Citation
A Methodology for Evaluating Spillback from Freeway Diverge Segments for Application in the HCM

Material Information

Title:
A Methodology for Evaluating Spillback from Freeway Diverge Segments for Application in the HCM
Creator:
Armstrong, Michael D
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (82 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Civil Engineering
Civil and Coastal Engineering
Committee Chair:
ELEFTERIADOU,AGELIKI
Committee Co-Chair:
SRINIVASAN,SIVARAMAKRISHNAN
Committee Members:
SAMPSON,WILLIAM M
Graduation Date:
12/18/2015

Subjects

Subjects / Keywords:
Average travel speed ( jstor )
Data collection ( jstor )
Flow velocity ( jstor )
Freeways ( jstor )
Highway ramps ( jstor )
Modeling ( jstor )
Roads ( jstor )
Speed ( jstor )
Travel ( jstor )
Vehicles ( jstor )
Civil and Coastal Engineering -- Dissertations, Academic -- UF
arterial -- capacity -- density -- freeway -- hcm -- highway -- intersection -- queueing -- regression -- speed -- spillback
City of St. Augustine ( local )
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Civil Engineering thesis, M.S.

Notes

Abstract:
The methodologies in the current edition of the Highway Capacity Manual (HCM 2010) are capable of predicting performance measures and level of service (LOS) for freeway facilities and surface streets - Volume 2: Uninterrupted Flow and Volume 3: Interrupted Flow, respectively - in a macroscopic and deterministic environment. However, these two types of facilities often interact with one another in physical transportation systems, so the assumption that they are mutually exclusive is not realistic. Transportation agencies at the federal and state levels are interested in evaluating the long-term performance and reliability of freeway facilities that provide the necessary connectivity to inter-city commuters, especially in urbanized areas. Without the use of microscopic simulation tools, however, there currently exists no methodology to consider the potential effects of operations at adjacent surface streets. The purpose of this project is to develop a new method for evaluating traffic operations within a freeway diverge segment while considering the effects of excessive queuing and spillback at an off-ramp. Volumes and speed data were collected at several isolated diverge segments featuring three mainline lanes that experienced spillback conditions due to excessive demand at the downstream interchange ramp terminal. Based on the qualitative observations obtained from the data collection efforts, the improved framework is proposed - the most important of these concepts being the variable influence area dictated by the pre-defined regime thresholds. The concepts within the framework are then quantitatively compared against the collected field data, adjusted to facilitate compliance with standard HCM-based metrics. Finally, the inferences taken from this comparison, as well as the recommendations for future research, are discussed. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (M.S.)--University of Florida, 2015.
Local:
Adviser: ELEFTERIADOU,AGELIKI.
Local:
Co-adviser: SRINIVASAN,SIVARAMAKRISHNAN.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2016-12-31
Statement of Responsibility:
by Michael D Armstrong.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
12/31/2016
Classification:
LD1780 2015 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

A METHODOLOGY FOR EVALUATING SPILLBACK FROM FREEWAY DIVERGE SEGMENTS FOR APPLICATION IN THE HCM By MICHAEL D. ARMSTRONG A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2015

PAGE 2

© 2015 Michael D. Armstrong

PAGE 3

To my parents and extended family thank you for your unwavering support of my academic and career goals!

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my advisor, Dr. Lily Elefteriadou, for her guidance and reviewing efforts throughout the writing process, and to my committee members, Professor Bill Sampson, Dr. Yafeng Yin and Dr. Siva Srinivasan, for their insightful advice and rec ommendations. I would also like to thank Max Elliott (City of Gainesville), Ryan Crist, Jesse Gilmour and Michael Harper (FDOT Jacksonville Urban Office), and Michelle Young, Joey Gordon and Derek Vollmer (FDOT Central Office) for their assistance in the d ata collection process. Finally, special thanks to Jeff rey Dayton, Kirk Stull and Jonathan Henderson (HDR Engineering ) for their patience and understanding , and allowing me to use company owned computers to complete this project.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTI ON ................................ ................................ ................................ .... 13 1.1 Problem Statement ................................ ................................ ........................... 13 1.2 Research Objectives ................................ ................................ ......................... 13 1.3 Organization ................................ ................................ ................................ ..... 14 2 LITERATURE REVIEW ................................ ................................ .......................... 15 3 DATA COLLECTION ................................ ................................ .............................. 21 3.1 Overview of Study Sites ................................ ................................ .................... 21 3.2 Qualitative Trends ................................ ................................ ............................. 23 3.3 Data Preparation ................................ ................................ ............................... 25 4 METHODOLOGY ................................ ................................ ................................ ... 33 4.1 Intersection Queues Affecting Diverge Segments ................................ ............ 33 4.2 Additional Queue Length ................................ ................................ .................. 36 4.3 Spillback Effects on Operational Measures within Diverge Segments .............. 37 4.3.1 Queue Regimes ................................ ................................ ...................... 38 4.3.2 Equilibrium Separation Distance ................................ ............................. 40 4.3.3 Lane Utilization ................................ ................................ ........................ 42 4.3.4 Average Travel Speed ................................ ................................ ............. 43 4.3.5 Density and Level of Service (LOS) ................................ ......................... 45 4.4 Capacity Checks and Adjustments ................................ ................................ ... 46 4.4.1 Base Capacity ................................ ................................ ......................... 47 4.4.2 Capacity Adjustment Factor ( CAF ) ................................ .......................... 47 4.4.3 Probability of Lane Blockage ( P B ) ................................ ............................ 48 4.4.4 Final Adjusted Capacity ................................ ................................ ........... 49

PAGE 6

6 5 QUANTITATIVE ANALYSIS ................................ ................................ ................... 55 5.1 Summary of Data Collected ................................ ................................ .............. 55 5.2 Comparison of Performance Measures ................................ ............................ 57 5.2.1 Comparison of Lane Utilization of Non Exiting Vehicles .......................... 58 5.2.2 Comparison of Average Travel Speed ................................ ..................... 58 5.2.3 Comparison of Density ................................ ................................ ............ 59 5.3 Speed Flow Relationship ................................ ................................ .................. 59 5.4 Analysis of Lane Utilization Trends ................................ ................................ ... 60 5.5 Probability of Lane Blockage ( P B ) ................................ ................................ ..... 62 6 CONCLUSIONS AND RECOMMENDATIONS ................................ ....................... 76 LIST OF REFERENCES ................................ ................................ ............................... 81 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 82

PAGE 7

7 LIST OF TABLES Table page 3 1 Study sites ................................ ................................ ................................ .......... 21 3 2 Values used for the proportions of heavy vehicles in the traffic stream .............. 26 3 3 Site specific measurement values used in data processing ............................... 27 4 1 Appropri ate procedures to use to quantify queue length ................................ .... 34 5 1 Summary of data collection by regime category ................................ ................. 55 5 2 Data collected at I 75 S B / SR 26 (Newberry Road) (11/5/2014) ....................... 56 5 3 Data collected at I 75 S B / SR 26 (Newberry Road) (11/6/2014) ....................... 56 5 4 Data col lected at I 95 S B / Old St. Augustine Road (5/14/2015) ........................ 56 5 5 Data collected at I 95 S B / Old St. Augustine Road (7/9/2015) .......................... 56 5 6 Data collected at I 95 S B / I 295 East West Split (7/9/2015) .............................. 57 5 7 Data collected at I 95 S B / I 295 East West Split (8/13/2015) ............................ 57 5 8 Data collected at I 95 S B / SR 202 (JTB Boulevard) (8/13/2015) ....................... 57 5 9 Comparison of P FD re gression model coefficients ................................ .............. 61 5 10 Comparison of P L1 regression model coefficients ................................ ............... 61 5 11 Comparison of P L1 regression model statistical measures ................................ . 61 5 12 Comparison of demands under lane blockage conditions ................................ .. 62 5 13 Comparison of non exiting lane utilization under lane blockage conditions ........ 62 5 14 Comparison of flow rates under lane blockage conditions ................................ .. 63 5 15 Comparison of speed and density under lane blockage conditions .................... 63 6 1 Critical metrics needed for spillback conditions to be developed in future research ................................ ................................ ................................ ............. 79

PAGE 8

8 LIST OF FIGURES Figure page 3 1 Map of the I 75 S B / SR 26 (Newberry Road) study site ................................ .... 28 3 2 I 75 S B / SR 26 (Newberry Road) study site ................................ ...................... 29 3 3 Overview map of Jacksonville study sites ................................ .......................... 29 3 4 I 95 S B / SR 202 (JTB Boulevard) study sit e ................................ ...................... 30 3 5 I 95 S B / I 295 East West Split study site ................................ ........................... 30 3 6 I 95 S B / Old St. Augustine Road study site ................................ ....................... 31 3 7 Still image captured from video feed of queued drivers ignoring lane striping denoting the end of the deceleration lane ................................ ........................... 31 3 8 Still image captured from video feed of drivers attempting to merge into the queue on the shoulder, temporarily blocking Lane 1 ................................ .......... 32 4 1 The upstream limit of the available queue storage distance ............................... 49 4 2 Extended queue storage length ................................ ................................ .......... 50 4 3 Regime 1 spillback and influence areas at a diverge segment ........................... 50 4 4 Regime 2 spillback and influence areas at a diverge segment ........................... 51 4 5 Regime 3 spillba ck and influence areas at a diverge segment ........................... 51 4 6 Regime 4 spillback and influence areas at a diverge segment ........................... 51 4 7 Possible scenarios under which L EQ is evaluated ................................ ............... 52 4 8 Expected lane utilization te ndencies of non exiting drivers within the influence area ................................ ................................ ................................ ..... 52 4 9 Capacity of Ramp Freeway Junctions, pc/h (HCM 2010 Exhibit 13 8) ............... 53 4 10 Distance to the adjacent downstream off ramp ( L DOWN ) ................................ ..... 53 4 11 Additional queue length ( Q a ) as it relates to capacity reduction .......................... 54 5 1 P FD at I 75 S B / SR 26 (Newberry Road) study site (11/5/2014) ......................... 63 5 2 P FD at I 75 S B / SR 26 (Newberry Road) study site (11/6/2014) ......................... 64

PAGE 9

9 5 3 P FD at I 95 S B / Old St. Augustine Road study site (5/14/2015) ......................... 64 5 4 P FD at I 95 S B / Old St. Augustine Road study site (7/9/2015 ) ........................... 65 5 5 P FD at I 95 S B / I 295 East West Split study site (7/9/2015) ............................... 65 5 6 P FD at I 95 S B / I 295 East West Split study site (8/13/2015) ............................. 66 5 7 S at I 75 S B / SR 26 (Newberry Road) study site (11/5/2014) ............................ 66 5 8 S at I 75 S B / SR 26 (Newberry Road) study site (11/6/2014) ............................ 67 5 9 S at I 95 S B / Old St. Augustine Road study site (5/14/2015) ............................ 67 5 10 S at I 95 S B / Old St. Augustine Road study site (7/9/2015) .............................. 68 5 11 S at I 95 S B / I 295 East West Split study site (7/9/2015) ................................ .. 68 5 12 S at I 95 S B / I 295 East West Split study site (8/13/2015) ................................ 69 5 13 D at I 75 S B / SR 26 (Newberry Road) study site (11/5/2014) ........................... 69 5 14 D at I 75 S B / SR 26 (Newberry Road) study site (11/6/2014) ........................... 70 5 15 D at I 95 S B / Old St. Augustine Road study site (5/14/2015) ............................ 70 5 16 D at I 95 S B / Old St. Augustine Road study site (7/9/2015) .............................. 71 5 17 D at I 95 S B / I 295 East West Split study site (7/9/2015) ................................ .. 71 5 18 D at I 95 S B / I 295 East West Split study site (8/13/2015) ................................ 72 5 19 Speed flow plot at I 75 S B / SR 26 (Newberry Road) study site (11/5/2015) ..... 72 5 20 Speed flow plot at I 75 S B / SR 26 (Newberry Road) study site (11/6/2015) ..... 73 5 21 Speed flow plot at I 95 S B / Old St. Augustine Road study site (5/14/2015) ...... 73 5 22 Speed flow plot at I 95 S B / Old St. Augustine Road study site (7/9/2015) ........ 74 5 23 Speed flow plot at I 95 S B / I 295 East West Split study site (7/9/2015) ............ 74 5 24 Speed flow plot at I 95 S B / I 295 East West Split study site (8/13/2015) .......... 75 5 25 Speed flow plot at I 95 S B / SR 202 (JTB Boulevard) study site (8/13/2015) ..... 75

PAGE 10

10 LIST OF ABBREVIATIONS AWSC All Way Stop Controlled CAF Capacity Adjustment Factor CATT Center for Advanced Transportation Technology FDOT Florida Department of Transportation FFS Free Flow Speed FTO Florida Traffic Online HCM Highway Capacity Manual LOS Level of Service O D Origin Destination PCE Passenger Car Equivalent PHF Peak Hour Factor QSR Queue Storage Ratio RITIS Regional Integrated Transportation Information System RV Recreational Vehicle SR State Road TRD Total Ramp Density TWSC Two Way Stop Controlled

PAGE 11

11 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science A METHODOLOGY FOR EVALUATING SPILLBACK FROM FREEWAY DIVERGE SEGMENTS FOR APPLICATION IN THE HCM By Michael Armstrong December 2015 Chair: Lily A. Elefteriadou Major: Civil Engineering The methodologies in the current edition of the Highway Capacity Manual (HCM 2010) are capable of predicting performance measures and level of service (LOS) for freeway facilities and surface streets Volume 2: Uninterrupted Flow and Volume 3: Interrupted Flow , respectivel y in a macroscopic and deterministic environment . However, these two types of facilities often interact with one another in physical transportation systems, so the assumption that they are mutually exclusive is not realistic. Transportation agencies at the federal and state level s are interested in evaluating the long term performance and reliability of freeway facilities that provide the necessary connectivity to inter city commuters , especially in urbanized areas . Without the use of microscopic simulation tools, however, there currently exists no methodology to consider the potential effects of operations at adjacent surface streets. The purpose of this project is to develop a new method for evaluating traffic ope rations within a freeway diverge segment while considering the effects of excessive queuing and spillback at an off ramp.

PAGE 12

12 Volumes and speed data were collected at several isolated diverge segments featuring three mainline lanes that experienced spillback conditions due to excessive demand at the downstream interchange ramp terminal. Based on the qualitative observations obtained from the data collection efforts, the improved framework is proposed the most important of these concepts being the variable in fluence area dictated by the pre defined regime thresholds. The concepts within the framework are then quantitatively compared against the collected field data, adjusted to facilitate compliance with standard HCM based metrics. Finally, the inferences take n from this comparison, as well as the recommendations for future research, are discussed.

PAGE 13

13 CHAPTER 1 INTRODUCTION 1.1 Problem Statement The procedures detailed in the current version of the Highway Capacity Manual (HCM 2010) estimate capacity and other performance measures dictating level of service (LOS) for freeway facilities as well as surface streets. However, the existing methods do not consider cases in which interaction s in the form of spillback occur from one type of facility to another. One case of these interactions is excessive demand at a signalized intersection leading to spillback at an adjoining off ramp and , eventually , the freeway. The existing procedure for Si gnalized Intersections (HCM 2010 Chapters 18 and 31) predicts both the average and maximum expected queue length at an approach within a specified analysis period , given any combination of geometric or traffic related variable inputs within the scope of t he methodology. As previously implied, the effects of these excessive queues that propagate upstream onto a freeway mainline are not accounted for in the existing Diverge Segments procedure (HCM 2010 Chapter 13) . 1.2 Research Objectives The purpose of this thesis is to propose a new framework to improve the existing Diverge Segments procedure , as detailed in HCM 2010 Chapter 13 , to address spillback conditions. This involves qualitatively observing driver behavior, developing quantitative metrics to model these behaviors and exploring what factors influence performance under spillback conditions. These modifications consider lane utilization along the freeway mainline , average travel speed and density by lane and the corresponding capacity drop . They consist of restructuring existing equations and reference tables as well as development of new equations and procedures. D ata

PAGE 14

14 collection (detailed in Chapter 3 ) is conducted at four different freeway diverge segments to validate the proposed modifications and obtain quantitative estimates of various parameters. 1.3 Organization Chapter 2 provides a literatu re review related to this topic, Chapter 3 details the data collection performed and Chapter 4 d escribes the proposed queue length estimation procedure followed by the proposed d iverge r amp analysis procedure . C hapter 5 describes the analysis techniques used to evaluate the proposed d iverge s egment s methods and presents the quantitative results. Chapter 6 discusses the conclusions and recommendations att ained from this project.

PAGE 15

15 CHAPTER 2 LITERATURE REVIEW A review of the literature regarding operational effects of freeway arterial interactions yielded a limited number of publications. The existing HCM 2010 procedure for Diverge Segments predicts lane uti lization, average travel speed, capacity and density (which dictates LOS) based on geometric and demand flow data for for effects from or elicited upon adjoining upstream and downstream segments of the freeway facility and surface street network. Although much of the currently published research on this topic does not directly address performance measures in a manner consistent with HCM 2010 procedures and LOS metrics, the concepts and framework presented in the research are useful in developing the app ropriate procedural adjustments. For example, previous research has quantified the average delay incurred to freeway mainline vehicles as a result of spillback from an off ra mp, but the HCM 2010 Freeway Facilities procedure does not estimate delay explicitly. Rather, average travel speed is estimated ; the difference between this flow speed could conceivably be used to estimate delay instead . Shockwave an alysis, which considers the effects of queuing upstream of a bottleneck on a freeway facility, has been used to evaluate operational impacts as a function of capacity deficiency at an off ramp. For general types of bottlenecks, Lighthill, et al. (1956) pro posed examining operations in time and space for freeway segments. Expanding on this idea while using a series of simplifying assumptions, Newell (1993) described a method of predicting the resultant delay and cumulative queued vehicle

PAGE 16

16 count given any comb capacity using the principles outlined in the Lighthill Whitham Richards (L W R) model. Daganzo (1997) later pointed out that the original L W R model assumes that operational effects and capac ity deficiency originated from only one source: bottlenecks and queuing at the downstream location on the freeway mainline. He further postulated that capacity deficiency and queuing at a downstream off ramp can also cause operational effects on the freewa y mainline. Thus, t wo separate sets of traffic conditions can exist on a single freeway segment , and queued vehicles can be evaluated separately from un queued mainline vehicles in a macroscopic shockwave analysis. He pipe regime , in which that are restricted to the lane adjacent to the off ramp. Newell (1999) constructed a graphical olygons that quantify the evolution of operational changes to freeway operations for both categories of vehicles namely, delay (perhaps implying a reduction in travel speed) and cumulative queued vehicle count over the ramp queue. Newell additionally posited several potential driver from switching between co near and accept a minimal perceived delay before continuing past the diverge point if the queue is relatively shor be evaluated under the First In First Out (FIFO) queuing theory method, since drivers at the back of a particularly

PAGE 17

17 long queue may attempt to bypass parts of the queue and squeeze back in closer to the exit. 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 caused a significant reduction in discharge flows downstream of the exit. During the queued hour on a three lane freeway mainline (only 62.8% o f per hour, according to the Basic Freeway Segments procedure) was observed. Note that this rate should be while the current version of the HCM uses a generic capacity value that does not consider the two capacity phenomenon. The proportion of exiting vehicles on the freeway also proved to have a s ignificant effect on capacity. O n average, discharge rates increased from 4,520 to 5,720 vehicles per hour (an increase of 26.5%) when the proportion of exiting vehicles decreased from 29% to 24%, even though the actual flow rate of exiting vehicles remain ed nearly constant. V ehicles were found to transition from free flow speeds to queuing the queue. It was also observed that drivers tend to adopt larger headway spacing over time , were collectively realized, driver aggression as a whole subsided. In a separate paper based on the data obtained in the previously described study, Munoz and Daganzo (2002) focused inst was found that the variation in speeds across the three mainline lanes was greatest

PAGE 18

18 closest to the diverge point, whereas occupancy detectors positioned further upstream indicated less variation. Furthermore, the righ tmost lanes closer to the off ramp queue were more influence d by the spillback and had reduced travel speed, whereas the leftmost lane(s) showed little to no difference between queued conditions and free flow conditions. N on exiting vehicles traveling in t he rightmost lane(s) in the vicinity of the spacing Regarding capacity reduction, an average discharge flow of 1,500 vehicles per hour per lane was recorded immediately beyond the diverge point estimated by the authors) could potentially accommodate. A similar study by Cassidy et al . (2002) also found that, in general, l onger exit ing queues from an 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 obst ructed non exiting vehicles by attempting to force their way into the queue from the adjacent lane further downstream rather than wait ing to be serviced in the queue . The authors do not discuss the possible correlation between queue length and forced queue entries. The presence of a queue also affected the non queuing 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. It is conceivable that traffic operations during an incident may be similar to operations when a queue is present, and thus we briefly review here studies related to

PAGE 19

19 the capacity an d traffic operational quality during incidents. In a comprehensive comparison between various incident related studies in the literature, Lu and conditions. According to the regression analysis developed based on past studies, capacity is reduced by 320 veh icles per hour simply by the occ urrence of congestion, 1,948 vehicles per hour from only the mainline shoulder being affected, 2,329 veh icles per h our from one lane being a ffected and 2,608 veh icles per hour from two lanes being blocked. A similar regression structure could conceivably be developed for queuing related capacity reduction, albeit one that incorporates the probability of further lane blockage in consideration o f the random and fluid nature of queue fluctuation. 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 potential driver behaviors. 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 pres ence of an off ramp queue, with one study observing 4,520 vehicles per hour on a three lane freeway mainline (approximately 1,500 veh icles per hour per lane ) Discharge rates along the mainline increase with decreasing off ramp flow; they were also found t o increase with decreasing queue lengths, which are correlated with off ramp demands Rightmost lanes are more affected regarding speed reduction, whereas the leftmost lane(s) show very little difference between the presence of a queue and free flow conditi ons 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

PAGE 20

20 The presence of a queue at the off ramp reduces the m with values observed as low as 25 kilometers per hour (15.5 miles per hour) 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

PAGE 21

21 CHAPTER 3 DATA COLLECTION T o gain insight to how traffic operations within diverge segments change under spillback conditions, data was collected at four different diverge junctions during time periods in which spillback occurred. The next section provides an overview of the study s ites and data collection effort while the second section s ummariz es observed trends in the data and the final section discusses the methods used for data preparation. 3.1 Overview of Study Sites T o gain insight to how traffic operations within diverge segm ents change under spillback conditions, data is collected at the diverge points of four different interchange junctions during time periods in which spillback occurred. The first study site is located on I 75 in Gainesville, FL, while the other three site s are all located on I 95 in Jacksonville, FL. The spillback conditions observed at the study site in Gainesville were considered to be unusual and possibly unexpected, whereas those observed in Jacksonville occur on a regular basis as a result of excessiv e demands on I 95 exiting the downtown Jacksonville area during the weekday afternoon peak period s . A summary of the dates and times when data collection took place is shown in Table 3 1 . Table 3 1. Study sites Freeway/Direction Surface Street Date Time Period Location I 75/Southbound SR 26 ( Newberry Rd ) 11/5/2014 7:01 8:15AM Gainesville, FL I 75/Southbound SR 26 ( Newberry Rd ) 11/6/2014 7:01 8:15AM Gainesville, FL I 95/Southbound Old St. Augustine Rd 5/14/2015 4:15 6:17PM Jacksonville, FL I 95/Southbound Old St. Augustine Rd 7/9/2015 4:40 6:43PM Jacksonville, FL I 95/Southbound I 295 East W est Split 7/9/2015 5:24 6:43PM Jacksonville, FL I 95/Southbound I 295 E ast W est Split 8/13/2015 4:26 5:27PM Jacksonville, FL I 95/Southbound SR 202 ( JTB B lvd) 8/13/2015 4:26 5:27PM Jacksonville, FL

PAGE 22

22 Figure 3 1 shows a map of the I 75 Southbound / SR 26 (Newberry Road) stu dy site in Gainesville , FL , while Figure 3 2 shows the geometric layout of the diverge segment. During the two time periods that data collection efforts took place, spillback occurred for 19 minutes and 4 minutes, respectively. Volumes along I 75 were relatively low (approximately 2,400 vehicles per hour), and it did not appear that operations along the freeway mainline were severely compromised in spite of the queuing . Figure 3 3 shows a map of the three study sites in Jacksonville along I 95, while Figures 3 4, 3 5 and 3 6 show the respective geometric layouts of the diverge segments. At the I 95 S outhbound / SR 202 (JTB Boulevard) study site (Figure 3 4), just one period was observed, and spillback occurred for the entire 62 minutes. This interchange junction is regarded as one of the most severe bottlenecks throughout the Jacksonville urban area a s a result of its proximity to downtown Jacksonville and connectivity to Jacksonville Beach heading eastbound (Figure 3 3). High volumes were observed (approximately 1,600 vehicles per hour per lane), and the off ramp queue remained very lon g throughout th e study period. At the I 95 Southbound / I 295 East West Split study site (Figure 3 5), spillback conditions occurred during only one of the two time periods observed. Volumes along the mainline were very high ( 1 , 5 00 vehicles per hour per lane and 1 , 9 00 vehicles per hour per lane , respectively) and the off ramp queue remained so long throughout the study periods that exiting demand could not be accurately quantified. At the Old St. Augustine Road study site, spillback conditions were observed during both periods for a total of 73 minutes and 50 minutes, respectively. Volumes

PAGE 23

23 were moderately high along the mainline (approximately 1,400 vehicles per hour per visibility of anything beyond 150 feet upstream of the diverge point (Figure 3 6), it was impossible to quantify accurately exiting demands. Data for average travel speeds were collected externally for this project. Travel speed data for the study periods downloaded an d assembled in one minute intervals from the Regional Integrated Transportation Information System (RITIS). RITIS is an automated data sharing, dissemination, and archiving system maintained by the n Technology (CATT) Laboratory, collected using HERE® (formerly NAVTEQ) speed detection technology. 3. 2 Qualitative Trends During the data collection process, the following qualitative trends were observed at the study sites: At the I 75 Southbound / SR 26 ( Newberry Road ) study site where the back of queue was visible throughout the study period, e xiting drivers entering the back deceleration lane . Instead, they used the availa shoulder when the queue of vehicles became longer than the allotted deceleration lane, rather than using the right most through lane on the freeway mainline and causing blockage (Figure 3 7 ) At the I 75 Southbound / SR 26 (Newberry Road) study site, d rivers appeared to travel at lower speeds when using the right most lane when the queue was present in the adjacent deceleration lane and/or shoulder . At the three study sites in Jacksonville, drivers appeared to be more aggre ssive in maintaining their desired speed s in the right most lane in spite of the queuing along the deceleration lane I n the first I 75 Southbound / SR 26 (Newberry Road) study period and the first I 95 Southbound / I 295 E ast W est Split study period, sever al drivers were unable to make the appropriate lane change(s) to enter the back of the queue . Instead, they blocking the right most through lane of the freeway mainline (Figure 3 8 )

PAGE 24

24 In the same two study periods as referenced in the previous observation, s ome drivers towards the back of the queue would exit the queue and attempt to re enter at a different point in the queue further downstream, again effectively blocking the right most through lane of the freeway mainline (Figure 3 8 ) In several study periods involving lengthy intervals of spillback conditions, s ome drivers, particularly at times when the queue grew to exceptionally long lengths, resorted to exiting the queue and (presumably) opting for an exit located further downstream Some admonitions must be considered regarding the quality of the data collected for this analysis: At the three study sites in Jacksonville, the vantage point of the camera recording the video feed did not allow for observation of operations upstream of the diverge area when queues were extensive. For example, the bridge overpass immediately upstream of the diverge point at Old St. Augustine Road only allowed for approximately 150 feet of the diverg e segment to be observed (the deceleration lane at this site is approximately 750 f ee t long) At all three study sites in Jacksonville, queue s extended v antage point , thus queue lengths or demands to the off ramp could not be observed . This is critical because HCM freeway analysis requires each demand to be used (the hourly rate of vehicles intending to use a particular facility within the defined time interval), rather than the throughput T he travel speed data from RITIS is a measurement of speed for the entire diverge segment along the freeway (i.e., all three mainline lanes) . Therefore, this data cannot be compared directly the HCM prediction of speed for the influence area ( S R ), but rather to the total travel speed ( S ) Spe ed data from RITIS wa s not recorded for every individual minute interval within the study periods . For the minute i ntervals not associated with a given speed from RITIS , the average of the previous and subsequent time interval is assumed T he fixed physical areas where travel speed data are collected by RITIS do not precisely coincide with the HCM defined influence area (1,500 feet upstream of the diverge point). For example, one data collection interval may encompass the first 400 feet of t he influence area while the next adjacent data collection interval encompasses the remaining 1,100 feet of the influence area. The data collection interval that encompassed a greater proportion of the HCM defined influence area was chosen to obtain the tra vel speed data for the given study site

PAGE 25

25 3.3 Data Preparation Lane specific vehicle counts are recorded in one minute intervals at each of the study sites, which are converted to hourly flow rates expressed in units of passenger cars. These lane specific fl ow rates are then organized into standard HCM flow rate metrics: total ( v F ) and exiting ( v R ) flow rate, and flow entering the influence area ( v 12 ) and the outer lanes ( v OA ). Using these flow rate metrics, the following HCM performance measures were derived : L ane utilization of non exiting vehicles ( P FD ), average travel speed ( S ) and density ( D ) . These were obtained to facilitate a comparison of HCM prediction models versus observed conditions P roportion of exiting vehicles ( v R / v F ) . This measure was obtaine d to evaluate the effects of ramp demand on lane utilization and average travel speed in the influence area The lane specific counts within each one minute interval are converted to hourly flow rates using E quation 3 1, a replication of HCM 2010 Equation 13 1 . ( 3 1 ) where v i = demand flow rate for movement i (pc/h), V i = demand volume for movement i (veh/ mine ), PHF = peak hour factor, f HV = adjustment for heavy vehicle presence (veh/pc), and f p = adjustment factor for driver population The adjustment factor for driver population ( f p ) is assumed to be equal to 1.00 to reflect the fact that the majority of drivers in the study interval were regular commuters. The adjustme nt factor for heavy vehicle presence ( f HV ) is calculated using E quation 3 2, a replication of HCM 2010 Equation 11 3 .

PAGE 26

26 (3 2) where f HV = heavy vehicle adjustment factor (veh/pc) , P T = proportion of trucks and buses in the traffic stream, P R = proportion of RVs in traffic stream , E T = PCE of one truck or bus in traffic stream (pc/veh) , and E R = PCE of one RV in traffic stream (pc/veh) It is assumed that the terrain at all four study sites is L evel, typical of roadways that do not have a significant amount of elevation change throughout the study segment. Therefore, based on HCM 2010 Exhibit 11 10, the passenger car equivalent (PCE) values in this equation are 1.5 passenger cars per v ehicle and 1.2 passenger cars per vehicle for trucks and buses ( E T E R ), respectively . The proportion of trucks and buses ( P T P R ) in the traffic stream are based on data acquired Traffic Online (FTO) application, publicly accessible on the FDOT website. The values used for each of the study sites are shown in Table 3 2 . Table 3 2 . Values used for the proportions of hea vy vehicles in the traffic stream Freeway/Direction Surface Street P T P R I 75/Southbound SR 26 ( Newberry R oa d ) 0.185 0.000 I 95/Southbound Old St. Augustine R oad 0.121 0.000 I 95/Southbound I 295 East W est Split 0.079 0.000 I 95/Southbound SR 202 ( JTB B ou l e v ar d ) 0.079 0.000 S everal important site specific measurements are made to ensure the collected data were compliant with the HCM methodology . T he free flow speeds of the off ramps ( S FR ) are assumed to be 5 mi/h greater than the posted advisory speed limit , obtained from Google Maps . The free flow speed of the freeway mainlines ( FFS ) are calculated

PAGE 27

27 using Equation 3 3, a replication of HCM 2010 Equati on 11 1 ; Total Ramp Density ( TRD ) and Decel eration Lane Length ( L D ) values were obtained using Google Maps. These measurement values are shown in Table 3 3 . (3 3) where f LW = adjustment factor for lane width, assumed to be 0 mi/ h, f LC = adjustment factor for lateral clearance, assumed to be 0 mi/h, and TRD = total ramp density (ramps/mi) Table 3 3 . Site specific measurement values used in data processing Freeway/Direction Surface Street Total Ramp Density FFS, Mainline FFS, Ramp Deceleration Lane Length I 75/Southbound SR 26/Newberry Rd 0.7 ramps/mi 73.1 mi/h 25.0 mi/h 756 feet I 95/Southbound Old St. Augustine Rd 1.2 ramps/mi 71.7 mi/h 30.0 mi/h 763 feet I 95/Southbound I 295 E/W Split 1.2 ramps/mi 71.7 mi/h 50.0 mi/h 1,965 feet I 95/Southbound SR 202/JTB Blvd 1.2 ramps/mi 71.7 mi/h 50.0 mi/h 2,470 feet Finally, three performance measures predicted in the Diverge Segments procedure are calculated using the collected data as follows: Lane utilization of non exiting vehicles ( P FD ) is calculated using Equation 3 4, a replication of HCM 2010 Equation 13 9 (3 4) O verall average travel speed ( S ) is calculated using the prediction model represented in HCM Exhibits 13 12 and 13 13 T otal density ( D ) is ca lculated using the fundamental speed flow density relationship (Equation 3 5 ) , using the above speed prediction model (3 5) The following HCM metrics lane utilization and density were derived directly from field measured data.

PAGE 28

28 The proportion of non exiting vehicles traveling in Lanes 1 and 2 ( P FD ) was calculated with Equation 3 6, a re arrangement of HCM 2010 Equation 13 8 (3 6) Total density ( D ) was derived again by using the fundamental speed flow density relationship in Equation 3 5 , instead using field measured speed data A total of 601 minutes, or approximately 10 hours, of data were recorded and analyzed in this project. Results and subsequent analysis of the data collection efforts are discussed in Chapter 5 while suggestions for future data collection efforts are discussed in Chapter 6. Figure 3 1 . M ap of the I 75 Southbound / SR 26 (Newberry Road) study site . Image courtesy of Bing Maps

PAGE 29

29 Figure 3 2 . I 75 Southbound / SR 26 ( Newberry Road ) study site . Image courtesy of Bing Maps Figure 3 3. Overview map of Jacksonville study sites . Image courtesy of Bing Maps

PAGE 30

30 Figure 3 4. I 95 Southbound / SR 202 (JT B Boulevard ) study site . Image courtesy of B ing Maps Figure 3 5. I 95 Southbou nd / I 295 East West Split study site . Image courtesy of Bing Maps

PAGE 31

31 Figure 3 6 . I 95 Southbound / Old St. Augustine Road study site . Image courtesy of Bing Maps Figure 3 7. Still image captured from video feed of queued drivers ignoring lane striping denoting the end of the deceleration lane (I 75 SB / SR 26 study site) , April 20, 2 0 15. Courtesy of the author, Michael Armstrong

PAGE 32

32 Figure 3 8. Still image captured from video feed of drivers attempting to mer ge into the queue on the shoulder, temporarily blocking Lane 1 (I 75 SB / SR 26 study site) , April 20, 2015. Courtesy of the author, Michael Armstrong

PAGE 33

33 CHAPTER 4 METHODOLOGY This chapter discusses the various shortcomings of the existing Diverge Segments procedure with regards to spillback from adjoining surface street facilities, as well as the corresponding proposed modifications to address them. Some modifications are supported by qualitative observations (Chapter 3) and quantitative analysis (Chapter 5) while others have yet to be confirmed or could potentially be expanded upon by future data collection efforts . Section 4 . 1 outlines the existing procedures to estimate the corresponding maximum queue length, depending on the intersection configuration. Section 4 . 2 provides guidance in determining whether the resultant queue is long enough to affect operations on the upstream freeway diverge segment. Sections 4 . 3 (Operational Measures) and 4 . 4 (Capacity) discuss the specific ways in which this excessive queue length can be considered in evaluating the facility performance when spillback occurs. 4 . 1 Intersection Queues Affecting Diverge Segments Spillback at an off ramp may occur due to either inadequate capacity of the ramp proper or inadequate capacity at the ramp terminal (typically the signal at the downstream interchange). The c apacity of the ramp proper is defined as the of f maximum allowable hourly flow rate, which is based only on its geometric characteristics (number o f lanes, free flow speed, etc.) values can be found in HCM 2010 Exhibit 13 10. The capacity of the ramp terminal is defined as the collective maxim um achievable flow rate at the signalized or unsignalized approach to the surface street, based on a variety of geometric and signal phasing conditions. Evaluation of capacity and queue length in the latter case depends on the type of intersection that

PAGE 34

34 exi sts at the surface street on the downstream end of the off ramp. These intersection types and corresponding procedures are shown in Table 4 1. Table 4 1. Appr opriate procedures to use to quantify queue length Intersection Type HCM 2010 Reference Procedure Signalized Chapt er 31: Signalized Intersections Supplemental Section 3: Queue Accumulation Polygon TWSC Intersection Chapter 19: Two Way Stop Controlled Intersections Section 2: Methodology (Automobile Mode) AWSC Intersection Chapter 20: All Way Stop Controlled Intersections Section 2: Methodology (Automobile Mode) Roundabout Chapter 21: Roundabouts Section 2: Methodology (Automobile Mode) To compare the predicted queue length ( Q ) directly to the available queue storage distance ( L a ) and determine whether or not spillback is expected to occur, the queue storage ratio ( R Q ) is calculated. In the case of signalized intersections, the R Q calculation is included within the Queue Accumulation Polygon p rocedure. However, this is not the case in the unsignalized intersection procedures (T WSC, AWSC, Roundabouts), so R Q must be calculated separately. This value is based on the queue length ( Q ) obtained from the respective procedure along with several other input variables. If R Q is less than or e qual to 1.00, spillback is not expected to occur, and no further adjustments to the freeway facilities analysis are necessary. On the other hand, if R Q exceeds 1.00, the recommended modifications described in Section 4 . 3 and Section 4. 4 should be used . Not e that if the capacity of the off ramp proper is exceeded, spillback is certain to occur and , therefore , calculating R Q is redundant. Queue storag e rat io is estimated using Equation 4 1 . (4 1)

PAGE 35

35 where R Q = queue storage ratio, L h = average vehicle spacing in stationary queue (feet /veh), Q = predicted queue length (veh), L a = avai lable queue storage distance (feet /ln), and N = number of full lanes at approach (ln) Average vehicle spacing can be estimated by Equation 4 2 , a replicat ion of HCM 2010 Equation 31 149. (4 2) where L pc = average stored passenger car l ength (default value of 25) (feet ), L HV = average stored heavy vehicle l ength (default value of 45) (feet ), and P HV = percentage of heavy vehicles in the movement group (%) Since a single period analysis (15 minutes by default) predicts a single queue length value ( Q ), a multi period analysis of the intersection approach associated with the off ramp should be performe d to account for residual queuing. Given that a spillback is expected to occur within a single analysis period ( R Q > 1.00), analyses for the previous and subsequent time periods should be performed, if the appropriate data is available. The HCM FREEVAL 201 0 computational engine associated with the Freeway Facilities procedure (Chapters 10 and 25) considers residual effects of capacity reduction over any specified sequence of time periods. Therefore, even if data for only urface street approach being evaluated is available, the effects of that given period would be considered, propagating along multiple time periods and upstream freeway segments.

PAGE 36

36 4 . 2 Additional Queue Length T he available queue storage length ( L a ) along the off ramp begins at the intersection approach stop bar and extends upstream to the diverge point, where the off ramp meets the freeway mainline. The length of the deceleration lane(s) ( L D ) directly adjacent to the freeway mainline are not part of the av ailable queue storage distance (Figure 4 1 ) . This is because there is still some level of friction expected to occur between the queue and the freeway mainline if the queue spills out into the deceleration lane even without directly forming on Lane 1 of the freeway mainline. Posted speeds on freeways are typically much greater than posted advisory speeds at off ramps, so decrease travel speeds. Under spillback conditions, th is available space for deceleration is decreased, and the difference in speeds between that lane and adjacent freeway lanes becomes significant, which causes the friction mentioned above . Assuming that R Q exceeds 1.00 in the previous step, the additional queue length (denoted Q a ) which extends beyond the queue storage length within the off ramp is calculated. This is done simply by multiplying the queue storage length by the extent that R Q exceeds 1.00 , as shown in Equation 4 3 . (4 3) where Q a = additional queue length (feet ), R Q = queue storage ratio, and L a = available queue storage d istance within the off ramp (feet )

PAGE 37

37 If demand exceeds the capacity of the off ramp proper (according to HCM 2010 Exhibit 13 10), queuing is predicted to occur but is not explicitly quantified in any as shown in Equation 4 4 (all terms are previously defined). (4 4) When the additional queue length ( Q a ) exceeds the deceleration lane length ( L D ), drivers intending to enter the back of the queue are likely to position themselves on the shoulder (if sufficient la teral clearance is available). This is done to avoid dangerous situations in which high speed rear end collisions at the back of the queue are likely to occur. The shoulder is normally reserved for disabled vehic les, but can effectively certain geometric conditions found upstream (a narrow bridge, jersey walls bordering construction zones, etc.) may limit this extension, so an addi tional variable describing this distance must be defined denoted as L E . If the additional queue length ( Q a ) is predicted to exceed the extended queue storage length ( L E ), the growing queue will, in theory, be forced to block Lane 1 of the freeway mainlin e (Figure 4 2 ). 4 . 3 Spillback Effects on Operational Measures within Diverge Segments This section provides the recommended procedures for evaluating traffic operations at a diverge segment under spillback conditions. These recommendations would be implem ented as an enhancement of the existing Diverge Segments (HCM 2010 Chapter 13) and/or the Freeway Facilities (HCM 2010 Chapter 10) procedures.

PAGE 38

38 measures average travel speed an d density Facilities procedure. As such, it is highly recommended that this fully segmented, multi period procedure is used in favor of the single period Diverge Segments procedure to obtain more comprehensive resul ts and observe how the queue from off ramp spillback affects subsequent periods and/or upstream segments. 4 . 3 .1 Queue Regimes In undersaturated conditions, an off turbulence caused by exiting vehicles has sho wn to have the most pronounced effects o n freeway mainline operations. It is defined as the surface area encompassing the two rightmost mainline lanes (Lanes 1 and 2) and the deceleration lane, extending upstream from the diverge point by 1,500 feet. Key p erformance measures lane utilization ( P FD ), average travel speed ( S R ) and density ( D R ) are predicted within these influence area boundaries . When spillback occurs, this influence area is expected to be significantly altered , var ying laterally (across l anes) as well as longitudinally upstream as a function of the additional queue length ( Q a ). Under spillback conditions, interactions between exiting vehicles and mainline vehicles are expected to occur further upstream, as noted in several studies of diverge segments that regularly experienced spillback (Daganzo and Munoz, 2000). This section discusses the determination of the influence area thresholds when spillback occurs. There are four queue regimes, each of which may occur based on the estimated additional queue length ( Q a ) compared to the effective queue storage distance upstream of the diverge point ( L E ), determined using the procedure outlined in Section 4.3 . These regime conditions designate the influence area boundaries laterally and longitudinally across the diverge segmen t. They are defined as follows:

PAGE 39

39 Regime 1 ( Q a L D ): Under this regime, the additional queue ends within the deceleration lane and does not spill over into the mainline freeway. In between free flow speeds on the mainline (typically 55 75 m i/h) and advisory speeds posted on the off ramp proper (typically 25 40 mi/h). As a result, drivers expect to travel at moderately high speeds when traveling along the deceleration lane. When spillback occurs, the available deceleration distance is reduced . Under Regime 1, minimal turbulence is expected along the freeway mainline, and it is likely that only Lane 1 is affected (Figure 4 3 ) Regime 2 ( L E Q a > L D ): Under this regime, the additional queue extends upstream beyond the deceleration lane, but suff icient lateral clearance on the shoulder allows for an extended queue storage distance ( L E ). If no deceleration lane exists, Regime 2 becomes the initial spillback conditions threshold. As previously mentioned, exiting drivers expect to travel at relativel y high speeds while approaching the off ramp in the deceleration lane. However, in this case , assumed that drivers do not expect to travel at high speed s along the shoulder s ince they are usually equipped with rumble strips and are used for disabled vehicles or law enforcement vehicles. Consequently, it is expected that drivers will decelerate and join the back of the queue more abruptly, potentially causing turbulence in Lane 1. Although the spillback influence area (Figure 4 4 ) is identical to that of Regime 1 in terms of lanes affected, the average travel speed and lane changing behavior of non exiting vehicles in Lane 1 are expected to differ from that of Regime 1. Note tha t if no lateral clearance exists immediately upstream of the deceleration lane ( L D = L E ), Regime 2 conditions are not poss ible. In that case, the segment is expected to incur blockage in the right most lane and should be analyzed ac cording to Regime 3 cond itions Regime 3 ( Q a > L E , Lane 1 blocked): Under this regime, the queue occupies all available queue storage length along the deceleration lane and/or shoulder, and subsequently blocks the rightmost mainline lane (Lane 1). Alternatively, this condition may occur when drivers choose to remain in Lane 1 rather than using the shoulder once the deceleration lane is entirely occupied (if one exists). In this regime, non exiting vehicles using Lane 1 are forced to change lanes to avoid delay, causing additional t urbulence in Lane 2 (Figure 4 5 ). As a result of Lane 1 being blocked, speeds in Lane 2 are expected to be significantly lower, and the number of lanes outside of the spillback influence area, if any more exist, are reduced Regime 4 ( Q a > L E , Lanes 1 and 2 blocked): Under this regime, the queue occupies all available queue storage along the deceleration lane and/or shoulder and subsequently blocks Lane 1, just as in Regime 3. As noted in observational studies (Cassidy et al . , 2002), each exiting driver may be un able to change lanes in time to join the back of the queue in Lane 1 . As a result, exiting drivers will often attempt to force their way into a more advantageous position within the queue, thus bl ocking Lane 2 for some time (Figure 4 6 ). This conditio n likely

PAGE 40

40 occurs as a result of an excessive flow rate of exiting vehicles, a high proportion of exiting vehicles and/or insufficient queue storage length along Lane 1 upstream of the diverge point. Under these conditions, additional lane changing activity and turbulence are expected to occur in Lanes 2 and 3, and thus average travel speed and capacity within the influence area are expected to be significantly reduced. Note that in cases of 2 lane freeway mainline segments, capacity reduction in this scenari o could cause severe congestion along upstream freeway segments within the study facility. Further empirical research is needed to determine what factors cause the transition between Regime 3 and 4 blockage s to occur (queue length, prevailing driver b ehavi or, etc.) and for how long The length of the influence area does not explicitly affect any performance measure estimates in the existing Diverge Segments procedure it simply defines the physical area within which the estimated performance measures are ex pected to occur. W hen spillback occurs, the estimated performance measures are expected to occur further upstream th a n in undersaturated conditions, as noted in observational studies (Daganzo and Munoz, 2000). For the procedures described here, the followi ng assumption shown in Equation 4 5 is used. (4 5) 4 . 3 .2 Equilibrium Separation Distance Where an adjacent upstream on ramp exists, the existing Diverge Segments procedure evaluates whether or not the upstream on ramp is expected to affect operations at the subject off ramp. This adjustment applies only to f reeways with six or more lanes and is based on the proportion of through traffic within the influence area ( P FD ). When the actual distance between the adjacent ramps ( L UP ) is greater than or equal to the equilibrium separation distance ( L EQ ), the off ramp is still considered L UP is less than L EQ , there is some degree of influence on t he study diverge segment operations, and a different prediction model for P FD is used.

PAGE 41

41 If L EQ L UP , use HCM 2010 Equation 13 9 If L EQ > L UP , use HCM 2010 Equation 13 10 When spillback occurs, the available maneuvering distance for appro aching vehicles becomes shorter. Thus, under spillback conditions, L EQ should instead be compared to the distance between the back of the queue and the upstream ramp ( L UP minus Q a ), rather than the actual distance between the ramps. Drivers entering the fr eeway from the upstream on ramp are confronted with the same lane changing decisions as in undersaturated conditions, but further upstream. As a conservative method of evaluating influence from upstream on ramps, the following logic should be employed in t he case of spillback at an off ramp for all Regimes: If L EQ L UP Q a If L EQ > ( L UP Q a ), influence from the upstream on ramp is expected It is conceivable that the conditions found under Regimes 1 and 2 may not affect the influence area length . As stated in Section 4. 3 .1 , under these conditions, there is not expected to be any significant blockage in the right most lane, so ramp to freeway vehicles would not be affected by off ramp spillback ; ramp to ramp vehicles, however, would be influenced by the reduced available space for lane changing maneuvers. In summary, there are two possible ways to evaluate the equili brium separation distance ( L EQ ) i n cases of adjace nt upstream on ramps (Figure 4 7 ) : No spillback occurs: L EQ is compared to the actual distance between the study off ramp and the adjacent upstream on ramp ( L UP ) Spillback occurs, Regimes 1 through 4: L EQ is compared to the distance between the back of the queue to the upstream on ramp ( L UP Q a ) A numerical example is provided here to help illustrate the recommended procedures. Consider an off ramp that is predicted to experience spillback during a

PAGE 42

42 given 15 minute analysis period. Based on traffic conditions at the downstream intersection, the queue length ( Q ) is predicted to be 1,600 feet and an on ramp is located 2,000 feet upstream along the freeway mainline ( L UP ). The equilibrium separation distance ( L EQ ) is determined to be 1,800 feet : Case 1 : The off ramp provides 1,800 feet of queue storage distance ( L a ). Therefore, since spillback is not predicted to occur ( Q L a ), the equilibrium separation distance is compared to the actual distance between the adjacent ramps. Since L UP is greater than L EQ (2,000 feet > 1,800 feet) , the study off ramp Case 2 : Th e off ramp provides only 900 feet of queue storage distance ( L a ). A d eceleration lane that is 400 feet long exists upstream of th e diverge point, and jersey walls from a construction zone exist 200 feet upstream of the deceleration lane. The queue beyond the ramp storage distance ( Q a ) is 700 feet (1 ,600 feet 900 feet ). This exceeds the extended queue storage distance upstream of the diverge ( L E ), which is 600 feet (400 feet + 200 feet ). Therefore, L EQ is compared to the distance between the back of the queue and the upstream on ramp ( L UP Q a , or 2,000 feet 700 feet ). Since L EQ is greater than L UP Q a (1,800 f ee t > 1,300), infl uence from the upstream on ramp is expected to occur It should be noted that the total predicted queue length ( Q ) and corresponding additional queue length ( Q a ) are likely to change between successive analysis periods as a result of evolving conditions over time at the adjacent surface street. Therefore, this method of evaluating the equilibrium separation distance ( L EQ ) in the case of upstream on ramps must be repeated for each period in a multi period f reeway f acilities analysis. Thus , the on ramp may have an influence on diverge segment operations in some periods, while in other periods it may not. 4 . 3 .3 Lane Utilization In the existing Diverge Segments procedure, the flow rate of all vehicles entering the influence area ( v 12 ) is defined as the flow r ate of all exiting vehicles ( v R ), plus a certain proportion ( P FD ) of non exiting vehicles ( v F v R ), as shown in Equation 4 6, a replication of HCM 2010 Equation 13 8 .

PAGE 43

43 (4 6) In the case of six lane freeways in undersaturated c onditions, P FD is a function of the following parameters: the freeway mainline flow rate ( v F ) the exiting flow rate ( v R ) the flow rate and separation distance of an adjacent upstream on ramp ( v U , L UP ) When spillback from an off ramp occurs on six lane freeways, it is expected that, overall, a smaller proportion of non exiting vehicles will use Lanes 1 and 2 (Figure 4 8 ). On the contrary, a study (Daganzo, 1999) has speculated that if the off ramp queue is relatively short and/or queued vehicles use the shoulder, non exiting mainline drivers may be willing to remain in Lane 1 albeit at a significantly reduced travel speed and accept a small amount of delay. Under spillback conditions, the influence a rea is no longer defined as strictly Lanes 1 and 2. Specifically, Lane 1 is considered to be within the influence area in Regimes 1 and 2, while Lane 2 is considered to be within the influence area in Regimes 3 and 4. There fore , the parameter of greatest i nterest under spillback conditions is the proportion of non exiting vehicles that use Lane 1 or Lane 2, depending on which Regime is expected to occur these parameters are denoted P L1 and P L2 , respectively. Assuming the hourly non exiting flow rate ( v F v R ) is known, the flow rate traveling within the influence area can then be determined. 4 . 3 .4 Average Travel Speed The existing d iverge s egments procedure provides a series of equations used to determine the average speed of vehicles in both the ramp infl uence area (Lanes 1 and

PAGE 44

44 that spillback at an off ramp can reduce speeds in the adjacent mainline lanes (Cassidy et al . , 2002). To determine the extent to which queuing a ffects the speed of each lane, empirical data collection and analysis of lane specific travel speeds in diverge segments before, during and after queuing from spillback is necessary. In lanes within the influence area, the average travel speed of vehicle s in a diverge segment is expected to be influenced by: Proportion of exiting demand flow rate ( v R / v F ) : studies have suggested that, as the proportion of vehicles attempting to change lanes and exit the freeway mainline increases, so too does the occurrences per unit of time of vehicles crossing paths (Daganzo and Munoz, 2000) Demand flow rate of vehicles traveling in the influence area ( v F v R ) P L1 or ( v F v R ) P L2 : as the fundamental speed flow relationship cur ve suggests, in general, the expected speed of any given segment eventually diminishes as demand flow rate increases A s in the case of predicting lane utilization of non ex iting vehicles, it is of greatest interest to predict the average travel speed of n on exiting vehicles within the influence area ( Lane 1 or Lane 2, depending on which Regime is expected to occur ) under spillback conditions these parameters are denoted S L1 and S L2 , respectively. Further empirical research is needed to determine what fa ctors influence the average travel speed of vehicles in b locked lanes under spillback conditions, as vehicles are expected to transition over time from free they approach the back of the queue. Of the possible range of values, the stopped threshold speed ( S s ), defined as the maximum speed a vehicle can travel while still (t ypically assumed to be 5 mi/h ), represented the lowest possible value.

PAGE 45

45 It is unlikely that spi llback not involving blockage of mainline lanes (Regimes 1 and 2) will affect average travel speed in lanes outside of the spillback area. Therefore, the prediction model previously presented in Figure 10 is likely still valid in these cases. However, in c ases involving lane blockage (Regimes 3 and 4), more unexpected lane changing within the inf luence area is likely. Thus , the proportion of the exiting demand flow rate compared to the entire freeway demand flow rate ( v R / v F ) likely also affects the speed ch oice of drivers using lanes outside of the spillback area, as shown in a past study (Daganzo and Munoz, 2000). Other factors that may affect average travel speed under spillback conditions are explored in further detail in Chapter 5 . LOS in the existing Di verge Segments procedure is dictated by density, which is estimated using a regression equation. As previously implied, i t is u nlikely that this equation i s also applicable to spillback conditions. It is more appropriate instead to use the fundamental spee d flow density relationship to estimate density to facilitate unbiased lane specific calculations of density . Therefore, the average travel speeds must be estimated before determining density, unlike in the existing procedure. Specifically, Step 5 must pre cede Step 4 (Figure 4 8) . 4 . 3 .5 Density and Level of Service (LOS) As previously mentioned, in the existing Diverge Segments procedure, density is predicted using a regression that was developed for predicting density in undersaturated conditions, and it is unlikely it is also applicable to spillback conditions. Aside from the expectation that queued vehicles will exhibit exceptionally high density (i.e., nearly jam density) vehicles within the influence area are expected to travel slower than usual , resulting in higher density. Vehicles traveling in lanes outside of the influence area , however, are not likely to experience higher density furthering the

PAGE 46

46 argument against using the current regression prediction, as it is meant to predict conditions s trictly within the influence area. Therefore, density should be determined separately for each lane group based on the estimated ave rage travel speed (Section 4 . 3 .4 ), calculated directly using the fundamental speed flow density relationship: flow rate divi ded by speed. 4 . 4 Capacity Checks and Adjustments When spillback occurs, the capacity of diverge segments is determined by t he prevailing extent of spillback and the expected trends in driver behavior . The capacity of the off ramp proper can be determined as a function of the type of intersection approach, as indicated earlier in this chapter. As discussed earlier, when a diverge area is affected by spillback, each of its mainline lanes will operate under one of three regimes: One or more of the rightmost l anes, the shoulder , and/or deceleration lane(s) will be completely blocked because of spillback mainline freeway capacity in these lanes is essentially zero The lane adjacent to the queued lane may have temporary blockages as exiting vehicles attempt to force their way int o the queue and queued vehicles unexpectedly exit the queue . The capacity of this lane depends on these two driver behaviors, considered by the probability of lane blockage ( P B ) and the capacity adjustment factor ( CAF ) , respectively The left most lanes (under the assumption of a right hand off ramp) will operate mostly uninhibited , and thus their capacity will remain mostly unchanged The following sections describe the process of evaluating capacity at a diverge seg ment under spillback conditions. C apacity is initially estimated by the default values as outlined in the current Diverge Segments procedure, which depend on free flow speed ( FFS ) and the total number of lanes in one direction along the segment ( N ) ; t his is discussed in Section 4 . 4 .1 . A capacity adjustment factor ( CAF ) is estimated based on

PAGE 47

47 the expected lower lane utilization of the right most lane by through freeway traffic ; t his is discussed in Section 4 . 4 .2 . The c apacity of lanes within the spillbac k area is further adjusted based on the probability of being blocked at any point during the 15 minute analysis period ( P B ) . This blockage may be temporary or may result in additional spillback; t his is discussed in Section 4. 4 .3 . F inally, capacity adjustm ents as outlined above are incorporated into the final determination of capacity in Section 4 . 4 .4 . 4 . 4 .1 Base Capacity c d ) is determined based on HCM 2010 Exhibit 13 8, replicated in Figure 4 9 . 4 . 4 . 2 Capacity Adjustment Factor ( CAF ) The capacity adjustment factor ( CAF ) with long static queues and the possibility that a queued vehicle could unexpectedly exit onto the freeway mainline, varies between 0.00 and 1.00, with a value of 1.00 corresponding to effectively no adjustment made (i.e., full capacity available). It is expected that the CAF is influenced by the following factors: Additional queue length, Q a (feet ) : the longer the additional queue len gth becomes, the higher the likelihood that a queued vehicle will exit the queue in an attempt to gain a more advantageous spot in the queue further downstream, or exit the diverge segment for another one altogether Distance to closest adjacent downstream off ramp, L DOWN (feet ) : it is likely that the closer downstream the next available off ramp is located relative to the subject off ramp, the more likely that a queued vehicle intending to exit at the study off ramp will exit the queue without warning and travel further downstre am to exit the freeway. Note that off ramps leading to rest areas or other unconnected destinations do not qualify as an adjacent downstream off ramp (Figure 4 10 ) Empirical data collection and qualitative observations at sites experiencing spillback condi tions before, during and after queuing would be necessary to

PAGE 48

48 establish accurate prediction models for the capacity adjustment factor ( CAF ) based on friction from unexpectedly exiting vehicles. 4 . 4 .3 Probability of Lane Blockage ( P B ) The probability of lane blockage ( P B ) , which considers the possibility of a vehicle queue, varies from 0.00 to 1.00. A value of 1.00 corresponds to a 100% certainty that the lane in the in fluence area w ill become blocked for the entire analysis period . Likewise, a P B of 0.50 corresponds to the expectancy that the lane will be block ed for half of the analysis period (7.5 minutes given the standard 15 minute analysis period). It is expected t hat P B is influenced by the following factors: Proportion of exiting vehicles in the traffic stream, v R / v F : past studies (Daganzo and Munoz, 2000) have shown that the proportion exiting vehicles within the whole traffic stream, rather than the flow rate of exiting vehicles itself, has a strong correlation with capacity reduction along the freeway mainline. It is conceivable to believe that higher proportions of vehicles changing lanes across the mainline could cause enoug h turbulence to divert vehicles attempting to join the back of the queue, consequently forcing them to join further downstream Additional queue length, Q a (feet ) : the longer the additional queue length becomes, the higher the likelihood that an exiting ve hicle will either be unable to switch lanes in time to join the back of the queue, or that a queued vehicle will exit the queue and attempt to gain a more advantageous spot in the queue further downstream, thus temporarily blocking an additional lane (Figu re 4 11 ) Distance between back of the queue and adjacent upstream on ramp, L UP Q a (feet ) : As discussed in Section 4. 3 . 2 , when the distance between the upstream on ramp (if one exists) and the back of the queue from the subject off ramp decreases, the intensity of lane changing activity is likely to increase. As a result, exiting vehicles may not find the desired spacing and lane changing gaps to enter the back of the queue in time Just as in Section 4 . 4 .2, empirical data collection and qualitative obse rvations at sites regularly experiencing spillback conditions before, during and after queuing are necessary to establish accurate prediction models for P B .

PAGE 49

49 4 . 4 .4 Final Adjusted Capacity The adjusted diverge segment capacity ( c ), is determined as follo ws: (4 7) where c = diverge segment capacity (pc/h), c d = base capacity as determined by HCM 2010 Exhibit 13 8 (pc/h), N = total number of mainline lanes in the diverge segment (ln), N O = number of lanes outside of t he influence area (ln), CAF = capacity adjustment factor , and P B = probability of blockage in the influence area Figure 4 1. The upstream limit of the available queue storage distance

PAGE 50

50 Figure 4 2. Extended queue storage length Figure 4 3. Regime 1 spillback and influence areas at a diverge segment

PAGE 51

51 Figure 4 4. Regime 2 spillback and influence areas at a diverge segment Figure 4 5. Regime 3 spillback and influ ence areas at a diverge segment Figure 4 6. Regime 4 spillback and influ en ce areas at a diverge segment

PAGE 52

52 Figure 4 7. Possible scenarios under which L EQ is evaluated Figure 4 8. Expected lane utilization tendencies of non exiting drivers within the influence area

PAGE 53

53 Figure 4 8. Re ordering of Diverge Segments procedure (HCM 2010 Exhibit 13 4) Figure 4 9. Capacity of Ramp Freeway Junctions, pc/h (HCM 2010 Exhibit 13 8) Figure 4 10. Distance to the adjacent downstream off ramp ( L DOWN )

PAGE 54

54 Figure 4 11. Additional queue length ( Q a ) as it relates to capacity reduction

PAGE 55

55 C HAPTER 5 QUANTITATIVE ANALYSIS This chapter outlines the primary findings of this study. Section 5. 1 summarizes the lane specific flow rates and non exiting lane utilization trends observed at each study site. Section 5 . 2 provides a comparison of the HCM prediction model results versus the field measured results in undersaturated conditions for lane utilization ( P FD ), average travel speed ( S ) , and density ( D ) . Section 5. 3 provides the speed flow plots observed at each study site compared to the fundamental speed flow curve s defined by the HCM. S ection 5 . 4 outlines a new framework to predict more accurately performance measures in Regime 1 condit ions , including a regression analysis to predict the term P L1 . Finally, Section 5.5 provides a quantitative analysis of the factors that influence the probability of lane blockage ( P B ), as previously discussed in Section 4. 4 .3. 5. 1 Summary of Data Collected Table 5 1 summarizes the data collected in terms of the number of one minute intervals th at were observed, organized by the r egime. Table s 5 2 through 5 8 present the raw data collected from each of the study sites. Average d emands (left hand side of the tables) were obtained in one minute Table 5 1 . Summary of data collection by regime category Freeway/ Direction Surface Street Date Reg . 0 (min) Reg . 1 (min) Reg . 2 (min) Reg . 3 (min) Reg . 4 (min) I 75/SB SR 26 (Newberry Rd.) 11/5/2014 56 4 13 2 0 I 75/SB SR 26 (Newberry Rd.) 11/6/2014 71 1 3 0 0 I 95/SB Old St. Augustine R d. 5/14/2015 50 73 0 0 0 I 95/SB Old St. Augustine R d. 7/9/2015 74 50 0 0 0 I 95/SB I 295 East W est Split 7/9/2015 13 64 0 3 0 I 95/SB I 295 East W est Split 8/13/2015 62 0 0 0 0 I 95/SB SR 202 ( JTB B lvd.) 8/13/2015 0 62 0 0 0

PAGE 56

56 intervals , and p roportions of non exiting vehicles traveling on the freeway mainline lanes (right hand side of the tables) are counted at the diverge points . Table 5 2. D ata collected at I 75 Southbound / SR 26 ( Newberry Road ) (11 / 5 / 2014) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 11.7 4.5 13.5 9.1 0.166 0.499 0.335 0 10.9 5.1 13.1 8.3 0.194 0.493 0.313 1 14.5 3.8 14.3 11.3 0.128 0.487 0.385 2 14.5 2.7 15.2 11.6 0.091 0.514 0.394 3 11.0 0.0 13.0 10.0 0.000 0.565 0.435 Table 5 3. D ata collected at I 75 Southbound / SR 26 ( Newberry Road ) (11 / 6 / 2014) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 11.9 5.0 13.4 9.1 0.182 0.487 0.331 0 11.8 5.2 13.2 8.8 0.190 0.485 0.325 1 15.0 3.0 18.0 17.0 0.079 0.474 0.447 2 13.0 1.7 16.7 12.7 0.054 0.538 0.409 3 Table 5 4. D ata collected at I 95 Southbound / Old St. Augustine Road (5 / 14 / 2015) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 26.1 5.9 23.9 35.6 0.091 0.365 0.544 0 25.9 7.0 24.2 35.0 0.106 0.365 0.529 1 26.3 5.2 23.7 36.0 0.081 0.365 0.555 2 3 Table 5 5. D ata collected at I 95 Southbound / Old St. Augustine Road (7 / 9 / 2015) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 25.5 6.7 24.2 35.4 0.101 0.365 0.534 0 25.8 8.0 25.3 36.1 0.116 0.364 0.520 1 25.1 4.7 22.7 34.4 0.076 0.367 0.557 2 3

PAGE 57

57 Table 5 6 . D ata collected at I 95 Southbound / I 295 E ast W est Split (7 / 9 / 2015) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 27.1 16.5 24.9 28.5 0.236 0.356 0.408 0 27.2 11.9 17.5 18.4 0.250 0.366 0.385 1 27.2 17.5 26.4 30.5 0.235 0.355 0.410 2 3 25.3 15.7 24.0 28.7 0.229 0.351 0.420 Table 5 7 . D ata collected at I 95 Southbound / I 295 E ast W est Split (8 / 13 / 2015) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 50.8 16.6 25.2 32.7 0.223 0.338 0.439 0 50.8 16.6 25.2 32.7 0.223 0.338 0.439 1 2 3 Table 5 8 . Dat a collected at I 95 Southbound / SR 202 ( JTB Boulevard ) ( 8 / 13 / 2015) Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) Lane 1 Lane 2 Lane 3 Overall 34.1 13.3 25.9 31.0 0.189 0.368 0.442 0 1 34.1 13.3 25.9 31.0 0.189 0.368 0.442 2 3 5 . 2 Comparison of Performance Measures This section provides a comparison between the performance measures according to H CM prediction models as found in the existing diverge segments procedure and the actual conditions as observed from the data collection efforts. S ection 5. 2 .1 outlines the comparison of lane utilization of non exiting vehicles, Section 5. 2 . 2 outlines the com parison of average travel speed and Section 5. 2 .3 outlines the comparison of density in the freeway segment.

PAGE 58

58 5. 2 . 1 Comparison of Lane Utilization of Non Exiting Vehicles The following plots (Figures 5 1 through 5 6 ) compare the HCM prediction model for lane utilization of non exiting vehicles ( P FD ) versus the observed ones at each of the study sites in undersaturated (Regime 0) conditions . The line with a 1.0 slope at a 45 conditions observed. Thus, a given data point under the line represents an under prediction, whereas a given data point over the line represents an over p rediction. In general, it appears that the HCM prediction model for P FD over predicted the proportion of mainline vehicles traveling in the influence area at lower values and vice versa for higher values. The transition between over/under predictions of P FD occurs at approximately 0.650 for the I 75 Southbound / SR 26 (Newberry Road) study site (Figures 5 1 and 5 2 ) and approximately 0.500 for the I 95 Southbound / Old St. Augustine Road study site (Figures 5 3 and 5 4 ) . The prediction model outputs a cons tant value of 0.450 for P FD in cases of two lane off ramps, as in the I 95 Southbound / I 295 E ast W est Split study site (Figures 5 5 and 5 6 ). 5. 2 . 2 Comparison of Average Travel Speed The following plots (Figures 5 7 through 5 12 ) compare the HCM prediction model for the average travel speed of the entire freeway segment ( S ) versus the actual observed speeds , according to RITIS data in the same format at the previous section , for undersaturated conditions only . The plot format represents the same c onvention as shown in the plots in the previous section. In general, the HCM prediction model under predicted speed s at the a t I 75 Southbound / SR 26 ( Newberry Road ) study site (Figures 5 7 and 5 8 ), versus over predicting speeds at the I 95 Southbound / Old St. Augustine Road study site (Figures

PAGE 59

59 5 9 and 5 10 ). Mixed results in terms of prediction accuracy were apparent at the I 95 Southbound / I 295 East West Split study site (Figures 5 11 and 5 12 ). 5. 2 . 3 Comparison of Density The following plots (Figures 5 13 th r ough 5 18 ) illustrate the comparison between density predictions for the whole freeway segment ( D ) , calculated by using the fundamental speed flow density relationship , for each of the study sites . Only undersaturated conditions are shown. The HCM density calculation is based on speeds acquired from the regression based speed prediction method, whereas the observed density calculation is based on the speed data obtained from RITIS. According to the fundamental speed flow density relationship, density is inversely proportional to speed. Therefore, it is expected that (given the same flow rate) this calculation method will over predict density to the same extent that it under predicts travel sp eed and vice versa. This trend is reflected accurately in each of these graphs. 5. 3 Speed Flow Relationship The following plots (Figures 5 19 through 5 25 ) show the speed flow relationship observed at each of study site. Ideally, the plotted data points would compare favorably to the fundamental speed flow relationship parabolic curves as defined in the Traffic Flow and Capacity Concepts chapter of the HCM 2010 (Chapter 4 ) . It is expected that in congested conditions that, at a given free flow speed, the average travel speed of vehicles in the traffic stream will diminish at lower flow rates compared to undersatur at ed conditions . In general, it appears as the ideal speed flow curve patterns were observed at the study sites. At lower flow rates, such as those observed at the I 75 Southbound / SR 26 (Newberry Road) study site (Figures 5 19 and 5 20 ), speeds remained relatively

PAGE 60

60 constant. At higher flow rates, such as those seen in the five Jacksonville area study sites (Figures 5 21 through 5 25 ), speeds tended to diminsh at flow rate increased. At study sites where multiple regime conditions were observed, the plots tended to e graph according to the specific regime under which they are classified. 5 . 4 Analysis of Lane Utilization Trends T o obtain the total density of diverge segme nts under Regime 1 conditions, the d ensity of the influence area (Lane 1) and d ensity of the lanes outside of the influence area (Lanes 2, 3 and any additional lanes) must be determined separately . Using the fundamental speed flow density rel ationship to calculate d en sity under spillback conditions requires two important metric s not included in the exi sting procedure. Proportion of non exiting vehicles ( v F v R ) traveling in Lane 1, P L1 Average travel speed of vehicles using Lane 1, S L1 The calculation of d ensity is as shown in Equation 5 1. (5 1) Data collected under Regime 1 conditions from the three study sites in Jacksonville were subjected to a regression analysis in an attempt to understand which factors influence the lane utilization tendencies of non exiting vehicles . The existing Diverge Segments procedure uses total flow ( v F ) and exiting flow ( v R ) to predict the proportion of non exiting vehicles travel ing in the influence area ( P FD ) for three lane freeways with a single, isolated off ramp. Table 5 9 compares the coefficients used for those two predictor terms versus the coefficients yielded from the data collection . Two additional regression analyses were performed with the proporti on of non exiting

PAGE 61

61 vehicles using Lane 1 ( P L1 ) as the resp onse variable. T he first iteration using total flow ( v F ) and exiting flow ( v F ) as the predictor variables, the second iteration using the proportion of exiting vehicles within the whole traffic strea m ( v R / v F ) in addition to the original two as the predictor variable. T able 5 10 compares the coefficients for each of the prediction var iables in the respective models while Table 5 11 compares the results of these two regression models using traditional s tatistics metrics. Table 5 9. Comparison of P FD regression model coefficients Regression Model Constant Total Flow, pc/h ( v F ) Exiting Flow, pc/h ( v R ) HCM 2010 Equation 13 9 0.760 0.000025 0.000046 Data Collection 0. 527 0.000023 0.000064 Table 5 10. Comparison of P L1 regression model coefficients Regression Model Constant Total Flow, pc/h ( v F ) Exiting Flow, pc/h ( v R ) Proportion of Exiting Flow ( v R / v F ) First Iteration 0.19 3 0.000066 0.000046 N/A Second Iteration 0.01 1 0.00003 5 0.000061 0.70 5 Table 5 11. Comparison of P L1 regression model statistical measures Regression Model Adjusted R 2 Standard Error F Test First Iteration 0.3 37 0.064 5 64. 1 Second Iteration 0.3 43 0.064 2 44. 2 The regression model generated from data collection shows similar prediction trends to the equation used in the Diverge Segments procedure, with the major difference being the negative correlation between exiting flow ( v R ) and non exiting vehicles entering the influence area ( P FD ). Regarding the predict ion the proportion of non exiting vehicles using Lane 1 in Regime 1 conditions, both models perform similarly, and it is not clear whether or not including the proportion of exiting vehicles ( v R / v F cy. Larger, more comprehensive datasets are needed to investigate further these trends.

PAGE 62

62 To determine the Density of the lanes outside of the influence area ( D O ), Equation 5 2 is used, this time multiplying the non exiting flow rate ( v F v R ) by the P L1 ): (5 2) The average travel speed data collected in this study via RITIS was based on the entire freeway mainline, making it analogous to the HCM metric S , rather than lane specific travel speed S R , S L1 or S O . Therefore, lane specific travel speed data would need to be collected to develop regression based prediction models for average travel speed of vehicles in the influence area, S L1 . 5. 5 Probability of Lane Blockage ( P B ) As discussed in Section 4.5.3, the probability of lane blockage ( P B ) is likely to be the most important factor in the determination of capacity. Regime 3 conditions (i.e., blockage of Lane 1) occurred in several instances during the da ta collection effort. Tables 5 1 2 through 5 1 5 show the difference between several metrics between unblocked and blocked conditions. Table 5 12. Comparison of demands unde r lane blockage conditions Regime Ramp (veh/min) Lane 1 (veh/min) Lane 2 (veh/min) Lane 3 (veh/min) 1, 2 21.1 6.3 20.2 22.2 3 19.6 9.4 19.6 21.2 (% Diff.) 7.1% 48.6% 1.9% 4.6% Table 5 13. Comparison of non exiting lane utilization under lane blockage conditions Regime Lane 1 Lane 2 Lane 3 1, 2 0.114 0.446 0.440 3 0.137 0.497 0.965 (% Diff.) 19.9% 11.5% 119.4%

PAGE 63

63 Table 5 14. Comparison of flow rates under lane blockage conditions Regime v F (pc/h) v R (pc/h) v R / v F (pc/h) 1, 2 4,804 1,457 0.31 3 4,836 1,361 0.29 (% Diff.) 0.7% 6.6% 7.1% Table 5 15. Comparison of speed and density under lane blockage conditions Regime S (observed) D (observed) 1, 2 53.7 mi/h 36.4 pc/mi/ln 3 37.2 mi/h 70.2 pc/mi/ln (% Diff.) 30.8% 93.0% Based on the se comparisons , lane utilization and observed density best indicate the occurrence of lane blockage. Regime 3 conditions occurred for only 5 minutes of all the data collected, so definitive trends are difficult to assert. Specifically , at the I 75 Southbound / SR 26 (Newberry Road) study site, lane blockage occurred for 2 of the 19 minutes (approximately 11%) in which spillback conditions were observed. Likewise, at the I 95 Southbound / I 295 East West Split study site, lane blockage occurred for 3 of the 67 minutes (approximately 4 %) in which spillback conditions were observed. Figure 5 1. P FD at I 75 Southbound / SR 26 ( Newberry R oa d ) study site (11 / 5 / 2014)

PAGE 64

64 Figure 5 2. P FD at I 75 Southbound / SR 26 ( Newberry R oa d ) study site (11 / 6 / 2014) Figure 5 3. P FD at I 95 Southbound / Old St. Augustine Road study site (5 /1 4 / 2015)

PAGE 65

65 Figure 5 4. P FD at I 95 Southbound / Old St. Augustine Road study site (7 / 9 / 2015) Figure 5 5. P FD at I 95 Southbound / I 295 E ast W est Split study site (7 / 9 / 2015)

PAGE 66

66 Figure 5 6. P FD at I 95 Southbound / I 295 E ast W est Split study site (8 / 13 / 2015) Figure 5 7. S at I 75 Southbound / SR 26 (Newberry Road) study site (11/5/2014)

PAGE 67

67 Figure 5 8. S a t I 75 Southbound / SR 26 ( Newberry Road ) study site (11 /6/ 2014) Figure 5 9. S at I 95 Southbound / Old St. Augustine Road study site (5/14/ 201 5 )

PAGE 68

68 Figure 5 10. S at I 95 Southbound / Old St. Augustine Road study site (7/9/ 2015) Figure 5 11. S at I 95 Southbound / I 295 East West Split study site (7 / 9 / 2015)

PAGE 69

69 Figure 5 12. S at I 95 Southbound / I 295 East West Split study site (8/13/ 2015) Figure 5 13. D a t I 75 Southbound / SR 26 ( Newberry Road ) study site (11 / 5 / 2014)

PAGE 70

70 Figure 5 14. D at I 75 Southbound / SR 26 ( Newberry Road ) study site (11 / 6 / 2014) Figure 5 15. D at I 95 Southbound / Old St. Augustine Road study site (5 / 14 / 2015)

PAGE 71

71 Figure 5 16. D at I 95 Southbound / Old St. Augustine Road study site (7 / 9 / 2015) Figure 5 17. D at I 95 Southbound / I 295 E ast W est Split study site (7 / 9 / 2015)

PAGE 72

72 Figure 5 18. D at I 95 Southbound / I 295 E ast W est Split study site (8 / 13 / 2015) Figure 5 19. Speed flow plot at I 75 Southbound / SR 26 (Newberry Road) study site (11/5/2015)

PAGE 73

73 Figure 5 20. Speed flow plot at I 75 Southbound / SR 26 (Newberry Road) study site (11/6/2015) Figure 5 21. Speed fl ow plot at I 95 Southbound / Old St. Augustine Road study site ( 5/14/ 2015)

PAGE 74

74 Figure 5 22. Speed f low p lot at I 95 Southbound / Old St. Augustine Road study site ( 7/9/ 2015) Figure 5 23. Speed flow plot at I 95 Southbound / I 295 East West Split study site (7/9/2015)

PAGE 75

75 Figure 5 24. Speed f low p lot at I 95 Southbound / I 295 East West Split study site ( 8/13/ 2015) Figure 5 25. Speed flow plot at I 95 Southbound / SR 202 (JTB Boulevard) study site (8/13/2015)

PAGE 76

76 CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS In general, in cases of congested conditions when the volume to capacity ratio of a given facility ( v / c ) exceeds 1.00 the HCM suggests the use of a microsimulation tool . While possessing the ability to provid e in depth insight in to the underlying causes of the spillback, microsimulation tools are not only typically time consuming and costly, but are also customarily employed in cases of existing or short term operational studies. On the other hand, a macroscopic tool is sufficient for long term planning purposes, as it is unrealistic to calibrate driver behaviors and code detailed signal timing settings for future year scenari os. The underlying purpose of this project is to investigate the effects of spillback from a n oversaturated surface street onto a freeway facility in a macroscopic environment , and to understand better driver behavior in response to spillback conditions. T he framework presented in this thesis is valid only for a specifi c freeway segment configuration: an isolated diverge segment along a three lane mainline freeway. Thus, further empirical investigation s of other configurations (two lane and four lane freewa y mainlines, diverge ramps with an upstream merge ramp, weaving segments, etc.) are needed to elucidate the effects of spillback upon freeway facilities more fully . It is of interest to government planning agencies at the federal and state level s to evaluate the long term performance and reliability of freeway facilities in urban areas that feature many interchange ramp terminals serving as connections to adjacent surface streets. Moreover, it is of highest priority to address deficiencies within segments along a given freeway that experience spillback conditions on a regular basis (i.e., wee kday morning and afternoon peak hours when most inter city commuters are

PAGE 77

77 using them), as opposed to spillback conditions that result from irregular occurrences , such as concerts or festivals. Thus, the data collected in the Jacksonville area sites are likely more applicable to future As previously implied, the separation of Uninterrupted Flow and Interrupted Flow into separate sections (HCM 2010 Volumes II and I II, respectively) unrealistically implies that the two types of facilities are mutually exclusive. The concepts presented in this thesis should s erve to open up the discussion of freeway surface street interactions to a variety of additional topics, such a s oversaturated freeway facilities causing operational deficiencies at adjacent surface streets. The findings of the aforementioned forthcoming investigations could conceivably serve as a guide in developing a computation engine to implement the adjustment s to the existing procedure so that practitioners serving interested government agencies can deliver recommendations for planning projects in a time effective manner. The speed flow plots shown in Figures 5 19 through 5 25 lend to the argument that a regim e based framework is best suited to model freeway arterial interactions, as the regime graph . Data collected at these study sites suggest that the proportion of non exiting veh icles traveling in the influence area (Lane 1 under Regime 1 conditions, P L1 ) is most strongly correlated with the proportion of exiting vehicles within the entire traffic stream ( v R / v F ) . This trend is compliant with several of the referenced past studies in Chapter 2 . According to the fundamental speed flow curves, in undersaturated conditions, any given speed remains constant as the flow rate increases starting at zero vehicles

PAGE 78

78 fl ow speed, intuition would suggest this occurs at a lower flow rate at any given speed. New fundamental speed flow curves should be developed at a time when more empirical data is available . Based on the comparison of lane utilization between the HCM prediction models versus field observations at diverge segments with two lane off ramps (Figu res 30 and 31), it appears that the current model of assigning a singular value is not sufficiently accurate for quantifying the proportion of non exiting vehicles traveling in the influence area ( P FD ). This proportion is directly correlated to the influen ce area flow rate ( v 12 ), which is a critical value in the subsequent steps in the diverge segments procedure. Therefore, it is feasible to invest time in generating a more accurate regression based prediction model for cases of two lane off ramps . A site specific comparison of average travel speed plots (Figures 5 7 through 5 12 ) and density plots (Figure 5 13 through 5 18 ) conforms to the fundamental speed flow density relationship. Specifically, given a fixed flow rate, density is inversely propor tional to speed (i.e., at sites where the HCM model over predicts speed, it under predicts density to the same extent, and vice versa). Thus, under the proposed based pre diction), the prediction of lane specific average travel speed becomes a critical step, rather than a formality as in the existing procedure.

PAGE 79

79 Table 6 1 summarizes the current needs for r egime specific p rediction models based on the proposed framework in t his report. This thesis developed a prediction model for P L1 under Regime 1, simply as a framework guide for future data collection. Additional data need to be collected to obtain the remaining metrics. Table 6 1. Critical metrics needed for spillback co nditions to be developed in future research Regime P L1 P L2 S L1 S L2 CAF P B Regime 1 Regime 2 Regime 3 Regime 4 There are a number of other critical shortcomings discussed in this report that can be addressed with future research. In order to properly collect the required data, traffic counts under different r egime specific conditions should be recorded in such a way that t he entire extent of the deceleration lane ( L D ) and any extended storage length beyond that ( L E ) are visible. Th e vantage point is essential in determining the operating Regime (according to the definitions in Section 4 . 3 .1) and is essential in observing demand according to its HCM definition. Additionally, f ield measured travel speed should be recorded on a lane specific basis to differentiate between blocked lanes, lanes within the influence area and lanes outside of the influence area. A nalysis of such data could reveal different influence area thresholds t han those defined in this report, which are based on a limited sample size . In terms of specific topics not explored in this project, several items in particular are worth bringing attention to. Namely, e mpirical observations should be made to determine h ow the Regime (or exact additional queue length, Q a ) is related to the 1,500 f eet increases under spillback conditions). Further, a specific term should be

PAGE 80

80 defined to describe the length of this increased influenced area to see if it significantly affects any important performance measures in subsequent regression analysis . Finally, i nteractions between off ramps experiencing spillback conditions and adjacent upstre am on ramps should be investigated to determine whether or not spillback conditions change the way in which the existing HCM models for lane utilization are used (i.e., how L EQ is evaluated in comparison to L UP ) . Rather than representing an authoritative methodology to evaluate the effects of freeway arterial interactio ns, this thesis serves to present new concepts and ideas to contribute towards developing a more comprehensive framework. This project features only a small dataset of empirical observations in a limited scope in terms of infrastructure configuration and geographic region. As such, research should be conducted in a variety of settings, and the use of big data with automated collection methods to generate greater sample sizes is ideal. Ultimat ely, a framework that is capable of accurately and reliably evaluating freeway arterial interactions in a macroscopic and deterministic setting will save transportation engineers and stakeholders at the state and federal levels considerable time and financ ial resources.

PAGE 81

81 LIST OF REFERENCES 1. Lighthill, M. J. and Whitham, G. B. On Kinematic Waves: II. A Theory of Traffic Flow on Long Crowded Roads. Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences . London, U. K., 19 95 . 2. Newell, G. F. A Simplified Theory of Kinematic Waves in Highway Traffic, Part II: Queuing at Freeway Bottlenecks . Transportation Research Part B: Methodological, 27(4) . Amsterdam, NL, 1 993 . 3. Daganzo, C. F. The cell transmission model, part II: network traffic . Transportation Research Par t B: Methodological 29 ( 2 ) , Amsterdam, NL, 1995 . 4. Daganzo, C. F. A Continuum Theory of Traffic Dynamics for Freeways with Special Lanes. Transportation Res earch Part B: Methodological 31( 2 ) , Amsterdam, NL, 1997 . 5. Newell, G. F. Delays Caused by a Queue at a Freeway Exit Ramp . Transportation Research Part B: Methodological, 33(5) , Amsterdam, NL , 1999 . 6. Daganzo, C. F. The lagged cell transmission model , University of California Berkeley Department of Civil and Environmental Engi neering, Berkeley, CA, 1999 . 7. Munoz, J. C. and Daganzo, C. F. Experimental Characterization of Multi Lane Freeway Traffic Upstream of an Off Ramp Bottleneck . California Partners fo r Advanced Transit and Highways, Berkeley, CA, 2000 . 8. Cassidy, M. J., Anani, S . B., and Haigwood, J. M. Study of freeway traffic near an off ramp . Transportation Research P art A: Policy and Practice 36(6) , Amsterdam, NL , 2002 . 9. Daganzo, C. F. A behavioral theory of multi lane traffic flow. Part II: Merges and the onset of congestion . Transportation Res earch Part B: Methodological 36( 2 ), Amsterdam, NL , 2002 . 10. Munoz, J. C., and C. F. Daganzo . The bottleneck mechanism of a freeway diverge . Transportation Research Part A: Po licy and Practice 36(6), Amsterdam, NL , 2002 . 11. Transportation Research Board of the National Academies. Highway Capacity Manua l, 5 th Edition . Washington, D.C. 2010 . 12. Lu, C., & Elefteriadou, L. An investigation of freeway capacity before and during incidents . Transportation letters , 5 (3), Leeds, England, U. K., 2013.

PAGE 82

82 BIOGRAPHICAL SKETCH Michael was born in Oca la, Florida and began school at the University of Florida in fall 2009. He graduated with a Bachelor of Science in Civil Engineering in s pring 2014 with Magna Cum Laude honors, and a Master of Science in c ivil e ngineering in f all 2015, both at the University of Florida. While in graduate school, h e served as a graduate assistant at the Mc Trans ® Center at the University of Florida Transporta tion Institute (UFTI) , guiding the development of Highway Capacity Software (HCS) and assisting in teaching Traffic Engineering and Highway Capacity Analysis, two graduate level courses in the Department of Civil and Coastal Engineering. His primarily inte rests within the field of transportation engineering include the simulation of traffic operations, route choice modeling and intelligent transportation systems (ITS). Upon completion of graduate school , Michael accepted a job with Henningson, Durham and Ri chardson, Inc. (HDR) as a transportation engineer in Raleigh, North Carolina.