.
Step 7C – Plot queue accumulation polygon for the on-ramp and unsignalized
movements
The procedure described earlier is applied but with an initial on-ramp queue
of 23.32 vehicles, which is the estimated queue at the end of time period 2. The
analysis begins at the middle of the cycle (t= 60s), which is the end of the
previous time period. Exhibit 38-61 illustrates the queue accumulation polygon
for both the on-ramp and the NBR movement.
Exhibit 38-61
Estimated Queue Lengths and
Merge Capacities – Time
Period 3
Queue spillback occurs during the third cycle (SBL queue service time),
when the on-ramp queue reaches the maximum storage LONR = 35.5 vehicles. At
this time, the maximum flow rate that can enter the on-ramp is constrained by
the merge capacity cmerge. In other words, the maximum number of vehicles
allowed to enter the ramp is equal to the number of vehicles that are able to
merge to the freeway mainline. Also, the queues developed in the NBR are
longer during cycles 3 through 8, causing an increased delay on this movement
due to the queue spillback conditions at the on-ramp.
The on-ramp queue at the start of cycle 3 is 27.9 vehicles. The cycle starts
with the SBL movement, with an effective green time g1 = 47.3s. Since this
movement already operates with v/c > 1, the queue service time gs1 is equal to g1,
and no green extension time is available (ge1 = 0). The protected movement then
discharges at saturation flow rate sSBL = 0.483 veh/s, while the NBR movement
discharges at a permitted saturation flow rate sNBR = 0.078 veh/s. At the same
time, the on-ramp discharges to the freeway at a rate cmerge = 1,142 veh/h = 0.317
veh/s. Therefore, the on-ramp queue grows at the following rate:
Spillback is then expected to occur within 31.2 seconds of the onset of g1. The
total effective green g1 value of 47.3s is then divided in two portions:
• gs1* (31.2s): discharging at saturation flow rate
• gs1,sp (16.1s): the remainder of g1 will be affected by queue spillback,
limiting the maximum discharge to the on-ramp to the merge capacity
cmerge = 0.317 veh/s. Note that this constraint is shared by two movements
entering the on-ramp (SBL and NBR).
The effect of queue spillback on the intersection capacity during gs1,sp is then
measured by the capacity reduction factor β1,sp, defined as the ratio between the
maximum on-ramp capacity during queue spillback and the throughput from
the intersection movements (SBL and NBR):
𝑐 0.317
𝛽 , = = = 𝟎. 𝟓𝟔𝟓
𝜆 + 𝜆 0.483 + 0.078
A capacity reduction factor β1,sp= 0.565 means that only 56.5% of the expected
intersection throughput is able to enter the on-ramp when queue spillback occurs
during phase gs1,sp. This capacity adjustment factor is applied to each movement
to obtain their adjusted throughputs for this time period:
𝜆 , =𝜆 ×𝛽 , = 0.483 × 0.565 = 0.273 𝑣𝑒ℎ/𝑠
𝜆 , =𝜆 ×𝛽 , = 0.078 × 0.565 = 0.044 𝑣𝑒ℎ/𝑠
The procedure is then repeated for the remaining movements of the cycle, as
shown in Exhibit 38-62.
As shown, at time t = 31.2 s the maximum storage length of the on-ramp is
reached and spillback occurs. From this time through t = 83.3s, the throughput
from intersection movements to the on-ramp λONR is greater than the merge
capacity cmerge. Therefore, the maximum allowed throughput λONR,ajd is
constrained by the on-ramp discharge capacity cmerge = 0.137 veh/s. For these
cases, the spillback capacity reduction factor fsp is computed as the ratio of
λONR,ajd and λONR. Note that for this time range the on-ramp queue is kept constant
at the maximum storage of 35.54 vehicles.
From t = 83.3s, the on-ramp queue begins to discharge at a rate of 0.142 veh/s,
followed by a small increase during the green time of phase 7 (EBT), but it is not
sufficient to cause spillback. At the end of the cycle, the residual on-ramp queue
is 33.51 vehicles.
The subsequent cycles follow a recurring pattern, with the on-ramp reaching
maximum storage early in the cycle and slightly diminishing at the end of the
cycle.
Protected
Permitted movement On-ramp analysis
Exhibit 38-62
movement Discharge Flow Rates Into the
Active Duration QONR
t (s) λONR,adj - βsp On-Ramp for Each Phase
phase (s) (veh) λprot vNBR λNBR Q(NBR) λONR λONR,adj
cmerge
(veh/s) (veh/s) (veh/s) (veh) (veh/s) (veh/s) Throughout the Cycle – Time
(veh/s)
Period 3
gs1* 0.0 31.2 27.9 0.483 0.175 0.078 0 0.561 0.561 0.244 1
gs1,sp 31.2 16.1 35.5 0.483 0.175 0.078 3.01 0.561 0.317 0 0.565
g2* 53.0 30.3 35.5 0 0.175 0.43 4.31 0.43 0.317 0 0.739
gs7 106.0 6.3 32.3 0.503 0.175 0.073 0 0.576 0.576 0.259 1
ge7 112.3 2 33.9 0.027 0.175 0.366 0.64 0.393 0.393 0.076 1
respectively, by their duration. At the end of the table, the expected and actual
volumes are aggregated and a capacity reduction factor βsp,SBL = 0.704 is obtained
as the ratio of these values.
The capacity of the SBL movement without consideration of queue spillback
is 685 veh/h (Exhibit 38-53). The adjusted capacity is calculated by applying the
spillback capacity reduction factor βsp, calculated in Exhibit 38-63
In this example, this step is not required for the EBT movement, since this
movement does not experience effects of queue spillback. As shown in Exhibit
38-61, the on-ramp queue during the EBT green does not reach the maximum
storage length of 35.5 veh.
Exhibit 38-63
On-ramp analysis Spillback adjustment
Calculation of Spillback
Capacity Reduction Factor for On-ramp On-ramp
Active Duration expected actual
the SBL Movement for Time Cycle λONR,adj
phase (s) λONR
βsp discharge discharge
Period 3
(veh/s) (veh/s)
volume volume
(veh) (veh)
2 gs1 47.3 0.561 0.561 1 26.56 26.56
3 gs1* 31.2 0.561 0.561 1 17.51 17.51
3 gs1,sp 16.1 0.561 0.317 0.565 9.04 5.11
4 gs1 8.3 0.561 0.561 1 4.67 4.67
4 gs1,sp 39 0.561 0.317 0.565 21.89 12.37
5 gs1 5.1 0.561 0.561 1 2.87 2.87
5 gs1,sp 42.2 0.561 0.317 0.565 23.68 13.39
6 gs1 4.7 0.561 0.561 1 2.62 2.62
6 gs1,sp 42.6 0.561 0.317 0.565 23.93 13.53
7 gs1 4.6 0.561 0.561 1 2.59 2.59
7 gs1,sp 42.7 0.561 0.317 0.565 23.97 13.55
8 gs1 4.6 0.561 0.561 1 2.58 2.58
8 gs1,sp 42.7 0.561 0.317 0.565 23.97 13.55
Time Period 4
The same steps performed for time periods 2 and 3 are applied again in time
period 4.
The merge capacity for time period 4 has been previously determined, as
shown in Exhibit 38-57a. Since the congestion along the freeway mainline is
dissipating during this time period, the merge capacity is not constant: from time
steps 1 through 4, the merge capacity is 1,142 veh/h, consistent with
oversaturated conditions from previous time periods. After time step 5, the
merge capacity is set equal to the ramp roadway capacity (1,904 veh/h)
Step 7C – Plot queue accumulation polygon for the on-ramp and unsignalized
movements
The procedure described earlier is applied to plot the queue accumulation
polygons, shown in Exhibit 38-64. Queue spillback occurs during the first cycle,
due to the residual queue from the previous time period. However, due to low
volumes at the intersection and improvement of performance along the freeway
mainline, the on-ramp clears quickly. The queue has cleared by the end of the
second cycle.
Exhibit 38-64
Estimated Queue Lengths and
Merge Capacities – Time
Period 4
On-ramp
Exhibit 38-65 Spillback adjustment
analysis
Calculation of Spillback Active Duration QONR On-ramp On-ramp
Capacity Reduction Factor for Cycle
phase (s) (veh) λONR λONR,adj expected actual
the SBL Movement During βsp
(veh/s) (veh/s) throughput throughput
Time Period 4 (veh) (veh)
1 gs1 6 34.4 0.505 0.505 1 3.02 3.02
1 gs1,sp 29.9 35.5 0.505 0.317 0.628 15.12 9.5
1 ge1 0 35.5 0.388 0.317 0.818 0 0
2 gs1 31.2 13.2 0.505 0.505 1 15.79 15.79
2 ge1 4.7 19.1 0.095 0.095 1 0.44 0.44
3 gs1 31.2 0.0 0.505 0.505 1 15.79 15.79
3 ge1 4.7 5.9 0.058 0.058 1 0.27 0.27
4 gs1 40.2 0.0 0.561 0.561 1 22.55 22.55
4 ge1 3.7 9.8 0.392 0.392 1 1.46 1.46
5 gs1 40.2 0.0 0.561 0.561 1 22.55 22.55
5 ge1 3.7 1.3 0.392 0.392 1 1.46 1.46
6 gs1 40.2 0.0 0.561 0.561 1 22.55 22.55
6 ge1 3.7 1.3 0.392 0.392 1 1.46 1.46
7 gs1 40.2 0.0 0.561 0.561 1 22.55 22.55
7 ge1 3.7 1.3 0.392 0.392 1 1.46 1.46
8 gs1 40.2 0.0 0.561 0.561 1 22.55 22.55
8 ge1 3.7 1.3 0.392 0.392 1 1.46 1.46
With the adjusted capacity values obtained, the performance measures for
the intersection can be computed using the remaining steps from the Signalized
Intersections methodology (Chapter 19): compute the adjusted demand-to-
capacity ratio (Step 8) and compute control delay (Step 9). Exhibit 38-66
compares the performance measures for the affected movement (SBL) for the
cases with and without accounting for spillback effects. There is no change in the
performance measures in time period 2 even though the on-ramp demand is
greater than the merge capacity, as the queue can be stored in the on-ramp. Time
period 3 yields a significant increase in the SBL control delay due to the queue
spillback: 589.2 s/veh, while the intersection analysis without consideration of the
spillback effects would return a control delay of 293.5 s/veh. Time period 4 shows
a small increase in control delay, from 575.2 s/veh to 609.5 s/veh. Even though
spillback occurs for only a short time during this time period, the high value of
control delay obtained is due to the initial queue delay (d3), as a result of the
unmet demand at the end of time period 3.
Exhibit 38-67
TWSC Intersection Geometry
– Acadian Thruway @ I-10
EB.
Time Movements
Parameter Exhibit 38-68
Period EBT NBR SBL Calculation of the On-Ramp
Demand (vR) Based on the
Demand (veh/h) 8 315 652 TWSC Intersection Operation.
v/c 0.064 - 0.96
1 c (veh/h) 125 1547 677
min (v, c) 8 315 652
Merge demand vR (veh/h) 975
Demand (veh/h) 96 521 586
v/c 0.768 - 0.93
2 c (veh/h) 125 1547 630
min (v, c) 96 521 586
Merge demand vR (veh/h) 1203
Demand (veh/h) 96 630 1071
v/c 0.77 - 1.56
3 c (veh/h) 125 1547 685
min (v, c) 96 630 685
Merge demand vR (veh/h) 1411
Demand (veh/h) 24 80 463
v/c 0.39 - 0.62
4 c (veh/h) 62 1547 746
min (v, c) 24 80 463
Merge demand vR (veh/h) 567
The on-ramp demand estimates are then used as inputs for the freeway
facility analysis. Since the input demands for the freeway are identical to the
example provided in Part 1, it is already known that spillback will occur during
time period 3 (Exhibit 38-56).
Step 9B. Obtain merging capacity using the freeway facilities methodology
This step computes the merging capacity into the freeway cmerge. Since the
inputs of the freeway facility remain unchanged, the same values from the
previous case study are used:
Exhibit 38-70 illustrates the queue accumulation polygon for the on-ramp,
based on the table results.
Exhibit 38-70
Queue Accumulation Polygon
for the On-Ramp – TWSC
Intersection
𝑐 , + 𝑐 , +𝑐 , =𝑐 = 1,142 𝑣𝑒ℎ/ℎ
𝑐 ×𝑣 1,142 × 685
𝑐 , = = = 554.4 𝑣𝑒ℎ/ℎ
𝑣 +𝑣 +𝑣 685 + 708 + 18
𝑐 ×𝑣 1,142 × 708
𝑐 , = = = 573.0 𝑣𝑒ℎ/ℎ
𝑣 +𝑣 +𝑣 685 + 708 + 18
𝑐 ×𝑣 1,142 × 18
𝑐 , = = = 14.6 𝑣𝑒ℎ/ℎ
𝑣 +𝑣 +𝑣 685 + 708 + 18
The equivalent capacities cEQ,i for each movement i, aggregated for the 15-
min time period, are obtained proportionately to the spillback time TSB (Finally,
the adjusted capacity of each affected movement ci,EQ is obtained as a function of
the amount of time within the time period when spillback was present. The
adjusted capacity considers the):
, × × . × . × .
𝑐 , = = = 757 𝑣𝑒ℎ/ℎ
, × × × . × .
𝑐 , = = = 869 𝑣𝑒ℎ/ℎ
, × × × . . × .
𝑐 , = = = 24 𝑣𝑒ℎ/ℎ
With the adjusted capacity values obtained, the performance measures for
the intersection can be computed using the next step from the TWSC
methodology (Chapter 20): compute movement control delay (Step 11).
Exhibit 38-71 compares the performance measures of the affected intersection
movements for the cases with and without spillback effects during time period 3.
All three movements discharging to the on-ramp experienced significant increase
in the control delay.
Exhibit 38-71
Comparison of Performance Capacity (veh/h) Control delay (s/veh)
Measures in a TWSC Demand
Movement Without With With
Intersection – Time Period 3 - (veh/h) Without spillback
with and without Spillback spillback spillback spillback
Effects EBT 18 28 18.7 166.5 479.8
NBR 708 1547 868.9 0 24.5
SBL 685 1222 757.2 9.4 37.2
Exhibit 38-72
AWSC Intersection Geometry
– Acadian Thruway @ I-10 EB
The estimated on-ramp demand values are provided as inputs for the
freeway facility analysis. The freeway facility is then analyzed and the expected
on-ramp queues are provided in Exhibit 38-74.
Ramp Ramp
Exhibit 38-74 Time vR Ramp Spillback
queue storage
Check for Spillback period (veh/h) queue (ft) expected?
(veh) ratio (RQ)
Occurrence – AWSC
Intersection 1 834 0 0 0 No
2 984 14.9 21.9 0.62 No
3 1020 82.1 53.4 1.5 Yes
4 504 0 0 0 No
Since spillback will occur, the impacts on the operation of the intersection
must be evaluated. The next section illustrates the application of the evaluation
methodology at the AWSC intersection.
2 15 984 0.023 0 - - 21
3 15 1020 0.033 15.2 7.25 7.75 35.5
4 15 504 -0.11 35.5 - - 0
Exhibit 38-94 illustrates the queue accumulation polygon for the on-ramp,
based on the table results.
Exhibit 38-76
Queue Accumulation Polygon
for the On-Ramp – AWSC
Intersection
Capacity Spillback
Regular Equivalent
during departure
Movement Capacity Capacity (cEQ)
spillback (csp) headway (hsp)
(c) (veh/h) (veh/h)
Exhibit 38-77 (veh/h) (s)
Equivalent Capacities and
Headways for on-ramp – Time EBT 15 396 212.1 17
Period 3 – AWSC Intersection NBR 439 550 496.5 7.3
SBL 445 462 453.7 7.9
With the adjusted capacity values obtained, the performance measures for
the intersection can be computed using the remaining steps from the AWSC
methodology (Chapter 21): compute the service times (Step 13) and compute
control delay (Step 14).
Exhibit 38-78 compares the performance measures of the intersection
movements for the cases with and without spillback effects during time period 3.
The three movements that discharge into the on-ramp (EBT, NBR and SBL)
experience increased delay, while the remaining movements have their
performance measures unchanged.
Departure headway
Exhibit 38-78 Capacity (veh/h) Control delay (s/veh)
Demand (s)
Comparison of Performance Movement
(veh/h) Without With Without With Without With
Measures – Time Period 3 -
spillback spillback spillback spillback spillback spillback
with and without Spillback
Effects EBL 75 359 359 15.6 15.6 10 10
EBT 19 396 212 12.6 21.7 9.1 17
NBT 229 497 497 16.3 16.3 7.2 7.2
NBR 539 550 497 58.9 92.3 6.5 7.3
SBL 546 462 454 128 136.5 7.8 7.9
SBT 220 494 494 16 16 7.3 7.3
Exhibit 38-79
Study Site for Freeway-to-
Freeway Queue Spillback
Check, Miami, FL
Exhibit 38-80
Individual Freeway Facilities:
(a) I-75 SB and (b) SR-826 SB
Input data
Traffic demands for the freeway facilities and ramps are provided in Exhibit
38-1 in 15-minute time periods.
Freeway Facility 1 (I-75 SB) Freeway Facility 2 (SR-826 SB)
Time Mainline Diverge Mainline Merge demand
Exhibit 38-81 Period demand flow demand flow demand flow flow rate
Traffic Demands for the
rate (veh/h) rate (veh/h) rate (veh/h) (veh/h)
Subject Freeway Facilities
1 5400 1400 4000 1400
2 6200 3000 5700 3000
3 6000 3400 5600 3400
4 4500 800 4500 800
Exhibit 38-82
Time Segment 1 Segment 2 Segment 3 Segment 4
Performance Measures for I-
period Basic Basic Diverge Basic 75 (Freeway Facility 1)
1 C C B B
2 C C C A
3 C C C A
4 B B A B
Spillback check
The analysis of SR-826 using the Freeway Facilities Oversaturated Segment
Evaluation provides the expected on-ramp queue for every time period. The first
check compares the off-ramp demand to the ramp roadway capacity, as shown
in Exhibit 38-84. The ramp queue starts to develop during time period 2. At the
end of this time period, a ramp queue length of 1188 ft is expected, yielding a
queue storage ratio of 0.33. Therefore, spillback is not expected during time
period 2. During time period 3 a ramp queue length of 5160 ft is expected with a
queue storage ratio of 1.41. Therefore, spillback will occur during time period 3.
Total
Number of Average
Exhibit 38-84 number Queue Ramp Queue
queued vehicle
Estimation of Queue Length of length length storage
vehicles in spacing
and Storage Ratio at the SR- Time queued (ft) (ft) ratio Spillback
each lane (ft)
826 On-Ramp period vehicles occurs?
1 0 0 - 0 0.00 No
4 0 0 - 0 0.00 Yes
Spillback analysis
Since spillback is expected to occur, the methodology described in Appendix
A (Exhibit 38-A8 through Exhibit 38-A11) is applied to evaluate its impacts on I-
75 SB. The application of the methodology for each time period is presented
below.
Time period 1
No oversaturated conditions occur, therefore no additional calculations are
needed for this time period.
Time period 2
During time period 2, the downstream merge segment operates at LOS F and
the on-ramp capacity is expected to be reduced.
𝑆𝐶𝐸𝑄 𝑖, 𝑁, 𝑁𝑄 = 𝑆𝐶 𝑖, 𝑁 − 𝑁𝑄 × 𝐶𝐴𝐹
The capacity adjustment factor CAFBL is obtained from Exhibit 38-3. For a
segment with 5 directional lanes and 2 blocked lanes, an adjustment factor CAFBL
= 0.67 is applied. Therefore, the equivalent capacity of the unblocked portion is
given by:
The unblocked background density KBUB is calculated next. For time period
2, an expected demand of 4165.8 pc/h for the mainline is used in the calculations.
The KBUB parameter of the unblocked lanes is computed as the density of a 3-
lane basic segment with a capacity SCEQ = 7872 pc/h:
Exhibit 38-86
Queued Vehicles and Total
Number of Vehicles in the
Ramp – Time Period 2
Exhibit 38-87
Ramp Capacity and Ramp
Inputs – Time Period 2
Since spillback does not occur, no additional calculations for the mainline are
required.
𝑅𝐹 𝑖, 𝑝, 𝑘 = 4 × 𝑅𝐹 𝑖, 𝑡, 𝑝, 𝑘 = 1679.5 pc/h/ln
1
𝑅𝐾 𝑖, 𝑝, 𝑘 = × 𝑅𝑁𝑉 𝑖, 𝑡, 𝑝, 𝑘 = 71.6 pc/mi/ln
60
𝑅𝐹 𝑖, 𝑝, 𝑘 1679.5
𝑆𝑅 𝑖, 𝑝, 𝑘 = = = 31.9mi/h
𝑅𝐾 𝑖, 𝑝, 𝑘 71.6
Time period 3
The same steps are repeated for time period 3. The ramp analysis is
summarized in Exhibit 38-88. For this time period, the ramp demand is 15.4 pc/ts,
while the merge capacity is 13.9 pc/ts. Since demand is greater than capacity, the
number of vehicles increases gradually, causing the capacity constraint RSTG to
decrease each time step. At time step 14, the value of RSTG becomes equal to the
merge capacity (13.9 pc/ts), which implies that the ramp has reached jam density
and the maximum flow that can enter the ramp is equal to the flow that departs
the ramp. Therefore, queue spillback into the mainline starts at time step 15.
Exhibit 38-88
Ramp Capacities and Ramp
Inputs – Time Period 3
After the onset of queue spillback, the number of unserved vehicles at the
exit is computed every time step through the parameter OFRUV(i,t,p). Then, the
expected length of the mainline queue OFRLQ(i,t,p) is computed based on the
number of unserved vehicles and the ramp queue density RKQ, as shown in
Equation 38-A33:
𝑂𝐹𝑅𝑈𝑉 𝑖, 𝑡, 𝑝
𝑂𝐹𝑅𝐿𝑄 𝑖, 𝑡, 𝑝 =
𝑅𝐾𝑄 𝑖, 𝑡, 𝑝
The ramp queue density RKQ is obtained using Equation 38-A21:
𝐾𝐽 – 𝑅𝐾𝐶 𝑥 𝑅𝐹 𝑖, 𝑡 − 1, 𝑝
𝑅𝐾𝑄 𝑖, 𝑡, 𝑝, 𝑘 = 𝐾𝐽–
𝑅𝐶 𝑖, 𝑡, 𝑝
𝑅𝐾𝑄 𝑖, 𝑡, 𝑝, 𝑘 = 190– 190 – 46.5 × 13.87 / 18.33 = 81.4 pc/mi/ln
Exhibit 38-89 illustrates the expected spillback queue length during time
period 3.
Exhibit 38-89
Spillback Queue Length –
Segment 3 (Diverge) – I-75
SB
The parameter OFRLQ represents the length of the queue if all unserved
vehicles were queued in a single line. Given the segment geometry (Exhibit 38-
90), the operating regimes and flow modes can be obtained as a function of
OFRLQ:
Exhibit 38-90
Available Queue Storage –
Segment 3 (Diverge) – I-75
SB
Exhibit 38-91
Back of Queue Length,
Including QIA, at the End of
Time Period 3
𝑅𝐹 𝑖, 𝑝, 𝑘 = 4 × 𝑅𝐹 𝑖, 𝑡, 𝑝, 𝑘 = 1707 𝑝𝑐/ℎ/𝑙𝑛
1
𝑅𝐾 𝑖, 𝑝, 𝑘 = × 𝑅𝑁𝑉 𝑖, 𝑡, 𝑝, 𝑘 = 108.4 𝑝𝑐/𝑚𝑖/𝑙𝑛
60
𝑅𝐹 𝑖, 𝑝, 𝑘 1707
𝑆𝑅 𝑖, 𝑝, 𝑘 = = = 21.5 𝑚𝑖/ℎ
𝑅𝐾 𝑖, 𝑝, 𝑘 108.4
For the freeway facility, performance measures are computed for the blocked
and unblocked portions of each segment.
The average density is obtained as the sum of two separate components. The
average number of vehicles in the blocked portion of the segment is computed
as:
1
𝐾𝐵𝐿 𝑖, 𝑝 = × 𝑁𝑉(𝑖, 𝑡, 𝑝) = 51 𝑝𝑐/𝑚𝑖/𝑙𝑛
60
1
∆𝐾(𝑖, 𝑝) = ∆𝑁𝑉(𝑖, 𝑡, 𝑝) = 20.1 𝑝𝑐/𝑚𝑖/𝑙𝑛
𝑆×𝑁
Finally, the speed in the blocked lanes is obtained through the fundamental
equation:
𝑆𝐹𝐵𝐿(𝑖, 𝑝) 3030
𝑆𝐵𝐿(𝑖, 𝑝) = = = 21.2𝑚𝑖/ℎ
𝑁(𝑖, 𝑝) × 𝐾(𝑖, 𝑝) 2 × 70.1
Time period 4
Exhibit 38-92
Schematic of the Study
Interchange for Example
Problem 4
𝑣 , , = 0 𝑝𝑐/ℎ
𝑣 , , = 300 𝑝𝑐/ℎ
⎡ 3,600 ⎤
⎢ 0.22⎥ 1,380
1,380
𝑄 , = 900(0.25) ⎢0.22 − 1 + (1 − 0.22) + = 1 𝑣𝑒ℎ
150(0.25) ⎥ 3,600
⎢ ⎥
⎣ ⎦
Similarly,
𝑄 , = 6 𝑣𝑒ℎ
𝑄 , = 121 𝑣𝑒ℎ
These values are rounded to the nearest vehicle.
Exhibit 38-93 provides the flows and resulting queues at the roundabout.
Exhibit 38-94
Priority Order for the
Roundabout of Example
Problem 4
In this example, it is assumed that the exit lane towards the on-ramp can
reach an exit flow rate of 1,300veh/h.
Starting from the approach with Rank 1 (southbound approach), first the
maximum throughput for the movement that exits through the eastbound leg
(the on-ramp) is calculated as follows:
3,600 100 3,600
𝜆 , = 𝑚𝑖𝑛 𝑣 , ,𝑐 , ×𝑝 , = 𝑚𝑖𝑛 100,1380 × , Equation 38-18
ℎ 300 2.77
= 100𝑝𝑐/ℎ
where
𝜆 , = maximum throughput for the southbound-left movement (pc/h);
𝑣 , = flow rate for the southbound-left movement (pc/h);
𝑐 , = entry lane capacity for the southbound roundabout approach (pc/h);
𝑝 = percent of demand from SB approach into the on-ramp
This total on-ramp demand flow rate is lower than the exit demand
(Equation 22-12), which has a rate of:
𝑣 , , =𝑣 , +𝑣 , +𝑣 , = 100 + 500 + 1500 = 2,100 𝑝𝑐/ℎ
3600
3600 × 0.22
𝑑 = + 900(0.25) 0.22 − 1 + (0.22 − 1) + 1380 +5
1380 450(0.25)
2100
× 𝑚𝑖𝑛 ,1
800
𝑑 = 747.94 𝑠/𝑣𝑒ℎ
Therefore, the total average delay per approach is:
100
𝑑 , =𝑑 +𝑑 × = 65.05 𝑠/𝑣𝑒ℎ
1234
500
𝑑 . =𝑑 +𝑑 × = 317.54 𝑠/𝑣𝑒ℎ
1234
634
𝑑 , =𝑑 +𝑑 × = 1020.14 𝑠/𝑣𝑒ℎ
1234
5. REFERENCES
Exhibit 38-A1
Off-Ramp Queue Spillback
Check Flowchart
CAPACITY CHECKS
The procedure first determines whether capacity is exceeded at any of the
critical points along the diverge section.
where
𝑄 = accumulated queue length at the end of analysis period 𝑡 (pc);
𝑄 = queue growth during analysis period 𝑖 (pc); and
𝑡 = the current analysis period.
The Study Period is defined as “The time interval within a day for which
facility performance is evaluated, consisting of one or more consecutive analysis
periods” (Chapter 9, Glossary and Symbols). Therefore, the study period t refers
to the time boundaries defined in Step A-1 on the Freeway System Methodology,
and is composed of N analysis periods, which typically have a 15-min duration.
The maximum queue length 𝑄 during the entire analysis period is the
maximum value of 𝑄 obtained using Equation 38-A2 and is used as input for the
next stage of the spillback check procedure.
where
𝑣 = maximum entering flow rate for the intersection approach (veh/h);
𝑣 = off‐ramp demand for the period (pc/h);
𝑐 = capacity of the off-ramp roadway (pc/h);
𝑓 = adjustment factor for heavy vehicle presence; and
𝑃𝐻𝐹 = peak hour factor;
If the off-ramp demand exceeds its capacity, the ramp roadway acts as an
upstream bottleneck and limits the demand to the intersection approach. This
step ensures that the incoming demand at intersection does not exceed the
capacity of the ramp roadway. The calculations of throughput for each
intersection type are described below.
Signalized Intersections: The methodology of Chapters 19 and 31 evaluates
the performance of individual lane groups for a subject approach. It also
estimates the back of queue length 𝑄 (Equation 31-149) or a percentile back-of-
queue length 𝑄% (Equation 31-150). In some cases, only one high-demand
movement at the intersection approach is the bottleneck that results in spillback,
yielding an unbalanced lane usage pattern at the ramp. Field observations have
shown that urban street intersection failures may occur at one lane group. As
drivers position themselves in a specific lane at the ramp to anticipate the
downstream signal, the lane usage in the ramp becomes unbalanced, as shown in
Exhibit 38-A3.
Exhibit 38-A3
Examples of Unbalanced
Ramp Lane Usage: (a)
Norfolk/VA and (b) Tampa/FL
At off-ramps with two or more lanes, the estimated queue lengths for each
intersection lane group must be associated with specific ramp lanes. Exhibit 38-
A4 illustrates an example of a typical ramp terminal for a two-lane off-ramp.
Drivers that desire to take a left turn at the intersection will position themselves
in the leftmost lane (Ramp Lane 1), while drivers who intend to turn right will
likely choose the rightmost lane (Ramp Lane 2). Analyst judgement is required to
define the grouping of intersection lane groups into ramp lanes.
Exhibit 38-A4
Spillback Occurrence by Lane
at an Off-Ramp / Weaving
Segment
where
𝑄 , = number of queued vehicles in Ramp Lane k, during a 15-min interval;
𝑄 , = number of queued vehicles from Lane Group m associated with ramp
lane k, during a 15-min interval;
𝑄 %, = estimated back of queue length (nth percentile), as defined in Equation
31-150 (measured in veh/ln); and
𝑁 , = number of approaching lanes for Lane Group 𝑚.
Unsignalized Intersections
Each unsignalized intersection type has its own methodology to estimate
queue length. The TWSC methodology estimates the 95th percentile queue length
for minor movements with Equation 20-68, while the 95th percentile queue length
for AWSC approaches is estimated with Equation 21-33. For roundabouts, the
95th percentile queue length for a given lane is provided by Equation 22-20.
Regarding intersection lane groups and ramp lanes, the same procedure
discussed above for signalized intersections is applied.
the demand difference among ramp lanes can be considered negligible for the
purposes of this analysis.
Next, the procedure estimates the queue storage ratio (𝑅 ) for the ramp
roadway queues. If 𝑅 exceeds 1.00, then spillback is expected to occur. The
calculations for each of the three possible cases are provided below.
where
𝑄 = length of queue beyond ramp storage distance (ft);
𝑣 = demand for the off‐ramp (pc/h);
𝑐 = capacity of the off-ramp (pc/h);
𝑓 = adjustment factor for heavy vehicle presence;
𝑃𝐻𝐹 = peak hour factor;
𝐿 = average vehicle spacing in stationary queue (ft/veh); and
𝑡 = analysis period i (h).
where
𝑁 = number of lanes in section 𝑖; and
𝐿 = length in section 𝑖 (ft).
The individual ramp storage for each of the 𝑘 lanes in the off-ramp, 𝐿 , , can
be estimated by assigning the intersection lane groups to ramp lanes, as
previously described:
Equation 38-A8
𝐿 , = 𝑁, × 𝐿
where
𝑁, = number of lanes in section 𝑖 that are associated to ramp lane 𝑘; and
𝐿 = length in section 𝑖 (ft).
Finally, the ramp queue ratio for every ramp lane 𝑘 is obtained as:
𝑄 ,
Equation 38-A9 𝑅 , =
𝐿 ,
where
𝑄 , = queue length associated to ramp lane 𝑘; and
𝐿 , = available ramp storage for ramp lane 𝑘.
Next, the total storage length is calculated. The example from Exhibit 38-A4
illustrates a common off-ramp geometry with three different sections from the
stop bar to the gore point:
• Section 1: 4 lanes with length L1 - two lanes (LG1) are associated with
ramp lane 1, and two lanes (LG2) are associated with ramp lane 2
• Section 2: 3 lanes with length L2 - one lane (LG1) is associated with ramp
lane 1, and two lanes (LG2) are associated with ramp lane 2
• Section 3: 2 lanes with length L3 - one lane (LG1) is associated with ramp
lane 1, and one lane (LG2) is associated with ramp lane 2
Therefore, the available ramp storage LR is calculated as:
𝐿 = (4 × 𝐿 ) + (3 × 𝐿 ) + (2 × 𝐿 )
The ramp storage for each ramp lane is as follows:
𝐿 , = (2 × 𝐿 ) + (1 × 𝐿 ) + (1 × 𝐿 )
𝐿 , = (2 × 𝐿 ) + (2 × 𝐿 ) + (1 × 𝐿 )
Exhibit 38-A5
Expanded Link-Node Structure
to Evaluate the Off-Ramp
Segment
Exhibit 38-A6
Sample Geometry of an off-
Ramp Considering the Arterial
Intersection with Heavy
Demanded Left-Turn
The type of ramp terminal is an important input into the analysis. Signalized
intersections operate in cyclical patterns, and therefore those have fluctuating
queue lengths. For certain demand scenarios, this can result in queues backing
up into the freeway and then discharging multiples times within a 15-min time
period.
Stop-controlled intersections and downstream merge segments (in the case
of a freeway-to-freeway connection), on the other hand, have a more uniform
discharging rate. For cases other than signalized intersections, off-ramp queues
Regime 1
The queue ends within the deceleration lane and does not spill back into the
mainline freeway (Exhibit 38-A7 (a)). During undersaturated conditions, the
deceleration lane serves as a transition zone between speeds on the mainline
(typically 55 – 75 mi/h) and advisory speeds posted along the off-ramp (typically
20 – 50 mi/h). When queues begin to form on the deceleration lane, the available
deceleration distance is reduced and speeds along the rightmost lane are
affected.
Regime 2
The queue of vehicles extends upstream beyond the deceleration lane, but
sufficient lateral clearance on the right-hand shoulder allows for additional
queue storage. In this case there is no transition zone within the deceleration lane
and drivers decelerate and join the back of the queue more abruptly, resulting in
turbulence and reduced speeds in the rightmost lane (Exhibit 38-A7 (b)). If no
lateral clearance exists immediately upstream of the deceleration lane, Regime 2
conditions are not possible. In some cases, this regime does not occur even if
storage is available; this depends on local driver behavior and is site-specific.
Regime 3
The queue extends to the rightmost lane of the freeway mainline (Exhibit 38-
A7 (c)). This may occur either when there is no shoulder available for additional
queue storage, or when drivers choose to queue in the rightmost lane once the
deceleration lane is entirely occupied. Non-exiting vehicles on the rightmost lane
are delayed or change lanes, which causes increased turbulence and reduced
speeds in both rightmost lanes.
Regime 4
The queue blocks the rightmost lane, and drivers occasionally or often use
the adjacent freeway mainline lane next to the rightmost freeway mainline lane
to force their way into the queue, blocking thus an additional lane (Exhibit 38-
A7(d)). During this regime, speed and capacity are significantly reduced. The
effects of spillback vary from site to site and from time period to time period due
to driver behavior and site geometry. Data collection at locations around the US
has shown that at some sites drivers block the adjacent lane, while at other sites
they do not, regardless of the queue spillback length at the site.
Exhibit 38-A7
Off-Ramp Queue Spillback
Regimes
Facility variables
• QIA(i, p): Length of the queue influence area (ft) for segment i during time
period p, measured from the back of the queue.
Segment variables
• KBBL(i,j): background density (pc/mi/ln) at the blocked lanes in segment i,
when queue spillback occurs at a downstream segment j
• KBUB(i,j): background density (pc/mi/ln) at the unblocked lanes in
segment i, when queue spillback occurs at a downstream segment j
• LCR(i,t,p): rate of lane change maneuvers in the queue influence area
upstream of a queue from an off-ramp, for segment i during time period
p and time step t.
• LD(i,p): available deceleration lane length (ft) for segment i during time
period p. This variable is used to calculate performance measures for
ramp segments (Chapter 14 - LD.)
• MQ1(i,t,p): mainline queue length of off-ramp unserved vehicles in the
rightmost mainline lane, for segment i during time period p in time period
t.
• MQ2(i,t,p): mainline queue length of off-ramp unserved vehicles in the
rightmost mainline lane, for segment i during time period p in time period
t. If Regime 4 is not expected to occur, this parameter value is set to zero.
• NQ(i): number of blocked lanes if the off-ramp queue backs up into the
freeway mainline. This parameter is a function of the prevailing spillback
regime at segment i as provided by the analyst. The value for this
parameter is an input and can be either 1 (Regime 3 - one blocked lane) or
2 (Regime 4 – two blocked lanes);
• OFRFUP(i,t,p): flow that can exit at the closest off-ramp downstream of i
during time step t in time period p.
Node variables
• CAFBL(i,t,p): capacity adjustment when one or more lanes of segment i are
entirely blocked during time period p in time period t. This is used to
calculate friction effects that cause through vehicles to slow down due to
the presence of a queue in the rightmost lanes.
• CAFUP(i,t,p): capacity adjustment factor of node i during time step t in
time period. This capacity adjustment factor affects approaching vehicles
within the queue influence area (QIA) upstream of an off-ramp queue.
This factor accounts for the turbulence caused by intense lane changing
within the queue influence area as vehicles adjust their position when
there is a downstream off-ramp queue.
• MFBL(i,t,p): mainline flow rate that can cross the blocked portion of node i
during time step t in time period p.
• MFUB(i,t,p): mainline flow rate that can cross the unblocked portion of
node i during time step t in time period p.
• MIBL(i,t,p): maximum flow desiring to enter the blocked portion of node i
during time step t in time period p.
• MIUB(i,t,p): maximum flow desiring to enter the unblocked portion of
node i during time step t in time period p.
Ramp variables
• RC(i,p): capacity of the ramp proper (pc/h) during time period p in time
period t. Capacity values for the ramp proper are provided in HCM
Exhibit 14-12.
• RF(i,t,p,k): flow (pc/ts) that can enter the ramp proper at segment i during
time period p in time period t and level k.
• RI(i,t,p,k): maximum flow (pc/ts) desiring to enter the off-ramp on
segment i during time period p in time period t and level k, including
queues accumulated from previous time periods.
• RKB(i,t,p,k): ramp proper queue density (pc/mi/ln) for segment i during
time period p in time period t and level k.
• RL(i): length of ramp proper (ft) for segment i.
• RN(i): number of ramp lanes for segment i.
• RNV(i,t,p,k): maximum number of passenger cars within the ramp of
segment i at the end of time step t during time period p and level k. The
number of vehicles is initially based on the calculations of Chapters 12, 13,
and 14, but, as queues grow and dissipate, input–output analysis updates
these values during each time step.
• RSTG(i,t,p,k): maximum number of passenger cars that can enter the ramp
level k of segment i, during time step t and time period p, due to the
presence of a queue in the downstream ramp segment.
• RUV(i,t,p,k): number of unserved vehicles at the entrance of the ramp
proper of segment i at the end of time step t during time period p and
level k. Any values of RUV greater than zero indicate the occurrence of
queue spillback from an off-ramp.
Exhibit 38-A8
Freeway Facilities
Oversaturated Segment
Evaluation Procedure,
Adapted for Off-Ramp Queue
Spillback Evaluation
Exhibit 38-A9
Freeway Facilities
Oversaturated Segment
Evaluation Procedure,
Adapted for Off-Ramp Queue
Spillback Evaluation -
Continued
Exhibit 38-A10
Freeway Facilities
Oversaturated Segment
Evaluation Procedure,
Adapted for Off-Ramp Queue
Spillback Evaluation -
Continued
Exhibit 38-A11
Freeway Facilities
Oversaturated Segment
Evaluation Procedure,
Adapted for Off-Ramp Queue
Spillback Evaluation -
Continued
Equation 38-A12
𝑣 (𝑗)
𝑂𝐹𝑅𝑃𝐶𝑇(𝑗) =
𝑣 (𝑗)
For any segment i, upstream of segment j and affected by the off-ramp
spillback from segment j, the ratio of vehicles traveling towards the off-ramp at
segment i is given by OFRPCT(j), while the ratio of vehicles traveling through in
the unblocked lanes is given by (1- OFRPCT(j)). Therefore, the unblocked queue
density KBUB at any segment i upstream of an off-ramp spillback in a segment j
is given by:
Equation 38-A13 𝐾𝐵𝑈𝐵(𝑖, 𝑗) = 𝐾𝐵 𝐸𝐷(𝑖) × 1 − 𝑂𝐹𝑅𝑃𝐶𝑇(𝑖) , 𝑆𝐶𝐸𝑄(𝑗)
where
KBUB(I, j)= background density at the unblocked lanes in segment i, when queue
spillback occurs at the downstream segment
ED(i) = expected demand at segment i , as defined in HCM Chapter 25
𝑂𝐹𝑅𝑃𝐶𝑇(𝑖)= rate of off-ramp flow and mainline flow at segment i
KB(v, c) = density at a segment with demand flow rate v and capacity c, as
provided by HCM Chapters 12 (basic), 13 (weaving) and 14 (merge
and diverge)
The analyst should select one of these two regimes based on prevailing
driver behavior at the site and in the vicinity of the site.
Exhibit 38-A14
Maximum Off-Ramp Queue
Storage Length at Diverge
Segments with Occurrence of
(a) Regime 3 Queue Spillback
And (b) Regime 4 Queue
Spillback, when no Shoulder is
Available
Exhibit 38-A15
Maximum Off-Ramp Queue
Storage Length at Diverge
Segments with Occurrence of
(A) Regime 3a Queue
Spillback and (B) Regime 4a
Queue Spillback, when
Shoulder is Available
Exhibit 38-A16
Node Structure for Example 1
Exhibit 38-A17
Node Structure for Example 2
Exhibit 38-A18
Node Structure for Example 3
The length of Queue Influence Area is based on time needed for arriving
drivers to react to partial lane blockage and adjust their speeds and positions.
Research [1] has shown that traffic speeds upstream of the back of queue are
negatively affected at a headway distance of 10.95s. Therefore, the influence area
represents the distance traversed by a vehicle during 10.95s with a speed
consistent with the traffic stream.
The length of the QIA is estimated as a function of the segment free-flow
speed (FFS), as shown in Exhibit 38-A21. The exact location of the QIA varies as a
function of the queue length. The QIA values are shorter than the ramp
influence distance of 1,500 ft. However, the two concepts are very different and
are used differently in analyzing ramp operations: the ramp influence area is
used to analyze undersaturated conditions, while the QIA is used to analyze
oversaturated conditions. Since drivers can only detect a downstream queue
visually, they have shorter times to react when compared to the presence of
When Regimes 3 or 4 occur and lane blockage is present in the mainline, the
estimated QIA is added to the queue length to determine the extent of spillback
effects. If an upstream node is located within the combined length of the queue
and QIA, capacity adjustment factors must be applied to account for the
spillback effects.
Exhibit 38-A22
Capacity of Ramp Proper for
Off-Ramps
Exhibit 38-A23
Speed-flow Curves for
Freeway Ramps
where
Nm = number of lanes serving movement m at the intersection
Lm = storage length for movement m at the intersection (ft)
N = number of movements at the approach
Lh = average vehicle spacing in stationary queue (ft/veh) (HCM Equation
31-155)
where
𝐼𝑁𝑉(𝑖, 𝑡, 𝑝, 𝑘)=numberof vehicles at the intersection of segment 𝑖, for level 𝑘 at the
end of time step 𝑡 during time period 𝑝
𝐼𝑁(𝑖, 𝑘) = number of lanes serving the subject approach 𝑘; and
𝑄 = back-of-queue length for the subject approach 𝑘 (veh).
Exhibit 38-A26
Selection of a Cycle Reference
Point to Determine the Initial
Number of Vehicles Within the
Approach
Signalized Intersections
For a signalized intersection approach, the capacity for each movement at
each time step is a function of the signal phase sequence and the capacities of the
individual movements at the intersection. Exhibit 38-A27 illustrates a sample
signalized intersection approach from an off-ramp, with two lane groups: left-
turn (Phase 3) and right turn (Phase 8).
Exhibit 38-A27
Sample Signalized
Intersection Approach from an
Off-Ramp
Input Parameters
The required parameters to evaluate the capacity of a ramp terminal capacity
are generally the same required for standard signalized intersection analyses, as
listed in Exhibit 19-11.
Arrival type: Chapter 19 of the HCM (Exhibit 19-14) provides guidelines for
selecting the appropriate Arrival Types based on the characteristics of arterial
operations, such as quality of progression and coordination. For an off-ramp
approach to the intersection, vehicles arrivals can be considered random.
Therefore, Arrival Type 3 (random arrivals) is recommended to analyze the off-
ramp approach at a signalized ramp terminal.
Phase duration and effective green time: The duration of each phase at the signal
can be fixed (pre-timed control), or variable (semi-actuated or actuated control).
For the former case, phase duration is known. For the latter, an average phase
duration is estimated as described in Section 2 of HCM Chapter 31 – Signalized
Intersections Supplemental. The effective green time g for each phase can then be
computed according to HCM Equation 19-3:
𝑔 =𝐷 −𝑙 −𝑙 Equation 38-A17
where
𝑔 = effective green time (s)
𝐷 = phase duration (s)
𝑙 = start-up lost time = 2.0 (s)
𝑙 = clearance lost time = 𝑌 + 𝑅 – 𝑒 (s)
the difference is included in the first time-step of the next cycle. Then, green
times for each time step from 1 to n are computed. This procedure must be
repeated for every time step within the 15 minutes time period, resulting in a
total of 900/15 = 60 time-steps.
Exhibit 38-A28
Conversion of Green Times to
Time Steps
The capacity ID for each approach and for each time step, is then obtained by
multiplying its respective green time by its capacity, as shown:
Equation 38-A18 𝐼𝐷(𝑖, 𝑡, 𝑝, 𝑘) = 𝑁 𝑠 𝐺𝑇(𝑖, 𝑡, 𝑝, 𝑚) × 𝑓
Where
𝑁 = number of lanes serving movement k
𝑠 = saturation flow rate for movement k (veh/h/ln)
𝐺𝑇(𝑖, 𝑡, 𝑝, 𝑚) = green time for each movement m (s)
The green time parameter GT(i,t,p,m) measures the available green time for a
given intersection movement m, downstream of a freeway segment i, in time step
t and time period p. It can range from 0 (when the movement has red through the
entire time step length) to 15 (movement has green through the entire time step
length).
The heavy vehicle factor fHV needs to be applied to the equation for
intersection discharge to make the units used in intersection capacity (veh/h)
consistent with the flow rates used in uninterrupted flow methods (pc/h).
Exhibit 38-A29, where the node (i+2) represents a diverge segment with an off-
ramp flow 𝑣 . When the queue extends upstream to node i, the approaching flow
𝑣 is segregated into two groups: the exiting vehicles that will join the back of the
queue, and the through vehicles that will use the non-blocked lanes.
Exhibit 38-A29
Illustration of Mainline Flow
Rate Split into Blocked and
Unblocked Lanes
where
OFRF(I, t, p) = flow that can exit the off-ramp 𝑖 during time step 𝑡 in time period 𝑝
RUV(I, t, p, k)= number of unserved vehicles at the off-ramp exit at segment 𝑖, during
time step 𝑡 in time period 𝑝
takes into account the off-ramp flow RF and the number of unserved vehicles on
the approach from the previous time step IUV. II is calculated as:
𝐼𝐼(𝑖, 𝑡, 𝑝, 𝑘) = 𝑅𝐹(𝑖, 𝑡, 𝑝, 𝑘) + 𝐼𝑈𝑉(𝑖, 𝑡, 𝑝, 𝑘) Equation 38-A25
If the number of vehicles trying to enter the intersection exceeds the amount
of vehicles allowed to enter the intersection, then the number of total unserved
vehicles must be computed and considered in the intersection input II during the
next time period:
𝐼𝑈𝑉(𝑖, 𝑡, 𝑝, 𝑘) = 𝐼𝑈𝑉(𝑖, 𝑡 − 1, 𝑝, 𝑘) + 𝐼𝐼(𝑖, 𝑡, 𝑝, 𝑘) − 𝑅𝑂(𝑖, 𝑡, 𝑝, 𝑘) Equation 38-A28
length of the queue in the upstream segment. The length of the queue within the
subject segment will then be used to evaluate whether the capacity of any
upstream node is affected by the queue.
For upstream segments that may be affected by spillback, the queue length
within the segment (measured from its downstream end) must be computed and
stored in the parameter SBLQ. This check is performed for every node upstream
of a congested off-ramp (Exhibit 38-A30).
Exhibit 38-A30
Procedure for Evaluating the
Impact of Queue Spillback on
Upstream Nodes and
Determination of the Queue
Length within Upstream
Segments
Exhibit 38-A31
Illustration of Different
Impacts of an off-Ramp
Queue at Node i: (a) Lane
Blockage, (b) Increased
Turbulence and (c) No Effect
The parameter LCR estimates the rate of lane change maneuvers performed
by vehicles within the Queue Influence Area trying to adjust their position when
spillback occurs. Vehicles traveling towards the exit ramp will move to the
shoulder lane attempting to join the back of the queue, while vehicles traveling
through will move to the median lanes in order to avoid the queue. Therefore,
the lane change rate LCR is computed as:
𝑆𝐵𝐿𝐶(𝑖, 𝑡, 𝑝) Equation 38-A37
𝐿𝐶𝑅(𝑖, 𝑡, 𝑝) =
𝑆𝐹(𝑖, 𝑡, 𝑝)
𝐿𝐹𝑅
𝑝 , = Equation 38-A38
𝐿𝐹𝑅 + 𝐿𝐹𝑅
𝐿𝐹𝑅
𝑝 , =
𝐿𝐹𝑅 + 𝐿𝐹𝑅
(b) According to the guidance provided in HCM Chapter 14, the
influence of ramps rarely extends beyond 8,000 ft. Therefore, for any
nodes located beyond 8,000 from the off-ramp, the distribution of pi
is taken as equal among all N freeway lanes:
1 Equation 38-A39
𝑝 =
𝑁
(c) At intermediate distances from the off-ramp ranging between
1,500 ft and 8,000 ft, the distribution values of pi can be obtained
through linear interpolation between the cases previously described.
Exhibit 38-A32
Distribution of pi as Function
of Distance from the Off-
Ramp Exit, for a 3-Lane
Segment
The value of pi as function of the distance from off-ramp exit can then be
obtained through the following equation:
1
Equation 38-A40 − 𝑝 , 𝑅 × (𝑑 − 1,500)
𝑝 =𝑝, + 𝑁
6500
As the lane-by-lane distribution of the off-ramp flow is known, the number
of lane change maneuvers, SBLC, can then be estimated. For Regime 3 cases (one
blocked lane), the number of lane changes is obtained as follows:
The equation adds the number of through vehicles in lane 1 that move to
lane 2 to avoid the queue and the number of exiting vehicles in the remaining
lanes that adjust their position to join the back of the queue, multiplied by the
necessary number of lane changes. Exhibit 38-A33 illustrates an example of the
proposed equation applied to a 4-lane segment.
Exhibit 38-A33
Illustration of Lane Change
Maneuvers Within the Queue
Influence Area in a 4-Lane
Segment With Regime 3
Exhibit 38-A34
Illustration of Lane Change
Maneuvers Within the Queue
Influence Area in a 4-Lane
Segment With Regime 4
𝑂𝑁𝑅𝑂(𝑖, 𝑡, 𝑝)
Equation 38-A45
𝑅𝑀(𝑖, 𝑡, 𝑝)
⎧
𝑂𝑁𝑅𝐶(𝑖, 𝑡, 𝑝)
⎪
⎪ 𝑀𝐹(𝑖 + 1, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪ ⎧𝑚𝑖𝑛 − 𝑀𝐼(𝑖, 𝑡, 𝑝)
𝑀𝑂3(𝑖, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪ ⎪
⎪ 𝑆𝐶(𝑖, 𝑡, 𝑝)
= 𝑚𝑖𝑛( ⎪
⎨𝑚𝑎𝑥( 𝑚𝑖𝑛 𝑀𝐹(𝑖 + 1, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪ ⎨ 𝑀𝑂3(𝑖, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪ ⎪ 2𝑁(𝑖, 𝑝)
⎪ ⎪ 𝑹𝑭(𝑶𝑭𝑹𝑵𝑬𝑿𝑻(𝒊), 𝒕, 𝒑))
⎪ ⎪
⎩ ⎩ 𝟐 × 𝑵𝑸 𝑶𝑭𝑹𝑵𝑬𝑿𝑻(𝒊)
Exhibit 38-A35
Impact of a queue spillback
on the discharge capacity of
an upstream on-ramp
Exhibit 38-A36
Illustration of Different
Density Values Within One
Diverge Segment
If there are no spillback effects, the segment operates with a uniform density.
In this case, the constraints for the unblocked and blocked portions (MO2UB and
MO2BL, respectively) are calculated proportionately to the number of unblocked
and blocked lanes:
The queue density for the blocked portion is computed as equal to the ramp
queue density:
𝐾𝑄𝐵𝐿(𝑖, 𝑡, 𝑝) = 𝑅𝐾𝑄(𝑂𝐹𝑅𝑁𝐸𝑋𝑇(𝑖), 𝑡 − 1, 𝑝) Equation 38-A51
With the queue density values for both the blocked and unblocked portions
known, the MO2 components MO2BL and MO2UB can be computed:
𝑀𝑂2𝑈𝐵(𝑖, 𝑡, 𝑝) = 𝑆𝐹𝑈𝐵(𝑖, 𝑡 − 1, 𝑝) − 𝑂𝑁𝑅𝐹(𝑖, 𝑡, 𝑝) Equation 38-A52
component MFUB represents flow across the node in the unblocked lanes, while
the component MFBL represents the flow across the node in the blocked lanes.
For both components, the resulting flow is computed as the minimum value
between input and the maximum allowed flow.
For MFUB, the maximum allowed flow is equal to the capacity of unblocked
lanes in the downstream segment, represented by the parameter SCEQ as
computed in the initialization step:
Equation 38-A54 𝑀𝐹𝑈𝐵(𝑖) = 𝑚𝑖𝑛 𝑀𝐼𝑈𝐵(𝑖, 𝑡, 𝑝), 𝑆𝐶𝐸𝑄(𝑖, 𝑡, 𝑝), 𝑀𝑂2𝑈𝐵(𝑖, 𝑡, 𝑝)
For MFBL, the maximum allowed flow is equal to the flow allowed to enter
the nearest downstream off-ramp RF, as presented in the following equation:
Equation 38-A55 𝑀𝐹𝐵𝐿(𝑖) = 𝑚𝑖𝑛(𝑀𝐼𝐵𝐿(𝑖, 𝑡, 𝑝), 𝑅𝐹(𝑁𝐸𝑋𝑇𝑂𝐹𝑅(𝑖), 𝑡, 𝑝, 𝑀𝑂2𝐵𝐿(𝑖, 𝑡, 𝑝))
Next, the Mainline Flow MF through node i is computed as the sum of the
blocked and unblocked portions, as follows:
Equation 38-A56 𝑀𝐹(𝑖, 𝑡, 𝑝) = 𝑀𝐹𝑈𝐵(𝑖, 𝑡, 𝑝) + 𝑀𝐹𝐵𝐿(𝑖, 𝑡, 𝑝)
𝑇
Equation 38-A58 𝑆𝐹(𝑖, 𝑝) = 𝑆𝐹(𝑖, 𝑡, 𝑝)
𝑆
𝑇
Equation 38-A59 𝑂𝐹𝑅𝐹(𝑖, 𝑝) = 𝑂𝐹𝑅𝐹(𝑖, 𝑡, 𝑝)
𝑆
1
Equation 38-A60 ∆𝐾(𝑖, 𝑝) = ∆𝑁𝑉(𝑖, 𝑡, 𝑝)
𝑆×𝑁
Similar to the mainline, the flow in the ramp roadway is also aggregated:
𝑇
𝑅𝐹(𝑖, 𝑝, 𝑘) = 𝑅𝐹(𝑖, 𝑡, 𝑝, 𝑘) Equation 38-A61
𝑆
1
𝑅𝐾(𝑖, 𝑝, 𝑘) = 𝑅𝑁𝑉(𝑖, 𝑡, 𝑝, 𝑘) Equation 38-A62
𝑆
Finally, the speed at the ramp for a time period p is obtained by dividing the
total ramp flow in the time period by its average density:
𝑅𝐹(𝑖, 𝑝, 𝑘) Equation 38-A63
𝑆𝑅(𝑖, 𝑝, 𝑘) =
𝑅𝐾(𝑖, 𝑝, 𝑘)
Queue spillback into an urban street intersection may occur when the
freeway merge segment has insufficient capacity to process the ramp’s demand.
Spillback may also occur in cases of ramp metering. This appendix presents the
methodology for determining whether spillback will occur from an on-ramp into
the upstream intersection.
The methodology considers signalized intersections, two-way stop-
controlled intersections, all-way stop controlled intersections, and roundabouts.
The procedure first estimates the demand approaching the on-ramp (determined
based on the upstream intersection’s configuration), and then estimates the
capacity of the off-ramp. The Chapter 10, Freeway Facilities methodology for
oversaturated conditions can estimate the resulting queue length, however, the
user must input the on-ramp demand flow rate.
The methodology framework for conducting this spillback check is presented
Exhibit 38-B1 .
Exhibit 38-B1
Procedure for Detecting
Spillback Occurrence at an
On-Ramp
DEMAND ESTIMATION
The first step in the methodology calculates the entering demand flow rate at
the on-ramp (𝑣 ), as a function of the upstream intersection configuration and
operations. Under low demand conditions, the on-ramp demand flow rate is
calculated as the sum of the demands on each of the intersection approaches that
discharge into the ramp. However, if any of these movements is operating over
capacity, the total throughput to the ramp will be constrained by the capacity of
these oversaturated movements. Hence, this check ensures that the on-ramp
demand is not overestimated. The analysis approach for each of four intersection
types is presented next.
𝑣 = min(𝑣 , 𝑐 )
Equation 38-B1
where
𝑣 = on-ramp demand (veh/h);
𝑣 = demand for movement 𝑖 at the intersection (veh/h);
𝑐 = demand for movement 𝑖 at the intersection (veh/h);
𝑁 = number of intersection movements that discharge into the on-ramp
If all movements operate below capacity, the on-ramp demand is obtained as
the sum of the movement demands. If any of the ramp terminal movements that
discharge into the on-ramp operates over capacity, the total throughput to the
on-ramp will be lower than the sum of the corresponding intersection
movements.
Unsignalized Movements
In the case of unsignalized movements discharging into the on-ramp, the
demand for these movements must also be compared to their capacity. The
potential capacity cp,i of an unsignalized movement can be computed by
aggregating its saturation flow rates at different phases throughout a cycle.
If the unsignalized movement is free-flowing and there are no other
conflicting movements discharging to the on-ramp, its saturation flow rate sFF is
obtained by HCM Equation 19-8, with the applicable adjustment factors applied:
𝑠 =𝑠 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓 𝑓
Equation 38-B2
Where
𝑠 = saturation flow rate for unsignalized movement during free-flow
(veh/h/ln);
𝑠 = base saturation flow rate (pc/h/ln)
All other adjustment factors are as described in Equation 19-8.
If the unsignalized movement must yield to a conflicting movement
discharging to the on-ramp, the permitted saturation flow rate sp is calculated
based on HCM equation 31-100:
𝜆 𝑒 ,
𝑠 =
1−𝑒 ,
Where:
𝑠 = permitted saturation flow rate for unsignalized movement (veh/h/ln);
𝜆 = throughput of the conflicting movement (veh/h/ln)
𝑡 = critical headway = 4.5 (s);
𝑡 = follow-up headway = 2.5 (s)
according to their ranks, using the default numbering of Chapter 20 (Exhibit 20-
1).
Exhibit 38-B2
Schematic of Movements
Turning to an On-Ramp from
a TWSC Intersection
𝜆 = 𝑚𝑖𝑛(𝑣 , 𝑠 )
Equation 38-B3
where
𝜆 = departure rate from major street right turn into the on-ramp (veh/h);
𝑣 = demand flow rate for the major street right turn; and
𝑠 = saturation flow rate for a right-turn movement (veh/h).
2. Rank 2 Movement (Left Turn from the Major Street): The maximum
throughput for this movement is limited by its potential capacity (𝑐 , ), as
defined in Equation 20-36. Therefore, the maximum throughput (veh/h)
for this left-turn movement is given by:
𝜆 = 𝑚𝑖𝑛(𝑣 , 𝑐 , ) Equation 38-B4
where
𝜆 = departure rate from the major-street left-turn into the on-ramp (veh/h);
𝑣 = demand flow rate for the major street left turn; and
𝑐 , = potential capacity for the major street left turn (veh/h).
3. Rank 3 Movement (Through Movement from the Minor Street): Similar to rank
2 movements, the maximum throughput for this movement is limited by
its potential capacity (𝑐 , ), as defined in Equation 20-47. Therefore, the
maximum throughput λTh (veh/h) for this through movement is given by:
Equation 38-B5 𝜆 = 𝑚𝑖𝑛(𝑣 , 𝑐 , )
where
𝜆 = departure rate from the minor street through into the on-ramp (veh/h);
𝑣 = demand flow rate for the minor street through; and
𝑐 , = potential capacity for the minor street through (veh/h).
Finally, the total on-ramp demand flow rate 𝑣 can be estimated as follows:
𝑣 =𝜆 + 𝜆 + 𝜆
Equation 38-B6
where
𝜆 = departure rate from major street right turn into the on-ramp (veh/h)
(Equation 38-B3);
𝜆 = departure rate from the major-street left-turn into the on-ramp (veh/h)
(Equation 38-B4); and
𝜆 = departure rate from the minor street through into the on-ramp (veh/h)
(Equation 38-B5).
Exhibit 38-B3
Schematic of Movements
Turning to an On-Ramp from
an AWSC Intersection
The on-ramp demand flow rate can be obtained directly from the departure
headways of the three movements combined:
3600 3600 3600
𝑣 = + + Equation 38-B7
ℎ , ℎ , ℎ ,
where
𝑣 = on-ramp flow rate (veh/h);
ℎ , = departure headway for the major street right turn(s);
ℎ , = departure headway for the major street left turn(s); and
ℎ , = departure headway for the minor street through(s).
Case D: Roundabouts
The Roundabouts methodology is based on the calculation of the potential
capacities of each approach, based on three main variables: the critical and the
follow-up headways, and the circulating flow (Equation 22-21 through Equation
22-23). Both critical and follow-up headway values can be obtained from Chapter
33. The methodology considers each approach independently. To analyze
roundabouts within a system it is first necessary to estimate the on-ramp
throughput from a roundabout.
The procedure first identifies the movements that discharge to the on-ramp
and their respective ranks (priority orders). Exhibit 38-B4 illustrates a typical
roundabout, where movements discharging into the on-ramp are numbered
according to their ranks. In contrast to other types of intersections, the approach
furthest from the on-ramp has priority as it enters the circulating stream without
any significant conflicting traffic (other than occasional U-turns). The operation
of each of these movements is as follows:
Exhibit 38-B4
Schematic of Movements
Turning to an On-Ramp from
a Roundabout
Rank 1 Movement (Left-Turn from the Third Upstream Approach from the On-
Ramp):
This movement has priority over the other movements because it enters the
circulating stream first. Also, because the on-ramp does not have an approach
into the roundabout, this movement is most often unopposed by the circulating
stream (except for occasional U-turns in the intersection). Therefore, the
maximum throughput 𝜆 (veh/h) for this left-turn movement is given by:
Equation 38-B8 𝜆 = 𝑚𝑖𝑛(𝑣 ,𝑐 )
where
𝜆 = departure rate from the third upstream approach into the on-ramp
(veh/h);
𝑣 = demand flow rate for the third upstream approach into the on-ramp;
and
𝑐 = potential capacity for the approach (veh/h).
Rank 2 Movement (Through from the Second Upstream Approach, Most Likely an
Off-Ramp):
The maximum throughput for this movement is limited by the upstream
approach departure rate and its own potential lane capacity (𝑐 ), as defined in
Equations 22-21 through 22-23. Therefore, the maximum throughput 𝜆 veh/h)
for this through movement is given by:
Equation 38-B9 𝜆 = 𝑚𝑖𝑛(𝑣 ,𝑐 )
where
𝜆 = departure rate from the second upstream approach into the on-ramp
(veh/h);
𝑣 = demand flow rate for the second upstream approach into the on-ramp
(veh/h); and
𝑐 = potential capacity for the approach (veh/h).
𝑣 =𝜆 + 𝜆 + 𝜆 Equation 38-B11
The total on-ramp demand flow rate can be calculated by the same method
for roundabouts with a higher number of approaches.
CAPACITY ESTIMATION
Capacity at the on-ramp must be estimated in order to predict the occurrence
of queue spillback. Three cases may occur:
Signalized Intersections
Exhibit 38-B5 presents the core methodology for evaluating the performance
of signalized intersections, with proposed modifications to address impacts from
an on-ramp queue spillback. New steps and modified steps to the methodology
are described in the following paragraphs.
Exhibit 38-B5
Signalized Intersections
Methodology With
Adjustments to Address On-
Ramp Queue Spillback
Exhibit 38-B6
Typical Signalized Intersection
Ramp Terminal in a Diamond
Interchange
The total throughput from the intersection into the on-ramp λONR is the sum
of the throughput from each of the contributing movements:
Equation 38-B12 𝜆 =𝜆 + 𝜆 + 𝜆
The throughput for each movement i is the minimum value of its demand
and capacity:
Equation 38-B13 𝜆 = 𝑚𝑖𝑛(𝑣 , 𝑐 )
where
vi = demand flow rate for intersection movement i (veh/h)
ci = capacity for intersection movement i (veh/h), as provided by Equation
19-16
Unsignalized movements, which are common for right-turn movements to
the on-ramp, are unrestricted. The capacity of these movements can be estimated
as the saturation flow rate (Equation 19-8), with an adjustment factor for right
turns fRT (Equation 19-13).
If all movements at the intersection are undersaturated, (vi ≤ ci for every i),
then Equation 38-B12 is simplified and the total on-ramp demand throughput
λONR is as follows:
Equation 38-B14
𝜆 = 𝑣𝑖
𝑖
𝑇 Equation 38-B15
𝑐 = 𝑂𝑁𝑅𝑂(𝑖, 𝑡, 𝑝)
𝑆
where
ONRO(I, t, p) = maximum output flow rate that can enter the merge point from on-
ramp 𝑖 during time step t in time interval p
T = number of time steps in 1 h (integer). T is set as a constant of 240 in the
computational engine, or equal to four times the value of S;
S = number of computational time steps in an analysis period (integer).
𝑆 is set as a constant of 60 in the computational engine, corresponding
to a 15-s interval and allowing a minimum segment length of 300 ft;
and
t = time step index.
Exhibit 38-B7
Step 7B - Estimation of
Merging Capacity in a
Freeway Ramp
maintained and it is equal to LONR until the end of the green for the SBL
movement. At the end of the SBL green, the vertical difference between the
projected number of vehicles (dashed line) and the actual number of vehicles
inside the on-ramp represent the number of unserved vehicles for the SBL
approach. This additional queue can be considered in a multiperiod analysis for
the signalized intersection or interchange, using the methods provided in
Chapter 23 – Ramp Terminals and Alternative Intersections.
Exhibit 38-B8
Sample Intersection for
Calculation of a QAP for the
On-Ramp
The slope of the red line connecting the number of vehicles in the end and
start of the green represent the reduced capacity of the SBL movement due to
queue spillback. For the remainder of the cycle, the NBR movement discharges
at a constant rate into the on-ramp, as this is an unsignalized movement. Given
that the discharge capacity cmerge is greater than the on-ramp demand λNBR, the
vehicles along the on-ramp are discharged to the freeway until the on-ramp is
cleared. Therefore, the NBR movement does not have its capacity affected by
queue spillback.
This procedure can be applied for both pretimed and actuated control types,
since the core methodology can address both controller types. If the signal is
actuated, the average phase duration lengths are applied, as obtained in Step 6.
movement cSBL,SP can be obtained from the QAP as the slope of the red line (cSBL,SP
- cmerge) as follows:
𝑁(𝑔 ) − 𝑁(0)
Equation 38-B16 𝑐 −𝑐 −
,
𝑔
where
N(g ) = number of queued vehicles along the on-ramp at t = g1 (end of green
for phase 1);
N(0) = number of queued vehicles along the on-ramp at t = 0 (start of the
cycle);
G = effective green time for phase 1
Exhibit 38-B9
On-Ramp Queue
Accumulation Polygon During
Queue Spillback
The adjusted capacity of the SBL movement cSBL,SP is then computed as:
𝑁(𝑔 ) − 𝑁(0)
Equation 38-B17 𝑐 , =𝑐 +
𝑔
If the queue develops and fully discharges during every cycle, then
subsequent cycles will have the same discharge. However, if there are residual
queues at the on-ramp by the end of the cycle, the QAP must then be plotted
again for the following cycle with an initial queue equal to the number of queued
vehicles in the end of the present cycle. This process must be then repeated for a
number of cycles N= 900/C, sufficient to analyze the entire 15-minute period.
The adjusted capacity for each movement is estimated as the average of the
discharge rates during each cycle.
𝑣
𝑋 = Equation 38-B18
𝑐
Exhibit 38-B11
TWSC intersections Core
Methodology With
Adjustments to Address On-
Ramp Queue Spillback
i discharging into the on-ramp, the throughput is the minimum value of its
demand and its movement capacity:
𝜆 = 𝑚𝑖𝑛 𝑣 , 𝑐 , Equation 38-B19
where
vi = demand flow rate for movement i
cm,j = movement capacity for movement i (Equations 20-36, 20-37 and 20-40).
Step 9B. Obtain merging capacity using the freeway facilities methodology
This step computes the merging capacity into the freeway cmerge. The
procedure described in Step 7B of the queue spillback analysis for signalized
intersections (Exhibit 38-B5) is applied.
Exhibit 38-B12
On-ramp Queue Accumulation
Polygon – TWSC Intersection
From this relationship shown in Exhibit 38-B12 the spillback time TSB is
defined as the amount of time within a time period when spillback is active:
𝐿 − 𝑁(0)
Equation 38-B20 𝑇 =𝑇−
𝜆 −𝑐
where
TSB = time period with active spillback (minutes)
T = duration of analysis time period (minutes)
LONR = available queue storage at on-ramp (veh)
N(0) = number of queued vehicles along the on-ramp at t = 0 (start of the
cycle);
cmerge = merging capacity of the on-ramp (veh/h)
λ from the intersection into the on-ramp (veh/hr)
Estimating the spillback time TSB is critical to the methodology, as the
aggregated calculations of capacity for each movement depend on the amount of
time that the intersection operates under queue spillback.
where
c , = adjusted capacity for movement i (veh/h)
⎡ 3600 𝜆 ⎤
× Equation 38-B24
3600 ⎢ 𝑣 𝑣 𝑐 , 𝑐 , ⎥
𝑑 = + 900𝑇 ⎢ −1+ −1 + ⎥+5
𝑐 , 𝑐 , 𝑐 , 450𝑇
⎢ ⎥
⎣ ⎦
Exhibit 38-B13
AWSC Intersections Core
Methodology With
Adjustments to Address On-
Ramp Queue Spillback
The only step in the methodology that differs from the TWSC (13D) is
described below.
Exhibit 38-B14
Roundabouts Methodology
With Adjustments to Address
On-Ramp Queue Spillback
Exhibit 38-B16
Priority Order for a
Roundabout Upstream of an
On-Ramp
Next, the methodology calculates the capacity of the roundabout’s exit lane
into the on-ramp. Previous research ( [4] [5]) suggests that the capacity of an exit
lane, accounting for pedestrian and bicycle traffic in a typical urban area, is in the
range of 1,200 to 1,300 vehicles per hour. Starting from the approach with Rank
1, and proceeding counterclockwise with the rest of the approaches, the capacity
for each approach is used to determine the maximum throughput for every
movement discharging to the on-ramp.
Rank 1 – SB approach. The Rank 1 approach has priority over the other
movements connecting to the on-ramp because it enters the circulating stream
first. Also, because the on-ramp does not have an approach into the roundabout,
the Rank 1 movement is most often unopposed by the circulating stream (except
for occasional U-turns along the arterial). Therefore, the maximum throughput
λSB-ONR (veh/h) for this left-turn movement is limited by its own lane capacity (cSB)
and the maximum throughput to the on-ramp, and it is given by:
3,600
Equation 38-B26 𝜆 = 𝑚𝑖𝑛 𝑣 ,𝑐 ×𝑝 ,
ℎ
where
λSB- rate from the SB approach into the on-ramp (veh/h)
vSB- flow rate for the SB approach into the on-ramp (veh/h)
cSB = lane capacity for SB approach (veh/h) (HCM Equation 22-21)
pSB- of demand from SB approach into the on-ramp
The total on-ramp demand flow rate can be similarly calculated if there are
additional approaches to the roundabout.
𝜆
𝑄 , =𝑄 × +𝑄 , Equation 38-B34
𝜆
𝜆
𝑄 , =𝑄 × +𝑄 , Equation 38-B35
𝜆
𝜆
𝑄 , =𝑄 × +𝑄 , Equation 38-B36
𝜆
Where
𝑄 , = queue due to the on-ramp spillback on 𝑖 approach (veh)
λ, = maximum throughput for 𝑖 approach into the on-ramp (veh)
𝑄 , = 95th percentile queue on 𝑖 approach (veh)
⎡ 3600 𝜆 ⎤
×
3600 ⎢𝜆 𝜆 𝑐 𝑐 ⎥ Equation 38-B37
𝑑 = + 900𝑇 ⎢ −1+ −1 + ⎥+5
𝑐 𝑐 𝑐 450𝑇
⎢ ⎥
⎣ ⎦
𝜆
× 𝑚𝑖𝑛 ,1
𝑐
where
c =merging capacity of the on-ramp (veh/h);
λ = exit flow rate into the on-ramp (veh/h); and
t = time period (h) (T = 0.25 h for a 15-min analysis).
where
𝑎 = multiplicative calibration parameter (Equation 38-C3, Equation 38-C5,
and Equation 38-C7);
𝑏 = additive calibration parameter (Equation 38-C4, Equation 38-C6, and
Equation 38-C8);
𝐿𝐹𝑅 = share of the total flow on lane 𝑖, where 𝑖 ranges from 1 to n-1 (n = total
number of segment lanes);
𝐿𝐹𝑅 = share of the total flow on the leftmost lane (lane n); and
𝑣/𝑐 = volume/capacity ratio 0 < ≤ 1 .
The model in Equation 38-C1 and Equation 38-C2 can be applied for basic,
merge, diverge and weaving segments. For merge and diverge segments, the
share of flow is estimated at the area upstream of the ramp. For weaving
segments, the share of flow is estimated at the mainline upstream the on-ramp.
Volume and capacity are given in veh/h. The calibration parameters 𝑎 and 𝑏
applicable in the analysis of basic segments are as follows:
For merge and diverge segments, the 𝑎 and 𝑏 parameters are as follows, with
additional coefficients 𝑎 and 𝑏 to address ramp demand:
𝑣
Equation 38-C5 𝑎 =𝑎 +𝐺×𝑎 +𝑡×𝑎 +𝑛×𝑎 + ×𝑎
1,000
𝑣
Equation 38-C6 𝑏 =𝑏 +𝐺×𝑏 +𝑡×𝑏 +𝑛×𝑏 + ×𝑏
1,000
where
𝑎 = multiplicative calibration parameter;
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 214 Version 1.0
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 215
The calibrated values for weaving segments are presented in Exhibit 38-C2.
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 216 Version 1.0
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 217
Exhibit 38-C3
LFR Distribution for a Sample
2-Lane Segment
(Minneapolis/MN)
Next, Exhibit 38-C4 illustrates the LFR distribution for a 3-lane freeway
segment. At low demand most of the flow of 3-lane segments is concentrated in
the center lane (lane 2), followed by lanes 1 and lane 3. As demand increases, lane
flow distribution increases in lane 3, while decreasing in lanes 1 and 2.
Exhibit 38-C4
LFR Distribution for a Sample
3-Lane Segment (Tampa/FL)
Exhibit 38-C5 shows the LFR distribution for 4-lane segments. At free-flow
conditions, lanes 2 and 3 carry the majority of flow. Lane 4 is typically underused
during undersaturated conditions, but for higher demands it carries the majority
of flow.
Exhibit 38-C5
LFR Distribution for a Sample
4-Lane Segment (Tampa/FL)
The flow distribution patterns shown in the previously exhibits for basic
segments are also observed in merge, diverge and weaving segments. Additional
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 218 Version 1.0
factors such as ramp volume, grade, truck percentage influence the boundary
values and slopes of the curves, but does not change the typical LFR distribution
as function of v/c.
Exhibit 38-C6
Check for Negative Lane
Flows
The second check compares the estimated flow by lane with the respective
lane capacities to ensure no lane operates with a demand-to-capacity ratio
greater than 1. The procedure is illustrated in Exhibit 38-C7. If any lane is
estimated to operate above its capacity, the flow in this given lane is constrained
by the capacity value and the exceeding demand is moved to the adjacent lane.
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 219
Exhibit 38-C7
Check for Lane Capacity
Lane FFS
Field observations have shown that speeds differ among lanes, and they are
typically lower in shoulder lanes and higher in median lanes.
Models were developed to estimate individual lane FFS by applying a
multiplying factor xFFS to the segment FFS. Exhibit 38-C8 summarizes the
recommended multipliers which are provided as a function of the segment type
and the number of lanes in the segment. As shown, when the number of lanes
increases, the range of FFS multipliers increase as well (i.e. there are lower
speeds in the shoulder lanes and higher speeds on the median lanes). For 2-lane
segments, merge and diverge segments have a higher difference in FFS between
the two lanes when compared to basic segments. For 3-lane segments, basic
segments show the highest FFS range, while merge segments have more uniform
lane FFS. As for 4-lane segments, merge segments show the highest FFS range,
followed by basic and merge segments yield similar results.
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 220 Version 1.0
where
𝐹𝐹𝑆 = free-flow speed for lane i (mi/h);
𝐹𝐹𝑆 = adjusted free-flow speed for the segment average (mi/h) (Equation 12-
5).
𝑥 = FFS multiplier (Exhibit 38-C8);
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 221
The segment capacities measured from field data may not be equal to the
estimated capacities using HCM methodologies. According to the HCM
Equation 12-6, the base capacity can be estimated as:
Equation 38-C11 c = min[2200 + 10 × (FFS − 50), 2400]
The adjusted capacity of a segment is obtained through Equation 12-8:
Equation 38-C12 c = 𝑐 × 𝐶𝐴𝐹
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 222 Version 1.0
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 223
APPLICATION EXAMPLES
The flow ratio for lane 1 (right lane) is obtained by the following equation:
𝑣
𝐿𝐹𝑅 = 𝑎 × 𝑙𝑛 +𝑏
𝑐
The calibration parameters 𝑎 and 𝑏 for lane 1 are obtained as follows:
𝑣
𝑎 =𝑎 +𝐺×𝑎 +𝑡×𝑎 +𝑛×𝑎 + ×𝑎
1000
960
𝑎 = −0.075 + 1 × 0.0077 + 2 × 0.0008 + 0 × 0.014 + × (−0.067)
1000
𝑎 = −0.116
𝑣
𝑏 =𝑏 +𝐺×𝑏 +𝑡×𝑏 +𝑛×𝑏 + ×𝑏
1000
960
𝑏 = 0.26667 + 1 × (−0.00810) + 2 × 0.00140 + 1 × 0.03129 +
1000
× 0.01324
𝑏 = 0.296
The flow rate on lane 1 can then be obtained by:
5003
𝐿𝐹𝑅1 = −0.116 × 𝑙𝑛 + 0.296 = 𝟎. 𝟑𝟓𝟎
7200
The same procedure is applied to obtain the flow rate on lane 2, using the
respective coefficients from Exhibit 38-C1:
𝑣
𝑎 =𝑎 +𝐺×𝑎 +𝑡×𝑎 +𝑛×𝑎 + ×𝑎
1000
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 224 Version 1.0
960
𝑎 = 0.0096 + 1 × (−0.00960) + 2 × (−0.00054) + 1 × (−0.0096) +
1,000
× (−0.048)
𝑎 = −0.0568
𝑣
𝑏 =𝑏 +𝐺×𝑏 +𝑡×𝑏 +𝑛×𝑏 + ×𝑏
1000
960
𝑏 = 0.34 + 1 × (−0.0019) + 2 × (0.00089) + 1 × 0.0052 + × (−0.073)
1,000
𝑏 = 0.275
5003
𝐿𝐹𝑅 = −0.0568 × 𝑙𝑛 + 0.275 = 𝟎. 𝟐𝟗𝟔
7200
Finally, the flow rate on the leftmost lane (lane 3) can be obtained as:
𝐿𝐹𝑅 = 1 − 𝐿𝐹𝑅 − 𝐿𝐹𝑅 = 1 − 0.350 − 0.296
𝐿𝐹𝑅 = 𝟎. 𝟑𝟓𝟒
Speed calculations
Individual free-flow speeds for each lane can be obtained by multiplying the
segment FFS (75.4 mi/h) by the corresponding multipliers (Exhibit 38-C6) as
follows:
𝐹𝐹𝑆 = 𝐹𝐹𝑆 × 0.943 = 75.4 × 0.943 = 71.1 mph
𝐹𝐹𝑆 = 𝐹𝐹𝑆 × 1.024 = 75.4 × 1.024 = 77.2 mph
𝐹𝐹𝑆 = 𝐹𝐹𝑆 × 1.064 = 75.4 × 1.068 = 80.5 mph
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 225
𝑐
𝐹𝐹𝑆 − (𝑣 − 𝐵𝑃 )
𝑆 = 𝐹𝐹𝑆 − 45
(𝑐 − 𝐵𝑃 )
1872
71.1 − (1753.2 − 1156)
𝑆 = 71.1 − 45 = 50.6 mi/h
(1872 − 1156)
2448
77.2 − (1482.6 − 911.6)
𝑆 = 77.2 − 45 = 74.1 mi/h
(2448 − 911.6)
2880
80.5 − (1767.3 − 778.9)
𝑆 = 80.5 − 45 = 76.8 mi/h
(2880 − 778.9)
The obtained speed-flow curves for each lane are presented and compared to
the segment-wise curve in Exhibit 38-C10:
Exhibit 38-C10
Comparison of Speed-Flow
Curves for Each Lane and for
the Segment
Exhibit 38-C11
Example of LFR Calculation for
a Weaving Segment
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 226 Version 1.0
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 227
The estimated capacity is 1,788.8 veh/h/ln. The flow ratio for lane 1 (right
lane) is obtained by the following equation:
𝑣
𝐿𝐹𝑅 = 𝑎 × 𝑙𝑛 +𝑏
𝑐
The calibration parameters 𝑎 and 𝑏 for lane 1 are obtained as follows:
𝑣 , 𝑣 , 𝐿
𝑎 = 𝑎 + 𝐺 × 𝑎 + 𝑡 × 𝑎 + 𝐼𝐷 × 𝑎 + ×𝑎 + ×𝑎 + ×𝑎
1000 1000 1000
+ 𝑉𝑅 × 𝑎
𝑎 = −0.13 + (−0.5) × 0.13 + (3.3) × (−0.012) + 0.67 × (−0.0025) + ×
,
, ,
0.072 + × (−0.13) + × (0.056) + 0.325 × (−0.11)
, ,
𝑎 = −0.1843
𝑣 , 𝑣 , 𝐿
𝑏 = 𝑏 + 𝐺 × 𝑏 + 𝑡 × 𝑏 + 𝐼𝐷 × 𝑏 + ×𝑏 + ×𝑏 + ×𝑏
1000 1000 1000
+ 𝑉𝑅 × 𝑏
𝑏 = 0.24 + (−0.5) × (−0.03) + (3.3) × (−0.0043) + 0.67 × (−0.0067) +
, ,
× 0.065 + × 0.063 + × (−0.03) + 0.32 × (−0.14)
, , ,
𝑏 = 0.1790
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 228 Version 1.0
𝑎 = 0.0176
𝑣 , 𝑣 , 𝐿
𝑏 = 𝑏 + 𝐺 × 𝑏 + 𝑡 × 𝑏 + 𝐼𝐷 × 𝑏 + ×𝑏 + ×𝑏 + ×𝑏
1000 1000 1000
+ 𝑉𝑅 × 𝑏
𝑏 = 0.26 + (−0.5) × (0.045) + (3.3) × (−0.011) + 0.67 × (−0.005) + ×
𝑏 = 0.2271
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 229
• Grade: 3% (rolling).
By applying the multiplying factors obtained in Exhibit 38-C8 to the segment
FFS, individual FFS can be obtained as follows:
𝐹𝐹𝑆1 = 𝐹𝐹𝑆 × 0.965 = 69.1 × 0.965 = 66.68 𝑚𝑝ℎ
𝐹𝐹𝑆2 = 𝐹𝐹𝑆 × 1.032 = 69.1 × 1.032 = 71.31 𝑚𝑝ℎ
Next, lane capacities are obtained by applying the multiplying the factors
obtained in Exhibit 38-C9 to the capacity as follows:
𝑐1 = 𝑐 × 44% = 3993 × 44% = 1757 𝑣𝑒ℎ/ℎ
𝑐2 = 𝑐 × 56% = 3993 × 56% = 2236 𝑣𝑒ℎ/ℎ
For comparison purposes, HCM methods would obtain the following
theoretical capacity:
𝑐 = [2200 + 10 × (𝐹𝐹𝑆 – 50)] × 𝑓𝐻𝑉 = [2200 + 10 × (69.1 − 50) )] × 0.967
= 2312 𝑣𝑒ℎ/ℎ/𝑙𝑛
Therefore, the recommended CAF for this location is obtained by
dividing the field-measured by the theoretical values of capacity:
𝑐 1996.5
𝐶𝐴𝐹 = = = 0.864
𝑐 2312
Next, the breakpoint values for each lane can be obtained:
𝐵𝑃1 = [1000 + 40 × (75 − 𝐹𝐹𝑆1)] × 𝐶𝐴𝐹 = [1000 + 40 × (75 − 66.68)] × 0.864
𝐵𝑃1 = 995 𝑣𝑒ℎ/ℎ
Appendix C: Lane-by-Lane Analysis for freeway facilities Chapter 38 System Analyses (Draft)
Page 230 Version 1.0
Exhibit 38-C12
Field × Predicted Speed-Flow
Curve for (a) Lane 1 and (b)
Lane 2 (CA-1 NB – Santa
Cruz/CA)
Chapter 38 System Analyses (Draft) Appendix C: Lane-by-Lane Analysis for freeway facilities
Version 1.0 Page 231
APPENDIX B
The current methodology for Freeway Facilities analysis (HCM Chapter 10) evaluates the performance
of each segment individually using standard 15-minute time periods. If any segment within the facility
yields a LOS F and/or a v/c ratio greater than 1.0, the analysis continues with the oversaturated procedure,
using smaller time steps.
Similarly, in order to determine whether there is queue spillback from a freeway off-ramp, the analysis
is first conducted using 15-minute time periods. If the analysis shows that any of the ramps are expected to
experience queue spillback, the oversaturated procedure must be used to estimate the spillback impacts on
the freeway mainline lanes, even if the segment-wide performance is not at a LOS F and/or a v/c ratio
greater than 1.0.
The methodology framework for conducting a spillback check at diverge critical points is presented in
Figure B-1 and described in more detail in the remainder of this section.
232
233
Signalized Intersections. The operation of a signalized intersection will yield queues even when the
operation is undersaturated. Although an oversaturated approach is expected to create longer queues that
are growing in time and are more likely to spill back into the freeway diverge, it cannot be guaranteed that
the queues at an undersaturated approach will not affect the freeway mainline. Therefore, the methodology
estimates the queue length and compares it to the available storage length for each analysis period. The
arriving demand at the intersection may be constrained by the ramp proper capacity, and for this reason the
ramp proper capacity check must be conducted first.
234
Where:
v = maximum entering flow rate for the intersection approach (veh/h)
vR = off‐ramp demand for the period (pc/h)
cR = capacity of the off-ramp roadway (pc/h)
fHV = adjustment factor for heavy vehicle presence
fp = adjustment factor for driver population
If the off-ramp demand exceeds its capacity, the ramp proper acts as an upstream bottleneck, and limits
the demand to the intersection approach. This step ensures that the incoming demand at intersection does
not exceed the capacity of ramp proper. The calculations of throughput for each intersection type are
described below.
Signalized Intersections. The current methodology described in Chapters 19 and 31 evaluates the
performance of individual lane groups for a subject approach. It also estimates the back of queue length Q
(HCM Equation 31-149) or a percentile back-of-queue length Q% (HCM Equation 31-150). In some cases,
only one high-demand movement at the intersection approach is the bottleneck that results in spillback,
yielding an unbalanced lane usage pattern at the ramp. Field observations have shown that arterial
intersection failures may occur at one lane group as drivers position themselves in a specific lane at the
ramp to anticipate the downstream signal, the lane usage in the ramp becomes unbalanced, as shown in
Figure B-3.
235
Figure B-3. Examples of unbalanced ramp lane usage: (a) Norfolk, VA and (b) Tampa, FL
At off-ramps with two or more lanes, the estimated queue lengths for each intersection lane group must
be associated with specific ramp lanes. Figure B-4 illustrates an example of a typical ramp terminal. It is
expected that drivers that desire to take a left turn at the intersection will position themselves in the leftmost
lane ramp (Ramp Lane 2), while drivers who intend to turn right will likely choose the rightmost lane at the
ramp (Ramp Lane 1). Analyst judgement is required to define the grouping of intersection lane groups into
ramp lanes.
Figure B-4. Off-ramp geometry with additional lanes at the arterial approach
By using the results of the queue estimation procedure, the number of queued vehicles in a given ramp
lane n is estimated as follows:
Where:
QL,k = number of queued vehicles in ramp lane k, during a 15-min interval
QLG,m = number of queued vehicles from lane group m associated with ramp lane k, during a 15-min
interval
Qn%,LGm = estimated back of queue length (nth percentile), as defined in HCM Equation 31-150 (measured
in veh/ln)
NLGm = number of approaching lanes for lane group m
236
Unsignalized Intersections. Each unsignalized intersection type has its own methodology to estimate
queue length. The TWSC methodology estimates the 95th percentile queue length for minor movements
with Equation 20-68, while the 95th percentile queue length for AWSC approaches is estimated with
Equation 21-33. For roundabouts, the 95th percentile queue length for a given lane is provided by Equation
22-20. Regarding intersection lane groups and ramp lanes, the same procedure discussed above for
signalized intersections is applied.
Where:
QSP = length of queue beyond ramp storage distance (ft)
vR = off‐ramp demand for the period (pc/h)
cR = capacity of the off-ramp roadway (pc/h)
fHV = adjustment factor for heavy vehicle presence
PHF = peak hour factor
fp = adjustment factor for driver population
ti = analysis period i (h)
𝐿 = ∑ 𝑁 𝑥𝐿 (Equation B-4)
Where:
Ni = number of lanes in section i
Li = section i length (ft)
237
The individual ramp storage for each of the k lanes in the off-ramp, LR,k, can be estimated by assigning
the intersection lane groups to ramp lanes, as previously described:
𝐿 , = ∑ 𝑁, 𝑥𝐿 (Equation B-5)
Where:
Ni,k = Number of lanes in section i that are associated to ramp lane k
Li = Section i length (ft)
Finally, the ramp queue ratio for every ramp lane k is obtained as:
,
𝑅 , = (Equation B-6)
,
Where:
QL,k = queue length associated to ramp lane k
LR,k = available ramp storage for ramp lane k
Next, the total storage length is calculated. Figure B-4 illustrates a common off-ramp geometry with three
different sections from the stop bar to the gore point:
• Section 1: 4 lanes with length L1: two lanes (LG1) are associated with ramp lane 1, and two lanes
(LG2) are associated with ramp lane 2
• Section 2: 3 lanes with length L2: two lanes (LG1) are associated with ramp lane 1 and one lane
(LG2) is associated with ramp lane 2, and
• Section 3: 2 lanes with length L3: one lane (LG1) is associated with ramp lane 1, and one lane
(LG2) is associated with ramp lane 2
Where:
onramp queue length (veh)
LR = available queue storage distance (ft/ln)
Lh = average vehicle spacing in stationary queue (ft/veh)
N = number of lanes in the diverge ramp
238
The exit ramp at I-95 SB to SW 25th Rd (Miami, FL) has a signalized intersection ramp terminal (Figure
A-5). The off-ramp has two lanes and the signalized approach from the ramp (WB) has three lanes (one
shared left-through, one through, and one shared through-right), as shown in Figure B-5.
Figure B-5. Study site for example problem 1 (off-ramp queue spillback check - Miami, FL)
The geometry of the approach, the channelization at the stop bar and the segment lengths are shown in
Figure B-6.
Figure B-6. Signalized approach geometry for example problem 1 (off-ramp queue spillback check
- Miami, FL)
Queues from each lane group are assigned as indicated in Table B-1:
239
The signalized intersection performance was estimated using HCM methods (Chapter 19 – Signalized
Intersections), and the 95th percentile back-of-queue lengths were calculated as shown in Figure B-7.
Figure B-7. Back-of-queue length estimation for intersection approach using HCM methodology
(off-ramp queue spillback check – Miami, FL)
240
The ramp storage ratio for each of the two ramp lanes is:
RQ,1 = 1,196/1,400 = 0.85 < 1 → No spillback expected
RQ,2 = 2,733/1,800 = 1.51 > 1 → Spillback is expected
It is concluded that spillback will occur due to the higher demand on ramp lane 2 (connected to the left
and through movements). The expected queue length beyond the gore is 2733 – 1,800 = 933 ft.
Given the length of the deceleration lane (LD = 450 ft), the queue will extend to the freeway mainline.
241
A freeway-to-freeway two-lane ramp is evaluated for queue spillback (I-75 SB to SR-826 SB – Miami,
FL). The schematic of the study site is shown in Figure B-8.
Figure B-8. Study site for example problem 2 (freeway-to-freeway queue spillback check, Miami-FL)
This freeway-to-freeway connector is modeled as two separate freeway facilities. The upstream freeway
(I-75) is modeled with a diverge section that is connected to the downstream freeway (SR-826). The detailed
geometry, including turning movements and segment lengths, are shown in Figure B-9.
Figure B-9. Geometry for the study site of example problem 2: (a) I-75 and (b) SR-826
A multi-period analysis considering AM peak hour traffic volume (6.00 AM-7.30 AM) is conducted for
both facilities for six time periods (15 min each). The free-flow speed was measured as 63.5 mph for I-75
and 67.3 mph for SR-826. There are 12.1% trucks along the ramp, and the ramp length is 3,588 ft. The
exiting demand from the diverge is equal to the entering demand at the merge for undersaturated conditions;
however, the throughput at the two locations may not be the same if demand exceeds capacity at either
location. The results of the analysis for the SR-826, where the queue originates, are shown in Figure B-10.
242
243
As shown, segments 1 and 3 (boundary basic segments) in SR-826 are operating at LOS C and D
respectively for the entire analysis period. On the other hand, the merge section operates at LOS F each
during the third and fourth time periods. Thus, there is a queue present at the ramp starting from the third
time period. The estimated on-ramp queue for each time period is provided using the Freeway Facilities
method (Table B-2). The queue is assumed to be distributed evenly among the ramp lanes. The length of
the queue along the ramp during each time period is estimated to be greater than the length of the ramp. As
shown in Table B-2, the available queue storage length is insufficient (Queue storage ratio > 1), therefore
queue spillback is expected to occur at the diverge segment of I-75.
244
APPENDIX C
1. Introduction
The HCM (Chapter 14) provides three LOS checks for diverge segments, and failure (LOS F) may occur
in any of the following two cases:
• the total demand flow rate on the approaching upstream freeway segment exceeds the capacity of
the upstream freeway segment;
• the off-ramp demand exceeds the capacity of the off-ramp.
The HCM methodology also provides a LOS evaluation based on the density of the ramp influence area
(Exhibit 14-3), but it only yields a LOS range of A through E; failure due to excessive density is not
considered in the methodology. The first case of LOS F is addressed by the Oversaturated Segment
Evaluation procedure (HCM Chapter 10) and is not the focus of this methodology. The Queue Spillback
Analysis, described in this document targets the second case of LOS F, when the off-ramp demand exceeds
the capacity of the off-ramp. The methodology of this appendix also addresses cases of spillback due to
insufficient capacity at the ramp terminal downstream of an off-ramp.
The methodology described in Appendix A - Off-ramp Spillback Check presents the necessary steps to
determine whether spillback from an off-ramp is expected to occur, based on a standard 15-min period
analysis. This appendix provides the methodology for evaluating operations when spillback occurs. The
approach is based on the Freeway Facilities Oversaturated Segment Evaluation (HCM Chapter 25), where
performance measures are computed at the 15-s time step level.
Section 2 introduces the basic link-node structure that is applied to model off-ramp segments in the
methodology. Section 3 presents the concept of spillback regimes as a function of the off-ramp queue.
Section 4 presents a glossary with the definition of all parameters used in the procedure. Section 5 presents
the methodology to evaluate the impacts of an off-ramp queue spillback, and discusses each step and the
respective calculations.
245
• Ramp node 3.3: the last node in the off-ramp represents the discharge capacity of the arterial
intersection approach. The volume that flows through this node is equivalent to the amount of vehicles
that are able to enter the intersection;
The geometry of an off-ramp is seldom a homogenous road segment, and additional lanes are frequently
added closer to the arterial intersection approach. Figure C-2 illustrates a sample off-ramp, considering its
entire length from the deceleration lane to the stop bar at the downstream signalized intersection. The ramp
proper is the uniform ramp segment with a downstream boundary defined by the point where additional
lanes are provided. When modeling the off-ramp geometry, the method considers the channelization at the
approach as imbalances in the turning movements may cause queues on a subset of lanes. Figure C-2 shows
a typical queue formation resulting from a left-turn movement that operates with insufficient capacity. In
this scenario, the approaching left-turn vehicles are positioned in the leftmost lane and spillback may occur
even if not all lanes of the approach are oversaturated.
Figure C-2 – Sample geometry of an off-ramp considering the arterial intersection with heavy
demanded left-turn
246
The type of ramp terminal is an important input into the analysis. Signalized intersections operate in
cyclical patterns, and therefore those have fluctuating queue lengths. For certain demand scenarios, this
can result in queues backing up into the freeway and then discharging multiples times within a 15-min time
period.
Stop-controlled intersections and downstream merge segments (in the case of a freeway-to-freeway
connection), on the other hand, have a more uniform discharging rate. For cases other than signalized
intersections, off-ramp queues are assumed to develop or discharge linearly based on the relationship
between demand and capacity.
Regime 1
The queue ends within the deceleration lane and does not spill back into the mainline freeway (Figure C-
3 (a)). During undersaturated conditions, the deceleration lane serves as a transition zone between speeds
on the mainline (typically 55 – 75 mi/h) and advisory speeds posted along the off-ramp (typically 20 – 50
mi/h). When queues begin to form on the deceleration lane, the available deceleration distance is reduced
and speeds along the rightmost lane are affected.
Regime 2
The queue of vehicles extends upstream beyond the deceleration lane, but sufficient lateral clearance on
the right-hand shoulder allows for additional queue storage. In this case there is no transition zone within
the deceleration lane and drivers decelerate and join the back of the queue more abruptly, resulting in
turbulence and reduced speeds in the rightmost lane (Figure C-3 (b)). If no lateral clearance exists
immediately upstream of the deceleration lane, Regime 2 conditions are not possible. In some cases, this
regime does not occur even if storage is available; this depends on local driver behavior and is site-specific.
Regime 3
The queue extends to the rightmost lane of the freeway mainline (Figure C-3 (c)). This may occur either
when there is no shoulder available for additional queue storage, or when drivers choose to queue in the
rightmost lane once the deceleration lane is entirely occupied. Non-exiting vehicles on the rightmost lane
are delayed or change lanes, which causes increased turbulence and reduced speeds in both rightmost lanes.
Regime 4
The queue blocks the rightmost lane, and drivers occasionally or often use the adjacent freeway mainline
lane next to the rightmost freeway mainline lane to force their way into the queue, blocking thus an
additional lane (Figure C-3(d)). During this regime, speed and capacity are significantly reduced. The
effects of spillback vary from site to site and from time period to time period due to driver behavior and
site geometry. Data collection at locations around the US has shown that at some sites drivers block the
adjacent lane, while at other sites they do not, regardless of the queue spillback length at the site.
247
Facility variables
• QIA(i, p): Length of the queue influence area (ft) for segment i during time period p, measured
from the back of the queue.
Segment variables
• KBBL(i,j): background density (pc/mi/ln) at the blocked lanes in segment i, when queue spillback
occurs at a downstream segment j
• KBUB(i,j): background density (pc/mi/ln) at the unblocked lanes in segment i, when queue
spillback occurs at a downstream segment j
• LCR(i,t,p): rate of lane change maneuvers in the queue influence area upstream of a queue from an
off-ramp, for segment i during time period p and time step t.
• LD(i,p): available deceleration lane length (ft) for segment i during time period p. This variable is
used to calculate performance measures for ramp segments (Chapter 14 - LD.)
• MQ1(i,t,p): mainline queue length of off-ramp unserved vehicles in the rightmost mainline lane,
for segment i during time period p in time period t.
• MQ2(i,t,p): mainline queue length of off-ramp unserved vehicles in the rightmost mainline lane,
for segment i during time period p in time period t. If Regime 4 is not expected to occur, this parameter
value is set to zero.
• NQ(i): number of blocked lanes if the off-ramp queue backs up into the freeway mainline. This
parameter is a function of the prevailing spillback regime at segment i as provided by the analyst. The
value for this parameter is an input and can be either 1 (Regime 3 - one blocked lane) or 2 (Regime 4 –
two blocked lanes);
• OFRFUP(i,t,p): flow that can exit at the closest off-ramp downstream of i during time step t in time
period p.
• OFRLQ(i,t,p): queue length of off-ramp unserved vehicles for diverge segment i during time period
p in time period t.
248
• OFRUV(i,t,p): number of off-ramp unserved vehicles for segment i during time period p in time
period t.
• SBKQ (i,t,p): spillback queue density for segment i during time period p in time period t.
• SBLC(i,t,p): number of lane change maneuvers within the Queue Influence Area at node i, during
time step t in time period p.
• SBLQ(i,t,p): queue length within segment i during time period p in time period t, caused by a
downstream off-ramp bottleneck.
• SBQS(i,p): total available off-ramp queue storage (ft) for a diverge segment i during time period p,
if the subject segment has an off-ramp bottleneck. It is calculated as a function of the available storage
lengths in the deceleration lane, shoulder and prevailing spillback regime.
• SCEQ(i,N,NQ): equivalent capacity of the unblocked portion of a segment i with N total lanes and
NQ blocked lanes.
• SL(i,p): available shoulder length (ft) for segment i during time period p. If the value of SL is
greater than zero, any off-ramp queues that exceed the deceleration lane will occupy the shoulder before
blocking mainline lanes.
• TIA(i,p): total influence area (ft) of a queue from an off-ramp bottleneck on segment i, during time
period p in time period t. It is calculated as the sum of parameters QIA(i,t,p) and MQ(i,t,p).
Node variables
• CAFBL(i,t,p): capacity adjustment when one or more lanes of segment i are entirely blocked during
time period p in time period t. This is used to calculate friction effects that cause through vehicles to
slow down due to the presence of a queue in the rightmost lanes.
• CAFUP(i,t,p): capacity adjustment factor of node i during time step t in time period. This capacity
adjustment factor affects approaching vehicles within the queue influence area (QIA) upstream of an
off-ramp queue. This factor accounts for the turbulence caused by intense lane changing within the
queue influence area as vehicles adjust their position when there is a downstream off-ramp queue.
• MFBL(i,t,p): mainline flow rate that can cross the blocked portion of node i during time step t in
time period p.
• MFUB(i,t,p): mainline flow rate that can cross the unblocked portion of node i during time step t
in time period p.
• MIBL(i,t,p): maximum flow desiring to enter the blocked portion of node i during time step t in
time period p.
• MIUB(i,t,p): maximum flow desiring to enter the unblocked portion of node i during time step t in
time period p.
• MO2BL(i,t,p): maximum number of passenger cars that can enter the blocked portion of segment
i, during time step t and time period p, due to the presence of a queue in the downstream ramp segment.
• MO2UB(i,t,p): maximum number of passenger cars that can enter the unblocked portion of segment
i, during time step t and time period p, due to the presence of a queue in the downstream ramp segment.
• NEXTOFR(i): index of the nearest downstream diverge segment relative to subject node i.
• OFRDIST(i): distance (ft) from node i to the start of the deceleration lane at the nearest downstream
off-ramp.
• OFRPCT(i,j): percent of the off-ramp demand at segment j over the mainline entering volume at
segment i.
Ramp variables
• RC(i,p): capacity of the ramp proper (pc/h) during time period p in time period t. Capacity values
for the ramp proper are provided in HCM Exhibit 14-12.
249
• RF(i,t,p,k): flow (pc/ts) that can enter the ramp proper at segment i during time period p in time
period t and level k.
• RI(i,t,p,k): maximum flow (pc/ts) desiring to enter the off-ramp on segment i during time period p
in time period t and level k, including queues accumulated from previous time periods.
• RKB(i,t,p,k): ramp proper queue density (pc/mi/ln) for segment i during time period p in time period
t and level k.
• RL(i): length of ramp proper (ft) for segment i.
• RN(i): number of ramp lanes for segment i.
• RNV(i,t,p,k): maximum number of passenger cars within the ramp of segment i at the end of time
step t during time period p and level k. The number of vehicles is initially based on the calculations of
Chapters 12, 13, and 14, but, as queues grow and dissipate, input–output analysis updates these values
during each time step.
• RSTG(i,t,p,k): maximum number of passenger cars that can enter the ramp level k of segment i,
during time step t and time period p, due to the presence of a queue in the downstream ramp segment.
• RUV(i,t,p,k): number of unserved vehicles at the entrance of the ramp proper of segment i at the
end of time step t during time period p and level k. Any values of RUV greater than zero indicate the
occurrence of queue spillback from an off-ramp.
• ID (i,t,p,k): discharge capacity (pc/ts) for intersection movement k in segment i during time period
p in time period t.
• IF(i,t,p): flow (pc/ts) that can enter the intersection on segment i, level k, during time period p in
time period t.
• II(i,t,p,k): maximum flow (veh/ts) desiring to enter the intersection on segment i, level k, during
time period p in time period t, including queues accumulated from previous time periods.
• IL(i,k): storage length of movements at intersection of segment i, for level k (ft)
• INV(i,t,p,k): number of vehicles at the intersection of segment i, for level k at the end of time step
t during time period p
• IO(i,t,p): flow (pc/ts) that can be discharged from the intersection on segment i, level k, during
time period p in time period t.
• ISTG(i,k): total available storage length at intersection of segment i, for level k (ft)
• IUV (i,t,p,k): number of unserved vehicles at the entrance of the intersection of segment i, for level
k, at the end of time step t during time period p
250
251
252
253
254
Step 1 - Calculate background density for unblocked lanes on each segment in the case of queue
spillback
The first step in the Oversaturated Segment Evaluation procedure computes a background density (KB),
for each segment at the start of each time period, defined as the expected density when there is no queueing
on the segment. It is used as a reference to estimate how many vehicles occupy a given segment at
undersaturated conditions, creating an initial reference point for oversaturated analyses.
When Regime 3 or Regime 4 occur, there is blockage of one or more freeway lanes in the affected
segments, and the through vehicles aim to move to the unblocked lanes. The capacity of the unblocked
lanes must be calculated at the initialization step, to be used as a reference value.
For a segment i with N lanes, a subset NQ of lanes will be blocked when spillback occurs (NQ = 1 for
Regime 3 and NQ = 2 for Regime 4). Therefore, the capacity of the unblocked lanes will be equivalent to
a similar segment with (N - NQ) lanes, adjusted for the impact of the blockage using a capacity adjustment
factor CAFBL. The values of CAFBL are equal to the Incident Capacity Adjustment Factors of Chapter 11,
Freeway Reliability Analysis (Exhibit 11-23), as there are currently no data available to accurately assess
the impacts of blockage due to spillback. These values may be conservative, as during incidents capacities
may be further reduced due to the presence of police vehicles. Table C-1 presents the recommended values
for CAFBL.
Table C-1. Capacity adjustment factors for lane blockage (CAFBL) as a function of the number of
directional lanes and the number of blocked lanes
The equivalent capacity SCEQ of segment i, with N lanes and NQ blocked lanes, is estimated as:
Figure C-8 presents an example of a basic 4-lane directional segment operating in Regime 4 (2 blocked
lanes). The capacity of the unblocked lanes will be equivalent to the capacity of a 2-lane basic segment with
a capacity adjustment factor CAFBL = 0.50 (4 directional lanes with 2 blocked lanes).
Figure C-8 – Equivalent segment capacity for unblocked lanes when lane blockage occurs
255
When Regime 4 occurs (2 blocked lanes), the equivalent capacity SCEQ is obtained as the equivalent
capacity of a 2-lane segment multiplied by a corresponding CAFBL of 0.5 (Table C-1):
Next, the unblocked background density KBUB is calculated. This parameter estimates the background
density of the uncongested portion of a given segment operating under a two-pipe regime due to a queue
spillback from a downstream off-ramp. To estimate this value, the method first determines the ratio of the
Expected Demand (ED) that will move to the uncongested side of the segment. When queue spillback
occurs in a diverge segment j, the parameter OFRPCT(i,j) is defined as the percent of the off-ramp demand
over the mainline entering volume vf:
𝑣 𝑗
𝑂𝐹𝑅𝑃𝐶𝑇 𝑖, 𝑗 = (Equation C-2)
𝑣 (𝑖)
For any segment i, upstream of segment j and affected by the off-ramp spillback from segment j, the
ratio of vehicles traveling towards the off-ramp at segment i is given by OFRPCT(j), while the ratio of
vehicles traveling through in the unblocked lanes is given by (1- OFRPCT(j)). Therefore, the unblocked
background density KBUB at any segment i upstream of an off-ramp spillback in a segment j is given by:
Where:
KBUB(i,j)I = background density at the unblocked lanes in segment i, when queue spillback occurs at
the downstream segment
ED(i) = expected demand at segment i , as defined in HCM Chapter 25
OFRPCT(i) = rate of off-ramp flow and mainline flow at segment i
KB (v,c)I = density at a segment with demand flow rate v and capacity c, as provided by HCM
Chapters 12 (basic), 13 (weaving) and 14 (merge and diverge)
The number of mainline blocked lanes is stored in the parameter NQ(i) and is determined by the prevalent
queue spillback regime as provided by the analyst. If the back of an off-ramp queue is calculated to reach
the freeway mainline, two possible spillback regimes may occur:
• Regime 3: blockage of one lane in the freeway mainline → Set NQ(i) = 1
• Regime 4: blockage of two lanes in the freeway mainline → Set NQ(i) = 2
256
The analyst should select one of these two regimes based on prevailing driver behavior at the site and in
the vicinity of the site.
Shoulder length
The available shoulder length must be input by the analyst for queue spillback analysis, and is stored
under the parameter SL(i) for oversaturated calculations.
The deceleration lane length is provided by the analyst for the analyses of diverge segments, and is stored
under the parameter LD(i) for oversaturated calculations.
The maximum storage length for off-ramp queues on segment i is computed as a function of the segment
length L(i), the deceleration lane length LD(i) and the number of queued lanes NQ(i). Figure C-9 provides
guidance on measuring each of the components required for Regimes 3 and 4.
Figure C-9 – Maximum off-ramp queue storage length at diverge segments with occurrence of (a)
Regime 3 queue spillback and (b) Regime 4 queue spillback, when no shoulder is available
Figure C-10 illustrates queue length measurements for special cases of queue spillback, when a shoulder
is present but its storage length is not sufficient to accommodate the unserved vehicles. Regime 3A (Figure
C-10a) occurs when there is blockage of one mainline lane in addition to the shoulder. Regime 4A (Figure
C-10b) occurs when there is blockage of two mainline lanes in addition to the shoulder.
Figure C-10 – Maximum off-ramp queue storage length at diverge segments with occurrence of (a)
Regime 3A queue spillback and (b) Regime 4A queue spillback, when shoulder is available
257
The data structure used in the methodology computations should be adjusted according to this “branch”
structure. Most parameters in the Oversaturated Segment Evaluation methodology are computed as a 3-
dimensional array (i, t, p), where i is the segment index in the freeway facility and t refers to a specific time
step within a given time period p. In the case of two-lane ramps that need to be evaluated independently, a
new dimension k will be added to the ramp parameter arrays to account for the specific lane under analysis.
Lanes are numbered right from left; therefore, k=1 stands for the rightmost lane and k=2 for the leftmost
lane of the ramp.
Example 1 – In this example, there is only one lane connecting the freeway exit to the entry leg of the
downstream roundabout. Therefore, only one node is required at each location (single branch structure,
with k=1 in all nodes), as shown in Figure C-11.
Example 2 – A single-lane ramp connects with a stop-controlled T-intersection ramp terminal (Figure
C-12). The intersection node is comprised of two branches (k=2), while the ramp proper has only one lane
(k=1). Each movement of the intersection (LT and RT) is represented by a node, and when there is a queue
on either one of the movements, the one with the longest length will constrain the flow of vehicles from the
ramp proper.
258
Example 3 – A two-lane ramp connects with a signalized intersection ramp terminal (Figure C-13). Both
the intersection and the ramp proper nodes comprise of two branches each (k=2). At the downstream end,
one node is defined for each lane group at the intersection (LT and RT). According to the ramp geometry,
left-turning drivers will be positioned along ramp lane 2, while right-turning drivers will be located along
ramp lane 1. Therefore, two nodes are also defined at the upstream location. If the queue storage ratio for
any of the ramp lanes reaches 1, vehicle flow in the respective upstream node will be constrained, resulting
in queue spillback on the freeway mainline.
259
the off-ramp with a single lane change. Therefore, drivers are more likely to wait until they are closer to
the exit to change lanes, blocking the adjacent through lane. However, not all lane drop exits experience a
Regime 4 queue spillback. Regime 4 occurs more frequently in locations with more aggressive driver
behavior. Local information and driver behavior should be taken into consideration in determining the
prevailing regime at a given site.
For operational analyses of existing locations, it is recommended that the analyst provides the expected
spillback regime based on observed field conditions. For planning level purposes where no field data is
available, Table C-2 provides the expected queue spillback regime as a function of the number of exiting
lanes and driver aggressiveness.
Table C-2. Default spillback regimes as a function of ramp geometry and driver aggressiveness
Ramp Driver aggressiveness
geometry Low Medium High
Diverge Regime 3 Regime 3 Regime 3
Lane drop Regime 3 Regime 4 Regime 4
“For right-hand off-ramps, the ramp influence area includes the deceleration lane(s) and Lanes 1 and
2 of the freeway for a distance of 1,500 ft upstream of the diverge point.”
When there is queue spillback in one or more freeway lanes, drivers would react to the presence of the
queue further upstream resulting in increasing lane changes and additional turbulence upstream of the ramp
influence area (Figure C-14). In this step the methodology estimates the length of the Queue Influence Area
(QIA), measured upstream from the back of queue.
Field data (video observations and loop detector data) were used to estimate the length of the Queue
Influence Area. This measurement process is illustrated in Figure C-15. For a given off-ramp bottleneck,
the distance between the exit and the upstream loop detector is known and fixed. The back of queue length
Q(t) at time t was measured by video observations of congested diverge segments. Speed measurements
from the loop detector were obtained with resolutions ranging between 20s and 60s (depending on the
source) and analyzed to determine the onset of congestion. Oversaturated conditions were determined to
occur when a speed drop greater than 20% occurred in at least one lane, and sustained for at least 15 min.
260
Figure C-15 – Measurement of queue influence area length based on queue lengths
For the timestamp tbd when congestion begins in at least one lane, the back of queue length is also known
from video observations. The distance between the detector and the back of queue Q(tbd) at the congestion
initiation time is defined as the Queue Influence Area. The speed prior to congestion, sbd, was also measured
(Figure C-16), and used in further calculations.
Figure C-16 – Sample measurement of queue lengths and speeds at the time of breakdown
261
This process was performed for all data obtained. It was observed that different locations operate at
significantly different speeds prior to congestion, therefore the Queue Influence Area measurements were
normalized to estimate the reaction headway, defined as the travel time between detector location (where
breakdown occurred) and the back of the queue:
3600 × 𝑄𝐼𝐴
ℎ = (Equation C-4)
5280 × 𝑠
Where:
hR = reaction headway (s)
QIAbd = measured Queue Influence Area at time of breakdown (ft)
sbd = measured speed at the beginning of congestion at the upstream detector (mi/h)
After all queue spillback observations were analyzed, the measured values of the reaction headway can
be found in the histogram of Figure C-17, with a median value of 10.95 s.
Using this median value of 10.95s, the length of the QIA is estimated as a function of the segment free-
flow speed (FFS), as shown in Table C-3. The exact location of the QIA varies as a function of the queue
length. The QIA values are shorter than the ramp influence distance of 1,500 ft. However, the two concepts
are very different and are used differently in analyzing ramp operations: the ramp influence area is used to
analyze undersaturated conditions, while the QIA is used to analyze oversaturated conditions. Since drivers
can only detect a downstream queue visually, they have shorter times to react when compared to the
presence of undersaturated off-ramps, where signing and navigation information is provided in advance
and allows drivers to adjust their position earlier.
262
When Regimes 3 or 4 occur and lane blockage is present in the mainline, the estimated QIA is added to
the queue length to determine the extent of spillback effects. If an upstream node is located within the
combined length of the queue and QIA, capacity adjustment factors must be applied to account for the
spillback effects.
The RC is compared to the off-ramp demand, and if the demand-to-capacity ratio is greater than 1.0 then
spillback is expected to occur.
Determining the speed-flow relationship at the ramp proper is also critical for the analysis. Speed data
along off-ramps are scarce, but a few field observations at off-ramp speed-flow curves (Figure C-19) have
shown that speeds have little variation as a function of demand.
Figure C-19 – Sample speed-flow curves for: (a) I-694 EB to Silver Lake Rd. and (b) I-94 EB to
Brooklyn Blvd. Minneapolis/MN
Determining the speed-flow relationship at the ramp proper is also critical for the analysis. Ramp speeds
can be obtained through the following equation:
𝑣
𝑆 = 1 − 0.109 × ×𝑆 (Equation C-5)
1000
where
𝑆 = ramp speed (mi/h);
263
The ramp density at capacity (RKC) is not necessarily equal to 45 pc/mi/ln as assumed for freeway
mainline lanes. This parameter is required to evaluate the queue density at the ramp proper when operating
in oversaturated conditions. The ramp density at capacity can be found by dividing the capacity by speed.
Table C-4 summarizes the values of RKC as a function of the Ramp FFS.
264
one oversaturated movement may extend upstream and block the throughput of all movements at the off-
ramp. ISTG is estimated as:
Where:
Nm = number of lanes serving movement m at the intersection
Lm = storage length for movement m at the intersection (ft)
N = number of movements at the approach
Lh = average vehicle spacing in stationary queue (ft/veh) (HCM Equation 31-155)
The initial number of vehicles in the intersection approach are also determined as an initial reference
point, as follows:
Where:
INV(i,t,p,k) = number of vehicles at the intersection of segment i, for level k at the end of time
step t during time period p
IN(i,k) = number of lanes serving the subject approach k
Qk = back-of-queue length for the subject approach k (veh)
The back-of-queue length Qk is estimated using equations corresponding to the intersection type at the
ramp terminal (Table C-5).
265
Signalized 31-149
TWSC 20-68
AWSC 21-33
Roundabout 22-20
At signalized intersections, due to their cyclic nature, queues form and discharge at different times for
different movements. Therefore, a reference point within the cycle must be selected as a starting point in
the methodology. The methodology assumes pretimed signalization, or converts actuated control to the
equivalent pretimed pattern. Typical signalized intersections at ramp terminals have the off-ramp approach
as the minor movement, with a start of green on the right side of the barrier (Figure C-21). It is
recommended setting a reference point at the onset of green for phases 3 and 7, as the back-of-queue lengths
at this time can be easily estimated using the methodology of Section 4, HCM Chapter 31.
Figure C-21 – Selection of a cycle reference point to determine the initial number of vehicles within
the approach
Signalized Intersections
For a signalized intersection approach, the capacity for each movement at each time step is a function of
the signal phase sequence and the capacities of the individual movements at the intersection. Figure C-22
illustrates a sample signalized intersection approach from an off-ramp, with two lane groups: left-turn
(Phase 3) and right turn (Phase 8).
266
Input Parameters
The required parameters to evaluate the capacity of a ramp terminal capacity are generally the same
required for standard signalized intersection analyses, as listed in Exhibit 19-11.
Arrival type: Chapter 19 of the HCM (Exhibit 19-14) provides guidelines for selecting the appropriate
Arrival Types based on the characteristics of arterial operations, such as quality of progression and
coordination. For an off-ramp approach to the intersection, vehicles arrivals can be considered random.
Therefore, Arrival Type 3 (random arrivals) is recommended to analyze the off-ramp approach at a
signalized ramp terminal.
Phase duration and effective green time: The duration of each phase at the signal can be fixed (pre-timed
control), or variable (semi-actuated or actuated control). For the former case, phase duration is known. For
the latter, an average phase duration is estimated as described in Section 2 of HCM Chapter 31 – Signalized
Intersections Supplemental. The effective green time g for each phase can then be computed according to
HCM Equation 19-3:
𝑔 =𝐷 −𝑙 −𝑙 (Equation C-9)
Where:
g = effective green time (s)
Dp = phase duration (s)
l1 = start-up lost time = 2.0 (s)
l2 = clearance lost time = Y + Rc – e (s)
Y = yellow clearance interval
Rc = red clearance interval
e = extension of effective green = 2.0 (s)
The standard signalized intersection analysis is performed in 15-min periods, while the queue spillback
evaluation requires a 15-second approach compatible with the Freeway Facilities oversaturated
267
methodology. Therefore, an adjustment is necessary to calculate the capacities of each movement in 15-
second intervals.
The cycle length C can be divided into n time steps, with a duration of 15 s each (Figure 23). If an integer
number of time steps is not obtained, the difference is included in the first time step of the next cycle. Then,
green times for each time step from 1 to n are computed. This procedure must be repeated for every time
step within the 15 minutes time period, resulting in a total of 900/15 = 60 time steps.
The capacity ID for each approach and for each time step, is then obtained by multiplying its respective
green time by its capacity, as shown:
The green time parameter GT(i,t,p,m) measures the available green time for a given intersection
movement m, downstream of a freeway segment i, in time step t and time period p. It can range from 0
(when the movement has red through the entire time step length) to 15 (movement has green through the
entire time step length).
The heavy vehicle factor fHV needs to be applied to the equation for intersection discharge to make the
units used in intersection capacity (veh/h) consistent with the flow rates used in uninterrupted flow methods
(pc/h).
268
When two freeway facilities are connected through a ramp junction, the merge segment at the
downstream facility becomes the ramp terminal. In this case, the capacity of the ramp terminal is equal to
the merge capacity at the downstream freeway.
If the downstream segment operates at undersaturated conditions, there is no data available in the
literature to estimate the capacity of the merge segment. In this case, the merge capacity is considered
unrestricted (for a computational engine, a very high capacity value such as 9999 pc/h can be assumed) and
the only constraint at the freeway-to-freeway interaction will be the capacity of the ramp roadway, provided
by HCM Exhibit 14-12.
If the downstream segment operates at oversaturated conditions, the merge capacity is constrained by the
congested conditions in the mainline. The Freeway Facility Oversaturated Segment Evaluation procedure
computes, for every time step, the parameter ONRO as the maximum number of vehicles that can merge
into the freeway in a given time period. Therefore, in these conditions the merge capacity can be obtained
by analyzing the downstream freeway facility and aggregating the yielded values of ONRO to an hourly
flow rate.
Figure C-24 – Illustration of mainline flow rate split into blocked and unblocked lanes
For nodes i and i+1, the closest downstream off-ramp is located at node (i+2), therefore the following
parameter is computed:
𝑁𝐸𝑋𝑇𝑂𝐹𝑅(𝑖) = 𝑖 2
The use of the parameter NEXTOFR facilitates referencing diverge segments downstream of a given
segment i and will be used for the spillback analysis procedure described over the next section.
269
The ramp input, RI, represents demand, and it is the number of vehicles that wish to travel through the
ramp proper node during a given time step. It takes into account the off-ramp demand, OFRF (as defined
in the Freeway Facilities Oversaturated methodology) and the number of off-ramp unserved vehicles from
the previous time step, RUV. The OFRF parameter already takes into consideration any bottleneck segments
upstream of the diverge that may meter the off-ramp demand (HCM Equations 25-23 through 25-25). The
RI is calculated as:
The ramp maximum flow RF represents capacity, i.e., the number of vehicles that are able to enter the
ramp proper by crossing the boundary node between the diverge segment and the ramp proper. It is
calculated as the minimum of three variables: RI, RC and RSTG.
The parameters RI and RC have been previously defined. The parameter RSTG represents the maximum
number of vehicles that can enter the ramp due to a queue inside the ramp proper. The calculations follow
the same approach taken by the Mainline Output 2 (MO2) parameter (Equation 25-11). It starts by
calculating the maximum number of vehicles allowed on the ramp at a given ramp queue density RKQ:
The calculation of RKQ takes an approach similar to the calculation of the mainline queue density KQ
(Equation 25-10), with the following remarks on the inputs:
• The jam density parameter KJ uses the same value adopted for the mainline calculations
• The ramp density at capacity RKC is determined based on the ramp FFS (Table C-4)
• The parameters SF (segment flow) and SC (segment capacity) from Equation 25-10 are replaced
with RF (ramp flow, previously defined) and RC (ramp capacity, previously defined)
The maximum ramp storage constraint RSTG is then calculated using an approach similar to the Mainline
Output 2 (MO2) parameter from the Oversaturated segment evaluation procedure. This constraint limits the
number of vehicles able to enter the off-ramp due to the presence of a queue within the ramp proper. RSTG
is calculated as:
270
Next, the number of unserved vehicles at the ramp entrance RUV is calculated. For each time step, the
number of unserved vehicles is computed as the value from the previous time step, plus the difference
between demand (RI) and throughput (RF) at the ramp node. RUV is calculated as:
Where:
K = number of different branches at the intersection
If there are multiple branches k at the ramp proper (two lane ramps), RI and RF are compared for each
branch k to obtain RUV for each branch k. The total number of unserved vehicles at the ramp RUV(i,t,p) is
then obtained as the sum of RUV for each lane:
The intersection approach input II is the number of vehicles that wish to travel through the intersection
node during a given time step, i.e., its demand. It takes into account the off-ramp flow RF and the number
of unserved vehicles on the approach from the previous time step IUV. II is calculated as:
The maximum allowable ramp output (RO) is calculated as a function of the available storage space
within the intersection approach, minus the number of vehicles present at the previous time step and the
number of vehicles discharged (IDC) in the present time period. RO is estimated as:
The intersection flow IF represents the number of vehicles that are able to cross the boundary node
between the ramp proper and the intersection (i.e., its capacity). It is computed as the minimum value
between the number of vehicles that wish to enter the intersection and the maximum number of vehicles
allowed to enter the intersection due to the available queue storage in the intersection:
If the number of vehicles trying to enter the intersection exceeds the amount of vehicles allowed to enter
the intersection, then the number of total unserved vehicles must be computed and considered in the
intersection input II during the next time period:
271
The number of vehicles at the intersection, INV, is updated every time step based on the NV from the
previous time step, plus the number of vehicles that enter the intersection approach minus the number of
vehicles that are discharged. The maximum allowable total number of vehicles is function of the available
storage at the intersection, ISTG. NV is calculated as:
The number of unserved vehicles, OFRUV, at the entrance of the ramp proper is updated every time step
as the difference between the number of vehicles that wish to enter the ramp proper (RI) and the flow
through the ramp node (RF):
The intersection flow, IO, represents the actual number of vehicles discharging from the intersection
approach. It is computed as the minimum value between the intersection discharge capacity and the sum of
number of vehicles present in the intersection and the intersection input demand:
The number of vehicles at the ramp proper, RNV, at the end of each time step is calculated based on the
number of vehicles from the previous time step plus the number of vehicles that entered the ramp minus
the number of vehicles that left the ramp:
Field observations have shown that off-ramp queues blocking mainline lanes are typically not stationary.
These queues usually consist of a platoon of closely-spaced vehicles moving at very low speeds (< 15mph).
The spacing between vehicles is also longer than the average vehicle spacing in stationary queues,
represented in the HCM by Lh (Equation 31-155). Therefore, the density of the spillback queue follows the
queue density at the ramp (RKQ, as previously defined), which allows the estimation of the queue length
OFRLQ. This parameter estimates the total queue length upstream of the off-ramp if all unserved vehicles
formed a single queue:
𝑂𝐹𝑅𝑈𝑉(𝑖, 𝑡, 𝑝)
𝑂𝐹𝑅𝐿𝑄(𝑖, 𝑡, 𝑝) = (Equation C- 25)
𝑅𝐾𝑄 (𝑖, 𝑡, 𝑝)
272
Next, the mainline queue length, SBLQ, is compared to the available spillback queue storage for the
prevalent spillback regime for the given time step, as follows:
If OFRLQ = 0 → Regime 0
If 0 < OFRLQ ≤ LD → Regime 1
If SBLQ > LD :
If SL(i,p) > 0:
If OFRLQ < (LD + SL) → Regime 2
Else: Regime 3 / 4
Finally, the queue length in the mainline lanes MQ1 (lane 1) and MQ2 (lane 2) are obtained as a function
of the expected spillback regime. The total queue length OFRLQ minus the available storage lengths at the
deceleration lane and shoulder computes the queue length that the associated blockage.
The freeway nodes upstream of a congested off-ramp may be affected by spillback as queues grow.
When this occurs, the methodology calculates the length of the queue in the upstream segment. The length
of the queue within the subject segment will then be used to evaluate whether the capacity of any upstream
node is affected by the queue.
For upstream segments that may be affected by spillback, the queue length within the segment (measured
from its downstream end) must be computed and stored in the parameter SBLQ. This check is performed
for every node upstream of a congested off-ramp (Figure C-25).
273
Figure C-25 – Procedure for evaluating the impact of queue spillback on upstream nodes and
determination of the queue length within upstream segments
When queue spillback occurs in a downstream off-ramp, the length of the mainline queue measured from
the start of the deceleration lane is known from the previous step. If a given segment has any queues
blocking one or more lanes, three possible scenarios may occur at the node (Figure C-26):
(a) Lane blockage: Queues extend through the entire segment and reach the upstream node, causing the
subject node to operate in a two-pipe regime. The blocked lanes operate in a congested regime, with
their capacity constrained by the off-ramp capacity. The unblocked lanes, on the other hand, operate
at uncongested conditions with a small reduction in capacity due to the friction of through vehicles
passing along congested lanes. For the through lanes, an adjustment factor CAFBL is applied. This
condition occurs when the Spillback Queue length SBLQ(i) is equal or greater than the Segment
Length L(i).
(b) Increased turbulence: Queues extend partially through the segment and the upstream node is located
within the Queue Influence Area (QIA). This region is characterized by intense turbulence as vehicles
quickly perform lane changes to adjust their position reacting to the queue ahead, and all lanes in
node i have their capacity reduced by an adjustment factor CAFUP. This condition occurs when the
sum of the Spillback Queue length SBLQ(i) and the Queue Influence Area QIA(i) is equal or greater
than the segment length L(i).
(c) No effect: Queues extend partially through the segment but the upstream node is located within the
Queue Influence Area (QIA). For this condition, no capacity adjustment factors are applied to the
node i. This condition occurs when the sum of the Spillback Queue length SBLQ(i) and the Queue
Influence Area QIA(i) is smaller than the segment length L(i).
274
Figure C-26 – Illustration of different impacts of an off-ramp queue at node i: (a) lane blockage, (b)
increased turbulence and (c) no effect
Based on how upstream nodes are affected as described under Step 6B (Lane Blockage, Increased
Turbulence or No Effect), the corresponding impacts on capacity are computed in this step. This section
describes the calculations of capacity adjustments depending on how upstream nodes are affected.
When one or more lanes are blocked, the subject node is analyzed as a two-pipe operation, with a
congested flow in one or more lanes of the ramp side and uncongested flow in the remaining lanes.
The capacity of these lanes is equal to the number of queued vehicles discharged at the downstream
segment. The flow rate attempting to cross the node through the congested lanes is equal to the off-ramp
flow rate (OFRF) at the closest downstream off-ramp
When a node falls under the Increased Turbulence case (Figure C-26b), all lanes are affected by the
turbulence caused by the intense lane changing. In this case, an adjustment factor CAFUP is applied
uniformly to the node capacity:
The calibration adjustments α and β were calibrated to match field conditions. Based on observed field
data, the recommended values for the parameters are α = 0.52 and β = 0.81.
The parameter LCR estimates the rate of lane change maneuvers performed by vehicles within the Queue
Influence Area trying to adjust their position when spillback occurs. Vehicles traveling towards the exit
ramp will move to the shoulder lane attempting to join the back of the queue, while vehicles traveling
through will move to the median lanes in order to avoid the queue. Therefore, the lane change rate LCR is
computed as:
275
𝑆𝐵𝐿𝐶(𝑖, 𝑡, 𝑝)
𝐿𝐶𝑅(𝑖, 𝑡, 𝑝) = (Equation C- 29)
𝑆𝐹(𝑖, 𝑡, 𝑝)
The parameter SBLC represents the number of lane change maneuvers performed in the queue influence
area. In order to compute SBLC for a given node, the number of vehicles driving toward the off-ramp must
be estimated for each freeway lane. For each lane i, the parameter pi represents the percent of the off-ramp
demand vR traveling on the subject lane. In order to estimate the values of pi as a function of the distance
from the off-ramp to the subject node, the following steps and assumptions are used:
a) Within the influence area (1,500 ft from the exit point), the off-ramp demand flow rate vR
is entirely positioned in the two rightmost lanes, according to the guidance provided in
HCM Chapter 14. Therefore, the sum of the off-ramp flow rate percentages in the ramp
influence area p1,R and p2,R is equal to 1. The methodology to estimate lane-by-lane flow
distribution in freeway segments allows the estimation of the Lane Flow Ratio (LFR) for
lanes 1 and 2. The proportion between p1,R and p2,R can then be estimated as follows:
𝑝 , = , 𝑝 , = (Equation C- 30)
b) According to the guidance provided in HCM Chapter 14, the influence of ramps rarely
extend beyond 8,000 ft. Therefore, for any nodes located beyond 8,000 from the off-ramp,
the distribution of pi is taken as equal among all N freeway lanes:
1
𝑝 = (Equation C- 31)
𝑁
c) At intermediate distances from the off-ramp dOFR ranging between 1,500 ft and 8,000 ft,
the distribution values of pi can be obtained through linear interpolation between the cases
previously described.
Figure C-27 – Distribution of pi as function of distance from the off-ramp exit, for a 3-lane segment
The value of pi as function of the distance from off-ramp exit dOFR can then be obtained through the
following equation:
276
1
− 𝑝 , 𝑅 × (𝑑 − 1,500)
𝑝 =𝑝, + 𝑁 (Equation C- 32)
6500
As the lane-by-lane distribution of the off-ramp flow is known, the number of lane change maneuvers,
SBLC, can then be estimated. For Regime 3 cases (one blocked lane), the number of lane changes is
obtained as follows:
The equation adds the number of through vehicles in lane 1 that move to lane 2 to avoid the queue and
the number of exiting vehicles in the remaining lanes that adjust their position to join the back of the queue,
multiplied by the necessary number of lane changes. Figure C-28 illustrates an example of the proposed
equation applied to a 4-lane segment.
Figure C-28 – Illustration of lane change maneuvers within the queue influence area in a 4-lane
segment with Regime 3
Figure C-29 illustrates an example of the proposed equation applied to a 4-lane segment.
Figure C-29 – Illustration of lane change maneuvers within the Queue Influence Area in a 4-lane
segment with Regime 4
277
The Oversaturated Segment Evaluation procedure computes the Mainline Input (MI) for each node, in
every time step. It is defined as the maximum flow desiring to enter the subject node during the current
time step.
An adjustment is necessary when the subject node is operating in a two-pipe regime, as the blocked and
unblocked portions will be subject to different input demands. Since exiting and through drivers segregate
when approaching a queue, the mainline input demand in the blocked side consists of the off-ramp demand,
while the remaining demand will move to the unblocked portion.
When node i operates in a two-pipe regime, the Mainline Input (MI) parameter is split into two
components: MIUB, representing the mainline input in the unblocked lanes, and MIBL, representing the
mainline input joining the back of the queue. These parameters are computed as follows:
𝑂𝑁𝑅𝑂(𝑖, 𝑡, 𝑝)
𝑅𝑀(𝑖, 𝑡, 𝑝)
⎧ 𝑂𝑁𝑅𝐶(𝑖, 𝑡, 𝑝)
⎪
𝑀𝐹(𝑖 + 1, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪ ⎧𝑚𝑖𝑛 − 𝑀𝐼(𝑖, 𝑡, 𝑝)
⎪ 𝑀𝑂3(𝑖, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪
= 𝑚𝑖𝑛( ⎪ 𝑆𝐶(𝑖, 𝑡, 𝑝) (Equation C- 37)
⎨𝑚𝑎𝑥( 𝑚𝑖𝑛 𝑀𝐹(𝑖 + 1, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝) /2𝑁(𝑖, 𝑝)
⎪ ⎨ 𝑀𝑂3(𝑖, 𝑡 − 1, 𝑝) + 𝑂𝑁𝑅𝐹(𝑖, 𝑡 − 1, 𝑝)
⎪ ⎪
⎪ ⎪ 𝑹𝑭(𝑶𝑭𝑹𝑵𝑬𝑿𝑻(𝒊), 𝒕, 𝒑))
⎩ ⎩ 𝟐 × 𝑵𝑸(𝑶𝑭𝑹𝑵𝑬𝑿𝑻(𝒊))
278
Figure C-30 – Impact of a queue spillback on the discharge capacity of an upstream on-ramp
If one or more lanes are blocked due to a downstream off-ramp bottleneck, the throughput in Lane 1 will
be equal to the maximum exit throughput in the congested off-ramp if the site operates in Regime 3, or 50%
of the maximum exit throughput in the off-ramp, if it operates in Regime 4. It is assumed that the on-ramp
and the flow arriving from the upstream on Lane 1 contribute equally to the downstream Lane 1 flow, and
thus the on-ramp maximum output, in this case, is assumed to be half of the downstream throughput in
Lane 1.
Figure C-31 – Illustration of different density values within one diverge segment
If there are no spillback effects, the segment operates with a uniform density. In this case, the constraints
for the unblocked and blocked portions (MO2UB and MO2BL, respectively) are calculated proportionately
to the number of unblocked and blocked lanes:
279
(1 − 𝑁𝑄(𝑖))
𝑀𝑂2𝑈𝐵(𝑖, 𝑡, 𝑝) = 𝑀𝑂2(𝑖, 𝑡, 𝑝) × (Equation C- 38)
𝑁(𝑖)
𝑁𝑄(𝑖)
𝑀𝑂2𝐵𝐿(𝑖, 𝑡, 𝑝) = 𝑀𝑂2(𝑖, 𝑡, 𝑝) × (Equation C- 39)
𝑁(𝑖)
If node i operates under Increased Turbulence (node is in the Queue Influence Area), the unblocked
portion of segment i will operate similar to a regular segment. Therefore, the component MO2UB is equal
to MO2 but proportional to the number of lanes in the unblocked portion:
(1 − 𝑁𝑄(𝑖))
𝑀𝑂2𝑈𝐵(𝑖, 𝑡, 𝑝) = 𝑀𝑂2(𝑖, 𝑡, 𝑝) × (Equation C- 40)
𝑁(𝑖)
For the blocked portion of segment i, the parameter is calculated as equal to MO2 proportional to the
number of lanes in the blocked portion plus an additional number of vehicles due to the presence of a partial
queue. This additional number of vehicles is obtained by the bold terms in the following equation, which
takes into account the difference between the queue spillback density (RKQ) and the segment queue density
(KQ), multiplied by the queue length:
𝑁𝑄(𝑖) 𝑺𝑩𝑳𝑸(𝒊, 𝒕 − 𝟏, 𝒑)
𝑀𝑂2𝐵𝐿(𝑖, 𝑡, 𝑝) = 𝑀𝑂2(𝑖, 𝑡, 𝑝) × + × 𝑵𝑸(𝒊, 𝒕 − 𝟏, 𝒑)
𝑁(𝑖) 𝟓𝟐𝟖𝟎 (Equation C- 41)
× [𝑹𝑲𝑸(𝑶𝑭𝑹𝑵𝑬𝑿𝑻(𝒊), 𝒕 − 𝟏, 𝒑) − 𝑲𝑸(𝒊 − 𝟏, 𝒕 − 𝟏, 𝒑)]
If node i experiences lane blockage, the values of queue density must be computed for both the unblocked
(KQUB) and blocked (KQBL) portions of segment i. For the unblocked portion, the queue density KQUB
is calculated similarly to Equation 25-10, but the inputs for segment flow (SF) and segment capacity (SC)
are replaced by their equivalent parameters SFUB and SCEQ:
The queue density for the blocked portion is computed as equal to the ramp queue density:
With the queue density values for both the blocked and unblocked portions known, the MO2 components
MO2BL and MO2UB can be computed:
280
For MFBL, the maximum allowed flow is equal to the flow allowed to enter the nearest downstream off-
ramp RF, as presented in the following equation:
𝑇
𝑆𝐹(𝑖, 𝑝) = 𝑆𝐹(𝑖, 𝑡, 𝑝) (Equation C- 51)
𝑆
281
The additional density in the queued lanes is obtained by aggregating the additional number of
vehicles ΔNV(i,t,p) in the off-ramp queue:
1
∆𝐾(𝑖, 𝑝) = ∆𝑁𝑉(𝑖, 𝑡, 𝑝) (Equation C- 53)
𝑆×𝑁
Similar to the mainline, the flow in the ramp roadway is also aggregated:
𝑇
𝑅𝐹(𝑖, 𝑝, 𝑘) = 𝑅𝐹(𝑖, 𝑡, 𝑝, 𝑘) (Equation C- 54)
𝑆
The aggregated density at the ramp is calculated as the average of the number of vehicles inside the
ramp along the time period:
1
𝑅𝐾(𝑖, 𝑝, 𝑘) = 𝑅𝑁𝑉(𝑖, 𝑡, 𝑝, 𝑘) (Equation C- 55)
𝑆
Finally, the speed at the ramp for a time period p is obtained by dividing the total ramp flow in the
time period by its average density:
𝑅𝐹(𝑖, 𝑝, 𝑘)
𝑆𝑅(𝑖, 𝑝, 𝑘) = (Equation C- 56)
𝑅𝐾(𝑖, 𝑝, 𝑘)
282
Figure C-32 – Study site for freeway-to-freeway queue spillback check, Miami, FL
This freeway-to-freeway connector is modeled as two separate freeway facilities. The upstream freeway
(Facility 1: I-75) is modeled as a diverge section that is connected to the downstream freeway (Facility 2:
SR-826). The system’s detailed geometry is shown in Figure C-33.
283
Figure C-33 – Individual freeway facilities: (a) I-75 SB and (b) SR-826 SB
Input data
Traffic demands for the freeway facilities and ramps are provided in Table C-6 in 15-minute time periods.
284
Spillback check
The analysis of SR-826 using the Freeway Facilities Oversaturated Segment Evaluation provides the
expected on-ramp queue for every time period. The first check compares the off-ramp demand to the ramp
roadway capacity, as shown in Table C-9. The ramp queue starts to develop during time period 2. At the
end of this time period, a ramp queue length of 1188 ft is expected, yielding a queue storage ratio of 0.33.
Therefore, spillback is not expected during time period 2. During time period 3 a ramp queue length of
5160 ft is expected with a queue storage ratio of 1.41. Therefore, spillback will occur during time period 3.
285
Table C-9. Estimation of queue length and storage ratio at the SR-826 on-ramp
Number of
Total number Average Queue Ramp Queue
queued
of queued vehicle length length storage
Time vehicles in Spillback
vehicles spacing (ft) (ft) (ft) ratio
period each lane occurs?
[D] = [F] =
[A] [B] = [A]/2 [C] [E]
[B]*[C] [D]/[E]
1 0 0 - 0 0.00 No
2 38.3 19.15 62 1188 0.33 No
3588
3 159.1 79.55 65 5160 1.44 Yes
4 0 0 - 0 0.00 Yes
Spillback analysis
Since spillback is expected to occur, the methodology of this chapter (Figure C-4 through Figure C-7) is
applied to evaluate its impacts on I-75 SB. The application of the methodology for each time period is
presented below.
Time period 1
No oversaturated conditions occur, therefore no additional calculations are needed for this time period.
Time period 2
During time period 2, the downstream merge segment operates at LOS F and the on-ramp capacity is
expected to be reduced.
Step 1 - Calculate background density for unblocked lanes on each segment in the case of queue
spillback
The diverge segment at I-75 has 5 lanes and Regime 4 (two blocked lanes) is expected. Therefore, when
queue spillback occurs this segment operates with two blocked lanes (lanes 1 and 2) and three unblocked
lanes (lanes 3 through 5).
The capacity per lane at the diverge segment SC(3) is 2,350 pc/h/ln or 11750 pc/h. For the time step level
analysis, the capacity is converted to 48.95 passenger cars per time step (ts), Therefore, the capacity for the
unblocked portion of the segment is given by (Equation C- 1):
The capacity adjustment factor CAFBL is obtained from Table C-1. For a segment with 5 directional lanes
and 2 blocked lanes, an adjustment factor CAFBL = 0.67 is applied. Therefore, the equivalent capacity of
the unblocked portion is given by:
286
The unblocked background density KBUB is calculated next. For time period 2, an expected demand of
4165.8 pc/h for the mainline is used in the calculations. The KBUB parameter of the unblocked lanes is
computed as the density of a 3-lane basic segment with a capacity SCEQ = 7872 pc/h:
287
3588
𝑅𝑁𝑉(3,0,2,1) = 37.4 × × 2 = 50.8 𝑝𝑐
5280
Figure C-35 – Queued vehicles and total number of vehicles in the ramp – time period 2
The capacity of the ramp roadway for a 2-lane ramp with FFS = 55mph, is equal to 4,400 pc/h or 18.3
pc/ts. Therefore, the capacity of the ramp roadway is not a constraint to ramp flow. The other potential
capacity constraint RSTG is calculated through (Equation C-14):
The constraint RSTG is dependent on the number of vehicles in the ramp RNV, which increases
progressively as the queue grows along the ramp. Figure C-36 compares the decreasing value of RSTG with
the ramp input RI during time period 2. At the end of the time period, the capacity is still greater than
demand, therefore no spillback occurs at the end of this time period as predicted by the queue spillback
check previously described.
288
Since spillback does not occur, no additional calculations for the mainline are required.
1
𝑅𝐾(𝑖, 𝑝, 𝑘) = × 𝑅𝑁𝑉(𝑖, 𝑡, 𝑝, 𝑘) = 71.6 pc/mi/ln
60
𝑅𝐹(𝑖, 𝑝, 𝑘) 1679.5
𝑆𝑅(𝑖, 𝑝, 𝑘) = = = 31.9mi/h
𝑅𝐾(𝑖, 𝑝, 𝑘) 71.6
Time period 3
The same steps are repeated for time period 3. The ramp analysis is summarized in Figure C-37. For this
time period, the ramp demand is 15.4 pc/ts, while the merge capacity is 13.9 pc/ts. Since demand is greater
than capacity, the number of vehicles increases gradually, causing the capacity constraint RSTG to decrease
each time step. At time step 14, the value of RSTG becomes equal to the merge capacity (13.9 pc/ts), which
implies that the ramp has reached jam density and the maximum flow that can enter the ramp is equal to
the flow that departs the ramp. Therefore, queue spillback into the mainline starts at time step 15.
289
After the onset of queue spillback, the number of unserved vehicles at the exit is computed every time
step through the parameter OFRUV(i,t,p). Then, the expected length of the mainline queue OFRLQ(i,t,p)
is computed based on the number of unserved vehicles and the ramp queue density RKQ, as shown in
(Equation C- 25):
𝑂𝐹𝑅𝑈𝑉(𝑖, 𝑡, 𝑝)
𝑂𝐹𝑅𝐿𝑄(𝑖, 𝑡, 𝑝) =
𝑅𝐾𝑄 (𝑖, 𝑡, 𝑝)
Figure C-38 illustrates the expected spillback queue length during time period 3.
290
The parameter OFRLQ represents the length of the queue if all unserved vehicles were queued in a single
line. Given the segment geometry (Figure C-39), the operating regimes and flow modes can be obtained as
a function of OFRLQ:
• 0 < OFRLQ ≤ 1,400 ft: Regime 1
• 1400 ft < OFRLQ ≤ 3000 ft: Regime 4, with increased turbulence
• 3000 ft < OFRLQ: Regime 4, with lane blockage (queue extends upstream beyond the diverge)
As previously shown in Figure C-38, the maximum queue length OFRLQ at time period 3
is equal to 4696 ft. Since queues develop along lanes 1 and 2, at the end of time period 3 the
back of queue will be located 848ft upstream of the boundary of segments 2 and 3. The length
of the queue influence area (QIA) is 1060 ft, and when it is added to the back of the queue it
does not reach the upstream node of segment 2. Therefore, segment 2 capacity is not affected by
the turbulence area upstream of the queue.
Figure C-40 – Back of queue length, including QIA, at the end of time period 3
291
1
𝑅𝐾(𝑖, 𝑝, 𝑘) = × 𝑅𝑁𝑉(𝑖, 𝑡, 𝑝, 𝑘) = 108.4 pc/mi/ln
60
𝑅𝐹(𝑖, 𝑝, 𝑘) 1707
𝑆𝑅(𝑖, 𝑝, 𝑘) = = = 21.5 mi/h
𝑅𝐾(𝑖, 𝑝, 𝑘) 108.4
For the freeway facility, performance measures are computed for the blocked and unblocked portions of
each segment.
Similar to the ramp, the flow through the blocked portion is aggregated for this time period:
The average density is obtained as the sum of two separate components. The average number of vehicles
in the blocked portion of the segment is computed as:
1
𝐾𝐵𝐿(𝑖, 𝑝) = × 𝑁𝑉(𝑖, 𝑡, 𝑝) = 51 pc/mi/ln
60
Finally, the speed in the blocked lanes is obtained through the fundamental equation:
𝑆𝐹𝐵𝐿(𝑖, 𝑝) 3030
𝑆𝐵𝐿(𝑖, 𝑝) = = = 21.2 𝑚𝑖/ℎ
𝑁(𝑖, 𝑝) × 𝐾(𝑖, 𝑝) 2 × 70.1
The same process is repeated for the unblocked portion of the segment, except the ΔK component is
omitted as no queues occur in these lanes:
Time period 4
During time period 4, the congestion at the downstream facility (SR-826) dissipates, which allows the
ramp to discharge at the ramp roadway capacity (4,400 pc/h, or 18.33 pc/ts). Given the low ramp demand
during this time period, the queue can be cleared quickly (9 time steps, or 135s). After the 10th time step,
the freeway facility returns to undersaturated conditions.
292
APPENDIX D
Queue spillback into an arterial intersection may occur when the freeway merge segment has insufficient
capacity to process the ramp’s demand. Spillback may also occur in cases of ramp metering. This Appendix
presents the methodology for determining whether spillback will occur from an on-ramp into the upstream
intersection.
The methodology considers signalized intersections, two-way stop controlled intersections, all-way stop
controlled intersections, and roundabouts. The first step of the proposed procedure estimates the demand
approaching the on-ramp (determined based on the upstream intersection’s configuration), while the second
step estimates the capacity of the off-ramp. The existing methodology for oversaturated conditions along
freeway facilities (HCM Chapter 10) can estimate the resulting queue length, however, the user must input
the on-ramp demand flow rate.
The methodology framework for conducting this spillback check is presented in Figure D-1 and described
in more detail in the remainder of this appendix.
293
294
Effective red time (r). During this period, no vehicles are discharged and queue is building at a rate qr
(arrival flow rate during the effective red time).
1. Queue service time (gs). During this period, the queue previously built discharges at the saturation
flow rate (sadj). Therefore, the total number of discharged vehicles for movement i during the queue
service time is given by:
𝑁 ,, = 𝑠 𝑥𝑔 (Equation D-1)
Where:
NR,i,gs = total number of vehicles discharged for movement i during the queue service time
si = saturation flow rate (veh/h/ln), as defined in HCM Equation 19-8, of movement i
gs = green service time (s) = Qr/(s – qg)
2. Extension green time (ge). Corresponds to the remaining portion of the effective green when the
queue has been completely discharged. During that time vehicles are discharged at the same rate they
arrive at the intersection. The extension green time calculation is only applicable to undersaturated
approaches, as its duration will be zero when the effective green time is insufficient to clear the
queue.
The total number of discharged vehicles for movement i during the extension green time is given by:
𝑁 ,, = 𝑞 𝑥𝑔 (Equation D-2)
Where:
NR,i,ge = total number of vehicles discharged for movementi during the green extension time
qg = arrival flow rate during the effective green time = P q C/g (veh/s)
ge = green extension time (s) = g - gs
The total number of vehicles discharged for a protected movement i during a cycle is then given by:
295
𝑁 , = 𝑁 ,, 𝑁 ,, (Equation D-3)
Figure D-2. Discharging patterns for a typically protected movement using a queue accumulation
polygon (QAP)
The next step is to estimate the permitted movements’ flow into the ramp. Right-turning movements
often operate as permitted (RTOR), however their effect is not addressed in the HCM Chapter 31
(Signalized Intersections Supplemental). Calculations to account for permitted movement flow are
increasingly more complex, given that each phasing combination requires the development of a specific
queue polygon.
Finally, the total on-ramp demand during a single cycle can be calculated as:
𝑁 , = ∑ 𝑁 , ∑ 𝑁 , (Equation D-4)
Where:
NR,i = total number of vehicles discharged from each protected movement i
NR,k = total number of vehicles discharged from each permitted movement
The last step is to convert the total demand to the on-ramp into an hourly flow rate. For a pre-timed
signal, the aggregated on-ramp flow can be estimated by Equation D-5:
𝟑𝟔𝟎𝟎
𝒗𝑹 = 𝑵𝑹,𝒕𝒐𝒕𝒂𝒍 𝒙 (Equation D-5)
𝑪
In the case of a semi-actuated or fully actuated intersection with unknown cycle length, the procedure
described in section 2 of HCM Chapter 31 can be applied to estimate phase durations and cycle lengths.
These can then be aggregated into an hourly flow rate using the same procedure.
296
With very few adjustments, estimating the on-ramp throughput from this intersection type is a relatively
straightforward task.
The first step is to identify the movements that discharge to the on-ramp and their respective ranks
(priority orders). The proposed methodology on freeway-arterial interactions assumes that, for TWSC
interchanges, the arterial will always be the major street. Figure D-3 illustrates a typical TWSC interchange,
where movements discharging into the on-ramp are numbered according to their ranks, using the default
numbering of the HCM methodology (Chapter 20, Exhibit 20-1).
Similarly to signalized intersections, there are three movements turning into the ramp, and their
respective flows are discussed below:
1. Rank 1 movement (right turn from the major street). This movement is considered unimpeded,
experiencing zero delay. The only physical constraint able to limit the throughput of this movement
is its saturation flow rate if demand is very high. Therefore, the maximum throughput λRT (veh/h) for
this right-turn movement is given by:
Where:
λRT = departure rate from major street right turn into the on-ramp (veh/h)
vRT = demand flow rate for the major street right turn
sRT = saturation flow rate for a right-turn movement (veh/h)
2. Rank 2 movement (left turn from the major street). The maximum throughput for this movement
is limited by its potential capacity (cp,j), as defined in HCM Equation 20-36. Therefore, the maximum
throughput λLT (veh/h) for this left-turn movement is given by:
297
Where:
λLT = departure rate from major street left turn into the on-ramp (veh/h)
vLT = demand flow rate for the major street left turn
cp,j = potential capacity for the major street left turn (veh/h)
3. Rank 3 movement (through from the minor street). Similar to rank 2 movements, the maximum
throughput for this movement is limited by its potential capacity (cm,k), as defined in HCM Equation
20-47. Therefore, the maximum throughput λTh (veh/h) for this through movement is given by:
Where:
λTh = departure rate from the minor street through into the on-ramp (veh/h)
vTh = demand flow rate for the minor street through
cm,k = potential capacity for the minor street through (veh/h)
Finally, the total on-ramp demand flow rate vR can be estimated as follows:
𝑣 =𝜆 𝜆 𝜆 (Equation D-9)
The onramp demand flow rate can be obtained directly from the departure headways of the three
movements combined:
298
𝑣 = + + (Equation D-10)
, , ,
Where:
vR = on-ramp flow rate (veh/h)
hd,RT = departure headway for the major street right turn (s)
hd,LT = departure headway for the major street left turn (s)
hd,Th = departure headway for the minor street through(s)
Case D – Roundabouts
The current HCM methodology for roundabouts is based on the calculation of the potential capacities of
each approach, based on three main variables: the critical and the follow-up headways, and the circulating
flow (Equation 22-21 through Equation 22-23). Both critical and follow-up headway values can be obtained
from HCM Chapter 33.
The procedure to evaluate the occurrence of queue spillback into roundabouts is highly integrated to the
evaluation of the impacts of queue spillback, given the interdependence of entering flows and the capacities
at the roundabout. Therefore, the methodology for this case is discussed in Appendix E - On-Ramp Queue
Spillback Analysis.
299
300
This example is based on the configuration used in the HCM Chapter 34 (Ramp Terminals and
Alternative Intersections Supplemental), as shown in Figure D-5.
Figure D-5. Schematic of the study interchange for the example problem
The diamond interchange has two closely-spaced signalized intersections, spaced 500 ft from each other.
The on-ramp connecting Intersection 2 to the freeway is metered at a rate of 650 veh/h and is being
evaluated for queue spillback. The signal controller is set to pre-timed operation, and the phasing sequence
and timings are presented in Figure D-6.
Figure D-6. Phasing sequences and timing for the study interchange
301
There are two movements that discharge directly to the on-ramp: WBR (Φ16 – Phase I) and EBL (Φ5 –
Phase III). The WBR movement is a protected-only movement. Additional parameters used in the analysis
are presented in Table D-1.
Table D-1 - Input parameters for the two movements discharging to the on-ramp
For the spillback check procedures, a 15-minute aggregation is recommended. Though the intersection
throughput volumes are calculated for every individual movement and on a cycle-by-cycle basis, they must
be aggregated to a 15-minute time period. Given a cycle length of 160s, the analysis can consider 900/160
= 5.6 cycles.
The calculations to estimate on-ramp throughput for the WBR movement are presented in
Table D-3. For each cycle, 24 vehicles are discharged considering the sum of queue serving time and
green extension. Therefore, for a 15-minute period it is expected that a total of 24 x 5 = 120 vehicles enter
the onramp from the WBR movement:
302
Finally, the on-ramp demand (sum of the two contributing movements) is compared to the freeway
discharge capacity (given by the ramp metering rate), and any unserved vehicles in the cycle are stored for
the start of the next cycle. As shown in Table D-4, during the fourth cycle the ramp storage ratio exceeds
one, indicating that spillback into the intersection is expected to occur at that time.
303
APPENDIX E
This document describes the methodological modifications required to address the occurrence
of queue spillback from an on-ramp. The occurrence of queue spillback affects each type of
intersection differently. The methods outlined here address signalized intersections, two-way
stop-controlled (TWSC) intersections, all-way stop-controlled (AWSC) intersections, and
roundabouts.
Signalized Intersections
Figure E-1 presents the methodology for evaluating the performance of signalized intersections,
with proposed modifications to address impacts from an on-ramp queue spillback. New steps and
modified steps to the methodology are described in the following paragraphs.
304
305
The volume of vehicles that enters a freeway on-ramp is a function of the demands and
capacities of each individual intersection movements that discharge into the ramp. A typical
signalized intersection within a diamond interchange is shown in Figure E-2, with three
movements discharging into the on-ramp (SBL, EBT and NBR).
The total throughput from the intersection into the on-ramp λONR is the sum of the throughput
from each of the contributing movements:
𝜆 𝜆 𝜆 𝜆 (Equation E-1)
The throughput for each movement i is the minimum value of its demand and capacity:
Unsignalized movements, which are common for right-turn movements to the on-ramp, are
unrestricted. The capacity of these movements can be estimated as the saturation flow rate (HCM
Equation 19-8), with an adjustment factor for right turns fRT (Equation 19-13).
If all movements at the intersection are undersaturated (vi ≤ ci for every i), then Equation E-1
is simplified and the total on-ramp demand throughput λONR is:
𝜆 𝑣𝑖 (Equation E- 3)
𝑖
306
The procedure to obtain cmerge is presented in Figure E-3. The freeway facility must be analyzed
using the Freeway Facilities methodology (HCM Chapter 10) to evaluate whether the merging
capacity is constrained by oversaturated conditions in the mainline. If the freeway facility is
undersaturated (LOS A-E), the merging capacity cmerge takes the minimum value between the on-
ramp capacity and the ramp metering rate, if present.
If the freeway facility is oversaturated (LOS F), the Oversaturated Segment Evaluation
procedure described in Chapter 25 can provide the maximum on-ramp output ONRO, computed
at a time-step level (15 seconds). The merging capacity cmerge can then be computed by
aggregating the parameter ONRO to an hourly flow rate:
𝑇 (Equation E- 4)
𝑐 = 𝑂𝑁𝑅𝑂 𝑖, 𝑡, 𝑝
𝑆
Where:
ONRO (i,t,p) = maximum output flow rate that can enter the merge point from on-ramp i during
time step t in time interval p
T = number of time steps in 1 h (integer). T is set as a constant of 240 in the
computational engine, or equal to four times the value of S
S = number of computational time steps in an analysis period (integer). S is set as a
constant of 60 in the computational engine, corresponding to a 15-s interval and
allowing a minimum segment length of 300 ft.
t = time step index
307
308
for the SBL movement. At the end of the SBL green, the vertical difference between the projected
number of vehicles (dashed line) and the actual number of vehicles inside the on-ramp represent
the number of unserved vehicles for the SBL approach. This additional queue can be considered
in a multiperiod analysis for the signalized intersection or interchange, using the methods
provided in Chapter 23 – Ramp Terminals and Alternative Intersections.
Figure E-4. Sample intersection for calculation of a QAP for the on-ramp
The slope of the red line connecting the number of vehicles in the end and start of the green
represent the reduced capacity of the SBL movement due to queue spillback. For the remainder
of the cycle, the NBR movement discharges at a constant rate into the on-ramp, as this is an
unsignalized movement. Given that the discharge capacity cmerge is greater than the on-ramp
demand λNBR, the vehicles along the on-ramp are discharged to the freeway until the on-ramp is
cleared. Therefore, the NBR movement does not have its capacity affected by queue spillback.
This procedure can be applied for both pretimed and actuated control types, since the core
methodology can address both controller types. If the signal is actuated, the average phase
duration lengths are applied, as obtained in Step 6.
𝑁 𝑔 −𝑁 0
𝑐 , −𝑐 − (Equation E-5)
𝑔
309
Where:
N(g1) = number of queued vehicles along the on-ramp at t = g1 (end of green for phase 1);
N(0) = number of queued vehicles along the on-ramp at t = 0 (start of the cycle);
g1 = effective green time for phase 1
The adjusted capacity of the SBL movement cSBL,SP is then computed as:
𝑁 𝑔 −𝑁 0
𝑐 , =𝑐 (Equation E-6)
𝑔
If the queue develops and fully discharges during every cycle, then subsequent cycles will have
the same discharge. However, if there are residual queues at the on-ramp by the end of the cycle,
the QAP must then be plotted again for the following cycle with an initial queue equal to the
number of queued vehicles in the end of the present cycle. This process must be then repeated for
a number of cycles N= 900/C, sufficient to analyze the entire 15-minute period. The adjusted
capacity for each movement is estimated as the average of the discharge rates during each cycle.
310
street through or right turning traffic, but they are not required to stop in the absence of oncoming
traffic.
The methodologies for evaluating the operations of TWSC intersections are based on gap
acceptance theory. Drivers from lower priority movements must select a suitable gap in order to
proceed through the intersection. According to previous research (Aakre & Aakre, 2017), during
oversaturated conditions and when queue spillback occurs drivers show cooperative behavior,
with higher priority vehicles often yielding to those with lower priority. The microsimulation
package AIMSUN, which was used to simulate the study sites for this project, includes the feature
Turn Cooperation Model to simulate this behavior, as illustrated in Figure E-6. In such cases, the
gap acceptance model is no longer valid and a new approach must be used to evaluate the
intersection performance.
When queue spillback occurs at a TWSC intersection the maximum throughput to the on-ramp
(exit capacity) is constrained by the discharge capacity of the freeway merge. It is assumed that
during oversaturated conditions the intersection movements that discharge to the on-ramp share
the exit capacity proportionately to their demands.
Figure E-7 presents the methodology for evaluating the performance of TWSC intersections,
with proposed modifications to address impacts from an on-ramp queue spillback. New steps and
modified steps to the methodology are described in the following paragraphs.
311
312
intersection into the on-ramp λONR is the sum of the throughput from each of the contributing
movements. For each movement i discharging into the on-ramp, the throughput is the minimum
value of its demand and its movement capacity:
Where:
vi = demand flow rate for movement i
cm,j = movement capacity for movement i (Equations 20-36, 20-37 and 20-40).
Step 9B. Obtain merging capacity using the freeway facilities methodology
This step computes the merging capacity into the freeway cmerge. The procedure described in
Step 7B of the queue spillback analysis for signalized intersections (Figure E-1) is applied.
From this relationship shown in Figure E-8 the spillback time TSB is defined as the amount of
time within a time period when spillback is active:
𝐿 −𝑁 0
𝑇 =𝑇− (Equation E-9)
𝜆 −𝑐
Where:
313
Estimating the spillback time TSB is critical to the methodology, as the aggregated calculations
of capacity for each movement depend on the amount of time that the intersection operates under
queue spillback.
Finally, the adjusted capacity of each affected movement ci,EQ is obtained as a function of the
amount of time within the time period when spillback was present. The adjusted capacity
considers the proportion of time there is blockage during queue spillback and consists of the
aggregation, at a time period level, of movement capacities cm,i (which is observed during
undersaturated conditions) and spillback capacities cSB,i,(which is observed during oversaturated
conditions):
When queue spillback lasts for the entire time period T (for example, in a multi-period
analysis), the spillback time TSB is equal to T, and the capacity of each movement i is obtained as
the capacity during spillback and (Equation E- 3) becomes:
𝑐 , =𝑐 , (Equation E-12)
314
The average control delay is obtained using Equation 20-64 replacing the movement capacity
cm,i by the adjusted capacity cEQ,i:
3600 𝜆 (Equation
⎡ × ⎤
3600 𝑐 , 𝑐 E-13)
⎢ 𝑣 𝑣 , ⎥
𝑑 = + 900𝑇 ⎢ −1+ −1 + ⎥+5
𝑐 , 𝑐 , 𝑐 , 450𝑇
⎢ ⎥
⎣ ⎦
315
The methodology to evaluate queue spillback into AWSC intersections follows the approach
developed for TWSC intersections. As shown in Figure E-9, after the capacities of individual
movements during undersaturated conditions are computed (Step 12), the process described for
TWSC intersections is performed by new steps 13A through D.
The only step in the methodology that differs from the TWSC (13D) is described below.
The AWSC methodology calculates the delay for each approach based on its departure
headway instead of capacity. The estimated spillback capacity (cSB,i) is converted to a spillback
headway hSB through the following equation:
316
317
318
Table E-1. Required data and potential data sources – roundabout spillback evaluation
Step 13 – Compute the maximum throughput into the on-ramp for every movement
The maximum throughput into the on-ramp per movement is calculated using the roundabout
priority order in a counterclockwise order starting from the most upstream approach from the on-
ramp exit leg. The Rank 1 approach (Figure E-11) is the one whose flow has the highest priority,
given it enters the circulating stream upstream of all other approaches). The next priority
movement is the Rank 2 approach, and the last is the Rank 3 approach.
Next, the methodology calculates the capacity of the roundabout’s exit lane into the on-ramp.
Previous research (Robinson et al, 2006; Rodegerts & Blackwelder, 2005) suggests that the
capacity of an exit lane, accounting for pedestrian and bicycle traffic in a typical urban area, is in
the range of 1,200 to 1,300 vehicles per hour. Starting from the approach with Rank 1, and
proceeding counterclockwise with the rest of the approaches, the capacity for each approach is
used to determine the maximum throughput for every movement discharging to the on-ramp.
Rank 1 – SB approach. The Rank 1 approach has priority over the other movements
connecting to the on-ramp because it enters the circulating stream first. Also, because the on-ramp
does not have an approach into the roundabout, the Rank 1 movement is most often unopposed
by the circulating stream (except for occasional U-turns along the arterial). Therefore, the
maximum throughput λSB-ONR (veh/h) for this left-turn movement is limited by its own lane
capacity (cSB) and the maximum throughput to the on-ramp, and it is given by:
3,600 (Equation E-15)
𝜆 = min 𝑣 ,𝑐 × 𝑝 ,
ℎ
Where:
λSB- rate from the SB approach into the on-ramp (veh/h)
vSB- flow rate for the SB approach into the on-ramp (veh/h)
319
Rank 2 – EB approach. The maximum throughput for this Rank 2 movement is limited by its
own lane capacity (cEB), as defined in HCM Equations 22-21 through 22-23, and the maximum
throughput after considering the departure rate of the upstream Leg 1. Therefore, the maximum
throughput λEB-ONR (veh/h) for this movement is given by:
Where:
λEB- rate from the EB approach into the on-ramp (veh/h)
vEB- flow rate for the EB approach into the on-ramp (veh/h)
cEB = lane capacity for EB approach (veh/h) (HCM Equation 22-21)
pEB- of demand from EB approach into the on-ramp
Rank 3 – NB approach. Similar to rank 2 movements, the maximum throughput for the NBR
(i.e., NB-ONR) movement is limited by its own lane capacity (cNB), as defined in HCM Equation
22-21 through Equation 22-23, and the maximum throughput to the on-ramp after considering
departure rates from the upstream approaches. Therefore, the maximum throughput (λNB-ONR) for
this right-turn movement is given by:
Where:
λNB- rate from the NB approach into the on-ramp (veh/h)
vNB- flow rate for the NB approach into the on-ramp (veh/h)
cNB = lane capacity for NB approach (veh/h) (HCM Equation 22-21)
pNB- of demand from NB approach into the on-ramp
The total on-ramp demand flow rate can be similarly calculated if there are additional
approaches to the roundabout.
𝜆 =𝜆 + 𝜆 + 𝜆 (Equation E-18)
Step 15 – Compute on-ramp merging capacity and compare to the maximum throughput to
the on-ramp
The calculation of the on-ramp merging capacity follows the exact same procedure used in Step
7B of the methodology developed for queue spillback into Signalized Intersections (Figure E-1).
The maximum number of vehicles that can merge into the on-ramp cmerge (estimated using
Equation 25-18) is compared to the maximum throughput from the roundabout to the on-ramp,
𝜆 . If cmerge ≤ λONR, then spillback is not expected to occur, and no adjustments are necessary in
the procedure. If cmerge > λONR, queues will develop along the on-ramp, and spillback may occur if
320
the queue storage is insufficient. The analyst must then proceed to Step 17 to evaluate the on-
ramp Queue Storage Ratio to evaluate whether spillback will occur.
Step 17 – Determine the on-ramp storage ratio and queue spillback length
With the exit flow rate into the on-ramp (λ ), the expected queue length QSP along the on-
ramp during a 15-minute period analysis is:
𝜆 −𝑐 (Equation E-19)
𝑄 =
4
If a multi-period analysis is performed, the queue length for the current time period p must be
added to the queue length obtained from the previous time period:
𝜆 , −𝑐 , (Equation E-20)
𝑄 , =𝑄 , +
4
The on-ramp storage ratio is calculated by dividing the available on-ramp storage LR (ft) by the
average vehicle spacing , Lh (Equation 31-155):
𝐿 ×𝑄 (Equation E-21)
𝑅 =
𝐿
If the on-ramp storage ratio (R ) is greater than 1, queues will form along each approach due
to spillback. The value of RQ corresponds to the specific analysis period. If congestion is
expected, but RQ < 1 for a single analysis period, multi-period analysis may have to be conducted.
𝑄 =𝑄 −𝐿 ×𝑄 (Equation E-22)
These queues are assumed to be distributed proportional to the demand flow rates to the on-
ramp per approach and added to the 95th percentile queues estimated for the undersaturated
conditions (Equation 22-20):
𝜆 (Equation E-23)
𝑄 , =𝑄 × +𝑄 ,
𝜆
𝜆 (Equation E-24)
𝑄 , =𝑄 × +𝑄 ,
𝜆
𝜆 (Equation E-25)
𝑄 , =𝑄 × +𝑄 ,
𝜆
Where:
Qsp,SB , Qsp,EB , Qsp,NB = queue due to the on-ramp spillback on SB, EB and NB approaches,
respectively (veh)
λSB-ONR, λEB-ONR, λNB- throughput for SB, EB, and NB approaches into the on-
ramp, respectively
Q95,SB , Q95,EB , Q95,NB = 95th percentile queue on SB, EB, and NB approaches, respectively
(veh))
321
⎡ 3600 𝜆 ⎤
×
3600 ⎢𝜆 𝜆 𝑐 𝑐 ⎥
𝑑 = + 900𝑇 ⎢ −1+ −1 + ⎥
𝑐 𝑐 𝑐 450𝑇
⎢ ⎥ (Equation
⎣ ⎦ E-26)
𝜆
+ 5 × 𝑚𝑖𝑛 ,1
𝑐
Where
cmerge = merging capacity of the on-ramp (veh/h)
λ flow rate into the on-ramp (veh/h)
t = time period (h) (T = 0.25 h for a 15-min analysis)
322
The freeway facility (I-10 EB) is modeled according to the Freeway Facilities methodology
(Chapter 10), while the ramp terminal is modeled according to its respective intersection
methodology. First a check is performed to confirm the occurrence of queue spillback. Next, the
respective spillback analysis is applied to evaluate the impacts of queue spillback in the capacity
of each movement at the intersection. With the estimated reduced capacities at the intersection,
the control delay values considering queue spillback are computed and compared to the delay
values without consideration of queue spillback.
323
Input data
Signalized Intersection
The geometry of the intersection connected to the I-10 EB on-ramp (I-10 EB) is shown in
Figure E-13. There are three movements leading into the on-ramp:
• NBR: One channelized, unsignalized right-turn lane;
• SBL: One exclusive left turn lane with a protected phase; and
• EBT: One through lane.
The phasing sequence of the subject intersection is presented in Figure E-14. The north-south
direction corresponds to the major street, while the minor streets correspond to the freeway off-
ramp and on-ramp. The intersection has a leading left turn phase with a protected left turn
movement (SBL).
The demand volumes for each time period are presented in Table E-2. Additional input data are
summarized in Table E-3.
324
325
The geometric features of the freeway facility are summarized in Table E-4.
326
throughput of the unsignalized turning movement is also assumed to be equal to its saturation
flow rate. Therefore:
𝑠 , =𝑠 , ×𝑓 ×𝑓
where
sNBR,FF = saturation flow rate of NBR movement at free-flow conditions (veh/h/ln)
s0,NBR = base saturation flow rate (1,900 pc/h/ln)
fRT = adjustment factor for right-turn vehicle presence in a lane group
fHVg = adjustment factor for heavy vehicles and grade
The adjustment factor for right-turn vehicle presence is computed using Equation 19-13:
1 1
𝑓 = =
𝐸 1.18
where
ET = equivalent number of through cars for a protected right-turning vehicle (1.18)
The adjustment factor for heavy vehicles and grade is computed using Equation 19-10:
1.18
0.961
Since there are conflicting movements discharging into the on-ramp (for example, a protected
left-turn), the NBR capacity is constrained as drivers yield to the higher priority movement. The
estimated discharge flow rate for the NBR movement with a conflicting protected flow vprot can
be obtained by the following equation, based on HCM equation 31-100:
/ ,
𝑣 𝑒
𝑠 = / ,
1−𝑒
Where:
sp = saturation flow rate of a permitted movement (veh/h/ln)
v0 = opposing demand flow rate (veh/h)
tcg = critical headway = 4.5 (s)
tfh = follow-up headway = 2.5 (s)
The computation of the permitted saturation flow rates must take into consideration that the
conflicting phase may have two distinct flow rates on signalized intersection operation, as
discussed in HCM Chapter 31 (Signalized Intersections Supplemental):
327
• During the queue service time (gs) portion of the conflicting phase green, the opposing
movement flow rate is equal to its saturation flow rate;
• During the green extension time (ge), the opposing movement flow rate is equal to its
arrival flow rate during the effective green (qg);
Table E-5 illustrates the calculation of the NBR potential capacity for a single cycle during
time period 1. For each active phase, the procedure identifies the respective conflicting flow to
the on-ramp along with its duration and flow rate. The NBR saturation flow rate is then computed
using HCM Equation 31-100. The last column computes the maximum number of vehicles that
can be discharged during each phase as the product of the NBR saturation flow rate and the phase
duration. Transition times between consecutive phases are also taken into consideration assuming
that they have no conflicting flow rate to the on-ramp.
Table E-5. Calculation of NBR potential capacity for a single cycle – Time Period 2
As shown, for a 120s cycle the capacity of the unsignalized NBR movement is 34.8 vehicles.
Aggregated to an hourly flow rate:
3600
𝑐 = 34.8 × = 1045 𝑣𝑒ℎ/ℎ
120
Because of the actuated control operation, the discharging rates to the on-ramp are different
during each time period since they depend on effective green times and flow profiles. Therefore,
this procedure must be repeated for every time period to compute the capacity of the NBR
unsignalized movement cNBR (Table E-6).
328
Table E-7 summarizes the calculations for this step. During time period 3, the SBL movement
operates at demand over capacity (v/c = 1.56), therefore its throughput to the ramp is constrained
by its capacity value (685 veh/h). For all other movements and time periods the throughput to the
on-ramp is equal to its demand because v/c < 1.
Table E-7. Calculation of the on-ramp demand (vR) based on the intersection operation.
Time Movements
Parameter
Period EBT NBR SBL
Demand (veh/h) 8 315 652
v/c 0.064 - 0.96
1 c (veh/h) 125 1213 677
min (v, c) 8 315 652
Merge demand vR (veh/h) 975
Demand (veh/h) 96 521 586
v/c 0.768 - 0.93
2 c (veh/h) 125 1045 630
min (v, c) 96 521 586
Merge demand vR (veh/h) 1203
Demand (veh/h) 96 630 1071
v/c 0.77 - 1.56
3 c (veh/h) 125 978 685
min (v, c) 96 630 685
Merge demand vR (veh/h) 1411
Demand (veh/h) 24 80 463
v/c 0.39 - 0.62
4 c (veh/h) 62 1182 746
min (v, c) 24 80 463
Merge demand vR (veh/h) 567
The calculated on-ramp demand is then provided as input into the freeway facility analysis
(Table E-8). As shown, the ramp flow rates for the merge segment (segment 5) are obtained from
Table E-7, and highlighted in bold.
The results of the freeway facility analysis are provided in Table E-9. Oversaturated conditions
occur during time periods 2 and 3, therefore queueing may occur along the on-ramp.
329
Table E-9. Performance measures for the freeway facility (I-10 EB)
The next step will estimate the on-ramp queue length compared to the available queue storage
length to determine whether spillback is expected to occur. Table E-10 shows the expected on-
ramp queues from the freeway facility analysis. For each time period, the ramp storage ratio (RQ)
is computed by dividing the ramp queue by the available storage length (924 ft). During time
period 2, a queue is expected on the ramp, but it is not long enough to cause queue spillback (RQ
< 1). During time period 3, however, the on-ramp is expected to have RQ = 2.31, which indicates
that spillback will occur at the intersection during this time period.
Since spillback will occur for at least one time period, the impacts on the operation of the
signalized intersection must be evaluated. The next section illustrates the application of the
methodology to evaluate spillback effects at a signalized intersection.
Time Period 2
The procedure to evaluate queue spillback into intersections is applied for time period 2, even
though spillback is not expected to occur during this time period. The application of the
methodology is presented for this time period to facilitate the understanding of the calculations.
330
Oversaturated Segment Evaluation procedure (HCM Chapter 25) computes the on-ramp queue
(ONRQ) and on-ramp capacity (ONRO) every 15 seconds. The merge capacity cmerge is then
obtained by aggregating the ONRO parameter into an hourly flow rate for each time period.
Figure E-16 shows the values of ONRQ and ONRO over the analysis period (60 minutes),
converted to hourly flow rates.
Figure E-16a compares the on-ramp capacity ONRO to the on-ramp demand. During the first
time period there are no oversaturated conditions along the freeway, thus the on-ramp capacity
ONRO equals 2,000 pc/h (corresponding to the ramp roadway capacity as provided by HCM
Exhibit 14-12), or 1,903 veh/h. During time periods 2 and 3, oversaturated conditions occur and
the on-ramp capacity drops to 5 pc per time step, corresponding to 1,142 veh/h. During the last
time period, the lower demand along the freeway allows the mainline queue to clear within 4 time
steps (60 seconds). Therefore, during the first 60 seconds the on-ramp capacity remains at 1,142
veh/h. From the fifth time step to the end of the time period, there is no congestion at the merge
and thus the on-ramp capacity is again 1,903 veh/h.
Figure E-16b provides the on-ramp queue as estimated by the Oversaturated Segment
Evaluation procedure. Since spillback is expected to occur, an adjustment to the Freeway
Facility Oversaturated Segment evaluation procedure is necessary to account for the
maximum ramp storage (35.5 vehicles). This value is the upper boundary of the on-ramp
queue length. At the end of time period 3, the predicted on-ramp queue length would be 82
vehicles if there were no storage constraints (black curve). The red curve represents the
adjusted queue profile for the on-ramp considering the maximum storage capacity. At the
start of time period 4, having an on-ramp queue of 35.5 vehicles instead of 82 results in a
shorter queue clearance time, with a slight positive impact on the freeway performance. In
other words, the intersection has a metering effect, which may improve operations along
the freeway.
Table E-11 compares the performance results of the freeway segments downstream of the
merge (see Figure E-15) with and without consideration of the maximum storage constraint.
331
Figure E-16. Freeway facility, segment 5 (merge) performance: (a) merge capacities and
(b) queue lengths
Table E-11. Freeway performance during time period 4 – with and without the queue
storage constraint
332
Step 7C – Plot queue accumulation polygon for the on-ramp and unsignalized movements
In this step, a queue accumulation polygon is plotted for the on-ramp as a function of all
protected and permitted movements entering the on-ramp, on a cycle-by-cycle basis. Since an
unsignalized movement (NBR) also discharges into the on-ramp, a queue accumulation polygon
must be developed for this movement as well. This is required to: (a) determine the discharge
pattern of the unsignalized movement throughout the cycle and (b) allow the estimation of control
delay for this movement.
Figure E-17 presents the queue accumulation profiles for (a) the on-ramp and (b) for the NBR
movement.
Figure E-17. Estimated queue lengths and merge capacities – time period 2
The cycle starts with a permitted left-turn movement (Φ1: SBL) discharging into the on-ramp
with a green time g1 = 43.9s, divided in a queue service time gs1 = 40.2s and a queue extension
time ge1 = 3.7s (as defined in HCM Chapter 31 – Signalized Intersections Supplemental). During
the green interval for SBL, the capacity of the NBR movement is constrained since drivers must
yield to the protected left-turn vehicles. The estimated saturation flow rate for the NBR movement
with a conflicting flow vSBL can be obtained by the following equation, based on HCM equation
31-100:
333
/ ,
𝜆 𝑒
𝑠 , = / ,
1−𝑒
Where:
sNBR,perm = saturation flow rate of the NBR movement (veh/h/ln)
λSBL = throughput of the opposing SBL movement(veh/h)
tcg = critical headway = 4.5 (s)
tfh = follow-up headway = 2.5 (s)
The saturation flow rates of the NBR movement during Φ1 are determined next. During the
SBL queue service time:
Where:
sSBL = saturation flow rate of the SBL movement (veh/h/ln)
sNBR,perm1 = saturation flow rate of the NBR movement during the SBL queue service time
(veh/h/ln)
The throughput for the NBR movement is obtained as the minimum of the demand and
saturation flow rate. Since the demand flow rate is greater than the saturation flow rate, a queue
will develop for the NBR movement:
Where:
λNBR,1 = throughput for the NBR movement during the SBL queue service time (veh/h/ln)
vNBR = demand flow rate of the NBR movement (veh/h)
During the SBL green extension time ge, the SBL throughput λSBL is equal to the arrival flow
rate during the effective green (qg,SBL, from Equation 19-32):
𝑣 𝐶
𝜆 =𝑞 , =𝑃 × ×
3600 𝑔
586 120
𝜆 = 0.08 × × = 0.0356 veh/s/ln = 128 veh/h/ln
3600 43.9
where
For this conflicting flow, therefore, the NBR saturation flow rate sNBR,perm2 is obtained using
Equation 31-100:
/ ,
𝜆 𝑒
𝑠 , = / ,
1−𝑒
× . / ,
128𝑒
𝑠 , = × . / ,
= 1282 𝑣𝑒ℎ/ℎ/𝑙𝑛
1−𝑒
334
Since a queue is present in the NBR movement, the throughput for the NBR movement is equal
to its saturation flow rate:
𝜆 , =𝑠 , = 1282 𝑣𝑒ℎ/ℎ
Where:
λNBR,2 = throughput for the NBR movement during the SBL green extension(veh/h/ln)
sNBR,perm2 = saturation flow rate of the NBR movement during the SBL green extension
time (veh/h/ln)
With the discharge patterns for the NBR determined, the queue profile in the on-ramp during
Φ1 can be determined. During the SBL queue service time (cycle time t = 0 to t = 40.2s), the
throughput to the on-ramp is given by:
Given the merge capacity cmerge = 1,142 veh/h for the current time period, the on-ramp queue
will grow at the following rate during the SBL queue service time:
Therefore, at the end of the SBL queue service time (t = 40.2s), the queue at the on-ramp
will be 0.244 x 40.2 = 9.8 vehicles (Figure E-17a).
This process is then repeated for all phases throughout the cycle. The results for a single cycle
(120 sec) are presented in Table E-12, where the maximum on-ramp queue occurs at t = 50.48s,
with 10.82 vehicles (t = 50.48s). The expected on-ramp queue at the end of the cycle is 2.02
vehicles. The remaining cycles within time period 2 show the same pattern, where the on-ramp
queue at the end of each cycle becomes the initial queue at the start of the next cycle.
Each row in Table E-12 describes a portion of the cycle, as follows:
• gs1: queue service time for SBL (Φ1), as previously discussed
• ge1: green extension time for SBL (Φ1). The NBR movement discharges at the permitted
saturation flow rate due to the queue that has developed during gs1, and the on-ramp queue
grows at a rate of 0.07 veh/s
• r1: effective red time for SBL (Φ1). There is no throughput from protected movements
and the NBR movement discharges freely at the saturation flow rate. The on-ramp queue
grows at a rate of 0.11 veh/s
• g2*: effective green for NBT (Φ2), with no throughput from protected movements. The
duration of 0.88s is calculated based on the queue service time of the NBR approach. The on-
ramp queue grows at a rate of 0.11 veh/s
• g2**: remaining effective green for NBT (Φ2). For this portion, no queue remains on
the NBR approach, therefore the NBR throughput is equal to its demand flow rate (vNBR). The
on-ramp queue discharges at a rate of 0.17 veh/s
• r2: effective red time for NBT (Φ2). There is no throughput from protected movements
and the NBR throughput is equal to its demand flow rate (vNBR). The on-ramp queue
discharges at a rate of 0.17 veh/s
• gs7: queue service time for EBT (Φ7). The EBT discharges into the on-ramp at the
saturation flow rate. The throughput of the NBR movement is restricted to the permitted
saturation flow rate, causing queues to develop in the NBR approach. The on-ramp queue
grows at a rate of 0.26 veh/s
• ge7*: green extension time for EBT (Φ7). The duration of 0.03s is calculated based on
the queue service time of the NBR approach. The NBR movement discharges at the permitted
saturation flow rate. The on-ramp queue grows at a rate of 0.08 veh/s
335
• ge7**: remaining extension time for EBT (Φ7). The EBT movement discharges at a rate
equal to its arrival flow rate during the effective green. For this portion, no queue remains on
the NBR approach, therefore the NBR throughput is equal to its demand flow rate (vNBR). The
on-ramp queue discharges at a rate of 0.15 veh/s
• r7: effective red time for EBT (Φ7). No throughput from protected movements and the
NBR throughput is equal to its demand flow rate (vNBR). The on-ramp queue discharges at a
rate of 0.17 veh/s.
Table E-12. Discharge flow rates into the on-ramp for each phase throughout the cycle –
time period 2
Protected
Permitted movement On-ramp analysis
movement
Active Duration
t (s) NBR λONR - On-ramp
phase (s) λprot vNBR λNBR λONR
queue cmerge queue
(veh/s) (veh/s) (veh/s) (veh/s)
(veh) (veh/s) (veh)
gs1 0.00 40.16 0.483 0.145 0.078 0.00 0.56 0.24 0.00
ge1 40.16 3.74 0.036 0.145 0.356 2.66 0.39 0.07 9.80
r1 43.90 5.70 0.000 0.145 0.430 1.87 0.43 0.11 10.08
g2* 49.60 0.88 0.000 0.145 0.430 0.25 0.43 0.11 10.72
g2** 50.48 49.82 0.000 0.145 0.145 0.00 0.14 -0.17 10.82
r2 100.30 5.70 0.000 0.145 0.145 0.00 0.14 -0.17 2.22
gs7 106.00 6.25 0.503 0.145 0.073 0.00 0.58 0.26 1.24
ge7* 112.25 2.02 0.027 0.145 0.366 0.45 0.39 0.08 2.85
ge7** 114.27 0.03 0.027 0.145 0.145 0.00 0.17 -0.15 3.01
r7 114.3 5.7 0.000 0.145 0.145 0.00 0.14 -0.17 3.01
Cycle
120 - 2.02
end
At the end of the time period, a residual queue of 23.32 vehicles is expected along the on-
ramp, and this value is carried to the start of the next time period. The time period length of
900s does not correspond to an exact number of signal cycles, and the last cycle is interrupted at
t = 60s. Therefore, the next time period will start the analysis from the same timestamp to
maintain consistency.
Time Period 3
The same steps performed for the analysis of time period 2 are applied again for the analysis of
time period 3.
336
Step 7C – Plot queue accumulation polygon for the on-ramp and unsignalized movements
The procedure described earlier is applied but with an initial on-ramp queue of 23.32 vehicles,
which is the estimated queue at the end of time period 2. The analysis begins at the middle of the
cycle (t= 60s), which is the end of the previous time period. Figure E-18 illustrates the queue
accumulation polygon for both the on-ramp and the NBR movement.
Figure E-18. Estimated queue lengths and merge capacities – time period 3
Queue spillback occurs during the third cycle (SBL queue service time), when the on-ramp
queue reaches the maximum storage LONR = 35.5 vehicles. At this time, the maximum flow rate
that can enter the on-ramp is constrained by the merge capacity cmerge. In other words, the
maximum number of vehicles allowed to enter the ramp is equal to the number of vehicles that
are able to merge to the freeway mainline. Also, the queues developed in the NBR are longer
during cycles 3 through 8, causing an increased delay on this movement due to the queue spillback
conditions at the on-ramp.
The on-ramp queue at the start of cycle 3 is 27.9 vehicles. The cycle starts with the SBL
movement, with an effective green time g1 = 47.3s. Since this movement already operates with
v/c > 1, the queue service time gs1 is equal to g1, and no green extension time is available (ge1 =
337
0). The protected movement then discharges at saturation flow rate sSBL = 0.483 veh/s, while the
NBR movement discharges at a permitted saturation flow rate sNBR = 0.078 veh/s. At the same
time, the on-ramp discharges to the freeway at a rate cmerge = 1,142 veh/h = 0.317 veh/s. Therefore,
the on-ramp queue grows at the following rate:
At this rate, the time remaining until spillback occurs is calculated by dividing the remaining
on-ramp queue storage by the growth rate:
35.5 − 27.9
𝑇𝑖𝑚𝑒 𝑡𝑜 𝑠𝑝𝑖𝑙𝑙𝑏𝑎𝑐𝑘 = = 31.2𝑠
0.244
Spillback is then expected to occur within 31.2 seconds of the onset of g1. The total effective
green g1 value of 47.3s is then divided in two portions:
• gs1* (31.2s): discharging at saturation flow rate
• gs1,sp (16.1s): the remainder of g1 will be affected by queue spillback, limiting the
maximum discharge to the on-ramp to the merge capacity cmerge = 0.317 veh/s. Note that this
constraint is shared by two movements entering the on-ramp (SBL and NBR).
The effect of queue spillback on the intersection capacity during gs1,sp is then measured by the
capacity reduction factor β1,sp, defined as the ratio between the maximum on-ramp capacity
during queue spillback and the throughput from the intersection movements (SBL and NBR):
𝑐 0.317
𝛽 ,
= = = 𝟎. 𝟓𝟔𝟓
𝜆 + 𝜆 (0.483 + 0.078)
A capacity reduction factor β1,sp= 0.565 means that only 56.5% of the expected intersection
throughput is able to enter the on-ramp when queue spillback occurs during phase gs1,sp. This
capacity adjustment factor is applied to each movement to obtain their adjusted throughputs for
this time period:
𝜆 , =𝜆 ×𝛽 ,
= 0.483 × 0.565 = 0.273 𝑣𝑒ℎ/𝑠
𝜆 , =𝜆 ×𝛽 ,
= 0.078 × 0.565 = 0.044 𝑣𝑒ℎ/𝑠
The procedure is then repeated for the remaining movements of the cycle, as shown in Table
E-13.
As shown, at time t = 31.2 s the maximum storage length of the on-ramp is reached and
spillback occurs. From this time through t = 83.3s, the throughput from intersection movements
to the on-ramp λONR is greater than the merge capacity cmerge. Therefore, the maximum allowed
throughput λONR,ajd is constrained by the on-ramp discharge capacity cmerge = 0.137 veh/s. For these
cases, the spillback capacity reduction factor fsp is computed as the ratio of λONR,ajd and λONR. Note
that for this time range the on-ramp queue is kept constant at the maximum storage of 35.54
vehicles.
From t = 83.3s, the on-ramp queue begins to discharge at a rate of 0.142 veh/s, followed by a
small increase during the green time of phase 7 (EBT), but it is not sufficient to cause spillback.
At the end of the cycle, the residual on-ramp queue is 33.51 vehicles.
The subsequent cycles follow a recurring pattern, with the on-ramp reaching maximum storage
early in the cycle and slightly diminishing at the end of the cycle.
338
Table E-13. Discharge flow rates into the on-ramp for each phase throughout the cycle –
time period 3
Capacity
Protected
On- movement Permitted movement On-ramp analysis reduction
Active Duration ramp adjustment
t (s)
phase (s) queue λONR,adj
(veh) λprot vNBR λNBR Q(NBR) λONR λONR,adj
- cmerge βsp
(veh/s) (veh/s) (veh/s) (veh) (veh/s) (veh/s)
(veh/s)
gs1* 0.0 31.2 27.92 0.483 0.175 0.078 0.00 0.561 0.561 0.244 1.000
gs1,sp 31.2 16.1 35.54 0.483 0.175 0.078 3.01 0.561 0.317 0.000 0.565
r1 47.3 5.7 35.54 0.000 0.175 0.430 5.12 0.430 0.317 0.000 0.739
g2* 53.0 30.3 35.54 0.000 0.175 0.430 4.31 0.430 0.317 0.000 0.739
g2** 83.3 17.0 35.54 0.000 0.175 0.175 0.00 0.175 0.175 -0.142 1.000
r2 100.3 5.7 33.11 0.000 0.175 0.175 0.00 0.175 0.175 -0.142 1.000
gs7 106.0 6.3 32.30 0.503 0.175 0.073 0.00 0.576 0.576 0.259 1.000
ge7 112.3 2.0 33.92 0.027 0.175 0.366 0.64 0.393 0.393 0.076 1.000
r7* 114.3 1.0 34.08 0.000 0.175 0.430 0.25 0.430 0.430 0.113 1.000
r7** 115.3 4.7 34.18 0.000 0.175 0.175 0.00 0.175 0.175 -0.142 1.000
Cycle
- - - - - -
end 120 0 33.51
The capacity of the SBL movement without consideration of queue spillback is 685 veh/h
(Table E-7). The adjusted capacity is calculated by applying the spillback capacity reduction
factor βsp, calculated in Table E-14:
In this example, this step is not required for the EBT movement, since this movement does not
experience effects of queue spillback. As shown in Figure E-18, the on-ramp queue during the
EBT green does not reach the maximum storage length of 35.5 veh.
339
Table E-14. Calculation of spillback capacity reduction factor for the SBL movement for
time period 3
On-ramp
Spillback adjustment
analysis
Active Duration
Cycle On-ramp expected On-ramp actual
phase (s) λONR λONR,adj
βsp discharge volume discharge volume
(veh/s) (veh/s)
(veh) (veh)
2 gs1 47.3 0.561 0.561 1.000 26.56 26.56
3 gs1* 31.2 0.561 0.561 1.000 17.51 17.51
3 gs1,sp 16.1 0.561 0.317 0.565 9.04 5.11
4 gs1 8.3 0.561 0.561 1.000 4.67 4.67
4 gs1,sp 39.0 0.561 0.317 0.565 21.89 12.37
5 gs1 5.1 0.561 0.561 1.000 2.87 2.87
5 gs1,sp 42.2 0.561 0.317 0.565 23.68 13.39
6 gs1 4.7 0.561 0.561 1.000 2.62 2.62
6 gs1,sp 42.6 0.561 0.317 0.565 23.93 13.53
7 gs1 4.6 0.561 0.561 1.000 2.59 2.59
7 gs1,sp 42.7 0.561 0.317 0.565 23.97 13.55
8 gs1 4.6 0.561 0.561 1.000 2.58 2.58
8 gs1,sp 42.7 0.561 0.317 0.565 23.97 13.55
Total: 185.89 130.89
Capacity reduction factor (βsp,SBL): 0.704
Time Period 4
The same steps performed for time periods 2 and 3 are applied again in time period 4.
Step 7C – Plot queue accumulation polygon for the on-ramp and unsignalized movements
The procedure described earlier is applied to plot the queue accumulation polygons, shown in
Figure E-19. Queue spillback occurs during the first cycle, due to the residual queue from the
previous time period. However, due to low volumes at the intersection and improvement of
performance along the freeway mainline, the on-ramp clears quickly. The queue has cleared by
the end of the second cycle.
340
Figure E-19. Estimated queue lengths and merge capacities – time period 4
Table E-15. Calculation of spillback capacity reduction factor for the SBL movement
during time period 4
On-ramp
Spillback adjustment
On-ramp analysis
Active Duration
Cycle queue On-ramp On-ramp actual
phase (s) λONR λONR,adj
(veh) βsp expected throughput
(veh/s) (veh/s)
throughput (veh) (veh)
1 gs1 6.0 34.42 0.505 0.505 1.000 3.02 3.02
1 gs1,sp 29.9 35.54 0.505 0.317 0.628 15.12 9.50
1 ge1 0.0 35.54 0.388 0.317 0.818 0.00 0.00
2 gs1 31.2 13.20 0.505 0.505 1.000 15.79 15.79
2 ge1 4.7 19.07 0.095 0.095 1.000 0.44 0.44
3 gs1 31.2 0.00 0.505 0.505 1.000 15.79 15.79
3 ge1 4.7 5.87 0.058 0.058 1.000 0.27 0.27
4 gs1 40.2 0.00 0.561 0.561 1.000 22.55 22.55
4 ge1 3.7 9.80 0.392 0.392 1.000 1.46 1.46
5 gs1 40.2 0.00 0.561 0.561 1.000 22.55 22.55
5 ge1 3.7 1.31 0.392 0.392 1.000 1.46 1.46
341
The adjusted capacity of the SBL movement is calculated by applying the spillback capacity
reduction factor βsp, calculated in Table E-15:
With the adjusted capacity values obtained, the performance measures for the intersection can
be computed using the remaining steps from the Signalized Intersections methodology (Chapter
19): compute the adjusted demand-to-capacity ratio (Step 8) and compute control delay (Step 9).
Table E-20 compares the performance measures for the affected movement (SBL) for the cases
with and without accounting for spillback effects. There is no change in the performance measures
in time period 2 even though the on-ramp demand is greater than the merge capacity, as the queue
can be stored in the on-ramp. Time period 3 yields a significant increase in the SBL control delay
due to the queue spillback: 589.2 s/veh, while the intersection analysis without consideration of
the spillback effects would return a control delay of 293.5 s/veh. Time period 4 shows a small
increase in control delay, from 575.2 s/veh to 609.5 s/veh. Even though spillback occurs for only
a short time during this time period, the high value of control delay obtained is due to the initial
queue delay (d3), as a result of the unmet demand at the end of time period 3.
Input Data
Figure E-20 shows the geometry of the TWSC intersection.
342
In this case there are no conflicting flows to the unsignalized right turn since it is a Rank 1
movement (highest priority). Therefore, the capacity for the NBR movement is equal to its
saturation flow rate. Table E-17 summarizes the calculations for this step.
343
Table E-17. Calculation of the on-ramp demand (vR) based on the TWSC intersection
operation
Time Movements
Parameter
Period EBT NBR SBL
Demand (veh/h) 8 315 652
v/c 0.06 - 0.96
1 c (veh/h) 125 1547 677
min (v, c) 8 315 652
Merge demand vR (veh/h) 975
Demand (veh/h) 4 608 591
v/c 0.10 - 0.48
2 c (veh/h) 42 1547 1222
min (v, c) 4 608 591
Merge demand vR (veh/h) 1203
Demand (veh/h) 18 708 685
v/c 0.64 - 0.56
3 c (veh/h) 28 1547 1222
min (v, c) 18 708 685
Merge demand vR (veh/h) 1411
Demand (veh/h) 24 80 463
v/c 1.00 - 0.60
4 c (veh/h) 24 1547 768
min (v, c) 24 80 463
Merge demand vR (veh/h) 567
The on-ramp demand estimates are then used as inputs for the freeway facility analysis. Since
the input demands for the freeway are identical to Part 1, it is already known that spillback will
occur during time period 3 (Table E-1).
Step 9B. Obtain merging capacity using the freeway facilities methodology
This step computes the merging capacity into the freeway cmerge. Since the inputs of the freeway
facility remain unchanged from Part 1, the same values are used:
• Time periods 2 and 3: 1,142 veh
• Time period 4: 1,142 veh/h during 4 time steps (60 seconds), then 1,903 veh/h.
344
calculated. Then, the time to spillback is obtained considering the queue growth and the available
queue storage. Time period 4 is split into two rows (4a and 4b), since the merge capacity changes
within this time period. For the first minute of the time period (4a), the merge capacity remains
at 1,142 veh/h due to existing oversaturated conditions along the freeway mainline. For the
remaining of the time period (4b), the merge capacity is equal to the ramp roadway capacity
(1,903 veh/h).
The results show that queue spillback occurs only during time period 3. The initial queue of
time period 2 is 15.2 vehicles, and it takes 4.55 minutes for the on-ramp to reach maximum storage
capacity. Therefore, the spillback time TSB is computed as 15 – 4.55 = 10.45 minutes.
Table E-18. Queue accumulation plot calculations for on-ramp – TWSC intersection
Figure E-21 illustrates the queue accumulation polygon for the on-ramp, based on the table
results.
Figure E-21. Queue accumulation polygon for the on-ramp – TWSC intersection
𝑐 , + 𝑐 , +𝑐 , =𝑐 = 1,142 𝑣𝑒ℎ/ℎ
345
The capacities during spillback conditions are then obtained proportionally to their demand
flow rates (Equation E-10):
𝑐 ×𝑣 1,142 × 685
𝑐 , = = = 554.4 𝑣𝑒ℎ/ℎ
𝑣 +𝑣 +𝑣 685 + 708 + 18
𝑐 ×𝑣 1,142 × 708
𝑐 , = = = 573.0 𝑣𝑒ℎ/ℎ
𝑣 +𝑣 +𝑣 685 + 708 + 18
𝑐 ×𝑣 1,142 × 18
𝑐 , = = = 14.6 𝑣𝑒ℎ/ℎ
𝑣 +𝑣 +𝑣 685 + 708 + 18
The equivalent capacities cEQ,i for each movement i, aggregated for the 15-min time period, are
obtained proportionately to the spillback time TSB (Equation E-11):
, × ×( ) . × . × .
𝑐 , = = = 757 𝑣𝑒ℎ/ℎ
, × ×( ) × . × .
𝑐 , = = = 869 𝑣𝑒ℎ/ℎ
, × ×( ) × . . × .
𝑐 , = = = 24 𝑣𝑒ℎ/ℎ
With the adjusted capacity values obtained, the performance measures for the intersection can
be computed using the next step from the TWSC methodology (Chapter 20): compute movement
control delay (Step 11).
Table E-23 compares the performance measures of the affected intersection movements for the
cases with and without consideration of spillback effects during time period 3. All three
movements discharging to the on-ramp have significantly higher delays when considering
spillback effects.
Input Data
Figure E-20 shows the geometry of the study intersection.
346
The estimated on-ramp demand values are provided as inputs for the freeway facility analysis.
The freeway facility is then analyzed and the expected on-ramp queues are provided in Table E-
21.
347
Table E-20. Calculation of the on-ramp demand (vR) based on the AWSC intersection
operation.
Time Movements
Parameter
Period EBT NBR SBL
Demand (veh/h) 54 467 313
Adjusted demand (veh/h) 54 467 313
v/c 0.143 - 0.672
1
c (veh/h) 377 539 466
min (v, c) 54 467 313
Merge demand vR (veh/h) 834
Demand (veh/h) 40 512 432
Adjusted demand (veh/h) 40 512 432
v/c 0.114 - 0.984
2
c (veh/h) 350 521 439
min (v, c) 40 512 432
Merge demand vR (veh/h) 984
Demand (veh/h) 19 539 546
Adjusted demand (veh/h) 19 539 546
v/c 0.048 - 1.18
3
c (veh/h) 396 550 462
min (v, c) 19 539 462
Merge demand vR (veh/h) 1020
Demand (veh/h) 28 160 316
Adjusted demand (veh/h) 28 160 316
v/c 0.062 - 0.618
4
c (veh/h) 455 619 511
min (v, c) 28 160 316
Merge demand vR (veh/h) 504
Since spillback will occur, the impacts on the operation of the intersection must be evaluated.
The next section illustrates the application of the evaluation methodology at the AWSC
intersection.
348
Table E-22. Queue accumulation plot calculations for on-ramp – AWSC intersection
On-ramp
On-ramp queue Initial Spillback Final
Time to
Time Duration demand growth rate ONR time ONR
spillback
Period (min) (vR) (λΟΝR - queue (TsB) queue
(min)
(veh/h) cmerge) (veh) (min) (veh)
(veh/s)
2 15 984 0.023 0.0 - - 21.0
3 15 1020 0.033 15.2 7.25 7.75 35.5
4 15 504 -0.110 35.5 - - 0.0
Figure E-21 illustrates the queue accumulation polygon for the on-ramp, based on the table
results.
Figure E-23. Queue accumulation polygon for the on-ramp – AWSC intersection
349
Table E-23. Equivalent capacities and headways for on-ramp – Time Period 3 – AWSC
intersection
Capacity during Regular Equivalent
Spillback departure
Movement spillback (csp) Capacity Capacity (cEQ)
headway (hsp) (s)
(veh/h) (c) (veh/h) (veh/h)
EBT 15 396.0 212.1 17.0
NBR 439 550.0 496.5 7.3
SBL 445 462.0 453.7 7.9
With the adjusted capacity values obtained, the performance measures for the intersection can
be computed using the remaining steps from the AWSC methodology (Chapter 21): compute the
service times (Step 13) and compute control delay (Step 14).
Table E-24Table E-20 compares the performance measures of the intersection movements for
the cases with and without consideration of spillback effects during time period 3. The three
movements that discharge into the on-ramp (EBT, NBR and SBL) experience increased delay,
while the remaining movements have the same performance measures.
Table E-24. Comparison of performance measures – time period 3 - with and without
spillback effects
Departure headway
Capacity (veh/h) Control delay (s/veh)
Demand (s)
Movement
(veh/h) Without With Without With Without With
spillback spillback spillback spillback spillback spillback
EBL 75 359 359 15.6 15.6 10.0 10.0
EBT 19 396 212 12.6 21.7 9.1 17.0
NBT 229 497 497 16.3 16.3 7.2 7.2
NBR 539 550 497 58.9 92.3 6.5 7.3
SBL 546 462 454 128.0 136.5 7.8 7.9
SBT 220 494 494 16.0 16.0 7.3 7.3
350
APPENDIX F
Data Collection
The data collected for this part of the project include speed and flow data from selected detector station
sensors in California, Virginia, Utah, Wisconsin, Minnesota, and Florida. These locations represent diverse
operational and design conditions across the US. Sites were selected based on the following criteria:
• Speed and flow data available for each lane, aggregated in 15-min intervals, for a period of at least
one year;
• Absence of freeway management strategies, such as express or high-occupancy vehicle (HOV)
lanes, ramp metering, speed harmonization, or demand shoulder use;
• For merge and diverge and segments, good health detector data available for the upstream,
downstream and ramp sections;
• Percentages of heavy vehicles were available.
The dataset includes 48 locations: 19 basic, 14 merge, 15 diverge and 16 weaving segments with 2, 3 or
4 lanes on each direction. There are no 5-lane segments in the database, as many of the identified locations
operate with HOV lanes. The number of required detector stations is different for each segment type (
Figure F-1). Basic segments require only one detector station. Diverge segments require two stations:
one at the ramp influence area (upstream of the exit) and one along the ramp. Merge segments require three
stations: one at the ramp influence area (downstream the merge), one along the ramp and one upstream of
the merge.
The list of data collection locations is provided in Table F-1 (basic segments), Table F-2 (merge
segments), Table F-3 (diverge segments) and Table F-4 (weaving segments).
351
Table F-1 – Database used for determination of lane-by-lane flows – Basic freeway segments
Detector #
# Observation %
State Road Station %HV adjacent
lanes Period Grade
IDs ramps
CA HWY 1 NB 2018 500014082 0.5 1.7 2
CA SR-132 EB 2018 10119910 -0.5 0.3 0
MN US-10 2017 - 18 946 0.3 1.4 1
MN I-694 NB 2017 1413 -1.2 8.6 1
2L
UT SR-67 NB 2018 820 -0.3 18 0
VA I-66 WB 2018 19002981 0.5 1 2
VA I-64 EB 2018 64239221 -0.9 9 1
WI I-43 SB 2017 - 18 692 -0.02 6.5 0
CA I-205 WB 2018 - 19 1027310 -0.9 4.2 0
CA SR-85 NB 2018 - 19 407336 -0.8 5.6 1
FL I-4 WB 2017 4712 -0.2 10.8 0
MN I-94 NB 2017 - 18 1356 -1.7 11.8 0
3L
UT I-215 SB 2016 - 17 82 -3.1 10.5 0
UT I-15 SB 2018- 19 963 1.45 32.9 1
VA I-64 EB 2017- 18 64055221 0.2 4 3
WI I-43 NB 2017 - 18 659 -0.6 10.2 1
CA I-80 2018 - 19 413373 0.4 6.8 1
CA SR-24 EB 2018 - 19 400532 1.7 14.1 0
CA I-80 EB 2018 - 19 413375 1.1 5 1
FL I-295W NB 2017 2220 0 13.6 1
4L
FL I-275 SB 2016 - 17 3829 -1.1 4.4 0
FL I-275 NB 2016 - 17 3408 0 4.4 0
UT I-215 CW 2018 - 19 50 -0.8 28.2 1
VA I-295 EB 2018 - 19 4044741 -0.1 10 1
352
Table F-2 – Database used for determination of lane-by-lane flows – Merge segments
#
# Observation %
State Road Detector Station IDs %HV adjacent
lanes Period Grade
ramps
UT SR-67 S 2018 872, 876, 1872 0 13 0
UT SR-67 N 2018 830, 834, 1830 0.1 17 0
2L
FL I-295E NB 2018 10821, 10852, 10890 2.7 11.8 0
FL I-295E NB 2018 10854, 10841, 10879 1.5 11.8 0
CA I-5 SB 2018 - 19 10121110, 10121310, 1083110 -0.9 11.9 0
CA CA-99 WB 2018 10109610, 10109810, 10109710 0.1 11.23 1
MN I-694 2017 174, 172, 753 1.8 8.6 1
UT SR201-W 2018 - 19 348, 350, 1348 0 27.9 1
3L
UT I-215 CCW 2017 173, 175, 1173 1.5 9 1
UT I-215S EB 2017 168, 169, 1168 0.2 11 0
UT I-215 2017 188, 190, 1188 0 10 0
UT I-80 2017 231, 232, 1231 1.32 14 0
CA I280 SB 2018 - 19 403908, 403328, 403909 0 1.3 1
4L
CA 1-8 WB 2017 403908, 403328, 403909 2.71 2.1 1
Table F-3 – Database used for determination of lane-by-lane flows – Diverge segments
#
# Observation %
State Road Detector Station IDs %HV adjacent
lanes Period Grade
ramps
CA US-101 SB 2018 – 19 406305, 406303 -0.3 5 1
CA I-8 WB 2016 - 17 1115624, 1122447 0.8 3.4 0
2L UT SR-67 NB 2017 810, 2810 0 9 0
FL I-295E NB 2017 10829, 10876 0 11.8 0
WI I-94 WB 2017 67663, 67662 -2 7.2 1
CA CA-242 2018 – 19 414251, 417124 0 2.43 0
CA I-5 NB 2018 – 19 10121410, 1090210 0 18.1 0
CA CA-73 2018 – 19 1208789, 1208940 -5.4 3.8 0
MN I-694 2017 151, 554 2.5 8.6 2
3L
MN I-694 2017 171, 755 2.5 8.6 2
UT I-215 CCW 2017 112, 2114 0 17 0
UT I-215 CW 2018 – 19 22, 2024 -2.2 61.6 0
UT I-215 CW 2018 – 19 86, 2087 0 8 0
CA SR-242 NB 2018 – 19 414252, 418256 1.5 2.21 4
4L
CA I-80 EB 2017 413375, 410766 0.7 5.1 1
353
Table F-4 – Database used for determination of lane-by-lane flows – Weaving segments
#
# Observation %
State Road Detector Station IDs %HV adjacent
lanes Period Grade
ramps
CA CA-56 WB 2018 1125546, 1125575, 1125543, 1126293 0 2.03 0
MN I-694 NB 2018 1027, 1410, 5120, 6226 0.6 14.8 3
MN Lafayette Hwy 2017 S1169, S1170, 5626, 5629 0 8.81 0
2L MN US 52 2017 S1442, S1444, 6061, 6069 0 8.01 0
MN Lafayette Hwy 2017 S1159, S1160, 5581, 5584 0 8.81 0
MN MN-36 WB 2017 S616, S617, 2361, 2362 0.8 8.81 0
MN MN-36 EB 2017 S591, S592, 2256, 2257 0.8 8.81 0
CA I-80 SB 2018-2019 409107, 404408, 409108, 419489 1 15.8 3
UT I-215 CW 2013-2014 320, 322, 1320, 2322 0 14 1
3
UT I-215 CW 2015-16 344, 346, 1344, 2346 0 14 0
UT Belt Route 2011-2012 162, 165, 1162, 2165 1 14 1
CA SR4-EB 2018 416930, 414707, 416931, 414708 -0.5 3.3 2
CA SR4-EB 2018 400049, 405269, 418869, 406635 -0.8 4.8 0
4 CA I-680 SB 2018 407177, 407179, 409059, 407178 0.51 4.4 4
CA I-880 SB 2018 400949, 400678, 403028, 403030 0.1 6.1 0
UT I-80 EB 2015 228, 240, 1228, 2231 0 16 0
Speed and flow detector data were collected from online sources from the respective state agencies. Data
were obtained by lane in 15-min intervals, over a 1-year period for each location. Erroneous speed and flow
data and those associated with crashes, lane closures, and work zones were removed. Holidays and
weekends were excluded. The heavy vehicle percentage (HV%) was collected from the respective state
agencies. Average truck percentages are typically reported on an annual basis by agencies; therefore the
speed-flow data from detectors were downloaded only for periods when HV% information was available.
The presence of ramps upstream or downstream of a site might cause significant impact on lane flow
distribution. Therefore, the analysis included the number of such ramps within a half-mile upstream and
downstream of the segment (access point density).
Where:
LFRi = share of the total flow on lane i, where i ranges from 1 to n-1 (n = total number of segment
lanes)
354
LFRn = share of the total flow on the leftmost lane (lane n);
fa = adjustment factor for 𝑎 (Equations F-3, F-5, F-7);
v/c = volume/capacity ratio (0 ≤ v/c ≤ 1)
fc = adjustment factor for 𝑐 (Equations F-4, F-6, F-8)
The model proposed in Equation F-1 can be applied for basic, merge, diverge and weaving segments.
For merge and diverge segments, the share of flow is estimated at the area upstream of the ramp. For
weaving segments, the share of flow is estimated at the mainline upstream the on-ramp.
For the proposed methodology, volume and capacity are given in veh/h. The adjustment factors fa and fc
applicable in the analysis of basic segments are as follows:
𝑓 =𝑎+𝐺∙𝑓 , +𝑡∙𝑓 , +𝑛∙𝑓, (Equation F-3)
𝑓 = 𝑐+𝐺 ∙𝑓, +𝑡∙𝑓, +𝑛∙𝑓, (Equation F-4)
For merge and diverge segments, the fa and fc factors are as follows, with additional coefficients f ,
and f , to address ramp demand:
𝑓 =𝑎+𝐺∙𝑓 , +𝑡∙𝑓 , +𝑛∙𝑓, + ∙𝑓 , (Equation F-5)
𝑓 = 𝑐+𝐺 ∙𝑓, +𝑡∙𝑓, +𝑛∙𝑓, + ∙𝑓, (Equation F-6)
where:
G = grade (%)
a = empirical constant
fa,G = adjustment factor for a due to impact of grade
fc,G = adjustment factor for c due to impact of grade
t = truck percentage (%)
fa,t = adjustment factor for a due to impact of trucks
fc,t = adjustment factor for c due to impact of trucks
n = access point density – number of ramps half a mile upstream and half mile downstream
fa,n = adjustment factor for a due to impact of access point density
c = empirical constant
fc,n = adjustment factor for c due to impact of access point density
vR = ramp flow (vph)
fa,vR = adjustment factor for a due to impact of ramp flow
fc,vR = adjustment factor for c due to impact of ramp flow
The adjustment factors for the weaving segments address the effect of weaving-specific properties:
𝑓 = 𝑎 + 𝐺 ∙ 𝑓 , + 𝑡 ∙ 𝑓 , + 𝐼𝐷 ∙ 𝑓 , + , ∙ 𝑓 , + , ∙ 𝑓 , + ∙ 𝑓 , + 𝑉𝑅 ∙ 𝑓 ,
(Equation F-7)
, ,
𝑓 = 𝑐 + 𝐺 ∙ 𝑓 , + 𝑡 ∙ 𝑓 , + 𝐼𝐷 ∙ 𝑓 , + ∙𝑓, + ∙𝑓, + ∙ 𝑓 , + 𝑉𝑅 ∙ 𝑓 ,
(Equation F-8)
where:
ID = interchange density, as defined in HCM Chapter 13
fa,I = adjustment factor for a due to impact of interchange density
fc,I = adjustment factor for c due to impact of interchange density
vR,m = on-ramp flow (veh/h)
fa,vm = adjustment factor for a due to on-ramp flow
fc,vm = adjustment factor for c due to on-ramp flow
vR,d = off-ramp flow (veh/h)
fa,vd = adjustment factor for a due to off-ramp flow
fc,vd = adjustment factor for c due to off-ramp flow
355
The empirical constants (a, c, and the adjustment factors f) were obtained by regression and are specific
for each combination of segment type, lane number and total number of lanes. The obtained values for
basic, merge and diverge segments are presented in Table F-5.
356
Table F-5 – Adjustment factors for lane flow distribution on basic, merge and diverge segments
357
The weaving LFRs are calculated at two locations: upstream of the on-ramp and within the weaving
segment (Figure F-1). The downstream LFRs are calculated as a function of the upstream LFRs.
The adjustment factors calculated for the upstream LFRs are presented in Table F-6. The number of lanes
refer to the freeway section immediately upstream of the weave, and any lanes connecting the on-ramp and
off-ramp are not considered as part of the total number of lanes.
Table F-6 – Adjustment factors for lane flow distribution on weaving segments
The LFRs within the weave are estimated considering the number of lanes involved in weaving.
According to the HCM Chapter 13, the number of weaving lanes (NWL) is the total number of lanes from
which a weaving maneuver may be completed with one lane change or no lane changes. The auxiliary lane
358
is included in the NWL. Figure F-3 provides two weaving example configurations with the respective NWL.
The number of weaving lanes on the mainline freeway, excluding the auxiliary lane(s) is denoted as NWUP,
also shown in Figure F-3 for the two example configurations.
Figure F-3. Notation and number of weaving lanes upstream (NWUP) and within the merge (NWL) for
two weaving configurations
The methodology assumes that all mandatory weaving lane change maneuvers are completed before the
midpoint of the short weaving length (LS). This assumption is based on the results reported by (Menendez,
and He, 2016; Ahmed at al., 2019). Menendez and He (2016) concluded that about 70% of lane change
completion in the weaving segment occurred within 19% of the weaving section length during all times and
for all traffic conditions. Another study (Ahmed at al., 2019) related to weaving lane changes at the US-
101 freeway indicated that 50% of all lane changes were completed within 16% of the short length of the
weave.
The methodology also assumes that for segments with only one weaving lane upstream (NWUP=1) the
entire freeway-to-ramp flow (vFR) will be positioned in the rightmost lane (L1,UP). For segments with two
upstream weaving lanes (NWUP=2), the methodology assumes that 80% of freeway-to-ramp flow (vFR) will
be on the rightmost lane (L1,UP) while the remaining 20% will be on the adjacent lane (L2,UP). This
assumption is based on recent work by Menendez, and He, (2016); and Ahmed at al. (2019). The
distribution of flows for non-weaving vehicles (freeway-to-freeway) in the middle of the weaving segment
is assumed to be equal to that upstream of the weave.
Therefore (Figure F-4), flows within the weave are calculated assuming the entire vFR flow will be on the
auxiliary lane when it reaches the midpoint of the weave. For traffic that is on the adjacent lane, it is assumed
that it will be located on L1 when it reaches the mid-point of the weave, i.e. it will make one lane change
toward the exit by the mid-point. If the VFR exceeds the calculated upstream flow in lane 1 v1,UP, then the
extra flow is allocated to the adjacent lane when estimating the mid-point lane allocations. Thus, the
methodology checks whether there is excess traffic from vFR (vEXFR) that needs to be allocated to L2 at the
midpoint of the weave. The sum of lane flows estimated within the weave should be equal to the sum of
the upstream lane flows (freeway and merge flows).
Figure F-4. Weaving flows lane allocation for two example weaving configurations
359
Step 1. Determine the number of upstream freeway weaving lanes (NWUP) based on geometry
For NWUP = 1:
𝑣 ≤𝑣 (Equation F-9)
Where:
v,FR = freeway to ramp flow (veh/h)
v1,UP = upstream weaving lane flow for lane 1 (veh/h)
v2,UP = upstream weaving lane flow for lane 2 (veh/h)
vEXFR1 = excess of freeway to ramp flow that needs to be accommodated in the lane adjacent to lane 1
upstream of the weaving segment when NWUP =1
vEXFR2 = excess of freeway to ramp flow that needs to be accommodated in the lane adjacent to lane 2
upstream of the weaving segment when NWUP = 2
vEXFR3 = excess of freeway to ramp flow that needs to be accommodated in the lane adjacent to lane 3
upstream of the weaving segment when NWUP = 2
where:
v0 = auxiliary lane flow (veh/h)
vFR = freeway-to-ramp flow (veh/h)
v1,UP = upstream lane 1 flow (veh/h)
vRR = ramp-to-ramp flow (veh/h)
360
where:
v1 = lane 1 flow (veh/h) within the weaving segment
v1,UP = upstream weaving lane flow for lane 1 (veh/h)
v2,UP = upstream weaving lane flow for lane 2 (veh/h)
vRF = ramp-to-freeway flow (veh/h)
vEXFR1 = excess of freeway to ramp flow that needs to be accommodated in the lane adjacent to lane 1
upstream of the weaving segment when NWUP =1
vEXFR2 = excess of freeway to ramp flow that needs to be accommodated in the lane adjacent to lane 2
upstream of the weaving segment when NWUP = 2
Step 5. Calculate the freeway weaving lane flow for lane 2 (v2)
𝑣 = 𝑣 , − 𝑣 (NWUP =1) or (Equation F-22)
𝑣 = 𝑣 , − (0.2 ∗ 𝑣 ) − 𝑣 (NWUP =2) 𝑖𝑓 (0.8 ∗ 𝑣 ) < 𝑣 , and (0.8 ∗ 𝑣 ) > 𝑣 ,
(Equation F-23)
𝑣 = 𝑣 (NWUP =2) 𝑖𝑓 (0.8 ∗ 𝑣 ) > 𝑣 , and (0.2 ∗ 𝑣 ) + 𝑣 >𝑣 ,
(Equation F-24)
where:
v2 = lane 2 flow (veh/h) within the weaving segment
Step 6. Calculate the freeway weaving lane flow for lane 3 (v3)
This step is only valid with the presence of vEXFR3 .
𝑣 = 𝑣 , − 𝑣 (Equation F-25)
where:
v3 = lane 3 flow (veh/h) within the weaving segment
v3,UP = upstream lane 3 (veh/h)
Step 7. Obtain flows for the remaining lanes, which will be equal to the respective upstream flow
values.
The flows of the remaining lanes will be same as their respective upstream lane flows.
Step 8. Check whether there are v/c ratios greater than 1 for each lane.
This step checks whether any of the lane v/c ratios are greater than 1. When that occurs, flows should be
adjusted based on the procedure described above.
361
For the weaving segments, capacity is estimated using the method proposed in HCM 6th Edition –
Chapter 13 (Step 5, Equations 13-5 through 13-9) in passenger-cars per hour (pc/h). It should be noticed
that capacity in this method is function of the volume ratio (VR). As such, the capacity for weaving
segments may vary for different time intervals.
Reasonableness checks
After lane flow ratios are obtained, a two-step reasonableness check must be performed to ensure the
obtained flow distribution remains under feasible constraints. The first step checks for any negative flows
that may occur in any segment lane – this issue is more likely to occur in the leftmost lane, as the flows on
this lane are obtained by the difference between the total segment flow and the sum of estimated flows in
the other lanes. Therefore, if flows on the remaining lanes are overestimated the resulting flow in the
leftmost lane may become negative. Figure F-5 illustrates the recommended procedure for the first check.
The second step of the reasonableness check compares the estimated flow by lane with the respective
lane capacities to make sure no lane operates with a demand-to-capacity ratio greater than 1.The procedure
illustrated in Figure F-5. If any lane is observed to operate above its capacity, the flow in this given lane is
constrained by the capacity value and the exceeding demand is rearranged to the adjacent lane.
362
Diverge segment
A practical application of the LFR model is presented next for a 3-lane diverge segment (single period
analysis), with the following input data:
• Grade (G): 3%
• Heavy vehicles (t): 4%
• Access point density (n): 2 adjacent ramps
• Mainline demand flow rate (v): 5500 veh/h
• Off-ramp demand (vR): 850 veh/h
• Measured segment capacity (c): 2050 veh/h/ln (6150 veh/h)
The flow ratio for lane 1 (right lane) is obtained by the following equation:
𝑣
𝐿𝐹𝑅 = 𝑓 ∙ ln +𝑓
𝑐
363
𝑓 = 0.3218
The same procedure is applied to obtain the lane flow ratio on lane 2, using the respective coefficients
from Table F-5:
𝑓 =𝑎+𝐺∙𝑓 , +𝑡∙𝑓 , +𝑛∙𝑓, + ∙𝑓 ,
𝑓 = 0.0096 + 3 ∙ (−0.00960) + 4 ∙ (−0.00054) + 2 ∙ (−0.0096) + ∙ (−0.04766)
𝑓 = −0.08107
5500
𝐿𝐹𝑅 = −0.08107 ∙ ln + 0.2854
3 ∙ 2050
𝐿𝐹𝑅 = 29.4%
Finally, the lane flow ratio on the leftmost lane (lane 3) can be obtained as follows:
𝐿𝐹𝑅 = 1 − 𝐿𝐹𝑅 − 𝐿𝐹𝑅 = 1 − 0.294 − 0.33
𝐿𝐹𝑅 = 37.6%
Weaving Segment
A practical application of the LFR model is presented next for a weaving segment (Figure F-7), where
the lane flow share among 5 lanes is estimated both upstream and within the weave (single period analysis).
Figure F-7. Study case on a weaving segment, with four mainline lanes upstream
364
LFRs and lane flows are first calculated for the upstream section of the weaving segment. The heavy-
vehicles adjustment factor can be estimated (adopting 𝐸 = 2) as:
1 1
𝑓 = = = 0.968
1 + 𝑃 (𝐸 − 1) 1 + 0.03(2 − 1)
The weaving and non-weaving demands can be adjusted to flow rates under ideal conditions. Because
the demands are estimated based on 15-minute volumes, PHF is equal to 1.
𝑉
𝑣=
𝑃𝐻𝐹 ∙ 𝑓
24
𝑣 = = 24.8 𝑝𝑐/ℎ
1 ∙ 0.968
404
𝑣 = = 417.3 𝑝𝑐/ℎ
1 ∙ 0.968
600
𝑣 = = 619.8 𝑝𝑐/ℎ
1 ∙ 0.968
3912
𝑣 = = 4041.3 𝑝𝑐/ℎ
1 ∙ 0.968
The capacity of the weaving segment is given by the minimum between the density-capacity (𝑐 ) and
weaving-demand-capacity (𝑐 ), which are:
𝑐′ = 𝑐 − 438.2(1 + 𝑉𝑅) . + (0.0765𝐿 ) + (119.8𝑁 )
𝑐 = 2400 𝑝𝑐/ℎ
𝑐′ = 2400 − 438.2(1 + 0.203) . + (0.0765 ∙ 3920) + (119.8 ∙ 2) = 2351 𝑝𝑐/ℎ/𝑙𝑛
𝑐 = 𝑐′ ∙ 𝑓 = 2350.5 ∙ 0.968 = 2275 𝑣𝑒ℎ/ℎ/𝑙𝑛
2400 2400
𝑐′ = = = 11822 𝑝𝑐/ℎ
𝑉𝑅 0.203
365
𝑐′ ∙𝑓 11822.5 . 0.968
𝑐 = = = 2861 𝑣𝑒ℎ/ℎ/𝑙𝑛
𝑁 4
𝑐 = min(𝑐 , 𝑐 ) = min(2275.3 ,2861.1 ) = 2275 𝑣𝑒ℎ/ℎ/𝑙𝑛
𝑣 , 𝑣 ,
𝐿
𝑓, =𝑐+ 𝐺∙𝑓, +𝑡∙𝑓, +𝐼 ∙𝑓, + ∙𝑓 + ∙𝑓 +
∙ 𝑓 + 𝑉𝑅 ∙ 𝑓 ,
1000 , 1000 ,
1000 ,
428
𝑓 , = 0.2434 + (−0.5) ∙ (−0.03) + (3.3) ∙ (−0.004) + 0.67 ∙ (−0.0067) + ∙ 0.0645
1000
624 3920
+ ∙ (0.0629) + ∙ (−0.0303) + 0.203 ∙ (−0.1432)
1000 1000
𝑓 = 0.1606
The same procedure is applied to obtain the lane flow ratio on lane 2, using the respective coefficients
from Table F-6:
𝑣 , 𝑣 , 𝐿
𝑓 , =𝑎+𝐺∙𝑓 , +𝑡∙𝑓 , +𝐼 ∙𝑓 , + ∙𝑓, + ∙𝑓 , + ∙ 𝑓 , + 𝑉𝑅 ∙ 𝑓 ,
1000 1000 1000
428
𝑓 , = 0.0048 + (−0.5) ∙ (−0.0048) + (3.3) ∙ (−0.00482) + 0.67 ∙ (−0.00482) +
1000
624 3920
∙ (−0.0313) + ∙ (0.03) + ∙ (0.0019) + 0.203 ∙ (−0.0044)
1000 1000
𝑓 = −0.00012
𝑣 , 𝑣 , 𝐿
𝑓, =𝑐+ 𝐺∙𝑓, +𝑡∙𝑓, +𝐼 ∙𝑓, + ∙𝑓 + ∙𝑓 + ∙𝑓
+ 𝑉𝑅 ∙ 𝑓 ,
1000 , 1000 , 1000 ,
428
𝑓, = 0.2571 + (−0.5) ∙ (0.0447) + (3.3) ∙ (−0.0112) + 0.67 ∙ (−0.0049) +
1000
624 3920
∙ (−0.0088) + ∙ (−0.0152) + ∙ (0.0107) + 0.203 ∙ (0.0401)
1000 1000
𝑓, = 0.2310
366
4512
𝐿𝐹𝑅 = − 0.00012 ∙ ln + 0.2310
4 ∙ 2275.3
𝐿𝐹𝑅 = 23.1%
The same procedure is applied to obtain the LFR on lane 3, using the respective coefficients from
Table F-6:
𝑣 , 𝑣 ,
𝑓 , = 𝑎 + 𝐺 ∙ 𝑓 , + 𝑡 ∙ 𝑓 , + 𝐼𝐷 ∙ 𝑓 , + ∙𝑓, + ∙ 𝑓 , + 𝐿 ∙ 𝑓 , + 𝑉𝑅 ∙ 𝑓 ,
1000 1000
428
𝑓 , = 0.12 + (−0.5) ∙ (−0.1199) + (3.3) ∙ (0.0185) + 0.67 ∙ (−0.1199) + ∙ (−0.0113)
1000
624 3920
+ ∙ (0.0509) + ∙ (−0.0405) + 0.203 ∙ (0.1199)
1000 1000
𝑓 , = 0.0524
𝑣 , 𝑣 ,
𝑓, =𝑐+ 𝐺∙𝑓, +𝑡∙𝑓, +𝐼 ∙𝑓, + ∙𝑓 + ∙𝑓 +𝐿 ∙𝑓, + 𝑉𝑅 ∙ 𝑓 ,
1000 , 1000 ,
428
𝑓 , = 0.2710 + (−0.5) ∙ (0.0410) + (3.3) ∙ (−0.0042) + 0.67 ∙ (−0.0026) +
1000
624 3920
∙ (−0.0377) + ∙ (−0.0372) + ∙ (0.0198) + 0.203 ∙ (0.1545)
1000 1000
𝑓 , = 0.3039
Finally, the LFR on the leftmost lane (lane 4) can be obtained as follows:
𝐿𝐹𝑅 = 1 − 𝐿𝐹𝑅 − 𝐿𝐹𝑅 − 𝐿𝐹𝑅 = 1 − 0.228 − 0.231 − 0.267
𝐿𝐹𝑅 = 27.4%
Based on these, we calculate the LFRs and lane flows within the weave.
Step 1: Determine the number of upstream freeway weaving lanes (NWUP) based on geometry
N =1, obtained from Figure F-7
𝑣 = 600 veh/h
Here, v < v ,
367
Hence, the entire freeway-to-ramp flow will be located at the upstream freeway-weaving lane and
v =0.
Step 7. Obtain flows for the remaining lanes, which will be equal to the respective upstream flow
values.
𝑣 = 𝑣 , = 1204 veh/h
𝑣 =𝑣 , = 1236 veh/h
Step 8. Check whether there are v/c ratios greater than 1 for each lane.
Given the following flows and capacities for each lane within the weave:
𝑣 = 624 veh/h, 𝑣 = 833 veh/h, 𝑣 = 1043 veh/h, 𝑣 = 1204 veh/h, 𝑣 = 1236 veh/h
𝑐 = 𝑐 = 𝑐 = 𝑐 = 𝑐 = 2275 veh/h
368
Lane FFS
Field observations have shown that operating speeds differ among lanes, and they are typically lower in
shoulder lanes and higher in median lanes. Free-flow speeds were measured as the average speed for
segment flow rates below 450 veh/h/ln. This criterion is consistent with HCM guidance, which recommends
measuring FFS for flows no greater than 500 pc/h/ln.
Next, lane FFS were modeled as a function of the segment FFS and as a function of the number of lanes
on the segment, as shown in Figure F-8. Due to the ramp influence on traffic flow, merge and diverge
segments are likely to have different distributions of FFS. Therefore, distinct models were developed by
segment type. Linear regression models were developed with the intercept set to zero.
As it can be observed in Figure F-8, there is a good correlation between segment and lane FFS, confirming
field observations: shoulder lanes’ FFS are lower than the segment average, while median lanes’ FFS are
higher. Center lanes typically have FFS values very close to the segment average.
Figure F-8. Segment FFS and lane FFS, by segment type and number of lanes
369
Based on the obtained results, models were developed to estimate individual lane FFS by applying a
multiplying factor to the segment FFS. These models are shown in Figure F-8 for each lane. Table F-7
summarizes the recommended multipliers which are provided as a function of the segment type and the
number of lanes in the segment. As shown, when the number of lanes increases, the range of FFS multipliers
increase as well (i.e. there are lower speeds in the shoulder lanes and higher speeds on the median lanes).
For 2-lane segments, merge and diverge segments have a higher difference in FFS between the two lanes
when compared to basic segments. For 3-lane segments, basic segments show the highest FFS range, while
merge segments have more uniform lane FFS. As for 4-lane segments, merge segments show the highest
FFS range, followed by basic and merge segments yield similar results.
FFS Multiplier
Segment Number of
type lanes L1 L2 L3 L4
2 lanes 0.965 1.032
Basic 3 lanes 0.934 1.010 1.087
4 lanes 0.924 0.989 1.028 1.079
2 lanes 0.964 1.044
Merge 3 lanes 0.955 1.015 1.045
4 lanes 0.935 0.991 1.036 1.091
2 lanes 0.961 1.035
Diverge 3 lanes 0.943 1.024 1.068
4 lanes 0.933 0.975 1.018 1.074
2 lanes 0.969 1.018
Weaving 3 lanes 0.968 1.023 1.062
4 lanes 0.910 0.988 1.053 1.110
370
Figure F-9. Example of lane capacity estimation (French Camp, CA): (a) lane flow distribution at
breakdown and (b) LFRs as a function of segment capacity
The same rationale was applied to all locations in the database. Figure F-10 shows the relationship
between the measured segment capacities and their respective capacities for individual lanes. As it can be
observed, capacity typically increases from the rightmost to the leftmost lanes, with center lanes showing
capacity values similar to the segment average.
371
Figure F-10. Relationship between segment capacity and individual lane capacity, by segment type
and number of lanes
For weaving segments, capacity distributions were observed to be significantly more complex and the
breakdown observation method was not capable of providing reliable results from the selected dataset.
Capacity is assumed uniform for all lanes within a weaving segment, obtained by HCM Equation 13-5
(based on a maximum density of 43 pc/h/ln):
𝑐 = 𝑐 − 438.2 (1 + 𝑉𝑅) . + (0.0765 𝐿 ) + (119.8𝑁 ) (Equation F-27)
372
Next, results were averaged by segment type and number of lanes. Figure F-11 presents the percent
distribution of the total segment capacity across lanes (the numbers below the whisker boxes represent the
average values of lane capacity).
Figure F-11. Capacity of individual lanes as a percentage of segment capacity, by segment type and
number of lanes
The segment capacities measured from field data may not be equal to the estimated capacities using HCM
methodologies. According to the HCM Equation 12-6, base capacity can be estimated as:
𝑐 = 𝑚𝑖𝑛 2200 + 10 𝑥 (𝐹𝐹𝑆 − 50), 2400 (Equation F-28)
373
For each location from the dataset, base capacity was calculated using Equation F-28, as FFS is available
from field measurements. Since this equation provides capacity values in passenger-car equivalents, a heavy
vehicle factor fHV (as defined in HCM Equation 12-10) was applied to convert the base capacity to veh/h
and then make the unit consistent with field data.
Figure F-12 shows the comparison of capacity values measured from the field against theoretical
estimates using the HCM methods. All observations yielded field measurements smaller than the estimated
capacities provided by the HCM. When different segment types are compared, however, no clear
conclusions can be drawn on which lanes have higher differences between field and estimated capacities.
The field measurements of capacity are, on average, 21.7% smaller than their respective HCM estimations.
It is a significant difference that can lead to inaccurate capacity analyses, as the HCM methodologies may
overestimate capacity and therefore overestimate the overall segment performance. For this reason, it is
recommended that capacity adjustment factors (CAFs) are applied to adjust the estimated capacities to local
conditions. Additional research is recommended to further investigate the calibration of CAFs.
374
Figure F-12. Field measured and HCM estimated capacity values, for (a) basic segments, (b) merge
segments and (c) diverge segments
375
With flow, capacity and FFS by lane determined, HCM equations can be used to estimate operating
speeds on individual lanes. Segment-wise inputs of flow, capacity and FFS are based on the field
measurements, and the developed methods previously described are applied to estimate their distribution
among individual lanes.
For basic segments, average speed is determined as:
𝑆 = 𝐹𝐹𝑆 − ( )
(Equation F-29)
This model is applied to individual lanes, as the three key parameters (FFS, c and vp) are input by lane.
The breakpoint value (BP) is also determined for each lane (Equation F-30).
𝐵𝑃 = 1000 + 40 𝑥 75 − 𝐹𝐹𝑆 𝑥 𝐶𝐴𝐹 (Equation F-30)
It is worth noting that a capacity adjustment factor (CAF) is considered in the estimation of the
breakpoint. The HCM method defines the adjusted capacity cadj as the product of the base capacity by a
capacity adjustment factor (CAF), which typically reflects impacts of weather, incident, work zone, driver
population, and calibration adjustments.
𝑐 = 𝑐 𝑥 𝐶𝐴𝐹 (Equation F-91)
As field values of segment capacities were obtained, these can be inserted into Equation F-31 as the value
of adjusted capacity. Therefore, CAFs become the single unknown in the equation and can be easily
obtained.
Practical example
A practical example was developed to verify and illustrate the developed methodology. A 2-lane basic
segment was modeled and the lane-by-lane performance is compared to field data (CA-1 NB – Santa Cruz,
CA). Field measured parameters are as follows:
• Free-flow speed: 69.1 mph
• Capacity: 3993 veh/h (1996.5 veh/h/ln)
• % heavy vehicles: 1.7
• Grade: 3% (rolling)
By applying the multiplying factors obtained in Table F-7 to the segment FFS, individual FFS can be
obtained as follows:
FFS1 = FFS x 0.965 = 69.1 x 0.965 = 66.68 mph
FFS2 = FFS x 1.032 = 69.1 x 1.032 = 71.31 mph
Next, lane capacities are obtained by applying the multiplying factors obtained in Figure F-11 to the
capacity as follows:
c1 = c x 44% = 3993 x 44% = 1757 veh/h
c2 = c x 56% = 3993 x 56% = 2236 veh/h
For comparison purposes, HCM methods would obtain the following theoretical capacity value:
c = [2200 + 10 x (FFS – 50)] x fHV = [2200 + 10 x (69.1 - 50) )] x 0.967 = 2312 veh/h/ln
Therefore, the recommended CAF for this location is obtained by dividing field measured by theoretical
values of capacity:
CAF = cadj/c = 1996.5/2312 = 0.864
376
Flows on each lane can be obtained by applying the model described in Equation F-1 to the flow rate
entering the segment. Next, speeds on individual lanes using the speed-flow relationship described in
Equation F-29. For this location, a sample of 14690 observations (15-min each) was randomly selected,
and then predicted values are compared to field data in Figure F-12.
Figure F-13. Field vs. predicted speed-flow curve for (a) Lane 1 and (b) Lane 2 (CA-1 NB – Santa
Cruz, CA)
As observed, the individual speed-flow models can replicate field conditions with good accuracy.
Naturally, the oversaturated portion of the curve cannot be addressed by the model, as this is already a
limitation of the existing method.
377
APPENDIX G
Glossary
General Acronyms
𝐴𝑇𝐷𝑀 = Active Traffic and Demand Management;
𝐴𝑊𝑆𝐶 = all-way stop controlled intersection;
𝐷𝐷𝐼 = diverging diamond interchange;
𝐷𝑂𝑇 = Department of Transportation;
𝐸𝐵 = eastbound;
𝐹𝐻𝑊𝐴 = Federal Highway Administration;
𝐻𝐶𝑀 = Highway Capacity Manual;
𝐼𝐶𝑀 = Integrated Corridor Management;
𝐼𝑅𝑇 = interchange ramp terminal;
𝑀𝑂𝐸 = Measures of Effectiveness;
𝑁𝐵 = northbound;
𝑄𝐴𝑃 = queue accumulation polygon;
𝑆𝐵 = southbound;
𝑆𝑃𝑈𝐼 = single-point urban interchange;
𝑇𝑊𝑆𝐶 = two-way stop controlled intersection;
𝑊𝐵 = westbound.
General Values
𝑑/𝑐 = demand to capacity ratio;
𝐸𝐷 = expected demand (veh/h);
𝑓 = adjustment factor for heavy vehicle presence;
𝐹𝐹𝑆 = free flow speed (mi/h);
𝐺 = grade (%);
𝐿𝑂𝑆 = level of service;
𝑂 − 𝐷 = origin-destination;
𝑃𝐻𝐹 = peak hour factor;
𝑃𝑀𝑇 = personal miles traveled (mi);
𝑣/𝑐 = volume/capacity ratio
𝑉𝑀𝑇 = vehicle miles traveled (mi).
378
1. Arterial Facilities
379
1.5. Roundabouts
𝜆( ), = maximum throughput for (a) movement (pc/h);
𝑐 ,( ) = entry lane capacity for roundabout approach, where (a) = approach direction (pc/h);
𝑑( ) = average control delay for (a) approach (sec/veh);
𝑑 = additional delay due to on-ramp spillback (sec/veh);
ℎ , = departure headway for the major street left turn (s);
ℎ , = departure headway for the major street right turn (s);
ℎ , = departure headway for the minor street through (s);
ℎ = departure saturation headway into the on-ramp (s/veh);
p(a)- of demand from NB approach a into the on-ramp
𝑄 = expected queue length QSP along the on-ramp during a 15-minute period analysis;
𝑄 = total number of vehicles queued during a 15-minute time period analysis;
𝑣( ), = flow rate for roundabout approach a (veh/h)
𝑣 ,( ), = circulating flow rate for roundabout approach a (veh/h)
𝑣 ,( ), = entry flow rate for roundabout approach a (veh/h)
total number of vehicles queued during a 15-minute time period analysis
2. Freeway Facilities
380
381
382
383