The final report of ITS Center project: Wet Weather Signalization

UVA Center for Transportation Studies

A Research Project Report

For the Center for ITS Implementation Research

A U.S. DOT University Transportation Center

INVESTIGATING THE IMPACTS OF RAINY WEATHER

AT ISOLATED SIGNALIZED INTERSECTIONS

Principal Investigators:

Dr. Michael J. Demetsky
Michael Tantillo

Center for Transportation Studies

University of Virginia

Thornton Hall

351 McCormick Road, P.O. Box 400742

Charlottesville, VA 22904-4742

804.924.6362

January 2006
Research Report No. UVACTS-15-13-90

Disclaimer

The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the Department of Transportation, University Transportation Centers Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof.

INVESTIGATING THE IMPACTS OF RAINY WEATHER

AT ISOLATED SIGNALIZED INTERSECTIONS

By: Michael Tantillo

Dr. Michael J. Demetsky

A Research Project Report for the Intelligent Transportation Systems Implementation Center (ITS)

A U.S. DOT University Transportation Center

Dr. Michael J. Demetsky

Department of Civil Engineering

Email: mjd@virginia.edu

Center for Transportation Studies at the University of Virginia produces outstanding transportation professionals, innovative research results and provides important public service. The Center for Transportation Studies is committed to academic excellence, multi-disciplinary research and to developing state-of-the-art facilities. Through a partnership with the Virginia Department of Transportation’s (VDOT) Research Council (VTRC), CTS faculty hold joint appointments, VTRC research scientists teach specialized courses, and graduate student work is supported through a Graduate Research Assistantship Program. CTS receives substantial financial support from two federal University Transportation Center Grants: the Mid-Atlantic Universities Transportation Center (MAUTC), and through the National ITS Implementation Research Center (ITS Center). Other related research activities of the faculty include funding through FHWA, NSF, US Department of Transportation, VDOT, other governmental agencies and private companies.

Disclaimer: The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the Department of Transportation, University Transportation Centers Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof.

1. Report No.	2. Government Accession No.			3. Recipient’s Catalog No.
UVACTS-15-13-90
4. Title and Subtitle				5. Report Date
INVESTIGATING THE IMPACTS OF RAINY WEATHER AT ISOLATED SIGNALIZED INTERSECTIONS				January, 2006
				6. Performing Organization Code
7. Author(s) Michael Tantillo Dr. Michael J. Demetsky				8. Performing Organization Report No.
9. Performing Organization and Address				10. Work Unit No. (TRAIS)
Center for Transportation Studies
University of Virginia				11. Contract or Grant No.
PO Box 400742 Charlottesville, VA 22904-7472
12. Sponsoring Agencies' Name and Address				13. Type of Report and Period Covered
Office of University Programs, Research Innovation and Technology Administration US Department of Transportation 400 Seventh Street, SW Washington DC 20590-0001				Final Report
				14. Sponsoring Agency Code
15. Supplementary Notes
16. Abstract
This thesis examines the impacts of wet weather on traffic flow and signalized intersections, and it also examines the feasibility of mitigating these impacts. First a literature review was completed. Various studies showed that the impact of weather on traffic flow is a topic that is of much interest to researchers and transportation officials. Inclement weather tends to cause drivers to behave differently than they would if the weather were clear, which may cause reductions in capacity along a roadway network. It was decided to focus on rainy weather since most other studies tended to examine a variety of different weather types. Various sources of data were examined, including obtaining data from the Smart Travel Laboratory, from the Smart Travel Van, and the city of Norfolk traffic signal system. None of these sources came to fruition, due to concerns over data quality, data accuracy, safety of people and equipment, and privacy concerns. It was therefore decided to video tape an intersection in the field to obtain data on the saturation flow rates during dry and wet weather conditions. Free flow speed data was also collected during wet and dry conditions to determine if a change in desired speed existed. The saturation flow rate decreased by approximately 4.7 percent, and the free flow speeds decreased by 9.4 percent during wet weather, and both were determined to be statistically significant differences. This data was used to calibrate the VISSIM microscopic simulation model, and allow an isolated intersection to be simulated in both dry and wet conditions. When simulated under a variety of input vehicle volumes, it was determined that there are reductions in the capacity of the intersection, and increases in delay, queue lengths, travel time, and number of stops during wet weather. These increases were most noticeable when the input volumes were near the intersection’s capacity. New signal timing plans were developed to mitigate the effects of wet weather. When an optimal timing plan for dry weather was developed, improvements were noted in almost all performance measures. Additionally, the percent deterioration in performance due to wet weather was lessened by implementing an optimized timing plan. However when special wet weather timing plans were developed, no system-wide benefits were realized. The best timing plan for both wet and dry conditions appeared to be the dry-weather optimal timing plan. Therefore it is unlikely to be feasible to develop special wet weather timing plans for isolated signalized intersections. The concept of developing wet weather timing plans for arterial networks should be investigated further using field data to support results and conclusions.
17 Key Words			18. Distribution Statement
IMPACTS OF RAINY WEATHER AT ISOLATED SIGNALIZED INTERSECTIONS			No restrictions. This document is available to the public.
19. Security Classif. (of this report)		20. Security Classif. (of this page)			21. No. of Pages	22. Price
Unclassified		Unclassified			190	N/A

ABSTRACT

This thesis examines the impacts of wet weather on traffic flow and signalized intersections, and it also examines the feasibility of mitigating these impacts. First a literature review was completed. Various studies showed that the impact of weather on traffic flow is a topic that is of much interest to researchers and transportation officials. Inclement weather tends to cause drivers to behave differently than they would if the weather were clear, which may cause reductions in capacity along a roadway network. It was decided to focus on rainy weather since most other studies tended to examine a variety of different weather types.

Various sources of data were examined, including obtaining data from the Smart Travel Laboratory, from the Smart Travel Van, and the city of Norfolk traffic signal system. None of these sources came to fruition, due to concerns over data quality, data accuracy, safety of people and equipment, and privacy concerns. It was therefore decided to video tape an intersection in the field to obtain data on the saturation flow rates during dry and wet weather conditions. Free flow speed data was also collected during wet and dry conditions to determine if a change in desired speed existed. The saturation flow rate decreased by approximately 4.7 percent, and the free flow speeds decreased by 9.4 percent during wet weather, and both were determined to be statistically significant differences. This data was used to calibrate the VISSIM microscopic simulation model, and allow an isolated intersection to be simulated in both dry and wet conditions. When simulated under a variety of input vehicle volumes, it was determined that there are reductions in the capacity of the intersection, and increases in delay, queue lengths, travel time, and number of stops during wet weather. These increases were most noticeable when the input volumes were near the intersection’s capacity. New signal timing plans were developed to mitigate the effects of wet weather.

When an optimal timing plan for dry weather was developed, improvements were noted in almost all performance measures. Additionally, the percent deterioration in performance due to wet weather was lessened by implementing an optimized timing plan. However when special wet weather timing plans were developed, no system-wide benefits were realized. The best timing plan for both wet and dry conditions appeared to be the dry-weather optimal timing plan. Therefore it is unlikely to be feasible to develop special wet weather timing plans for isolated signalized intersections. The concept of developing wet weather timing plans for arterial networks should be investigated further using field data to support results and conclusions.

ACKNOWLEDGEMENTS

I’d like to take this time to thank those people who have been extremely helpful and supportive of me throughout my time spent at the University of Virginia and throughout my entire academic career. First, I’d like to thank Professor Michael J. Demetsky for being an extremely kind, helpful, and supportive advisor throughout the course of this project. Next, I’d like to thank Professors Brian Park, Brian Smith, and Lester Hoel for being great mentors and for sharing their wealth of knowledge with the transportation students. I’d like to thank fellow graduate students Hongtu “Maggie” Qi, George Lu, Daniel Son, and Ilsoo Yun for advice and assistance with various aspects of this project. I’d like to thank the brothers of the Theta Chapter of Alpha Phi Omega for providing me with the opportunity to serve the Charlottesville community during my time at the University, but most especially Kathleen Saathoff, Rebecca Weber, Caroline Ward, Lindsay Woodson, Caroline Tran, Emily Pedneau, Sheela Narayanan, Mark Jensen, Caroline Dyar, Claire Winchester, Heather Skakandy, Will Hook, Kristen Hooper, Kari Browning, Kate Gubareva, Steven Longstreet, and Christina Luckett for providing much support and encouragement when I most needed it. I’d like to thank Ashraf Hossain, Elizabeth Huang, Shaena Ryan, Christie Ives, and Mamie Wise for being amazing friends as well. I’d like to especially thank Jason Broder and Salil Gulati for always being there for me for the past 20 years, because I wouldn’t be where I am today without both of you guys. I’d like to thank my Mom and Dad and sister Laura for being an amazing and supportive family. Last, but certainly not least, I’d like to thank the University of Virginia’s Center for Transportation Studies for its support of this project.

TABLE OF CONTENTS

ABSTRACT iii

ACKNOWLEDGEMENTS v

TABLE OF CONTENTS vi

LIST OF FIGURES xi

LIST OF TABLES xiii

CHAPTER 1: INTRODUCTION 1

1.1 Background 1

1.2 Purpose and Scope 3

1.3 Structure 5

CHAPTER 2: LITERATURE REVIEW 7

2.1 Introduction 7

2.2 Identifying Inclement Weather Types 7

2.3 Weather Information Systems 8

2.3.1 Virginia’s Use of Road Weather Information Systems 8

2.3.2 FORETELL in North America 9

2.4 Highway Capacity Manual References to Weather 10

2.5 Impacts of Rainy Weather on Freeways in Hampton Roads 12

2.6 Research Needs for Weather Responsive Traffic Management 14

2.7 Signalized Intersection Studies 16

2.7.1 Federal Highway Administration Study on Simulation

and the Benefits of Re-timing Signals 16

2.7.2 Salt Lake City Study 18

2.7.3 Minnesota Study 21

2.7.4 Anchorage Study 22

2.7.5 Burlington, Vermont Study 23

2.8 Summary 26

CHAPTER 3: DATABASE DEVELOPMENT 27

3.1 Data Sources 27

3.1.1 Smart Travel Laboratory 27

3.1.2 Smart Travel Van 29

3.1.3 City of Norfolk Signal System 30

3.1.4 Field Collection Using Video Camera 30

3.1.5 Speed Data 31

3.1.6 Fairfax County Red Light Enforcement System Data 31

3.1.7 Weather Data 32

3.1.8 Summary 33

3.2 Selection of Data Collection Locations 33

3.2.1 Intersection Data Collection 34

3.2.2 Arterial Free-flow Speed Data Collection 38

3.3 Summary of Findings 39

CHAPTER 4: DATA COLLECTION AND ANALYSIS 41

4.1 Intersection Data Collection 41

4.1.1 Field Determination of Saturation Flow Rate 41

4.1.2 Summary of Weather Conditions During Data Collection 43

4.2 Data Reduction and Results 44

4.3 Speed Data Collection 46

4.4 Significance Testing 48

4.5 Discussion of Results 51

4.6 Additional Field Observations 52

4.7 Data Collection Issues 55

CHAPTER 5: MICROSCOPIC SIMULATION NETWORK DEVELOPMENT 57

5.1 Selection of Simulation Software Package 57

5.2 Network Coding 58

5.2.1 Isolated Intersection 58

5.2.2 Three Intersection Network with Actuated-Coordinated

Signals 63

5.3 Network Calibration 65

5.3.1 Desired Speeds 66

5.3.2 Saturation Flow Calibration 69

CHAPTER 6: ANALYSIS OF THE EFFECTS OF WEATHER ON

TRAFFIC FLOW AND MITIGATION USING VISSIM 75

6.1 Performance Measures 75

6.2 Preparing the VISSIM Model 77

6.2.1 Volumes 78

6.2.2 Traffic Signal Timing Plan 79

6.2.3 Data Collection Points 80

6.2.4 Simulation Length 81

6.2.5 Random Seed and Number of Runs 82

6.2.6 Seminole Trail Network 83

6.3 Compilation and Discussion of the Effects of Weather on Traffic Flow 84

6.3.1 Traffic Volume 87

6.3.2 Travel Time and Delay 90

6.3.3 Stops 98

6.3.4 Queue Length 101

6.3.5 Summary of Findings 106

6.4 Mitigating the Effects of Rain Through Signal Timing 107

6.4.1 Development of Optimal Timing Plans 108

6.4.2 Traffic Volume 112

6.4.3 Travel Time and Delay 114

6.4.4 Number of Stops 121

6.4.5 Queue Length 123

6.4.6 Findings 127

6.5 Summary 129

CHAPTER 7: CONCLUSIONS AND RECOMMENDATIONS 131

7.1 Conclusions 131

7.2 Recommendations For Future Research 134

REFERENCES 136

APPENDIX A: Field Obtained Values for Saturation Headway and

Free-Flow Speed 138

APPENDIX B: Effects of Rain on Individual Intersection Approaches

(from VISSIM) 142

APPENDIX C: Effects of New Signal Timing Plans on Individual

Intersection Approaches (from VISSIM) 161

LIST OF FIGURES

Figure 1. Flowchart of Research Methodology 6

Figure 2. Lane Usage and Camera Angles for Ivy Road and Emmet Street 36

Figure 3. Lane Usage and Camera Angles for Barracks Road and Emmet St. 36

Figure 4. View of Emmet Street and Barracks Road Intersection from

Camera Location 38

Figure 5. Schematic of Speed Data Collection Location 47

Figure 6. Emmet Street & Barracks Road VISSIM Network 60

Figure 7. Close-up of Emmet Street & Barracks Road Showing

Network Elements 60

Figure 8. NEMA Editor User Interface and Signal Timing Plan 62

Figure 9. Seminole Trail Network 64

Figure 10. Saturation Flow as a Function of VISSIM Parameter Settings 71

Figure 11. Total Observed Vehicle Volume vs. Input Vehicle Volume 89

Figure 12. Percent Difference Due to Weather in Observed vs. Input

Vehicle Volumes 89

Figure 13. Average Travel Time vs. Input Vehicle Volume 93

Figure 14. Increase in Travel Time Due to Weather vs. Input Volume 94

Figure 15. Average Delay vs. Input Vehicle Volume 95

Figure 16. Percent Change Due to Weather in Average Delay vs. Input Volume 96

Figure 17. Average Stop Delay vs. Input Vehicle Volume 97

Figure 18. Percent Change Due to Weather in Average Stop

Delay vs. Input Volume 98

Figure 19. Average Number of Stops vs. Vehicle Input Volume 100

Figure 20. Percent Change Due to Weather in Number of Stops vs. Input Volume 100

Figure 21. Average Queue Length vs. Input Volumes 103

Figure 22. Percent Change Due to Weather in Average Queue Length 103

Figure 23. Maximum Queue Length vs. Input Volumes 105

Figure 24. Percent Change Due to Weather in Maximum

Queue Length vs. Input Volume 105

Figure 25. Total Volume vs. Signal Timing Plan 113

Figure 26. Average Travel Time vs. Signal Timing Plans 115

Figure 27. Average Delay vs. Signal Timing Plan 117

Figure 28. Average Stop Delay vs. Signal Timing Plan 119

Figure 29. Stops per Vehicle vs. Signal Timing Plan 122

Figure 30. Average Queue Length vs. Signal Timing Plan 124

Figure 31. Maximum Queue Length vs. Signal Timing Plan 126

LIST OF TABLES

Table 1. FHWA Defined Weather Types and Associated Speed Reductions 7

Table 2. Summary of Values Obtained in Vermont Study 24

Table 3. List of intersections considered for data collection 35

Table 4. Data Collection Dates, Times, and Weather Conditions 43

Table 5. Temperatures and Precipitation Amounts for Data Collection Sessions 44

Table 6. Table of Saturation Headway Results 45

Table 7. Compiled Results for Saturation Headway 45

Table 8. Dates, Times, and Weather Conditions for Speed Data Collection 48

Table 9. Temperatures and Precipitation Amounts for Speed Data Collection 48

Table 10. Speed Data Collection Results in Miles per Hour 48

Table 11. Null Hypothesis Test Parameters and Significant Differences in Means 51

Table 12. Traffic Volumes for Emmet Street and Barracks Road 61

Table 13. Selected Values for Saturation Headway Observed During Calibration 74

Table 14. VISSIM Volume Inputs 79

Table 15. Emmet Street and Barracks Road Current Timing Plan 80

Table 16. Mean Values of Performance Measures for Dry Weather Conditions 85

Table 17. Standard Deviation of Performance Measures for Dry Weather

Conditions 86

Table 18. Mean Values of Performance Measures for Wet Weather Conditions 86

Table 19. Standard Deviation of Performance Measures for Wet Weather

Conditions 87

Table 20. Timing Plans Obtained From Synchro 110

Table 21. Mean Values of Performance Measures with Optimal and

Special Timing Plans 111

Table 22. Standard Deviations of Performance Measures with Optimal

and Special Timing Plans 112

Table 23. Volume Percent Change, Benefits, and Costs 114

Table 24. Travel Time Percent Change, Benefits, and Costs 116

Table 25. Average Delay Percent Change, Benefits, and Costs 118

Table 26. Stop Delay Percent Change, Benefits, and Costs 120

Table 27. Number of Stops Percent Change, Benefits, and Costs 123

Table 28. Average Queue Length Percent Change, Benefits, and Costs 125

Table 29. Maximum Queue Length Percent Change, Benefits, and Costs 127

CHAPTER 1. INTRODUCTION

1.1 Background

Inclement weather is something that can affect peoples’ everyday lives in many ways. One such way is its effect on our transportation systems. While most people will readily recognize that weather has an impact on transportation, many of these impacts have not been examined in a quantitative sense until fairly recently.

The impacts of weather on our transportation systems are something that researchers, DOT’s, and the public are looking at increasingly. As our states and cities become more populated, our transportation systems are becoming overburdened with traffic. Congestion and delays are becoming more common. Due to environmental and political constraints, it is often very difficult to simply add more capacity to congested transportation facilities. Therefore the use of Intelligent Transportation Systems (ITS) and other methods for increasing the efficiency and maximizing the use of existing capacity are being developed and deployed in many areas.

ITS technology has a wide variety of forms. Variable Message Signs, Highway Advisory Radio, Advanced Traveler Information Systems, and traffic cameras are all being used in Virginia to alert motorists when congestion or other problems arise. Additionally, Road Weather Information Systems (RWIS) are also being deployed throughout the state to alert DOT authorities to possible problems on the Commonwealth’s roadways. These RWIS stations report quantitative weather data, such as wind velocity and direction, precipitation amounts, and pavement temperatures. This information is occasionally sent, via the ITS systems, to the general public when officials think it is necessary for travelers to know about adverse weather conditions, such as high winds or heavy snowfall accumulations. While weather data is readily available to DOT’s and travelers, and some of the general effects of inclement weather are known, little research has been done to quantify some of the lesser-known impacts of weather on our transportation systems. Knowing that roadways can become slippery when it snows is inherently obvious, but knowing how things such as accident rates and delays are affected is not as well known.

One area where inclement weather can have an impact on traffic flow is at signalized intersections on arterial roadways. As opposed to freeways which have un-interrupted traffic flows, arterial capacity is usually limited by individual intersection capacity and design. Therefore, knowing the effects of weather on intersection capacity is an important part of managing arterials during inclement weather incidents. Knowing the effects of weather on arterial capacity is an important step towards more effective management strategies might be able to be implemented along arterial corridors.

Perhaps the most dynamic component of signalized intersections that can easily be manipulated in order to mitigate the effects of weather is the signal-timing plan. Much time has been devoted to developing methods to optimize signal timing plans, however these methods almost always assume ideal weather conditions are present (i.e., dry pavement, unrestricted visibility). Due to the possible impacts of weather on driver behavior and traffic flow, the optimal signal timing plan at an intersection, or a network of intersections, could be different during inclement weather then it is during ideal weather conditions. If this is the case, and the timing plans present at many intersections are ideal for clear weather, then they might be sub-optimal for inclement weather. Therefore, an analysis of the effects of inclement weather at signalized intersections is necessary to quantify the effects of these sub-optimal timing plans. Furthermore, if new timing plans were developed for inclement weather, it is possible that some of the adverse effects of inclement weather can be mitigated.

1.2 Purpose and Scope

The purpose of this research project is to determine and assess the impacts of inclement weather on traffic flow at a signalized intersection. Some negative impact due to inclement weather is expected. After the impacts are determined, the feasibility of mitigating all or some of the impacts can be determined. This mitigation can occur using new signal timing plans or other forms of ITS technology.

While other similar research projects have been conducted in the past, there are two important reasons why this project is justified. First and foremost, driver behavior is subject to regional variations. This means that studies conducted in one area of the nation will often produce different results then studies conducted in other areas. Since no quantitative study of impacts of weather on traffic flow at signalized intersections has been done in central Virginia, this will be the first time that results are seen for this area of the country. Secondly, other studies examined a multitude of different inclement weather conditions, and drew conclusions based on the effects of many different weather conditions. While frozen precipitation does occur in Virginia, it is often confined to the more rural, mountainous areas of the state where traffic congestion is not as big of a concern as it is in more urban areas. Additionally, many drivers in Virginia’s cities try to avoid travel during frozen precipitation incidents, meaning that congestion is less of a concern. During rainy weather, people do not adjust their travel plans as much, so during rainy weather is when the largest weather impacts at signalized intersections will probably be seen in Virginia. Therefore, it was decided to focus exclusively on rainy weather, as opposed to any form of frozen precipitation.

The following are the expected results of this research:

• A literature review of pertinent weather related research
• Field data showing the effects of rainy weather on traffic flow parameters at signalized intersections
• A microscopic simulation model that is calibrated based on the field data collected to represent clear and rainy weather
• An analysis of the effects on intersection and network performance, using selected performance measures
• A determination of the feasibility of adjusting signal timing plans to mitigate the adverse effects of rainy weather on intersection and network performance
• Recommendations for future research

Field data collection was confined to the Charlottesville-Albemarle region of Virginia, and was collected when it was clear with dry pavement or rainy with no frozen precipitation.

1.3 Structure

Figure 1 shows a flow diagram of the research methodology used in this project. The remainder of this thesis contains the following sections. Chapter 2 is a literature review focusing on previous research conducted on inclement weather impacts on traffic flow, as well as road weather information systems currently in use. Chapter 3 presents the development of the data sources. Chapter 4 describes the data collection and results, which would determine if there were actually differences in driver behavior between wet and dry weather. Chapter 5 describes the methodology for coding and calibrating the microscopic simulation model. Chapter 6 presents the impacts of weather on signalized intersections based on the simulation results, along with the feasibility of adjusting signal timing to improve traffic flow and safety during rainy weather. First the effects of weather on various performance measures were tested. Next the signal timing plan was optimized to determine if conditions improved. Lastly, special wet weather timing plans were developed to determine if further improvement could be realized. Finally, Chapter 7 provides the relevant conclusions from this research, as well as recommendations for future research.

Figure 1. Flowchart of Research Methodology

CHAPTER 2. LITERATURE REVIEW

2.1 Introduction

The next step in completing this research project was a review of relevant literature. Literature examined included studies on weather information systems and weather related research needs, as well as previous studies about weather impacts on traffic flow. One of the main reasons for completing this task was to identify gaps in previous research that could be filled in this project.

2.2 Identifying Inclement Weather Types

A Federal Highway Administration study published in 1977 assessed impacts of weather on highways during various types of inclement weather. While the study mostly dealt with economic impacts, they did define six inclement weather conditions, which are referred to repeatedly in other literature. As a basis for supporting their research, the authors collected speed data from interstate highway facilities in the Washington DC area during these six inclement weather conditions (1). The weather conditions and the associated observed reductions in speed, compared to the dry case, are shown in Table 1.

	Weather Condition	Percent Reduction in Speed
1	Dry	(Base Case)
2	Wet	0%
3	Wet and Snowing	13%
4	Wet and Slushy	22%
5	Slushy in Wheel Paths	30%
6	Snowy and Sticking	35%
7	Snowing and Packed	42%

Table 1. FHWA Defined Weather Types and Associated Speed Reductions (1)

2.3 Weather Information Systems

Weather Information Systems are sprouting up across the country in an effort to provide accurate real-time conditions and forecasts to travelers and transportation planners. The following parts show the state of weather information systems in Virginia and in North America.

2.3.1 Virginia’s Use of Road Weather Information Systems

A report by the Virginia Transportation Research Council published in 1998 describes Virginia’s Road Weather Information Systems (RWIS) program, and some of the issues relating to it (2). There were 40 RWIS stations in place throughout the Commonwealth when this report was published, and two 24-hour weather forecasts were generated from these stations every day. The RWIS stations measure wind-speed and direction, precipitation amounts, humidity and temperature, and various road condition variables using pavement sensors. Virginia maintains its RWIS program with two contracts: one to maintain the RWIS stations themselves, and the other to provide the forecasts from the data. Based on the study, the RWIS program had much room for improvement. Five of the 10 stations examined had physical problems with the sensing equipment, and of the two stations where forecasts were examined, neither had accurate weather predictability rates. As much as half of all snow or ice events went unpredicted, and 27 percent of the precipitation cases that did occur were not predicted to occur. This is something that can definitely be problematic if VDOT relies on this data for road maintenance and operations purposes. Some other states have data availability requirements, with penalties imposed if the sensors malfunction, yet only one other state (New Hampshire) has any requirement for forecast accuracy. The report recommended that VDOT take steps to improve the RWIS program, and it highlights some of the issues with obtaining accurate weather data from RWIS stations (2).

RWIS stations would not provide data that is useful for studying the effects of rain at intersections, as the output data is not related to traffic flow. Therefore it is necessary to collect data from other sources to study the problem. However if the study concludes that solutions should be implemented based on real-time weather conditions, RWIS data would be necessary to support those solutions from a station located near the intersection. The ability of the RWIS system to provide accurate forecasts is not as important as the ability to provide real-time data that can be used to determine when to place rainy weather solutions into effect. Steps should be taken to ensure that real-time data is available, reliable, and accurate, especially if it will be used to develop guidelines for when to implement weather-specific solutions, and when to revert to normal operations.

2.3.2 FORETELL in North America

FORETELL is a US Department of Transportation funded initiative to link together various sources of weather data for use by transportation professionals. It is envisioned that FORETELL will be a key link between integrating weather data with Intelligent Transportation Systems. The data will be available to professionals for use in traffic operations, and eventually will be available to the general public as well. Right now, the project is in the preliminary phase, with Iowa, Wisconsin, Missouri, the Federal Highway Administration, and private firm Castle Rock involved. The data comes from a variety of sources, including Road Weather Information System stations, the National Weather Service, airport weather stations, and even some Canadian sources (3).

The stated goals for FORETELL are to create a weather information network throughout North America that can be used for transportation purposes, and to reduce winter weather related accidents. In addition to collecting weather data, weather events will be forecasted up to 24 hours in advance, using a 6-mile grid system instead of a 20-mile grid system that the National Weather Service traditionally uses (3). In the immediate future, FORETELL would be of little use to traffic planners who wish to examine the effects of rain on arterial roadways. However if it is determined that weather related traffic management strategies are feasible to implement, then having sufficient weather data collecting infrastructure will be crucial to the success of such strategies. Weather information systems can later be integrated with ITS technology if the need arises, as the weather information can be provided in real time.

2.4 Highway Capacity Manual References to Weather

The Highway Capacity Manual (4) has some references to weather related effects on traffic operations. In the chapter on freeway facilities, there is a section that specifically states that capacity and operating speed reductions can be seen as a result of inclement weather. That section has three subsections, which examine rain, snow, and fog.

The rain section states that speeds will not be affected unless visibility is reduced. Therefore the claim is made that light rain will not have an impact on operating speeds, but heavy rain will since it can affect visibility. A minimal reduction in the maximum observed flows is cited during light rain, while a much greater reduction was observed during heavier rain. Similar observations were made for operating speeds. During light rain, the speeds were reduced by 1.2 miles per hour in free-flow conditions, and by 4 to 8 miles per hour when 2400 vehicles per hour per lane were present on the freeway. During heavy rain, the drop in free flow speeds was 3 to 4 miles per hour, and the reduction in speeds at 2400 veh/hr/lane were 8 to 10 miles per hour. Unlike light rain conditions where little effect was seen on capacity, heavy rain is said to produce a 15 percent reduction in capacity (4). No indication is given as to what exactly constitutes light rain, and what constitutes heavy rain in scientific terms.

A research project in Germany attempted to apply the observations to speed-flow diagrams. What was interesting about this project was that it also investigated the effect of darkness and weekday versus weekend on capacity in addition to rain. During rainy conditions, a 12 percent reduction in capacity was observed on a 6-lane freeway during weekday daylight hours. In similar conditions, an 18 percent reduction was observed on a 4-lane freeway. The reductions in capacity for darkness were 13 and 19 percent for the 6 and 4 lane freeways respectively. Under combined dark and wet conditions, the capacity reductions were 38 and 47 percent. When weekend data was examined, an additional 7 to 10 percent reduction from the weekday capacities was observed. It is stated that American freeways and German Autobahns differ in terms of design standards. Therefore, lower capacities per lane are observed on German Autobahns, and the capacities would likely be different on American Freeways (4).

In determining the capacity of a signalized intersection, the one of the most important parameters is the saturation flow rate. The saturation flow rate is defined as: “the flow in vehicles per hour that can be accommodated by the lane group assuming that the green phase is displayed 100 percent of the time” (4). The saturation flow rate is computed starting with a base saturation flow, multiplied by the number of lanes, and then reduced by multiplying in various reduction factors. These reduction factors take into account items such as lane width, amount of heavy vehicles, approach grade, parking, transit stops, area types, lane utilization, turns, pedestrians, and bicycles. These factors should all have static values at any one point in time. Since weather is a random event, it is not a factor in computing the saturation flow rate, and is therefore not a factor that is taken into account when determining intersection capacity (4).

2.5 Impacts of Rainy Weather on Freeways in Hampton Roads

Professor Brian Smith and four Research Assistants completed a study in 2003 at the University of Virginia (5). This study was restricted to determining the impacts of rainy weather on freeway traffic flow using archived detector data from the Smart Travel Laboratory. This study was conducted with the goal of providing useful information on the impacts of weather to those who wish to conduct research on how to mitigate its effects on freeways. The paper suggests that large amounts of data need to be examined in order to account for traffic variability, that studies such as this one should be conducted in various locations to account for regional differences, and that there is a need to classify rainfall into various categories of intensity (5).

Volume, time-mean speed, and occupancy values from two reliable detector stations were compiled. The detectors were located on segments of freeway that often see congestion. The values were collected over 15 minute intervals, and were compared with hourly precipitation data from the Virginia State Climatology office. Throughout the study, it was assumed that the rainfall intensity amount was constant over the course of the hour. The rain was classified as follows: heavy if the intensity was greater than 0.25 inches per hour, and light if the intensity fell between 0.01 and 0.25 inches per hour. Speed-flow plots were created for each rainfall category to estimate capacity, to verify that demand had reached the capacity, and to see if changes in speeds were observed (5).

The maximum throughput was taken to be the capacity of the two freeway links. This was done by taking the top 5 percent of the flow rate values, and then averaging them together. To examine the operating speeds, the researchers took the average of the values in the non-congested regime, which they assumed to be any speed reading of over 50 miles per hour. A capacity reduction was evident for each rainfall intensity level (9.88 percent and 3.99 percent for light rain, and 29.22 percent and 25.55 percent for heavy rainfall at each of the two detector stations). A reduction in speed was present as well (5.07 percent and 3.55 percent for light rain, and 4.40 percent and 3.48 percent for heavy rain). These reductions in speed were statistically significant when compared to dry conditions; however there was no significant difference in the reduction when light rain was compared to heavy rain. The authors then compared their results to the guidance given in the Highway Capacity Manual, and produced the following conclusions: the impact of rain on capacity is greater then that stated in the manual, the impact of light rain on operating speed is similar to that observed in the manual, and that the manual overstates the impacts of heavy rain on operating speed. The researchers continue on to say that this study should be replicated in other locations to see if regional differences affect the impacts of rain on freeway traffic flow, and that this should be done before any updates to the Highway Capacity Manual are considered (5).

2.6 Research Needs for Weather Responsive Traffic Management

Weather-responsive traffic management is a new area of research that the Federal Highway Administration is examining. It views weather events as incidents that can be observed and predicted, and whose adverse effects can be mitigated. Weather impacts vehicle performance, driver behavior, and roadway conditions, and can lead to congestion due to reductions in capacity as a result of these impacts. Therefore, more systematic approaches are needed in order to better manage traffic during weather incidents. Weather-responsive management strategies face several hurdles, including the need for accurate prediction of incidents and dissemination of data to proper traffic managers. Also, the traffic must be controlled under less then ideal conditions. However the idea is to attempt to avoid congestion wherever possible, and to mitigate the effects when it is impossible to avoid weather related congestion (6).

The effects of weather on traffic flow as discussed in this study are as follows. First, reductions in capacity are often seen as a result of precipitation, water collecting on the roadway, fog, or even high wind conditions. These factors can affect the travel speeds, pavement traction, and variability in speed, volumes, and delays. If conditions are severe enough, portions of roadways can be blocked, and traffic control devices can fail. The authors cite previous studies showing a 10 percent decrease in freeway speeds during rainy weather, and a 16 to 40 percent decrease during heavy rain or snow when visibility is impacted, cars accelerate slower, and drivers leave more space between them and the next vehicle. The authors claim that arterial speeds can drop from 10 to 25 percent in rainy weather, and 30 to 40 percent in snowy or icy weather. One interesting claim made in this paper is that the traffic volumes often decrease (from 6 to 30 percent) during inclement weather on arterials; however this is most likely a location-specific, as opposed to a general claim (6).

The authors specifically mention that the effectiveness of signal timing plans developed for ideal conditions is reduced during inclement weather. A Washington, DC study that they cite shows a 14 percent increase in travel time during off peak periods on arterial roadways in the city when inclement weather was present, and a 24 percent increase during peak periods. Statistics showing that rain and wet pavement conditions have a greater impact on roadway safety and crash rates then snow or ice. Lastly, when visibility is impacted, some drivers slow down more than others, leading to greater variability in speeds, and greater risk of rear-end or overtaking related crashes (6).

It is expected that once Advanced Traffic Management Systems, Intelligent Transportation Systems, and Road Weather Information Systems are integrated, traffic managers will be able to take specific actions to mitigate the impacts of weather, using weather information as a decision support tool. This can allow a transportation agency to achieve three important goals: to improve safety, increase efficiency of the system, and increase mobility. The theory is that traffic managers would simply integrate weather information into their existing decision processes for managing networks through the use of advisories and control strategies. For rainy weather, traffic managers can advise motorists of problems on the roadway network, advise them of safe speeds, and implement control strategies (such as adjusting signal timing plans) to accommodate changes in driver behavior (6).

One very interesting feature of this report is the first of three case studies presented. In Clearwater Beach, FL, there are often afternoon thunderstorms. When these thunderstorms arrive, there is a mass exodus of vehicles from the beach back to the mainland (an example of a case where traffic flow increases during inclement weather). A special signal timing plan for the bridge and its approach roadways was designed to accommodate these increases in demand by increasing green times on the main roadway. The plan automatically goes into effect after a certain amount of rain has fallen, and the plan reverts back to normal when traffic volumes dip back below a certain threshold (6).

2.7 Signalized Intersection Studies

Several studies have examined the topic of adjusting signal timing during inclement weather. The first study in this segment is a theoretical study by the Federal Highway Administration examining how to evaluate the benefits of re-timing signals, as well as how to simulate weather conditions in simulation software. The next four studies examined the benefits of re-timing signals in various weather conditions, concentrating mostly on frozen precipitation.

2.7.1 Federal Highway Administration Study on Simulation and Benefits of Re- timing Signals

The Federal Highway Administration conducted a research study to assess the benefits of implementing specialized timing plans during inclement weather (7). Although this study examines many of the issues regarding weather related timing plan development, the authors did not collect any data of their own, and therefore the numerical results are only theoretical. However the guidelines they give are helpful. The authors state the need for their work due to the perception that the effects of weather cannot easily be mitigated, and that traffic simulations and analytical tools are all based on ideal weather conditions. The authors were merely interested in proving that simulation packages (in this case, CORSIM) can be used to assess the impacts of weather and the assessment of benefits of new timing plans, as opposed to actually using real data and real networks.

The researchers point out five parameters used in simulation which can easily be adjusted to simulate weather effects: free flow speeds, maximum acceleration and deceleration rates, gap acceptances, queue discharge headways (related to saturation flow rate), and start up lost times. In their study, they decided to exclusively focus on the effects of weather, and therefore used a homogenous set of driver behavior parameters (parameters other then the ones listed above). The test scenarios examined are the base case, the inclement weather case, and the inclement weather case with retimed signals. In order to obtain the weather case, the authors reduced the free-flow speed, discharge headway, and startup lost times by 20 percent each. Additionally, the volumes were reduced by 15 percent to account for people who chose not to travel during bad weather conditions (7).

When the authors examined the corridor speeds and demands, they determined that the weather had an impact on the speeds over a large range of demand. Even if demand drops by 15 percent, the speeds would still be less in most cases. With signal times being adjusted during inclement weather, improved speeds on the corridor were seen when the vehicle volumes were between the boundary values of 1100 vehicles per hour and 1700 vehicles per hour. They claim that when the road is experiencing congested conditions, improving the average speed by just a few miles per hour has a larger time savings benefit then when the speeds are much higher. A chart is presented showing the delay reduction per mile for each Level of Service category for an arterial roadway. The benefits seem to be realized when the level of service is between C and F (when speeds are between 7 and 10 miles per hour), with the greatest reduction at LOS E of 1.0 minutes per mile. At speeds below 7 miles per hour (in LOS F category), the arterial is gridlocked, and at LOS B and above, there would not be many benefits to retiming signals. The authors then go on to extrapolate the regional benefits of retiming traffic signals, and conclude that the weather specific management strategies can have a significant effect on improving traffic flow. However they concede that more research is needed, and that their numerical example would likely produce different results (including the “boundary values” above) with a different network (7).

2.7.2 Salt Lake City Study

A study was conducted in Salt Lake City, Utah during the winter of 1999-2000 to study the feasibility of changing signal-timing plans during inclement weather incidents (8). The authors decided to conduct the study because they noted drivers behave more cautiously during inclement weather, which can affect the carrying capacity of arterial roadways. Additionally, the Salt Lake City area traffic signal system was not an adaptive system that could automatically adjust to changing traffic conditions on its own, however the signals did have direct communication links to traffic control centers where the timing plans could be remotely updated as needed. Therefore it was determined that if inclement weather timing plans could be developed using modified traffic flow parameters, they could easily be implemented manually when needed.

Researchers collected data from two intersections during the peak periods. The data collected was the saturation flow rate, the free flow speed, and the startup lost time, and the data was collected over a period of fourteen different days. Additionally, data was collected during some dry days for comparison purposes. The data was collected for a total of six different inclement weather conditions: wet, wet and snowing, wet and slushy, slushy in wheel paths, snowy and sticking, and snowing and packed. One can easily see that the study primarily focuses on frozen precipitation, which most other studies do as well. While results are available for wet conditions, the results of this study are based on the “average” impacts of the various types of inclement weather listed above. Naturally, the study conclusions are therefore based on more severe weather impacts then a study focusing exclusively on rainy weather (8).

The saturation flow rate at an intersection decreases during inclement weather for several reasons. First, drivers are driving at reduced speeds due to loss of visibility and more cautious driving behavior. Drivers will be accelerating more slowly as well. Finally, drivers often leave more room between themselves and the vehicle in front, which increases the headways. The Salt Lake City study showed the reduction in saturation flow rate was proportional to the severity of the inclement weather. During rainy conditions, the reduction was 6 percent. The free flow vehicle speeds were also shown to decrease during inclement weather. The speed data was collected over 18 days over five of the seven weather conditions, and the decreases in speed were found to be proportional to the weather severity. The reduction in speed during wet weather was found to be 10 percent. The startup lost time decreases due to a reduction in tire traction between the vehicle and the roadway. In the Salt Lake City study, the startup lost times decreased during frozen precipitation incidents; however wet weather had only a negligible impact (8).

This study was the only one that gave consideration to the length of the amber and all-red phases, as well as pedestrians. For pedestrians, the authors cited an earlier study that determined that pedestrian crossing speeds increase during inclement weather. Therefore, the authors concluded that pedestrian phases did not need to be lengthened during weather incidents. Due to changes in traction (which the authors state might only be perceived), the amount of stopping distance required might increase. Since dilemma zones at signalized intersections are dependent on speed and stopping distance, the length of the amber phase might need to change in order to possibly eliminate dilemma zones during inclement weather. Additionally, the decrease in speed and increased likelihood of a dilemma zone existing might cause the all-red interval to be insufficient during weather incidents. The authors call for a 10 to 15 percent increase in Amber time, and a 1 second increase in all red time, but do not really have much information to back up how they arrived at these recommendations (8).

This study showed that the largest decrease in performance was seen when the weather conditions changed from “wet and snowing” to “wet and slushy”, and they used that as the boundary point for when they recommend changing the signal timing plans. The numbers used to describe the decrease in performance were the average values of the reductions of the three weather conditions that were more severe then “wet and snowing”. The authors cite unsuccessful “recent attempts” (8) at implementing signal timing plans through automated processes as the reason for suggesting that trained traffic management professionals decide when the new timing plans go into affect. The suggested parameters that should be considered when implementing the plans are the following: severity of the weather incident, duration of the weather incident, area of influence, and traffic flow (8).

The authors recommend changing the offsets, splits, and clearance intervals, but not the cycle lengths to accommodate inclement weather. They also cite the possibility of the traffic volumes changing during inclement weather as a possible factor impacting the performance of the system; however they have no data to support that hypothesis. The final results are shown as decreases in saturation flow, decreases in speeds, and increases in startup lost time, as well as the increases in amber and all-red times mentioned above. The authors plan to run simulations using Synchro/Simtraffic in order to verify that their findings are correct; however no information on the results of any further research is available at this time (8).

2.7.3 Minnesota Study

This study was done by the Minnesota Department of Transportation in order to evaluate the feasibility of implementing special timing plans during inclement weather (9). The study was done during the winter in Minneapolis, and data was collected on both clear weather days and inclement weather days. For the purposes of this study, inclement weather was defined as 3 inches of accumulated snow or more. The data was collected during the PM peak hours from a variety of sources, including a RWIS station, system detectors, and personal observation.

The data items used for formulating conclusions were the signal timing plans, turn counts, vehicle volumes, startup lost time, saturation flow rates, air and pavement temperatures, dew point, humidity, and roadway pavement conditions. The saturation flow rates decreased, but not as much as in Salt Lake City (11 percent, as compared to 21 percent respectively). Speeds decreased by as much as 40 percent, and startup lost times increased by 50%. Synchro was used to create optimized signal timing plans for the network. The study found that delays could be reduced by 13 percent when inclement weather signal timing plans were used, and that vehicle stops decreased by 6 percent on average as well. Lastly, the study examined the feasibility of using RWIS data to automatically implement the inclement weather signal timing plans when certain road condition threshold values were met. The authors concluded that the RWIS data and actual road conditions differed enough such that reliability was a problem, and therefore the timing plans would be manually implemented. This study also found that during inclement weather incidents, traffic demand decreases between 15 and 20 percent. This decrease in demand lessens the need for mitigation of the effects of inclement weather, as there are fewer vehicles in the system (9).

2.7.4 Anchorage Study

This study was done to assess the impacts of inclement weather on traffic flow on a 24-signal network in Anchorage, Alaska. Traffic flow parameters, including saturation flow, capacity, startup lost time, and vehicle speeds were measured during summer, winter, and during inclement winter weather. It was found that vehicle speeds decreased during the winter, and even more during inclement winter weather. Additionally, the saturation flow rate decreased by 12 percent during the winter. Using this information, new signal timing plans were developed using SIGNAL 85 and TRANSYT-7F, with the former used to generate the phase sequences and splits, and the latter to determine the offsets. The arterial network showed a decrease in delay of 23 percent, and a decrease in travel times of 13 percent when the new signal timing plans were tested (10).

2.7.5 Burlington, Vermont Study

This study was done in Burlington, Vermont by University of Vermont researchers. The study was done for similar reasons that the studies in other areas were done: the impact of weather on traffic flow at signalized intersections is great enough such that researchers are trying to find solutions to the problem, however the solutions must be able to be implemented without a massive upgrade to the traffic signal system. The authors stated that since no other research on weather impacts on traffic flow had been conducted in New England, that there was a regional need to conduct the study (11).

An isolated intersection near the University of Vermont was chosen for use in data collection. The data was collected over six different categories of surface conditions, which were the same categories used in the Salt Lake City study except that the packed snow condition was not examined. The data was collected between 7AM and 7PM, during the winter of 2002/2003, over a three-month period using a video camera. The saturation flow rates and startup lost times were then manually reduced from the videotapes (11).

A statistical analysis was performed on the data using an Analysis of Variance (ANOVA) test, which concluded that there were significant differences in the saturation flows, however only minor differences existed in the startup lost times. Using a multiple comparison test, it was determined that the differences between each of the weather conditions were significant except between Dry and Wet, between Wet and Wet + Snowy, and between Wet + Slushy and Snowy + Sticky. With the startup lost times, the multiple comparison test showed significant differences between only two of the condition pairs, and no significant differences between the rest, which indicates that weather is not a major factor in startup lost times. The values, shown below in Table 2, were comparable to the values seen in the Utah and Minnesota studies (11).

Road Condition	Startup Lost Time (s)		Saturation Headway (s)		% Reduction in Saturation Headway		Utah Avg % Reduction
	EB	WB	EB	WB	EB	WB
Dry	2.20	1.84	2.24	2.04	0	0	0
Wet	2.42	1.98	2.31	2.08	3	2	6
Wet and Snowy	2.18	2.28	2.42	2.13	7	4	11
Wet and Slushy	1.29	2.00	2.41	2.39	7	15	18
Slushy		1.90		2.58		21	18
Snowy and Sticky	3.04	2.20	2.67	2.44	16	16	20

Table 2. Summary of Values Obtained in Vermont Study (11)

For analysis of the benefits of implementing weather based traffic signal timing plans, two software packages were used. First, the microscopic simulation program CORSIM was used to determine the impacts, and secondly TRANSYT-7F was used to develop the timing plans. For calibration purposes, CORSIM’s saturation headways, free flow speeds, and startup lost times were adjusted, and the results were checked against travel times and queue lengths from the field for validation purposes. The model was calibrated for the dry condition only, and the parameters for the inclement saturation flows and speeds were simply reduced by the appropriate percentages (saturation flow reduction factors were obtained from the study, free flow speed reduction factors were taken from the Salt Lake City study) (11).

To assess the benefits, the authors compared performance measures between the dry (base) case with its optimal timing plan, and each of the inclement weather scenarios with each other their own optimal timing plans in effect on a network of several intersections in Burlington (note that this network was completely separate from the intersection used to collect data). Both TRANSYT-7F and CORSIM were used to determine the benefits. The three performance measures that were examined were average delay time, total delay time, and average speed. In both simulation programs, benefits were seen from implementing new signal timing plans during inclement weather. However CORSIM showed less reduction in delay and less of an increase in speed then TRANSYT-7F did. In order to validate their findings, the researchers compared their results to studies done in Minnesota and Anchorage, Alaska. The TRANSYT-7F results were similar to the results in the Minnesota study, while the CORSIM results showed less benefit to implementing new timing plans. The Alaska study showed greater benefits then either of the simulations in the Vermont study showed. Finally, the authors increased the volumes by 5 percent, and discovered that the benefits to retiming the traffic signals increased significantly, which led them to include a warning about over generalizing the results of their study until more research could be conducted (11).

2.8 Summary

This literature review points out several items of interest. First, impact of weather on traffic flow is a “hot topic” at the present time. Much work has been devoted to creating weather information systems, and researchers are searching for ways to use these systems for purposes other than winter roadway maintenance. Previous studies have been completed on the impacts of weather at signalized intersections and on freeways, however none of the intersection studies focus exclusively on rain. Since rainy weather conditions are more common than snow or icy conditions, especially in the mid-Atlantic region, a study focusing exclusively on rain would be more beneficial to Virginians than one that examined all weather conditions. The studies show that weather has impacts on traffic flow, however the results tend to differ in different sections of the country. The saturation flow rate and free-flow speed seem to be the two traffic flow parameters that are both affected by weather and easily measurable.

CHAPTER 3. DATABASE DEVELOPMENT

3.1 Data Sources

In order to learn more about how rainy weather affects traffic flow at signalized intersections, obtaining data from various sources was necessary. The original intent was to obtain archived data from the Smart Travel Laboratory and some field data from a signalized intersection. The data collection methods evolved over time, with the eventual sources being limited to field data collected at both an intersection and an arterial segment. This section will explain the data sources that were considered and the reasons for accepting or rejecting the sources.

3.1.1 Smart Travel Laboratory

The Smart Travel Laboratory at the University of Virginia is the result of a joint effort by the Virginia Transportation Research Council and the university’s Department of Civil and Environmental Engineering. The facilities’ computers are linked to the Virginia Department of Transportation’s Smart Traffic Center computers, which collect and archive ITS data for research, developmental, and educational purposes. The data archiving systems that supply data to the lab are the Smart Traffic Centers in Hampton Roads, Richmond, and Northern Virginia, which provide freeway detector and video data. Additionally, the Northern Virginia Smart Traffic Signal System (NVSTSS) supplies data from about 1,000 signalized intersections located in Fairfax, Prince William, and Loudon Counties (12). The data available from the NVSTSS includes volume, speed, and occupancy data from system detectors on arterial roadways, as well as data from the Management Information System for Transportation (MIST) on which timing plans were in effect at what times, and records of signal changes. This data is aggregated over 15 minute time periods and stored at the NVSTSS. Each night, the NVSTSS computers compile the 96 data sets from each detector station and signal, compile them, and send it to the Smart Travel Lab. Additionally, real time data is also available from the system detectors and MIST system for use in traffic management tasks.

At first the NVSTSS and MIST system data might seem to be an ideal source of data, however several concerns caused this source to be ruled out. In order to determine the effects of rain at signalized intersections, detailed data from actual intersections is needed. Calculating the saturation headway accurately requires a fine data resolution. Data that is aggregated every 15 minutes will definitely not provide this resolution. The NVSTSS system detector data could have been used in order to perform an analysis similar to the freeway study conducted by Professor Brian Smith and assistants. Three concerns prevented this from occurring. First, arterial speed and occupancy data can be very misleading, especially if queues back up onto the system detectors for portions of the signal cycle. Given the 15-minute aggregation, it would be difficult to discern if a reduction in speed was caused by actual network congestion or a long queue at an intersection. Without speed data, speed-flow diagrams cannot be created, and therefore knowing if the arterial is in the congested regime would not be possible. Secondly, a major assumption would be that demand on the network is constant. Some drivers might change their behavior and travel patterns during inclement weather, and therefore it would be difficult to tell if a change in volume was indicative of a change in travel patterns or a reduction in network capacity. Given that drivers can divert to alternate routes with much more ease on an arterial then on a freeway, the volume data might be misleading as well. Lastly, data quality and detector malfunctions are concerns with using Smart Travel Lab data. Detector malfunctions and data quality might even be directly related to the weather conditions as well. Therefore, it was decided to seek alternate data sources for this project.

3.1.2 Smart Travel Van

The Smart Travel Van allows researchers to collect detailed traffic data from areas that are not served by the Smart Travel Lab data networks. The van can be parked along any roadway in Virginia, and can collect very detailed data from freeways, signalized intersections, rural roads, or work zones. The vehicle is a cargo van that has a 42-foot long telescoping mast, on top of which sits a camera and a positioning unit. Connected to the camera is a state of the art video detection system. This allows data, such as volumes, speeds, occupancies, and headways to be collected. The camera output can be recorded as well for the purpose of manual observation and reduction of data (12).

Provided there is a suitable place to park the Smart Travel Van at an intersection, this would have been an ideal method for collecting data on saturation headways and speeds. One issue prevented the van from being used for data collection. Smart Travel Lab staff members were concerned that having the van out on the road with the mast up during rainy weather presented a risk of the van being struck by lightening. This is not only a safety risk for those collecting the data, but also the van’s equipment could be severely damaged as well. Therefore, lab staff would not allow the van to be used during rain events. Even if the van could have been used, Smart Travel Lab staff informed me that data quality would be poor during inclement weather.

3.1.3 City of Norfolk Signal System

The city of Norfolk recently upgraded their traffic signal system. The signals are controlled from a central traffic management center, similar to the NVSTSS. Conditions are monitored in real time using detector data and video camera images of traffic on arterials in downtown Norfolk. These cameras are permanently installed and guarded from the elements, so therefore the image quality would be better then that from the Smart Travel Van. At the time this project was in the data collection stage, the Norfolk signal system was not yet connected to the Smart Travel Lab, and the date on which this connection would occur was uncertain. Contact was made with Brian Stewart, one of the managers in Norfolk, and a request was made to see if taping the feeds from the video cameras was possible. The response was that taping the feeds was not allowed due to privacy concerns, which ruled out this source of data.

3.1.4 Field Collection Using Video Camera

The Smart Travel Lab has several digital video cameras available for use by students and staff. Tripods were available for use as well. Most importantly, the cameras were available on short notice, and were rarely reserved by others. Due to the ease of availability, it was decided to use these cameras to videotape traffic flow at a signalized intersection. From the video, the saturation headways could easily be reduced.

One concern with using these cameras was how to keep them dry while collecting data during rainy weather. Rain could cause the cameras to be damaged, as well as diminish the quality of the video if precipitation droplets accumulated on the lens. Additionally, collecting data for long periods of time in rainy weather would be tedious for the researchers. The solution to these problems was to run the video cameras from inside of an automobile. Using this method, the cameras and researchers are protected from the elements, and a clear view of the intersection could be maintained using the vehicle’s windshield wipers.

3.1.5 Speed Data

In addition to saturation flow data, collecting some free-flow speed data was also necessary for the completion of this project. This was desired to see if rainy weather had any impact on drivers’ desired speeds. Since the video cameras were not connected to any form of detection system, automatically obtaining speed data from the videos used for saturation flow data was not possible. Simple physics indicates that speed is distance traveled divided by the time it takes to travel that distance. Using that theory, which is the same theory that a pair of loop detectors use, speed can be obtained by observing the travel time over a known distance.

3.1.6 Fairfax County Red Light Enforcement System Data

An attempt was made to obtain data from the Fairfax County red light enforcement system, in order to determine if inclement weather causes the violation rate to increase. The data that was requested was either the number of violations within a specified time period (such as one hour), or the exact dates and times that each violation was generated. While this data certainly exists, it was unclear which entity archives the actual information. Contact was made with officials at VDOT, Fairfax County, and the camera vendor company, however the requests never led to any data being acquired. All agencies except the vendor only had access to monthly violation totals, whereas the vendor had the detailed data.

3.1.7 Weather Data

While collecting traffic data for this project, observation of the weather conditions was extremely important. Since the weather is subject to abrupt changes at any time, stopping daily data collection efforts prematurely might be necessary if precipitation stops falling. Specific numerical data on precipitation amounts and weather conditions was desired for several reasons. First, the temperature should be noted to ensure that conditions were above the freezing point. Secondly, the amount of precipitation that fell would be desirable to know. Lastly, other factors should be recorded as well, such as the relative humidity, visibility, and sky conditions. While the latter items have no direct numerical relevance to this study, having a record of conditions observed during data collection is desirable.

The National Climactic Data Center is a branch of the National Oceanic and Atmospheric Administration, and claims to be the world’s largest archive of weather data, and is the same database that the Virginia Climatology office uses. Weather stations, primarily located at airports around the country, report weather data items of interest, such as sky conditions, visibility, weather type, temperature, humidity, wind speed and direction, atmospheric pressure, and precipitation amounts. The nearest station to Charlottesville is at the Charlottesville-Albemarle County Airport, which is approximately 6 miles north of the city center and the University of Virginia. Data is available on an hourly basis, and archived data is available from November 1998 to present. While a fee is normally required to obtain this data via the web, educational users accessing the page from a .edu domain name are granted free access (13).

3.1.8 Summary

From this section, one can easily see some of the reasons why so little rainy weather traffic data exists. The rain not only affects traffic flow, but it can also affect data quality, data collection methods, and the safety of researchers and equipment. It was determined that several data sources were not suitable for this project, including Smart Travel Lab detector data which was not archived at a high enough resolution, and the Smart Travel Van, which might have produced poor quality data and be dangerous to researchers. Therefore the field data collection effort was conducted from inside vehicles using video cameras. Weather data was obtained to complement the traffic flow data.

3.2 Selection of Data Collection Locations

After the methods for collecting data had been decided, locations to collect the data from needed to be decided upon as well. As stated in the previous section, two types of data were collected: saturation headways from an intersection, and free-flow speed data from an arterial segment. Several considerations were taken into account when choosing data collection locations, which are described in the following sections.

3.2.1 Intersection Data Collection

According to Garber and Hoel, the maximum discharge rate, or the saturation flow rate, is typically achieved 10 to 14 seconds into the green phase. This is usually the time that the 4^th or 5^th vehicle enters the intersection. Therefore the saturation flow rate calculation starts with the headway between the 4^th and 5^th vehicle. Only vehicles that are initially queued up when the signal turns green are considered in the saturation flow calculation (14). What this means is that an intersection used for calculating saturation flow rates should have sufficient traffic volumes such that queues more than 5 vehicles long are present when the green phase begins most of the time. Two other requirements for an intersection were that it should be close to the grounds of the University of Virginia, and have a parking lot or other safe area where data collection can take place from inside a vehicle. Being close to the University was necessary to facilitate quick access, as most data collection was “spur of the moment” with little or no advance planning, due to the unpredictability of the weather. The parking requirement included a caveat that from the angle at which the car was parked, the signal head and stop bar had to be visible to the recorder.

Several intersections in Charlottesville, which are notorious for congestion, immediately came to mind. These intersections are: University Avenue/Ivy Road and Emmet Street, Emmet Street and Barracks Road, Emmet Street and Angus Road, Emmet Street and Hydraulic Road, West Main Street and Ridge/McIntire, and McIntire Road and Route 250 bypass. Table 3 shows the intersections with the number of corners where adequate parking was available, and their proximity to the University (with 1 being the closest and 6 being the farthest).

Intersection	Proximity to UVA (rank)	Number of Corners with Parking
Emmet/Ivy	1	1
Emmet/Barracks	2	3
McIntire/West Main	3	0
Emmet/Angus	4	3
Emmet/Hydraulic	5	1
McIntire/250 bypass	6	0

Table 3. List of intersections considered for data collection

From the list of intersections above, it is easy to eliminate the two located on McIntire Road, as there is no available place nearby to park a vehicle from which to run the video cameras. The intersection at Emmet Street and Angus Road was ruled out due to another intersection in very close proximity (the Route 250 bypass off-ramp). The intersection of Emmet Street and Hydraulic Road was eliminated from consideration due to the large intersection area (three through lanes, dual left turn lanes, and a 5-lane cross road), which would make videotaping difficult. Of the four intersections on Emmet Street, the best intersection in terms of number of places to park a vehicle and proximity to the University is Emmet Street and Barracks Road. The intersection of Ivy Road and Emmet Street was still a possibility though. Figures 2 and 3 are diagrams showing the lane usage and the possible camera angles for the intersections of Ivy Road/Emmet Street and Barracks Road/Emmet Street. It should be noted that in both cases, Emmet Street is the north to south roadway, while Ivy Road and Barracks Road both run from east to west.

Figure 2. Lane Usage and Camera Angles for Ivy Road and Emmet Street

Figure 3. Lane Usage and Camera Angles for Barracks Road and Emmet St.

After considering the camera angles that could be used at both intersections, Emmet Street and Barracks Road is clearly the best location to collect saturation flow data from. Due to the constrained space at the corner of Emmet Street and Ivy Road, viewing the stop bar and the signal head for any particular direction is not possible. Additionally, a shared through and right turn lane, such as that on southbound Emmet Street is not as desirable as a dedicated through lane. At the intersection of Emmet Street and Barracks Road, the camera angles are much more desirable. From both camera angles, one can see the signal heads and the stop bars on Emmet Street, which is the roadway with the dedicated through lanes. The camera angle on the northwest corner (camera angle # 2 in the diagram) gives a better view than the one on the southeast corner. Additionally, the camera angle on the northwest corner gives a view of southbound Emmet Street, which has two dedicated through lanes, as opposed to northbound Emmet Street, which has one shared through and right turn lane. Therefore, camera angle # 2, on the northwest corner of Emmet Street and Barracks Road was the preferred location to collect the video data from. The view from the camera location is shown in Figure 4.

Figure 4. View of Emmet Street and Barracks Road Intersection from Camera Location (Photo Credit: Kathleen Saathoff, May 1, 2005)

3.2.2 Arterial Free-flow Speed Data Collection

In order to determine the effect of rain on drivers’ desired speeds, cars were observed driving a set distance along an arterial roadway in free-flow conditions, and the speed was derived from the travel time. Free-flow conditions were necessary in order to ensure that a driver was traveling at the speed he desired based on the roadway and road conditions, as opposed to being affected by the speeds of other vehicles. In order to carry out this task, a long, straight, and uninterrupted segment of arterial roadway was needed. Much like the intersection selection, this task required an area to park a vehicle where one would have a view of the arterial segment in question. Two such segments that came to mind were Seminole Trail just north of Hilton Heights Road, and Fifth Street south of downtown Charlottesville. Both are arterial roadways with 45 mile per hour speed limits. Frontage roads on the east side of both arterials make it possible to park parallel to the arterial and look into the traffic stream, meaning vehicles can be observed over long distances.

Despite its location farther from the University of Virginia, the Seminole Trail segment was chosen for this portion of the data collection. The main reason was the higher traffic volumes on Seminole Trail would make it easier to collect larger amounts of data. A public frontage road runs parallel to Seminole Trail and serves several businesses. Only a grassy median separates the frontage road from Seminole Trail, so no trees or any other objects are present to obstruct the view of traffic from that vantage point, whereas on Fifth Street, several trees block the view of the main roadway.

3.3 Summary of Findings

While many sources of traffic flow data and accurate hourly weather data are available, combining the two together to draw conclusions about the effects of weather on traffic is more difficult. Data collection that involves researchers observing traffic almost exclusively occurs during dry weather. Archives of traffic data, such as in the Smart Travel Laboratory, are usually not precise enough to determine parameters that are of interest to weather researchers, such as the saturation flow rate. Furthermore, some archived data, such as archived speed and occupancy values, is not useful when collected in interrupted flow conditions, such as on arterial roadways. Two data collection sites were chosen, one for saturation flow data, and the other for speed data. The intersection of Emmet Street and Barracks Road is an ideal location to collect traffic data from under saturated conditions, however due to the close proximity of other traffic signals, it is not a suitable location for obtaining free flow speed data. The stretch of Seminole Trail north of Hilton Heights Road is a better location for collecting speed data, due to the unobstructed vantage point, free-flowing traffic with few driveways, and a large volume of traffic.

CHAPTER 4. DATA COLLECTION AND ANALYSIS

4.1 Intersection Data Collection

The saturation flow data collection occurred over the fall and winter of 2003/2004. Care was taken to ensure that no frozen precipitation occurred and that only rainy weather was considered for this project. The data collection occurred over 7 sessions of approximately 1 or 2 hours in duration. Four of these sessions occurred in fall, 2003, and the remaining three occurred in spring, 2004, with all sessions occurring during the work-week, approximately during the mid-day. This was done to ensure that driver characteristics would be similar from day to day.

4.1.1 Field Determination of Saturation Flow Rate

The method used to determine the saturation flow rate is the same method that is commonly used by traffic engineers calculating the flow rate by hand. The saturation flow rate is the maximum discharge flow rate during green time, which usually occurs between 10 and 14 seconds into the green phase. The flow immediately after the signal turns green is less then the saturation flow rate, as those vehicles are traveling at slower speeds. Therefore, the saturation flow is calculated by noting the time at which the rear of the 4^th vehicle in the queue crosses the stop bar, as well as the time at which the last vehicle in the queue crosses the stop bar. The difference in time divided by the number of headways counted is then taken to be the saturation headway (note that the number of headways is one less then the number of vehicles counted, since the 4^th vehicle is already in the intersection when the time starts). The saturation flow rate is obtained by dividing the saturation headway into 3600 (14).

saturation flow = 3600/[(t₄ – t_n)/(n – 4)]

where:

t₄ = time 4^th vehicle crosses stop bar

t_n = time last queued vehicle crosses stop bar

n = index of last queued vehicle

The digital video camera embeds a time stamp onto the tape, to 1/100^th of a second precision. Therefore, determination of the saturation flow rate would be relatively straightforward, using the timestamps on the videotape.

Before data collection could occur, the fine details of collecting data with the video cameras had to be worked out. The original intent was to use two video cameras, with two people. One camera would be focused on the stop bar and traffic signal head to catch the saturation headways. The other camera would be pointed upstream to note the length of the queue at the signal. Due to the very impromptu nature of data collection during rainy weather, it was often not possible to arrange to have assistance in advance. Therefore, on the first day of rainy weather data collection, a method for using one camera was improvised. The camera was aimed at the stop bar and the signal head, while the data collector looked in the other direction at the queue of vehicles. Immediately after the last queued vehicle cleared the stop bar, the data collector waved an object in front of the camera lens, signifying that no more vehicles should be considered in saturation flow calculations. This improvised method allowed the rest of the data collection effort to occur with only one person, often on the spur of the moment depending on weather conditions.

4.1.2 Summary of Weather Conditions During Data Collection

Weather conditions were noted before the data was collected by simple visual inspection of conditions. However the actual weather conditions were verified to check factors such as precipitation amount, temperature, and visibility conditions. The weather data was downloaded from the NOAA’s National Climactic Data Center’s (NCDC) website (13). Two things must be noted when interpreting NCDC data sets. First, for precipitation amounts, no entry means no precipitation occurred, “M” means missing data, “T” means trace precipitation (less then 0.01 inches), and numerical values equal precipitation amounts. The data number presented for a particular time is the amount of precipitation in the preceding hour (therefore 0.02 inches at 11PM means that the precipitation fell between 10:01 and 11:00. Table 4 lists the data collection dates and times, and Table 5 shows the observed weather conditions.

Session	Date	Day	Start Time	End Time	Weather Condition
1	10/27/03	Monday	11:45	12:45	Rain + Mist
2	11/17/2003	Monday	12:00	1:00	Clear
3	11/19/2003	Wednesday	11:30	1:30	Rain
4	11/20/2003	Thursday	1:00	3:00	Scattered Clouds
5	2/18/2004	Wednesday	11:15	1:15	Clear
6	2/19/2004	Thursday	11:45	12:45	Clear
7	3/16/2004	Tuesday	11:30	1:30	Rain + Mist

Table 4. Data Collection Dates, Times, and Weather Conditions

Date	Start Temp (Deg. F.)	End Temp (Deg. F.)	Start Precip (in.)	Mid Precip (in.)	End Precip (in.)
10/27	60	60	0.06		0.02
11/17	51	56	0		0
11/19	66	60	0.12	0.83	0.03
11/20	59	61	0	0	0
2/18	42	46	0	0	0
2/19	55	58	0		0
3/16	37	37	0.14	0.03	0.01

Table 5. Temperatures and Precipitation Amounts for Data Collection Sessions

In Table 5, the starting and ending temperatures in degrees Fahrenheit are noted for each session, as well as three values for precipitation in inches per hour. The wet weather data collection sessions all started and ended during the middle of an hour time block of NOAA weather data. The “Start Precip” and “End Precip” values represent the amount of rain that fell during the entire hour in which data collection started and ended respectively. Therefore, only part of that hour was actually captured in the field data, and the values for precipitation in the above table cannot be summed to determine the total amount of precipitation that fell during data collection. The above numbers were given to show the reader the average rate at which precipitation fell during that times when data collection started and ended. For the two hour data collection sessions, a “Mid Precip” value is also given to represent the amount of precipitation that fell during the entire middle hour of the data collection.

4.2 Data Reduction and Results

The saturation flow data was reduced from the video images gathered during the data collection. The previous section described the methodology for determining the saturation flow rate from the video. The reduced data was from the southbound right through lane on Emmet Street. The position of the camera during data collection made it potentially difficult to see the stop bar and the queue length for the left lane, so the right lane was observed to make the data consistent. The data was reduced by hand, watching the video in slow motion to determine the values for saturation headway as precisely as possible. The results from each individual session are shown in the Table 6 and Table 7.

Date	Weather	# of Observations	Mean (s)	Median (s)	Min (s)	Max (s)	Mean Saturation Flow (veh/hr)
11/17	Dry	18	2.02	2.00	1.59	2.71	1782
11/20	Dry	43	2.01	1.93	1.62	2.85	1791
2/18	Dry	44	2.02	1.98	1.60	2.60	1782
2/19	Dry	23	2.08	1.99	1.72	2.77	1731
10/27	Wet	23	2.07	2.00	1.60	3.10	1739
11/19	Wet	39	2.11	2.04	1.65	3.03	1706
3/16	Wet	42	2.19	2.03	1.64	2.97	1644

Table 6. Table of Saturation Headway Results (seconds)

Weather	# of Observations	Mean (s)	Median (s)	Standard Deviation (s)	Min (s)	Max (s)	Mean Sat Flow (veh/hr)
Dry	128	2.03	1.98	0.270	1.59	2.85	1773
Wet	104	2.13	2.03	0.367	1.60	3.1	1690

Table 7. Compiled Results for Saturation Headway

As seen in Table 7, the mean saturation headway increased from 2.03 seconds per vehicle to 2.13 seconds per vehicle during inclement weather. This is an increase of 4.9 percent, which leads to a reduction in saturation flow from 1,773 vehicles per hour per lane to 1,690 vehicles per hour per lane. This translates to a 4.7 percent reduction in the saturation flow rate and the capacity of the intersection.

4.3 Speed Data Collection

The desired free flow speed data was collected later in the spring of 2004, in two sessions lasting approximately one hour each. The location of the data collection was Seminole Trail north of Hilton Heights Road, where a vehicle could be parked along the frontage road facing into the stream of traffic. Using a scaled aerial photograph of the area, a distance of 528 feet (1/10^th of a mile) was measured from the stop bar at the intersection of Hilton Heights Road northward. The exact location was found in the field and verified using a digital vehicle odometer while driving northward on Seminole Trail. Once this location was verified, the data collection could be completed. The vehicle was parked along the frontage road, facing south, and vehicles were observed and timed driving from the Hilton Heights Road stop bar to the end point. Figure 5 shows a diagram of the data collection.

Figure 5. Schematic of Speed Data Collection Location

Vehicles that were clearly not following other vehicles were randomly chosen to have their travel time timed. This meant that one had to wait until after the signal at Hilton Heights Road turned green, and the platoon of queued vehicles proceeded through the intersection. After it was determined that the platoon was sufficiently far from the intersection, and vehicles heading north on Seminole Trail were doing so at their own desired speed (as opposed to following another vehicle), then randomly selected vehicles were timed driving this 1/10^th of a mile segment of roadway. The starting location (the stop bar at Hilton Heights Road) was visible and painted onto the roadway, and the end location was referenced by a tree planted in the median of Seminole Trail directly across from the observer. The travel time for the vehicles was noted, and the speeds were calculated from the travel time by multiplying the travel time by 10 and then dividing the result into a conversion factor of 3600. The weather conditions and results are shown below in Tables 8, 9, and 10.

Session	Date	Day	Start Time	End Time	Weather Condition
1	5/18/2004	Tuesday	5:15	6:00	Rain
2	5/21/2004	Friday	5:15	6:15	Clear

Table 8. Dates, Times, and Weather Conditions for Speed Data Collection

Date	Start Temp (deg. F.)	End Temp (deg. F.)	Start Precip (in.)	End Precip (in.)
5/18	80	66	0.31	Trace
5/21	76	76	0	0

Table 9. Temperatures and Precipitation Amounts for Speed Data Collection

Date	Weather	# of Observations	Mean (mi/hr)	Median (mi/hr)	Standard Deviation (mi/hr)	Min (mi/hr)	Max (mi/hr)
5/21/04	Dry	63	47.8	47.8	3.888	39.9	61.2
5/18/04	Wet	47	43.3	42.9	4.088	35.3	53.6

Table 10. Speed Data Collection Results in Miles per Hour

The average free flow speeds decreased from 47.8 miles per hour to 43.3 miles per hour during wet weather. This represents a decrease of 9.44 percent.

4.4 Significance Testing

From the above sections, it would appear to the casual observer that there are differences in the saturation flow rates and free flow speeds during wet and dry weather. An informed researcher will instead conclude that the mean saturation flow rate, startup lost time, and free flow speeds are actually the same during wet and dry weather, due to statistical variation in the randomly chosen sample that was observed and measured. This would indicate that wet weather has no effect on traffic flow parameters, and that the wet and dry weather data samples were really from the same whole population. This assumption is made until the researcher can statistically prove otherwise. This is called a null hypothesis (H₀), which has to be disproved. If it is disproved, then the researcher can conclude that the alternate hypothesis, (H_A), is true, meaning that wet and dry weather sample data measurements came from different populations, and that wet weather has a significant impact on traffic flow. A commonly used method for achieving this is the t-test (15).

A t-test computes a value for the t-statistic, which is based on the standard deviation of the difference in means. The t-statistic is computed using the following formula (and can also be done using a statistical software package as well).

After the t-statistic for the two samples of data is calculated, it is then compared to the critical t-value, t_c. As all statistics are based on probabilities, there is still a chance that the null hypothesis test can provide the wrong result. Therefore, a confidence interval is chosen. If a confidence interval of 95 percent is chosen, then the researcher will really be stating that he is 95 percent sure that the null hypothesis is invalid if the t-test proves so. The significance level, alpha, is computed by subtracting the confidence interval from 1. So a confidence interval of 95 percent yields an alpha value of 0.05. At that point, the critical t-value can be determined by looking at appropriate tables in statistical texts, using the alpha value and degrees of freedom (related to sample size) to determine which value to use. If the critical t-value is larger than the absolute value of the computed t-statistic, then the difference in the means is assumed to be from chance variation in the selected sample. If the computed t-statistic is greater than the critical t-value, then the differences in the means fall at one of the extreme ends of a normal probability curve, and therefore the null hypothesis is void and the means are considered to be “significantly different”. This would indicate that some external factors other than chance variation were present which caused the means to be different (15).

In this example, the following null and alternative hypotheses are presented:

H₀ = No differences in driver behavior due to weather

H_A = Differences in driver behavior exist due to weather

Table 11 shows the value of the t-statistic and the critical t-value for the saturation headway and free-flow speed data. The null hypothesis is disproved in both cases, and therefore, one can assume with 95% confidence that the alternative hypothesis is true, and wet weather has a significant impact on the saturation headway and free-flow speed.

Parameter	Dry Mean	Wet Mean	Percent Change	D.F.	Alpha	Critical-t	t	Null Hypothesis Valid	Significant Difference
Saturation Headway	2.03	2.13	4.93	230	0.05	1.96	2.32	NO	YES
Free Flow Speed	47.8	43.3	-9.41	108	0.05	2.00	5.83	NO	YES

Table 11. Null Hypothesis Test Parameters and Significant Differences in Means

4.5 Discussion of Results

The saturation headways were shown to increase 4.92 percent during rainy weather, from 2.03 to 2.13 seconds. This corresponds to a decrease in the saturation flow rate at the intersection from 1,773 to 1,690 vehicles per hour per lane. While this might not seem like a very large reduction, it should be noted that if the intersection is operating near capacity during dry weather, the wet weather could be the limiting factor that pushes the intersection into the congested regime. In a study done by Wilbur Smith Associates on the intersection of Emmet Street and Barracks Road, the saturation flow rates were calculated in Synchro, which uses the reduction factor method seen in the Highway Capacity Manual. The saturated flow rate for the southbound through lane group was listed as 3,505 vehicles per hour (16). Divide that by two, and the result is 1753 vehicles per hour per lane. This result is in between the values found in the field results, which indicates that the intersection actually has a higher capacity than that seen in the Highway Capacity Manual; however a reduction in capacity from the Highway Capacity value is still observed during rainy weather. In the Salt Lake City study, the saturation flow rate decreased 6 percent during rainy weather, which is slightly higher than the 4.92 percent observed in Charlottesville (8). The study done in Vermont showed only a 2.5 percent reduction in saturated flow rates. While there was a reduction shown, it was determined to be insignificant (11). Regional variation in results might be a factor in this. The results do seem to be fairly similar to those seen in other studies, which add some validity to them.

The Salt Lake City study was the only other study that took speeds into account. Their results indicated a 10 percent reduction in speed (8), which is similar to the 9.44 percent value seen on Seminole Trail. The FHWA study done in 1977 on Interstate highways, however, shows no reduction in speed during inclement weather (1). The Hampton Roads Freeway study showed a 3 to 5 percent decrease in operating speeds during light and heavy rain conditions (5). It appears that the arterial results for speed reductions are greater then those for freeways, based on the literature. One would intuitively expect there to be a reduction in speed, due to the reductions in visibility and drivers behaving more carefully during inclement weather conditions. It should be noted that the precipitation rate at the beginning of the speed data collection was 0.31 inches per hour. This was the highest rainfall rate observed during data collection (and the only instance of heavy rain), and may have exaggerated the effect of rain on free-flow speed.

4.6 Additional Field Observations

While the majority of the field work concentrated on obtaining numerical data, several observations were made that may be of interest to the reader. These observations generally relate to driver behavior and the functionality of traffic control devices during wet weather. These observations might be investigated further in additional studies.

First, the Highway Capacity Manual notes that changes in driver behavior due to wet weather are likely a result of reduced visibility (4). The manual is specifically referring to the drivers’ desired free flow speed; however reduced visibility would likely affect other parameters (such as acceleration rates and saturation headways) as well. It was noted that the reduction in visibility due to wet weather did not necessarily correspond to the times when precipitation was falling. When precipitation first begins to fall, a reduction in visibility takes place due to the accumulation of water on the vehicle’s windshield. Generally, the higher the rate of precipitation is, the greater the reduction in visibility. However when precipitation slows or stops falling, the visibility does not increase as much as one may expect if there is heavy traffic on the roadway. After heavy downpours, water accumulates on the roadway, and after precipitation stops falling, the water remains on the roadway for some time afterwards. The tires on vehicles passing over this wet roadway tend to “spray” the water on the road backwards and into the windshield of the following vehicle. This leads to a continued reduction in visibility. Therefore, when conducting research on driver behavior due to precipitation, pavement conditions should be taken into consideration in addition to precipitation rates obtained from weather data. Immediately after precipitation stops falling, the roadway will not be operating under “dry” conditions.

On a well designed roadway, water flows away from the center of the road and into drainage structures on either side. However imperfections in the roadway surface often cause water to accumulate in the travel lanes. Water can also accumulate in intersections, especially if a drainage gutter on one roadway crosses the intersecting roadway. As many vehicle drivers are cautious and fearful of hydroplaning, drivers will often slow down when approaching accumulations of water in a roadway. This behavior was noted at Emmet Street and Barracks Road, especially for through vehicles on westbound Barracks Road. Some minor depressions in the roadway surface near the western limits of the intersection caused water to accumulate and many drivers to reduce their speed upon seeing this puddle of water. Any reduction in speed can affect intersection capacity and cause disturbances in smooth traffic flow. It is therefore important to maintain the roadway surface properly to prevent water from accumulating in the roadway.

Traffic control devices can fail during inclement weather, which will merely add to any traffic problems caused by weather. It was observed that a detector failed during heavy rainfall while collecting speed data on Seminole Trail. The signals on Seminole Trail rely on video detection, with cameras mounted on the signal mast arms. The cameras send images to the signal controller, and the controller looks for changes in the camera image in certain locations. The heavy rainfall likely caused a change in the image significant enough that the controller interpreted it as a vehicle waiting to be served. The result was that the southbound left turn signal turned green for the maximum amount of allotted time on every cycle. Many times very few vehicles, or even no vehicles, were actually waiting to turn. Given that the peak direction of traffic during the evening rush hour is northbound, the northbound through movement was being given less green time as a result of this failing detector. When the heavy rainfall stopped, the signal returned to normal operation, with the left turn only being activated when vehicles were waiting, and then only long enough to service them. During dry data collection at the same location, the signal operated properly the entire time.

Lastly, red light running seemed to be a common occurrence at Emmet Street and Barracks Road, both during dry and wet weather. Arrival patterns and the lack of signal coordination often meant that the signal would turn red as a pack of vehicles arrived. Since drivers know that this signal is on a long cycle, many of them attempt to pass through before the onset of the red phase. Others blatantly enter the intersection after the signal has turned red. This may be indicative of an amber phase that is too short, or it may simply be a result of aggressive driving behavior. During one day of data collection, 5 consecutive signal cycles had red light runners when the light on southbound Emmet Street turned red. On even more occasions, northbound left turning vehicles were still in the intersection after southbound Emmet Street had a green signal indication.

4.7 Data Collection Issues

Blurry images and obstructed views were two main issues with the data collection effort. These concerns were especially true during wet weather data collection. Collecting data from inside of a vehicle was necessary to protect the researchers and the equipment, however video-taping through a windshield with wipers running is far from an ideal field environment, and data quality may have been affected as a result. Additionally, there were some issues with trying to locate the end of the queue while keeping the camera pointed at the stop bar. This was a result of glare and water accumulating on the car windows, and also a result of other parked vehicles obstructing the upstream view for short periods of time. Efforts were made to overcome these difficulties, however collecting data in this fashion is not recommended for future research. The best method for obtaining video data would likely involve setting the camera on a tripod with some form of temporary shelter over it. Therefore the camera would stay steady and dry, and the image would be clearer as there would not be glass between the camera and the data subjects.

CHAPTER 5. MICROSCOPIC SIMULATION NETWORK DEVELOPMENT

The next step in this research project was to develop a microscopic simulation model to simulate the conditions observed in the field, to determine how rainy weather affects various performance measures, and to check the feasibility of mitigation strategies such as adjusting the signal timing plans. It is much easier to obtain values for parameters such as travel time and delay time using a simulation model then in the field. In order to meet this goal, a simulation software package had to be chosen, the network and all network elements coded, and the model calibrated.

5.1 Selection of Simulation Software Package

The first step in this network development process was to select a simulation software package. Several software packages were available for use, including CORSIM, VISSIM, and Paramics. CORSIM is a simulation software package that has been developed by the Federal Highway Administration. It is a comprehensive package that allows the simulation of arterials and freeways and other large networks. It is a program that has been in use for several decades, and is constantly being refined to add more features and capabilities (17). Paramics is another microscopic simulation software package that has been used by researchers and students alike. While it too has many features, VISSIM appeared to be the best program to use in terms of providing detailed output. VISSIM is a very robust, behavior-based simulation package designed in Germany. It has the ability to be used in many different aspects of transportation, including freeways, arterials, pedestrian malls, and even rail transit lines. There are many different performance measures that VISSIM can evaluate and report to the user, which makes evaluation very comprehensive (18). Since the researcher had previous experience using VISSIM, and none using CORSIM or Paramics, VISSIM was used for the simulation software package.

5.2 Network Coding

In order to use VISSIM, the networks have to be coded into the simulation software package. Two networks were developed: an isolated intersection, and a network with three actuated-coordinated traffic signals. The isolated intersection was Emmet Street and Barracks Road, and the reason why this was used was so that the field data could be used to calibrate the simulation software to output the desired results. However examining the effects of changing the offsets of signal timing plans could not be evaluated using an isolated intersection. Therefore, another network was coded using three coordinated intersections on Seminole Trail north of Charlottesville.

5.2.1 Isolated Intersection

The intersection of Emmet Street and Barracks Road had to be coded from scratch, as no network file for that intersection existed previously. The first step for doing so was to obtain an aerial photograph of the intersection, which was then scaled and inserted into VISSIM as the background image. The roadway links were then drawn in on top of the background image in order to ensure that everything was at the proper scale. After the links were drawn, connectors were added between the segments where the number of lanes change (such as when a left turn bay starts) and for each of the 12 possible movements at the intersection. Routing decision points were then inserted such that vehicles would determine which route to take (left turn, right turn, or straight through) almost immediately upon entering the network. A traffic signal controller was created, and detectors laid on all approaches to the intersection. The traffic signal heads were then installed as well, using appropriate NEMA numbering conventions. The signal heads were associated with the appropriate detectors as this is an actuated intersection. Since right turns on red are allowed at this intersection, the right turn connectors were extended back behind the signal heads (meaning cars turning right do not stop for the traffic signal), and a stop sign was inserted directly over the signal head with the “only on red signal indication” box checked. Therefore, the vehicles will stop before turning right on red, and will not stop when the signal is green. Lastly, a yield sign and priority rule was created for the channelized southbound right turn lane. Figures 6 and 7 are diagrams showing the network geometry.

Figure 6. Emmet Street & Barracks Road VISSIM Network

Figure 7. Close-up of Emmet Street & Barracks Road Showing Network Elements

Next, vehicles were added to the network. The Wilbur Smith Associates study (16) of this intersection had the most recent traffic counts from the AM peak, midday peak, and PM peak hours. The counts and turning volumes for the midday peak were used, as this was when the field data was collected. Traffic volumes for all three turns on each approach were summed and the result used as the input volume at the end of the link. The 12 individual turning volumes could then be added as “relative flows” in the routing decisions parameters. A relative flow calculates the turning volume by summing up all of the relative flows from the same routing decision point and then taking the proportion of each route’s relative flow value to the total, and then that proportion of the vehicles entering the network will take that route, regardless of the input volume. Table 12 shows the total and turning volumes for Emmet Street and Barracks Road.

	Left (veh)	Through (veh)	Right (veh)	TOTAL (veh)
Southbound	265	1013	96	1374
Northbound	63	791	96	950
Eastbound	519	245	53	817
Westbound	247	180	216	643

Table 12. Traffic Volumes for Emmet Street and Barracks Road (15)

Lastly, the signal-timing plan had to be input into VISSIM. Since this is an actuated signal, entering the timing plans is slightly more involved then simply entering the start and end times for the green phases directly into VISSIM. Formerly, a VAP file had to be generated, which required programming code into a file. VISSIM has since developed a Java-based NEMA signal timing editor, which makes the process much more user-friendly. For an un-coordinated actuated signal, only the minimum and maximum green times, as well as the passage time, all-red time, and amber phase lengths need to be input on the left. On the right side, the phase orders need to be set inside the NEMA phase rings, with the starting phases selected. Once these steps were completed, the isolated intersection network was ready to be calibrated. Figure 8 shows the user interface and the signal timing plan for Emmet Street and Barracks Road.

Figure 8. NEMA Editor User Interface and Signal Timing Plan

5.2.2 Three Intersection Network with Actuated-Coordinated Signals

In order to examine how offsets play a part in developing weather related signal timing plans, a three-intersection network was used as well as the isolated intersection for analysis. The network was on Seminole Trail, at its intersections with Rio Road, Albemarle Square, and Woodbrook Drive. This network was already coded with all of the elements, and provided to me by Daniel Son. Three steps had to be taken to update the network for this research project. First, on the west side of the intersection at Albemarle Square, there was a driveway that was omitted on the original network that had to be added. Secondly, since the original network had pre-timed signals, no detectors were in place, so those needed to be added, and the signal timing plans input into the NEMA editor. Lastly, the appropriate input and turning volumes had to be added. The volumes and signal timing plans were provided by the Virginia Department of Transportation in the form of Synchro files of the entire Seminole Trail corridor. The network is shown in Figure 9, with Rio Road being the southern-most intersection, Albemarle Square in the middle, and Woodbrook Drive at the northern end.

Figure 9. Seminole Trail Network

In order to time the signals in the NEMA editor, the same steps were taken as for the isolated intersection. However additional items need values on the user interface as well. First, the cycle length and offset needs to be input, and the coordinated phases ticked. The lead phases in each NEMA ring need to be selected, and the coordinated phases ticked as well (the coordinated phases are always phases 2 and 6). Finally, the auto-calc splits option is ticked, allowing the user to input the splits, which will automatically calculate the permissive starts, permissive ends, and force off values for each intersection.

5.3 Network Calibration

In order to provide accurate results, the simulation program must be calibrated. The default parameters that impact the speed and the saturation flow rate do not necessarily provide results that are representative of real world conditions. Therefore, the simulation is run with the default parameters, and the output of results is checked. If the results differ from real world values, then the default parameters must be changed in order to make the simulation program represent the real world. In VISSIM, there are several parameters that deal with weather conditions that can be adjusted to calibrate the model.

These include the following:

• Desired Speed Profile
• Maximum and Desired Acceleration Rates
• Maximum and Desired Deceleration Rates
• Driver Behavior Parameters bx_add and bx_mult, which relate to saturation flow rate
• Lane Changing behavior
• Lateral Behavior
• Reaction to Amber Signal

The acceleration and deceleration rates are two parameters that are most likely affected by weather conditions. However no detailed information was available on exactly how these parameters were affected. No data on acceleration or deceleration was available for analysis either, as this data would be very difficult to collect. Finally, the acceleration and deceleration rates are not simply values that are input into the simulation model. Graphs are actually provided and the user must arrange the line deflection points to represent the acceleration and deceleration profile. Lane changing behavior, lateral behavior, and reaction to the amber signal were also parameters that could be calibrated if more data were available. The desired speed profiles and driver behavior parameters relating to the saturation flow rate were much easier to input into VISSIM, and they could be calibrated to show the conditions observed in the field. Therefore, it was decided to focus calibration efforts on those two parameters. In future research, it may be desirable to calibrate and validate additional parameters to better reflect field conditions.

5.3.1 Desired Speeds

The desired speed of a vehicle is determined by a speed profile and desired speed points which are placed on the network. A vehicle will then attempt to travel at its desired speed, however other vehicles may prevent it from doing so. Desired speed points are placed across particular lanes in a network, and any vehicle passing over that point is assigned a desired speed based on the desired speed profile associated with that point. The desired speed profile is a graph showing the speed distribution along that section of roadway. The shape and end points of that graph can be altered to change the probability that any vehicle passing over the point will be assigned a certain speed. The points placed on the graph are assigned two values. The y-value is a speed, and the x-value is a percentage. The percentage in the x-value represents what percentage of vehicles has a desired speed lower than the speed for that point.

For the Emmet Street and Barracks Road isolated intersection, 14 desired speed points were placed along the network. One was placed on each lane entering the network to assign desired speeds to entering vehicles. Additionally, one was placed on each lane leaving the intersection in order to assign a new speed to any vehicle which turns from Emmet Street onto Barracks Road and vice versa, as Barracks Road and Emmet Street have differing speed limits. A similar logic was applied to the Seminole Trail network, with desired speed points placed on entrances to the network and exits from the intersections.

The desired speed profile for Seminole Trail was created first, and the logic and assumptions that were used to create that were then applied to the other profiles. Seminole Trail has a speed limit of 45 miles per hour. The speed data collected indicates a higher average speed, though this is likely because the data collection location is on a downhill grade. Therefore, it was assumed that the average desired speed would be the speed limit of 45 miles per hour. The standard deviation of the speed data was roughly 4 miles per hour, indicating that approximately 66% of the traffic was within 4 miles per hour of the mean speed, and therefore within the range of 41 to 49 miles per hour. Approximately 95% of the traffic would fall within two standard deviations of the mean, or within the range of 37 to 53 miles per hour. The remaining 5% of the traffic falls outside the 37 to 53 miles per hour range.

It was decided to ignore this remaining 5% of the traffic, and just assume that two-thirds of the traffic falls within one standard deviation of the mean, and the remaining third falls within two standard deviations of the mean. The lower and upper bounds for the speed limit were placed at 37 and 53 miles per hour. Two additional points were created, one at 41 miles per hour and the other at 49 miles per hour. The 41 mile per hour point was located on the speed profile at an x-value of 16.5 percent, while the 49 mile per hour point was located at 83.5 percent. Thus the speed profile was centered about 45 miles per hour, but the vehicle speeds were not evenly distributed amongst the 37 to 53 mile per hour range. Two-thirds of the vehicles were evenly distributed among the 41 to 49 mile per hour range, while the remaining third was evenly distributed among the 37 to 41 and 49 to 53 mile per hour ranges.

Since Emmet Street has a 40 mile per hour speed limit, the profiles were adjusted to reflect that change. A standard deviation of 3.5 miles per hour was calculated using proportions, and it was used in lieu of the 4 miles per hour seen on Seminole Trail. Therefore the speed profile points were located as follows:

• Boundary points of 33 and 47 miles per hour
• 16.5% point at 36.5 miles per hour
• 83.5% point at 43.5 miles per hour
• Mean speed of 40 miles per hour

The Barracks Road speed profiles were created in the same manner, using a 2.5 mile per hour standard deviation, with a mean speed of 25 miles per hour.

Those speed profiles represented the desired speeds of vehicles under dry conditions. A speed reduction was observed in the field under wet conditions, and this speed reduction was applied to the desired speed profiles in order to simulate wet conditions. The mean desired speed was set to be 90% of the speed limit (signifying a 10% reduction in desired speed). The speed profile graph points were shifted in order to create the new speed profiles. For example, on Emmet Street, the desired speed changed from 40 to 36 miles per hour, thus each point on the speed profile was shifted by 4 miles per hour to create the wet weather profiles. The same process was used for Barracks Road and Seminole Trail.

Each desired speed point on a network is assigned desired speed profiles. Multiple profiles are assigned for different vehicle types. In this project, the same profile was used for the two default vehicle types in the network: autos and HGV’s (heavy vehicles, such as trucks and busses). First the “dry” speed profiles were assigned to the appropriate point to represent dry conditions. Then a copy of each file was made and the speed profiles changed to the “wet” profiles to represent inclement weather.

Lastly, the speed profiles were checked to ensure their accuracy. A copy of the simulation file was created and adjusted in two ways. First the traffic signal was removed such that vehicles could travel through the network at their desired speeds. Next the input volumes were drastically reduced to ensure that the few vehicles traveling through the network would not have their speeds impacted by other vehicles on the network. A detector was placed on the network to capture vehicle speeds, and a simulation run was performed for each network. In all cases, the aggregate of all vehicle speeds was within 0.5 miles per hour of the desired speed. It was therefore assumed that the speed profiles were properly set.

5.3.2 Saturation Flow Calibration

While setting the speed profiles was fairly straightforward, calibrating the saturation flow rate was a bit more complicated. The saturation flow rate cannot be set directly in VISSIM; it is instead controlled by two parameters called “bx_add” and “bx_mult” in the driver behavior model. In the VISSIM manual (18), the parameters bx_add and bx_mult are defined as the additive part of the desired safety distance and the multiplicative part of the desired safety distance, which are part of the Wiedemann 74 car following model. The parameters are used to compute the safety distance between vehicles. VISSIM uses a modified version of the car following model, and the inner workings of the software are not known, as the software is essentially a “black box”. Therefore, the only guidance as to how to set the parameter values was from the VISSIM manual, shown in Figure 10.

Figure 10. Saturation Flow as a Function of VISSIM Parameter Settings (18)

This advice is very vague and appears to be of little use. The values were calculated under very specific conditions, shown at the top of the figure. These values could at least be used as a starting point for trying to find the correct values using trial and error.

The value that VISSIM needs to be calibrated to is 1,773 vehicles per hour for dry weather, and 1,690 vehicles per hour for wet weather, which were the values observed in the field. Since the desired free flow speeds differ during wet and dry weather, and desired speed would very likely have an impact on the saturation flow rate, two sets of trial and error runs were done with the different speed profiles for wet and dry weather.

In order to determine the saturation flow rate from the simulations, a special evaluation had to be enabled and configured within VISSIM. The Emmet Street and Barracks Road network was taken, and a data collection point was inserted into the right lane at the stop bar, which represents the same location in the field that the field data was collected from. After special evaluations were enabled in the output evaluations dialog box, the VISSIM *.inp file had to be opened in a text editor to add the following lines of text:

EVALUATION TYPE DISCHARGE SCJ 1 SIGNAL_GROUP 2 COLLECTION_POINT 1 TIME FROM 0.0 TO 99999.0

With this text added to the VISSIM network file, then a special file would be created with a chart of all the discharge headways on the network at the data collection point. Using this file, the saturation flow rate can be determined using the same method used to calculate the value in the field. The output file contains one row for each green cycle, with the discharge headway for each vehicle listed in a column. Therefore, the second vehicle, third vehicle, etc. from each cycle are in their own column.

The vehicle volumes were increased to 2400 vehicles per hour on the southbound approach to the network in order to ensure that a sufficient number of vehicles would be queued up at the signal when it turns green. The assumption was made that at least 10 vehicles would always be queued up at the signal at the beginning of the green phase, and this was verified by watching some simulation runs. Once the simulation run is over, the file can be opened, and the saturation headway computed by summing all of the headway values in columns 5 through 10, and taking the average. The values in columns 1 through 4 are ignored as the headways for the first four vehicles are not used when calculating the saturation flow rate in the field. The values in columns greater than 10 are ignored as those vehicles might not have been queued up originally. The end result is taken to be the saturation headway. This method is mathematically equivalent to the method used in the field to determine saturation headway. When the first set of runs was done, it was determined that the vehicle volumes were causing the left turn bay on southbound Emmet Street to overflow, which might have affected the saturation flow rate calculations. Therefore, for the remainder of the calibration process, the left and right turning volumes on southbound Emmet Street were set to zero.

Four sets of runs were performed to determine the appropriate values for bx_add and bx_mult. Wet and dry values were determined separately, each using their own speed profiles during the calibration runs. In each scenario, values for bx_add and bx_mult were selected, and 4 runs were performed using each scenario. The values for each of the four runs were averaged to determine what saturation headway values the parameters were generating. The first set of runs was performed using the default values, and educated guesses based on the charts in the VISSIM manual. The results indicated that the saturation headways were far too low in both the wet and the dry scenarios. Much higher values for bx_add and bx_mult were then chosen for the second set of runs. The results produced saturation headway values that were both higher and lower then the desired values, so at this point, an approximate range of appropriate values was determined. The third set of runs attempted to “zoom in” on the correct values. Last, the fourth set fine-tuned the approximate values determined in the previous set of runs. The final values that were selected were 3.7 for bx_add and 4.7 for bx_mult in the dry weather case, and 4.1 for bx_add and 5.1 for bx_mult in the wet weather case. It should be noted that a large increase in the parameter values was required from the default values of 2.0 and 3.0. A summary of some of the saturation headway values observed with various values of bx_mult and bx_add are shown below in Table 13.

dry			wet
bx_add	bx_mult	sat. headway	bx_add	bx_mult	sat. headway
2	3	1.64	2	3	1.63
2.5	3	1.72	2.2	3.5	1.75
2.5	3.5	1.76	2.5	3.5	1.75
3.5	4.5	1.98	4	5	2.08
4	5	2.14	4.5	5.5	2.20
3.7	4.7	2.03	4.1	5.1	2.14

Table 13. Selected Values for Saturation Headway Observed During Calibration

In order to ensure the accuracy of the results, portions of two runs were observed manually to ensure that the discharge headways in the table matched the values seen in the simulation.
CHAPTER 6. ANALYSIS OF THE EFFECTS OF WEATHER ON TRAFFIC FLOW AND MITIGATION USING VISSIM

In order to understand how to improve traffic flow at signalized intersections during inclement weather, the effects of the weather must be fully understood. In order to do this, the VISSIM network that was calibrated in the previous section shall be used to examine the effect of rain on various performance measures. These performance measures should be ones in which differences are readily apparent to both traffic engineers and the general public. Similarly, the measures should be ones that are affected by the differences in saturation flow rate and speed that were observed in the field. After these performance measures are analyzed, new signal timing plans can be developed and tested to determine if the effects of inclement weather can successfully be mitigated.

6.1 Performance Measures

As stated above, the performance measures examined should be ones which are readily apparent to the general public and traffic engineers. The reasons for this are two-fold. First, the general public may notice performance measures in a real intersection if decreased performance affects their travel patterns. For example, if traffic backs up on rainy days such that an entire queue of vehicles at a signal cannot proceed through on one cycle, drivers may notice these effects. Secondly, there are features of traffic flow which drivers may not notice, but that are of interest to traffic planners. For example, if the queue length at an intersection increases significantly, drivers may not notice (so long as they can proceed through the intersection on the next green cycle), however if this queue backs up onto a freeway, or blocks a fire station driveway, or causes a left-turn bay to overflow, then transportation planners might see a need to mitigate these effects.

Within VISSIM, there are numerous options for output reports. These reports can be automatically generated or custom created, and often have the option of returning raw data or aggregating the results. Aggregated results would be of greater interest to the researcher, as the raw data would need to be aggregated anyway before any useful results could be extracted from it.

In the Vermont study (11), the researchers used speed and delay as their performance measures. As speed was one of the varied inputs that allowed VISSIM to simulate both dry and wet conditions, this would not be a valuable performance measure for this project. However delay would likely be a decent measure to examine. The delay at a signalized intersection is the amount of additional time that it takes a vehicle to proceed through the intersection than if the vehicle were able to proceed through at its desired free-flow speed with no impediments (including slowing down or stopping at the signal itself). Obviously if a signal is operating efficiently, the delays would be kept to a minimum. There are two components to delay the total delay, and the stop delay. The total delay is described above, while stop delay is only the portion of the delay in which the vehicle is stopped in a queue. Travel time is related to delay. The travel time is simply the delay plus the free-flow travel time. The travel time is something that will likely be noticed by drivers. Queue length is an important performance measure that should be examined in this project. Intersections, ramps, and most roadway networks are designed with a finite amount of storage capacity. If rain is causing queues to become excessively long, other transportation facilities may be blocked. Queues can also become a safety concern when drivers happen upon them un-expectedly. This would be especially true if the driver’s ability to brake is reduced by wet weather. Therefore, queue lengths shall be examined as well.

Lastly, vehicle volumes proceeding through an intersection should be examined. This measure is tied to an intersection’s capacity. If the intersection normally operates at or near capacity, and its capacity is reduced by inclement weather, then the intersection may fail without the addition of any more vehicles. The vehicle volumes processed through the intersection would therefore indicate if the capacity is being reached during dry or inclement weather.

In conclusion, the performance measures that will be examined include vehicle volumes, travel time, delay, and queue lengths.

6.2 Preparing the VISSIM Model

In order to extract the desired outputs from the VISSIM model, several inputs had to be set properly. The isolated intersection VISSIM model was coded as follows. The first two inputs were the proper signal timing plan and the vehicle input volumes. The signal timing plan was obtained from a signal optimization project assignment in Prof. Brian Park’s fall 2002 Traffic Operations class (19). The vehicle input volumes were obtained by summing the turning movement counts from a Wilbur Smith Associates study (16) done at Emmet Street and Barracks Road in January, 2002. This was the most recent vehicle volume information available in the City of Charlottesville Traffic Engineer’s office in 2004. Since the intersection often operates at a poor level of service during the PM peak hours (LOS E, according to Wilbur Smith), which is worse than what is seen during a majority of the day, the midday peak values were used as the default vehicle volumes in this project. The midday peak operates at LOS D, which is the same as the AM peak hour, however the AM peak hour only sees heavy traffic on the southbound approach, whereas the midday peak has a more even spread of traffic volumes.

6.2.1 Volumes

In the literature review, Lieu and Lin (7) conclude that there are “boundary” vehicle volumes beyond which the effects of rainy weather and the benefits of retiming signals would be minimal. They do stress that these flow rates are specific to the intersection, drivers, and other factors, and therefore their values of 1,100 veh/hr and 1,700 veh/hr should not be assumed to be true at all intersections. Rain is unlikely to have a large effect on the intersection performance when vehicle volumes are very low, as delays caused by other vehicles would be minimized, and there would be much excess green-time on each intersection approach. Similarly, at very high volumes where the entire intersection is already operating at a very poor level of service, rain may only have a small effect on performance. However in between lies a traffic volume where the intersection is likely to be affected the greatest by inclement weather. In the case of Emmet Street and Barracks Road, the boundary values were not known, thus the analysis of the effects of rainy weather was performed over a large range of vehicle volumes in the same manner that Lieu and Lin conducted their research. Ten ranges of vehicle volumes were used. The first set of input volumes was the actual, present midday peak hour volumes from the field. Then the volumes were multiplied by 25, 50, 75, 125, 150, 175, 200, 225, and 250 percent. This assumes that traffic volumes change as a whole, and that one approach will not have proportionately greater traffic growth than another. Since there was a high likelihood that the capacity of each approach is somewhere between 25% and 250% of the present volume, performance measures should be obtained at, below, and above the capacity. Below is Table 14 showing input volumes that were entered into the VISSIM model.

% actual volume	Northbound (veh)	Southbound (veh)	Eastbound (veh)	Westbound (veh)
25%	238	316	204	161
50%	475	631	409	322
75%	713	947	613	482
100%	950	1262	817	643
125%	1188	1578	1021	804
150%	1425	1893	1226	965
175%	1663	2209	1430	1125
200%	1900	2524	1634	1286
225%	2138	2840	1838	1447
250%	2375	3155	2043	1608

Table 14. VISSIM Volume Inputs

The relative flow numbers that were assigned during network calibration were left in place. Since they are relative rates, changing the total input volumes on a given approach will not affect the proportion of vehicles that turn left, turn right, or continue straight through the intersection during the simulation.

6.2.2 Traffic Signal Timing Plan

The traffic signal timing plan for Emmet Street and Barracks Road uses split phasing for Barracks Road, and concurrent phasing with leading protected left turns on Emmet Street. This signal has pedestrian signal heads and pedestrian push-buttons, however there are few pedestrians crossing this busy intersection. Therefore, pedestrians were not taken into account during this analysis. The timing plan, as obtained from Dr. Park, is shown in Table 15. This timing plan was entered into VISSIM using the Java-based NEMA signal editor.

	Phase 1	Phase 2	Phase 3	Phase 4	Phase 5	Phase 6	Phase 7	Phase 8
Max I	35	55	25	40	20	55	25	40
Walk	0	7	0	7	0	7	0	7
Flash DW	0	12	0	17	0	17	0	15
Max Initial	7	10	7	7	7	10	7	7
Min Green	7	10	7	7	7	10	7	7
TBR	10	10	10	10	10	10	10	10
TTR	5	10	5	10	5	10	5	10
Passage	3	3	3	3	3	3	3	3
Min Gap	2	3	2	3	2	3	2	3
Yellow	3.5	3.5	3.5	3.5	3.5	3.5	3.5	3.5
Red Clear	1.5	1.5	1.5	1.5	1.5	1.5	1.5	1.5

Table 15. Emmet Street and Barracks Road Current Timing Plan

6.2.3 Data Collection Points

In order to extract information from the simulation model, data collection points need to be inserted. Three types of data collection points were used in this simulation:

• Travel Time
• Data Collection
• Queue Counter

Travel time data collection points are used in pairs and measure the travel time of all vehicles crossing both data points in the pair. Typically, the “entry” and “exit” travel time data points are placed near the beginning and end of the network, respectively. In total, 24 travel time data points were placed (two for each of 12 possible movements in the intersection). Therefore, travel time data for each of the 12 movements was collected individually during the simulation runs.

Data collection points can collect a variety of different data items from the vehicles as they travel through the simulation network. In a microscopic simulation model, the computer “remembers” items such as the number of stops a vehicle makes, how much time the vehicle is delayed while waiting for a green signal indication, etc. This data is then compiled as the vehicle crosses the data collection point. Additionally, data collection points can serve the same function as a standard loop detector by collecting volume, speed, and occupancy data. Data collection points are placed across individual lanes on the network. Points in adjacent lanes can be “linked” such that the data collected at the two points is aggregated. Data collection points were placed and set to collect vehicle volumes, vehicle delays, stop delays, and the number of stops experienced for each of the 12 movements in the intersection.

Lastly, four queue counters were placed in the intersection. Queue counters count the average and maximum length of the vehicle queue and are typically placed directly behind the stop bar. Therefore, the queue data is collected by approach instead of by movement.

6.2.4 Simulation Length

The simulation length was an important factor that had to be determined. Running a simulation that is too short will not allow for enough signal cycles to be counted during the runs. Therefore, there might be more stochastic variation in the simulation runs than if a greater number of cycles were observed. If the simulation is run for too long, unrealistic results may be seen when modeling peak-hour traffic volumes. By definition, peak hour volumes only last for one hour, and should not be used for a multiple-hour simulation run. It was therefore decided to run the simulations for one hour, or 3,600 seconds.

Unfortunately, a model can give unexpected results if the data collection begins immediately after the simulation starts. When the simulation starts, the first vehicles enter the network from the end points and typically do not encounter any other vehicles impeding their desired speed and flow characteristics. Therefore, the results would not generally be considered realistic. In order to overcome this problem, the network is “primed” by running the simulation, with no data collection taking place, for a short period of time. Since the signal at Emmet Street and Barracks Road has a typical cycle length of approximately 2.5 minutes during peak periods, a 5 minute (2 cycle-lengths) priming period preceded the one hour of data collection. The simulation was run for 3,900 seconds, and the data collection occurred over the time period from 300 seconds into the simulation to the end of the simulation at 3,900 seconds. The simulation was run at the maximum resolution of 10 time steps per simulation second. The collected data was then aggregated over the entire hour and recorded in an output file.

6.2.5 Random Seed and Number of Runs

The input volumes were varied as described earlier in this chapter, however some stochastic variation within each input volume range was desired. In order to do this, the initial random seed number was changed in simulation runs with the same input volumes. The random seed is what generates the randomness in the simulation, and two runs of the simulation with the same random seed will generate exactly the same results. The first run used a random seed of 6, and the seed was incremented by 3 for each successive run. First, 5 runs were performed at each of the 10 input volumes for a total of 50 wet weather runs and 50 dry weather runs. The output was then compiled, and it was determined that some of the results were counter-intuitive (such as a large decrease in delay with an increased traffic volume). In order to obtain more realistic results, more stochastic variation needed to be introduced. Five more runs on each of the input volumes were performed for a grand total of 100 runs per weather condition. The random seed selected merely continued the pattern established above. Therefore, runs were performed using random seeds of 6, 9, 12, 15, 18, 21, 24, 27, 30, and 33.

6.2.6 Seminole Trail Network

The Seminole Trail VISSIM network was coded and had timing plans, vehicle volumes, data collection points, number of runs, and random seeds established in the same manner as the isolated intersection network. The volume and signal timing plan information came from Synchro files provided by Matthew Grimes at VDOT. The runs were performed, however it was observed that the signal at Woodbrook Drive would “freeze” in the all red mode beyond a certain point in the simulation run. The point at which this occurred varied depending on the input volumes and the random seed number, however it occurred in almost all simulation runs. Despite several attempts at altering the settings, this technical issue was unable to be resolved, which meant that accurate results from the three intersection network were unable to be obtained. The remainder of this project therefore focuses exclusively on the isolated intersection and its results.

6.3 Compilation and Discussion of the Effects of Weather on Traffic Flow

Several hundred data files were obtained through these simulations, and the results needed to be compiled in a format that could easily be understood by the readers. VISSIM sends the simulation output to text files, each with a unique name and file extension. The filename denotes if the run was done under wet or dry conditions, which set of input volumes were used, and which random seed was used for the run. The extension denotes the type of data contained in the file.

The data points in each of these files were exported to an Excel spreadsheet. The files contained aggregated data for one hour of simulation length, and consisted of one row of data numbers. When compiled, each column represented a piece of data (for example, travel time for the southbound left turn). The column headings represented a movement number, which corresponds loosely to the NEMA phase numbers assigned to the movement in VISSIM. These numbers were removed and replaced with a three letter code denoting the movement (SBL is southbound left turn, EBT is eastbound through, etc.). The output values for each of the 10 random seed inputs were averaged in order to obtain a single set of results for each possible scenario. It should be noted that the only difference between the 10 runs on any given scenario was the initial random seed. Therefore it was assumed that an average value for that scenario could be obtained by summing the values and dividing by 10. The end result was an Excel spreadsheet with data pieces in columns, and each scenario in rows for both wet and dry conditions. A percent change from dry weather to wet weather for each data value was computed, using the dry weather scenario as the base case. The spreadsheet of output results can be found in the appendix.

The results for each movement were then combined to obtain one set of results for each performance measure. The volumes were added together to obtain a total intersection vehicle volume. A volume weighted average was used to compute the average delay, stops, and travel time for the intersection. A simple average was taken to compute the average queue lengths on the intersection’s approaches. Line graphs were then created showing trends for each item of output data as the input volumes changed. Two lines, one for dry conditions and one for wet conditions, were shown on each graph. These graphs form the basis of the remainder of the discussion. The VISSIM numerical results, including the mean and standard deviation of each performance measure, are shown in Tables 16, 17, 18, and 19.

MEAN - DRY CONDITIONS
Volume	Delay (s)	Stop Del (s)	Stops	Volume (veh)	Travel Time (s)	Avg QL (veh)	Max QL (veh)
25%	13.84	9.32	0.65	904.50	40.94	10.50	81.33
50%	19.09	14.00	0.67	1827.80	46.73	27.18	145.30
75%	26.96	21.75	0.71	2752.40	55.78	57.58	235.20
100%	46.55	39.97	0.85	3657.80	78.20	135.98	482.90
125%	97.82	83.80	1.35	3876.10	133.41	313.13	582.25
150%	107.37	91.49	1.39	3893.10	144.46	355.73	583.78
175%	107.34	91.94	1.35	3924.20	145.05	364.28	582.70
200%	109.15	92.92	1.39	3872.10	146.70	361.78	583.00
225%	106.79	91.37	1.37	3937.00	144.99	354.70	582.98
250%	108.39	92.59	1.37	3900.00	146.42	346.38	583.45

Table 16. Mean Values of Performance Measures for Dry Weather Conditions

ST DEV - DRY CONDITIONS
Volume	Delay (s)	Stop Del (s)	Stops	Volume (veh)	Travel Time (s)	Avg QL (veh)	Max QL (veh)
25%	1.49	1.23	0.05	28.67	1.52	0.58	5.71
50%	1.46	1.25	0.04	39.01	1.47	0.89	12.02
75%	2.12	1.95	0.03	37.26	2.18	3.48	22.16
100%	6.92	5.67	0.07	32.24	6.55	12.76	58.20
125%	10.37	8.57	0.10	61.33	10.52	37.16	1.63
150%	5.92	4.98	0.10	65.18	6.16	28.34	2.05
175%	5.95	5.20	0.09	64.84	6.05	24.87	1.40
200%	5.17	4.51	0.09	30.18	5.09	18.36	2.60
225%	6.04	5.28	0.09	66.88	6.02	22.25	2.57
250%	5.68	4.97	0.07	66.52	5.57	12.22	2.71

Table 17. Standard Deviation of Performance Measures for Dry Weather Conditions

MEAN - WET CONDITIONS
Volume	Delay (s)	Stop Del (s)	Stops	Volume (veh)	Travel Time (s)	Avg QL (veh)	Max QL (veh)
25%	13.72	9.44	0.66	895.70	43.38	11.15	87.88
50%	19.10	14.20	0.68	1827.10	49.40	27.78	147.05
75%	28.11	22.85	0.72	2751.20	59.67	60.60	250.45
100%	60.52	51.87	1.02	3586.50	95.94	181.18	544.83
125%	105.56	90.16	1.41	3711.10	144.29	333.18	582.60
150%	110.39	94.28	1.43	3721.70	150.41	355.58	584.13
175%	110.64	94.34	1.43	3722.10	150.88	356.43	583.38
200%	110.81	94.50	1.44	3722.90	151.37	376.23	583.48
225%	111.84	94.96	1.43	3688.00	152.33	361.23	583.95
250%	112.12	94.89	1.45	3683.30	152.42	379.68	585.70

Table 18. Mean Values of Performance Measures for Wet Weather Conditions

ST DEV - WET CONDITIONS
Volume	Delay (s)	Stop Del (s)	Stops	Volume (veh)	Travel Time (s)	Avg QL (veh)	Max QL (veh)
25%	1.61	1.42	0.05	35.51	1.71	1.34	21.08
50%	1.31	1.11	0.03	37.81	1.30	1.02	10.70
75%	2.31	1.98	0.04	36.16	2.35	3.28	28.62
100%	14.39	10.56	0.18	85.67	14.36	26.67	42.78
125%	8.09	7.01	0.09	57.27	8.42	20.66	1.33
150%	5.88	5.10	0.08	51.28	6.09	24.46	3.59
175%	6.27	5.30	0.10	67.58	6.28	25.39	2.28
200%	6.61	5.59	0.10	51.44	6.29	34.10	3.00
225%	5.52	4.83	0.06	69.94	5.46	32.14	1.64
250%	7.78	6.85	0.10	77.67	7.95	34.96	2.62

Table 19. Standard Deviation of Performance Measures for Wet Weather Conditions

6.3.1 Traffic Volume

The traffic volumes through the simulation network were affected by the driver behavior characteristics, the desired speed settings, and the network input volumes. The network input volumes controlled the demand of the system, and they were varied between the 25 and 250 percent of the present field traffic volumes as described above. As the vehicle input volume increases, the traffic volume passing through the intersection will increase as well until capacity is reached. The car-following model and desired speed profile adjustments affected the supply of capacity in the network. Despite having exactly the same demand, the reductions in desired speed and saturation flow rate caused by wet weather reduced the capacity of the intersection, and in some cases, reduced the volume of vehicles that the intersection processed during the simulation run.

Figure 11 shows the total vehicle volume, on all approaches, through the simulation network vs. the input volume (shown as a percentage of the actual mid-day peak hour field volumes). One can see that when the demand (input volume) is lowest, the differences in the observed volumes are smaller. This tends to make intuitive sense, as there is excess capacity when the demand is low. Despite a slight reduction in capacity during wet weather conditions, the demand is still less than the supply of capacity. As the input volumes approach the capacity of the intersection, a larger reduction in the number of vehicles processed by the intersection during wet conditions is observed. The dry and wet weather lines in the graph begin to spread apart when the input volumes near the capacity, and the area between the two lines represents the impact of wet weather on capacity. Additionally, both lines begin to level off as the intersection reaches its maximum capacity. This capacity reduction is first observed when the simulation input volumes approach the present volumes. In other words, the intersection is operating close to capacity under current conditions. Figure 12 follows Figure 11 and shows the percent difference in vehicle volumes processed by the simulation network under wet and dry conditions at the same input volume.

Under current conditions, the intersection services 3,658 vehicles under dry conditions, and 3,587 vehicles under wet conditions, which is a 1.95% reduction. The capacity of the intersection is approximately 3,900 vehicles per hour during dry conditions, and 3,725 vehicles per hour during wet conditions. This is roughly a 4.5% to 5% reduction in capacity.

Figure 11. Total Observed Vehicle Volume vs. Input Vehicle Volume

Figure 12. Percent Difference Due to Weather in Observed vs. Input Vehicle Volumes

One particularly problematic movement at this intersection is the eastbound right turn. The lane configuration at this intersection is such that the right lane on the eastbound approach serves as the only through and right turn lane, while the other two lanes are exclusive left turn lanes. This movement shows a 10.2% difference in vehicle volumes carried between wet and dry conditions using the current input volumes, and a 15.5% difference at 125% of the present field volumes. The eastbound through and left turn movements also show a significant difference in vehicle volumes during given 125% of the current field volumes (7.3% and 9.3% respectively). Due to these large differences in volumes between wet and dry conditions, the eastbound approach seems to be affected heavily by rainy weather when traffic volumes are close to present volumes. This is likely due to the heavy traffic this approach sees, meaning a small disturbance, such as weather, can have a large effect on the flow of traffic through the intersection. Therefore, when mitigation strategies are examined this approach should be a priority for mitigation in order to lessen the negative effects of rain on traffic volumes processed by the intersection.

6.3.2 Travel Time and Delay

The next set of performance measures are related to the amount of time it takes for a vehicle to travel through the simulation network. Three sets of performance measures in this category were examined: travel time, delay, and stop delay. One would expect that each of these performance measures would show similar trends when vehicle volumes are varied and the weather conditions in the simulation were changed from dry to wet.

Travel time is the amount of time that elapses from when a vehicle enters the simulation network until it leaves the network. When more vehicles are added to the network, the travel time will likely increase since vehicles are “interacting” with one another more. The probability of a vehicle having to adjust its speed to let other vehicles change lanes in front of it or to wait for a gap in oncoming traffic increases with more vehicles in the system. The driver behavior model, which was adjusted to simulate a reduction in the saturation flow rate, can increase travel time to a slight degree by enforcing stricter vehicle spacing requirements. Additionally, decreasing the desired speeds increases the travel time through the network because vehicles are driving slower.

Delay is very closely related to travel time. The delay is the difference between the actual travel time for a particular vehicle and the travel time if the vehicle could drive through the network at its desired free-flow speed completely unimpeded by other vehicles or traffic control devices. Thus, two vehicles that have the same travel time may have a different delay if their desired free-flow speeds differ. The reason why travel time and delay were examined separately is because the desired speed profiles were different for wet and dry conditions. Therefore the “delay” experienced by a vehicle due to its reduced desired speed in inclement weather is not accounted for in the delay output, however it would appear as an increase in travel time. The delay output does account for the increases in travel time due to the congestion, traffic control devices, and car-following during wet weather, however.

Stop delay is the amount of time a vehicle spends completely stopped while within the simulation network, which usually occurs while waiting in a queue at a traffic signal or other control device. During light traffic conditions, stop delay is minimal, and only those vehicles that encounter a stop sign, yield sign, or red signal will experience this. Other vehicles will not experience any stop delay. During heavier traffic conditions, stop delay can increase dramatically as queue lengths increase and vehicles may have to sit through multiple cycles of a traffic signal before proceeding through the intersection. Therefore the stop delay is affected mostly by the traffic volumes, though stop delays may also increase due to additional network congestion caused by inclement weather.

Shown below are Figures 13 and 14 representing the average travel time under wet and dry conditions and the percent change in average travel time from dry to wet conditions respectively. Similar to the vehicle volume graphs, the area between the two lines on the average travel time graph represents the cumulative effect of wet weather on the travel time. The two lines never overlap one another, which is because of the increase in travel time due to the reduction in desired speed. The travel time increases dramatically between input volumes of 75% and 125% of the present field volumes. This makes sense, as it was determined earlier that the intersection was approaching its capacity with an input volume close to the actual traffic volume. The largest increase in travel time occurs when the input volumes are the same as the present field traffic volumes. The increase in average travel time due to weather is 17.73 seconds. At lower traffic volumes, the difference in travel time is in the range of 2 to 4 seconds. One will note that the travel time lines level off beyond an input volume of 125% of the present volume. Since the travel time is indicative of the time it takes a vehicle to travel the length of the network, if the intersection is inundated with traffic to the point where the queue backs up beyond the bounds of the network, any delay incurred beyond the bounds of the network will not be accounted for, and therefore will not be shown as an increase in travel time.

Figure 13. Average Travel Time vs. Input Vehicle Volume

Figure 14. Increase in Travel Time Due to Weather vs. Input Volume

Certain turning movements gave interesting travel time results. Most of them showed a “spike” in the difference in travel times at a particular input volume, very similar to the spike in the average travel time graph at the present field traffic volume. The travel time difference for the northbound left, through, and right peaked with an input volume of 125% of the present field volume. The time difference was between 17 and 19 seconds depending on the movement. This indicates that the northbound approaches in the field are still below capacity, and therefore wet weather will only increase travel times by approximately 3 to 6 seconds. However if the traffic volumes in the field were to increase, the effects of rain may become much more noticeable. Other problematic movements are the eastbound through and right turns, with a travel time increase of approximately 55 seconds when the input volume is the same as the present traffic volume, and the southbound left turn, with a travel time increase of approximately 45 seconds.

The trend in the average delay is similar to that observed in the travel time. The average delay starts out small, with little difference between dry and wet conditions. As the simulation network traffic volumes increase to 75% of the present traffic volume, the differences between dry and wet conditions start to become more apparent. The majority of the increase in average delay occurs when the network volumes are between 75% and 125% of the present field volumes. The largest difference in average delay occurs at the present field traffic volume, a difference of approximately 14 seconds, which is a 30% increase. Figures 15 and 16 show the average delay and percent difference due to weather in the average delay.

Figure 15. Average Delay vs. Input Vehicle Volume

Figure 16. Percent Change Due to Weather in Average Delay vs. Input Volume

The problematic individual approaches to the intersection are the same ones as for travel time. The eastbound through movement sees a 59% increase in delay (from 90 to 143 seconds), the eastbound right turn movement sees a 69% increase in delays (from 79 to 133 seconds), and the southbound left turn sees a 34% increase in delay (from 123 to 166 seconds) when input volumes set to the present field traffic volumes.

The stop delay should be very similar to the average delay. Since a majority of the delay time is spent stopped, it is no surprise that the trends for stop delay are nearly identical to the average delay ones. The maximum increase in stop delay is from 40 seconds to 52 seconds when the network volumes are the same as the field volumes. Likewise, for individual approaches, the eastbound right and through movements, as well as the southbound left turn, are the problematic ones. The eastbound through and right turn see a 58% and 71% increase in stop delay respectively, and the southbound left turn sees a 32% increase. Figures 17 and 18 are the graphs for the average stop delay under wet and dry conditions, as well as the percent increase in average stop delay from dry to wet conditions.

Figure 17. Average Stop Delay vs. Input Vehicle Volume

Figure 18. Percent Change Due to Weather in Average Stop Delay vs. Input Volume

6.3.3 Stops

The number of stops a vehicle makes during its trip through the simulation network is simply the number of times the vehicle comes to a complete stop, and therefore encounters stop delay. This is closely tied to the amount of congestion in the network. Typically with little network congestion, the average stop delay is between zero and one. This is representative of some vehicles traveling through the network non-stop, while others must stop at a traffic control device. When a greater proportion of vehicles make one stop at a traffic control device, the average number of stops increases. The number of stops increases significantly when traffic congestion increases, as some vehicles may make multiple stops if the queues at traffic control devices are long.

In the simulation network, the number of stops was nearly equal during dry and wet conditions at low input volumes. When the input volumes were 25% of the actual field traffic volumes, the number of stops increased from 0.65 to 0.66 during wet weather, which is approximately a 1% increase. When the input volumes grow from 75% to 125% of the present field volumes, the number of stops increase significantly as well. This is expected, as the intersection is nearing its capacity at those input volumes. Similar to the other performance measures, the number of stops increases faster under wet conditions than under dry conditions, creating a large difference in the number of stops observed at present field traffic volume. The increase is from 0.85 stops per vehicle to 1.02 stops per vehicle, which represents an increase of approximately 20%. Figures 19 and 20 show the average number of stops and the percent increase in number of stops from dry to wet conditions, respectively.

Figure 19. Average Number of Stops vs. Vehicle Input Volume

Figure 20. Percent Change Due to Weather in Number of Stops vs. Input Volume

Similar to the other performance measures, the problematic individual turning movements are the southbound left turn, and the eastbound through and right turns. The southbound right turn experiences a very small number of stops, due to the presence of a yield sign on the channellized right turn lane. Therefore, vehicles need not stop unless there is conflicting traffic entering from another approach. The southbound right turn sees a negligible increase in stops during wet weather from 0.11 to 0.19 stops per vehicle at the present traffic volume. In contrast, at the same input volume, the southbound left turn sees an increase from 1.46 to 1.91 stops per vehicle, which is an increase of 31%. The eastbound through movement sees an increase from 1.21 to 1.74 stops per vehicle, which is a 43% change. Note that in both cases, the number of stops per vehicle in dry weather is already greater than one, indicating that some vehicles already must wait through more than one signal cycle in order to proceed through the intersection. The wet weather simply makes an existing problem worse. The southbound left turn sees values as high as 3 stops per vehicle when the input volumes are increased to 125% of the present field traffic volumes, which is certainly indicative of a severe congestion problem on this approach.

6.3.4 Queue Length

The queue length performance measure simply counts the number of vehicles queued up behind the stop bar on a particular approach to the intersection. This includes all vehicles, regardless of their ultimate destination. Some queuing is to be expected at a signalized intersection, however congestion or bad signal timing can cause queuing to reach excessive levels. When this occurs, delays, the number of stops, and the travel time can increase significantly. However queue lengths often affect other road users as well. If a queue backs up sufficiently far, vehicles at other intersections may be blocked. This can cause delays to vehicles far away from the problematic intersection in a roadway network. Two queuing performance measures were examined. The average queue length is the average number of vehicles stopped behind the stop bar during the course of the simulation run. The maximum queue length is the maximum number of vehicles queued up at any point during the simulation run.

Figures 21 and 22 represent the average of the average queue length on each approach. As with the other performance measures, the queue lengths are similar between dry and wet conditions at lower traffic volumes, and they begin to differ more as volumes increase. The largest increase in average queue length is when the traffic volume increases from 75% of the present field volume to 125% of the present volume. The largest difference between the wet and dry weather average queue lengths occurs at the present field traffic volume. The dry weather average queue length is 136 vehicles, and the wet weather average length is 181 vehicles. This represents an increase of over 33%.

Figure 21. Average Queue Length vs. Input Volumes

Figure 22. Percent Change Due to Weather in Average Queue Length

It should be noted that the reason for this large increase in average queue length, as well as the reason for a high average number of vehicles queued on each approach, is that the southbound and eastbound approaches are overloaded with traffic. The eastbound approach sees a very large increase in average queue length under both wet and dry conditions when the input volume changes from 75% of the field traffic volume to 100% of the field traffic volume. At 75% of the field traffic volume, the eastbound approach has 62 vehicles queued under dry conditions, and 65 vehicles queued under wet conditions. When this input volume is increased to match the field traffic volume, the queued vehicles under dry conditions increase to 193, and under wet conditions to 307. Under the larger input volume, a 59% increase in average queue length due to wet weather is shown.

The maximum queue length shows a similar trend to the average queue length. The maximum queue lengths for each approach were averaged together to obtain the graphs below. One will note that the maximum queue length levels off at approximately 585 vehicles. This is the maximum number of vehicles that can be queued up in the simulation network. Beyond that point, the queue extends outside the bounds of the network, and the actual maximum queue length cannot be reported in the simulation output. In real-world conditions, the maximum queue length would be significantly higher given large traffic volumes. The greatest difference between the maximum queue lengths under wet and dry conditions occurs when the input volumes are equal to the field traffic volumes. The maximum queue length increases from 482 to 545 vehicles due to wet weather, which is an increase of 13%. Figures 23 and 24 show these differences.

Figure 23. Maximum Queue Length vs. Input Volumes

Figure 24. Percent Change Due to Weather in Maximum Queue Length vs. Input Volume

6.3.5 Summary of Findings

From the analysis of these performance measures, one can see the effects of increasing the traffic volume at an intersection, as well as how wet weather affects those trends. As traffic volumes increased, the travel time increased, the delay and stop delay increased, the number of stops increased, and the average and maximum queue lengths increased. Under the same input volumes but different weather conditions, wet weather had a similar effect on the above performance measures. Additionally, the vehicle volumes carried through the network were reduced under wet conditions, indicating a reduction in intersection capacity. The largest overall deterioration in performance measures under both weather conditions was typically seen when the traffic volumes increased from 75% of the field volume to 125% of the actual field volume, and the largest percent difference in performance measures between wet and dry conditions was seen at 100% of the field volume in most cases. This tends to indicate that the intersection of Emmet Street and Barracks Road is operating very close to its capacity. Additionally, the effects of weather on performance measures seem to be greatest when the intersection is operating very close to capacity. The specific turning movements that are most problematic at this intersection are the southbound left turn, and the eastbound through and right turning movements. Some performance measures indicate that these movements are operating above capacity under current conditions. If similar research is conducted in the future, it may be desirable to include more data points at volumes close to the intersection’s capacity (in this case, between 75% and 125% of the original field volume) in order to better understand how weather affects the parameters.

6.4 Mitigating the Effects of Rain Through Signal Timing

The last step in this project was to attempt to mitigate the effects of rain by creating new signal timing plans. The signal plan that gives optimal results during dry weather may be sub-optimal during wet weather. If this is the case, then creating new signal timing plans specifically for wet weather conditions may reduce or eliminate the weather related impacts. If it is not possible to completely reduce the impacts of weather at an intersection, wet weather timing plans could be developed to mitigate specific adverse impacts.

When the effects of rain were discussed above, the dry weather condition and wet weather condition were both compared using the existing signal timing plan. If traffic engineers were to take the time to develop optimal wet-weather timing plans for a traffic signal, surely they would want to develop an optimal dry-weather timing plan as well. Therefore, when comparing results, the base case scenario should be the dry weather optimal timing plan under dry weather conditions. The un-mitigated effects of rain would be the dry optimal timing plan under wet weather conditions. The improvement due to mitigation would be the percent difference in performance measures between the dry optimal plan and the wet optimal plan, both under wet weather conditions.

The same performance measures that were used to analyze the effects of wet weather were used to analyze the mitigation. Instead of using line graphs comparing the same timing plan over various volumes, bar charts were used to compare several timing plans at the same traffic input volume. The mitigation was performed using the actual field traffic volumes, as the effects of rain seemed greatest given those input volumes.

6.4.1 Development of Optimal Timing Plans

In order to assess the benefits of mitigation, new signal timing plans had to be developed. VISSIM does not have the capability to optimize signal timing plans, therefore another program had to be used. Synchro is a program that is often used by traffic engineers to develop optimized signal timing plans. It is a macroscopic simulation model. Therefore, data from individual vehicles cannot be recorded, though the system as a whole can be analyzed. A simple network with the intersection of Emmet Street and Barracks Road was created in Synchro, the lane layouts were input, and the signal was added. The current timing plan was then manually input.

Two scenarios were created in Synchro: a dry weather and wet weather scenario. This was done because weather may impact the optimal timing plan. Synchro normally computes the saturation flow rate automatically, however in order to optimize for wet and dry conditions, the saturation flow rates were manually over-ridden with new flow-rates calculated using reduction factors. On Emmet Street, the Synchro saturation flow rate was listed as 3,505 vehicles per hour. Since this accounts for two lanes, dividing this number by 2 gives a rate of 1,753 vehicles per hour per lane. The saturation flow rates used to calibrate VISSIM were 1,773 vehicles per hour per lane in dry conditions, and 1,690 vehicles per hour per lane in wet conditions, which yields a conversion co-efficient of 1.011 for dry weather and 0.964 for wet weather. The default saturation flow rates for each lane group were then multiplied by the appropriate coefficient in order to obtain the flow rates to be used during signal optimization in Synchro. Additionally, the link speeds were reduced by 10 percent during wet weather to reflect field observations. Emmet Street was reduced from 40 MPH to 36 MPH, and Barracks Road was reduced from 25 MPH to 22.5 MPH.

Creating optimized timing plans in Synchro involves optimizing the splits and the cycle lengths. The dry weather optimal timing plan was created first. Then the following wet weather timing plans were created using Synchro:

1. Optimizing the cycle length only
2. Optimizing splits only
3. The wet weather optimal timing plan (optimized splits and cycle lengths)
4. The wet weather optimal plan, with some extra green time on the eastbound approach (obtained by extending the cycle length)
5. The wet weather plan with optimized splits and a 120 second cycle length
6. The wet weather plan with optimized splits and a 200 second cycle length
7. Wet weather optimal plan with additional southbound green time (obtained by adjusting the splits)
8. Wet weather optimal plan with additional southbound green time (obtained by extending the cycle length)
9. Extra green time for both the southbound and eastbound approaches
10. The wet weather plan with optimized splits and a 60 second cycle length
11. Extra green time for the eastbound approach (obtained by adjusting the splits)

The logic in creating these plans was to first try to optimize the timing. Then the cycle lengths were adjusted to attempt to see the effects of cycle lengths. Lastly, additional green time was given to either the southbound or eastbound approach, or both, as these two approaches were the problematic ones where the effects of wet weather were felt the most.

Table 20 shows the splits and cycle lengths obtained from Synchro for each of the above referenced timing plans. The timing plan index numbers are used to refer to each of these timing plans throughout the remainder of this section.

Timing Plan Index		Splits (seconds)
	Timing Plan	SB Left	NB Through	NB Left	SB Through	EB	WB	Cycle Length
1	Optimal C/L	17	27	11	33	18	13	75
2	Optimal Split	35	58	16	77	38	29	160
3	Optimal	17	27	11	33	18	13	75
4	Extra green EB (C/L)	17	27	11	33	23	13	80
5	optimal - 120s C/L	26	45	13	58	28	21	120
6	optimal - 200s C/L	43	73	19	97	47	37	200
7	Extra green SB (splits)	20	24	13	31	18	13	75
8	Extra green SB (C/L)	20	29	14	35	18	13	80
9	Extra green SB & EB	20	31	12	39	20	14	85
10	optimal - 60s (C/L)	13	22	9	26	14	11	60
11	Extra green EB (splits)	16	23	10	29	23	13	75
	Optimal Dry weather	18	29	11	36	19	14	80

Table 20. Timing Plans Obtained From Synchro

After the timing plans were obtained, they were coded into VISSIM using the NEMA editor. The optimal dry weather plan was run under both dry and wet conditions, while each of the other above plans was run under wet conditions only. Each scenario was run 10 times for 3,900 seconds. Data was collected and aggregated from time = 300 seconds to the end of the simulation run, much like when determining the effects of rain in the previous segment. The results were then tabulated in the same way that the effects of rain were tabulated. When examining results, the following comparisons were made. First, the differences between the original timing plan and the dry optimal timing plan were examined for both wet and dry conditions to see what effect optimizing the signal had on intersection performance. Next, the effect of wet weather was compared for both of those timing plans. The effect of weather was determined to be the percent difference between wet and dry conditions using the same signal timing plan. Lastly, each wet weather timing plan was compared to the dry weather optimal timing plan under wet conditions. The mean and standard deviation of the VISSIM numerical results are shown in Tables 21 and 22. A discussion of those results follows.

MEAN
Timing Plan	Delay (s)	Stop Del (s)	Stops	Volume (veh)	Travel Time (s)	Avg QL (veh)	Max QL (veh)
dry optimal/dry weather	38.17	25.66	0.90	3673.20	63.08	108.80	358.60
1	49.71	33.44	1.15	3644.80	77.57	154.10	410.80
2	57.42	45.27	0.85	3646.40	85.57	164.40	500.70
3	49.71	33.44	1.15	3644.80	77.57	154.10	410.80
4	43.61	28.97	1.00	3650.30	71.50	116.60	430.50
5	45.70	33.80	0.85	3672.70	73.65	128.90	407.50
6	67.48	54.79	0.85	3606.00	95.71	186.10	541.60
7	43.53	28.74	1.04	3667.50	71.36	132.00	411.40
8	50.04	34.25	1.12	3651.90	77.87	153.20	401.30
9	44.41	30.39	0.98	3665.70	72.24	136.10	407.40
10	66.65	42.68	1.67	3523.10	95.14	199.00	456.80
11	56.23	35.62	1.30	3611.40	84.10	169.00	448.10
Dry optimal/wet weather	43.67	29.60	0.99	3659.60	71.43	124.80	403.00

Table 21. Mean Values of Performance Measures with Optimal and Special Timing Plans

STD DEV
Timing Plan	Delay (s)	Stop Del (s)	Stops	Volume (veh)	Travel Time (s)	Avg QL (veh)	Max QL (veh)
dry optimal/dry weather	5.87	4.32	0.10	27.71	6.12	21.55	36.40
1	12.22	8.76	0.24	37.50	12.21	28.11	51.05
2	4.80	4.14	0.04	27.85	4.86	16.96	47.47
3	12.22	8.76	0.24	37.50	12.21	28.11	51.05
4	6.91	4.75	0.12	23.73	7.09	11.82	33.21
5	5.16	4.02	0.06	31.62	5.32	20.25	51.55
6	5.32	5.01	0.04	15.60	5.32	11.79	29.31
7	6.81	4.76	0.13	24.42	6.81	18.48	28.00
8	11.05	8.06	0.20	31.18	11.05	23.92	29.97
9	8.66	5.92	0.14	23.38	8.76	23.26	43.80
10	20.12	13.31	0.45	85.55	18.71	41.53	53.81
11	9.55	5.74	0.19	33.26	9.55	23.90	53.73
Dry optimal/wet weather	8.76	6.08	0.15	22.19	8.80	22.14	58.31

Table 22. Standard Deviations of Performance Measures with Optimal and Special Timing Plans

6.4.2 Traffic Volume

Optimizing the signal timing plan led to an improvement in overall intersection capacity. During dry conditions, the total volume was increased from 3,658 vehicles to 3,673 vehicles using the dry weather optimal timing plan. During wet conditions, the total volume increased from 3,587 vehicles to 3,660 vehicles, also using the dry weather optimal timing plan. In addition to an overall improvement in travel time, the negative effect of weather on capacity is reduced from a 1.95 percent capacity decrease to a 0.37 percent decrease.

With two exceptions, the special wet weather timing plans did not show any increase in the overall capacity of the system as compared to wet conditions with the dry weather optimal timing plan. Any increase was limited to a maximum of 0.36 percent, or about 12 vehicles per hour. Many of the plans led to a decrease in capacity from the dry optimal timing plan. In most cases, adjusting the signal timing merely shifted capacity from one approach to another. This particularly applies to the timing plans where extra green time was assigned to one approach. This seems to imply that adjusting the timing plan will not have a net benefit from the perspective of the system, however the benefits to individual approaches can often be seen at the expense of other approaches. See Figure 25 and Table 23 for the total vehicle volumes for the network. In the figure, “dry orig” and “wet orig” refer to the original timing plan under dry and wet conditions. “Dry opt” and “wet opt” refer to the dry-weather optimal timing plan under dry and wet conditions. The numbered timing plans are the wet-weather optimized plans from Table 15.

Figure 25. Total Volume vs. Signal Timing Plan

Timing Plan Index	Total Volume	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	3644.8	-0.40	decrease in EB & SBL capacity
2	3646.4	-0.36
3	3644.8	-0.40
4	3650.3	-0.25		Increase in EB capacity
5	3672.7	0.36		Increase in net capacity
6	3606	-1.46		Increase in EBT capacity
7	3667.5	0.22		Increase in net capacity
8	3651.9	-0.21
9	3665.7	0.17
10	3523.1	-3.73	3% drop in capacity
11	3611.4	-1.32	10% reduction in SBL and EBL capacity
Wet weather with Dry Optimal	3659.6	0.00		Increase in EB capacity

Table 23. Volume Percent Change, Benefits, and Costs

6.4.3 Travel Time and Delay

Optimizing the signal timing plan led to an improvement in travel time. During dry conditions, the average travel time was reduced from 78.2 seconds to 63.08 seconds using the dry weather optimal timing plan. During wet conditions, the travel time dropped from 95.94 seconds to 71.43 seconds, also using the dry weather optimal timing plan. In addition to an overall improvement in travel time, the effect of weather on travel time is reduced from a 22.7 percent travel time increase to a 13.3 percent increase with optimization.

With one exception, the wet weather optimized timing plans did not reduce the average travel time over the network as compared to the dry weather optimized plan under wet conditions. This one exception was only a reduction of 0.11 percent. Many of the timing plans had travel times that were similar to the travel times seen under wet conditions with the dry optimal timing plan. Some of the plans, such as plan number 6 and plan number 10 had significant increases of over 20 seconds per vehicle. Similar to volume, increasing the green time has the effect of reducing the travel time on one approach at the expense of the others. Though adding extra green time to the eastbound approach, as in timing plan 4, had little effect on the system as a whole. Figure 26 and Table 24 below show the results.

Figure 26. Average Travel Time vs. Signal Timing Plans

Timing Plan Index	Avg Trav Time	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	77.57	8.59	increase in EB trav times
2	85.57	19.78	increase in system trav times	decrease in EB trav times
3	77.57	8.59	increase in EB trav times
4	71.50	0.09	increase in SBL trav time	decrease in EB trav times
5	73.65	3.10	increase in NB and WB trav time	decrease in EB trav times
6	95.71	33.98	large increase in system trav times
7	71.36	-0.11	increase in NB travel times	minor travel time improvement
8	77.87	9.01	increase in EB trav times
9	72.24	1.12		increase in WB trav times
10	95.14	33.18		large increase in EB trav times
11	84.10	17.73		very large NB increase
Wet weather with Dry Optimal	71.43	0.00

Table 24. Travel Time Percent Change, Benefits, and Costs

A significant improvement in delay was seen by optimizing the signal timing plan. During dry conditions, the average delay was reduced to 38.17 from 52.36 seconds per vehicle. During wet conditions, the delay was reduced to 43.67 from 67.83 seconds using the dry-optimal timing plan. The effects of wet weather on the delay were cut almost in half with the optimized signal timing plan, from a 29.55 percent increase to a 14.7 percent increase in travel times due to wet weather.

Two timing plans gave a reduction in overall system delay, which are the same two timing plans which gave either a reduction or very small increase in travel time. These plans gave extra eastbound green time at the expense of either the northbound or southbound left turn approach. Plans 6 and 10, which gave very large increases in travel times, also gave large increases in delays. Giving extra green to the southbound approach had the effect of increasing delays significantly for eastbound traffic. Increasing green time on both the southbound and eastbound approaches had little overall effect on the entire system, however it decreased southbound delays at the expense of westbound delays. Figure 27 and Table 25 show the results below.

Figure 27. Average Delay vs. Signal Timing Plan

Timing Plan Index	Average Delay	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	49.71	13.83	increase EB delay	reduce NB delay
2	57.42	31.50	significant delay increase	reduce EB delay
3	49.71	13.83	significant EB delay increase
4	43.61	-0.12	significant SBL increase in delay	significant EB reduction in delay
5	45.70	4.65	significant WB increase in delay	minor EB reduction in delay
6	67.48	54.52	significant delay increase
7	43.53	-0.30	increase NB delay	minor improvement in delay
8	50.04	14.59	significant EB delay increase
9	44.41	1.69	increase in WB delay	reduction in SBL delay
10	66.65	52.63	significant delay increase
11	56.23	28.77	significant increase in NB delay	reduction in EB delay
Wet weather with Dry Optimal	43.67	0.00

Table 25. Average Delay Percent Change, Benefits, and Costs

The stop delays were improved greatly by optimizing the signal. During dry conditions, the average stop delay was reduced to 25.66 from 39.97 seconds per vehicle. During wet conditions, the delay was reduced to 29.60 from 51.87 seconds using the dry-optimal timing plan. The effects of wet weather on the delay were cut almost in half with the optimized signal timing plan, from a 29.79 percent increase to a 15.37 percent increase.

The wet weather timing plans 4 and 7 reduced the stop delay by almost 2 percent, though the latter came with an increase in the northbound stop delay. Plans 2, 6, and 10 had significant increases in stop delays, with the increase being felt most by the eastbound approach in plan 10. Plan 9 shifted some stop delay from the westbound to the southbound approach, with little overall added stop delay to the system. The results are shown in Figure 28 and Table 26.

Figure 28. Average Stop Delay vs. Signal Timing Plan

Timing Plan Index	Average Stop Delay	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	33.44	12.98	significant increase in EB stop delay	reduction in NB stop delay
2	45.27	52.95	significant increase in stop delay
3	33.44	12.98	significant EB increase	SB decrease in stop delay
4	28.97	-2.12		minor improvement in stop delay
5	33.80	14.20	WB and SB increase in stop delay	EB decrease in stop delay
6	54.79	85.10	significant increase in stop delay
7	28.74	-2.89	increase in NB stop delay	minor improvement in stop delay
8	34.25	15.73	increase in WB stop delay
9	30.39	2.69	WB increase in stop delay	SB decrease in stop delay
10	42.68	44.21	significant increase in EB stop delay
11	35.62	20.35	increase in NB stop delay	decrease in EB stop delay
Wet weather with Dry Optimal	29.60	0.00

Table 26. Stop Delay Percent Change, Benefits, and Costs

Overall, the effects of the special wet weather timing plans were mirrored between delay, stop delay, and travel time, as those three performance measures are closely related to one another. In all three cases, minimal overall system benefits are seen with a couple of the special wet-weather timing plans, however in most cases, an increase in performance on one approach comes at the expense of another approach. These performance measures are the most likely to be noticed by roadway users.

6.4.4 Number of Stops

During dry conditions, the stops were increased to 0.895 from 0.853 stops per vehicle by optimizing the signal. During wet conditions, the number of stops was reduced to 0.990 from 1.024 stops per vehicle using the dry-weather optimal timing plan. The effects of wet weather on the number of stops using the optimized timing plan was cut almost in half with the optimized signal timing plan, from a 20.9 percent increase to a 10.6 percent change.

The number of stops is a difficult performance measure to quantify in terms of percent, as the number is less than one. A minor increase in the number of stops might only represent less than one hundredth of a stop per vehicle. With this performance measure, the number of stops on certain timing plans was actually reduced below that of the number of stops on the original timing plan. This was true in timing plans 2, 5, and 6. The number of stops is related closely to the cycle length of the traffic signal, so therefore the number of stops increased significantly in timing plan number 10, the one with the short 60 second cycle length. Figure 29 and Table 27 show the results below.

Figure 29. Stops per Vehicle vs. Signal Timing Plan

Timing Plan Index	Average Number of Stops	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	1.15	15.70	increase in overall stops
2	0.85	-14.61	increase in NBL stops	decrease in overall number of stops
3	1.15	15.70	increase in overall stops
4	1.00	1.30		decrease in EB number of stops
5	0.85	-14.13		decrease in overall number of stops
6	0.85	-14.18	increase in SB stops	decrease in overall number of stops
7	1.04	5.32
8	1.12	13.29	increase in overall stops
9	0.98	-0.77	WB increase in number of stops
10	1.67	68.65	significant increase in overall stops	reduction in EB number of stops
11	1.30	31.18	significant increase in overall stops
Wet weather with Dry Optimal	0.99	0.00

Table 27. Number of Stops Percent Change, Benefits, and Costs

6.4.5 Queue Length

The queue lengths showed some significant changes as a result of changing the signal timing plans. When optimizing from the original timing plan to the dry optimal plan, the average queue lengths decrease from 136 vehicles to 108.8 vehicles under dry conditions, and from 181.2 to 124.8 vehicles under wet conditions. The effect of rain is reduced to a 14.7 percent increase from a 33.2 percent increase when the signal is optimized.

Several of the timing plans show a significant increase in the average queue lengths, including timing plans 1, 3, 6, 8, 10, and 11. This increase in queue length is likely either the result of reduced efficiency of the intersection, or longer cycle lengths (which may not necessarily indicate reductions in other performance measures). Timing plan number 4 showed a reduction in overall system-wide queue lengths, however upon closer inspection of the individual numbers, the eastbound approach showed a very large decrease in queue lengths, while all of the other approaches showed small increases. The same applies to timing plan number 11 as well. Timing plans 5 and 7 show slight system-wide queue length increases, however small decreases on the southbound and eastbound approaches are seen with these timing plans. The results are shown in Figure 30 and Table 28.

Figure 30. Average Queue Length vs. Signal Timing Plan

Timing Plan Index	Average Queue Length	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	154.10	23.48	significant increase in queue lengths	reduction in NB and SB queue lengths
2	164.40	31.73	significant increase in queue lengths	decrease in EB queue lengths
3	154.10	23.48
4	116.63	-6.55	increase on all approaches except EB	decrease in overall queue lengths, especially EB
5	128.93	3.31		decrease in EB and SB queue lengths
6	186.10	49.12	significant increase in queue lengths	slight EB reduction in queue lengths
7	132.03	5.79		reduction in SB queue lengths
8	153.18	22.74	significant increase in queue lengths	reduction in NB and SB queue lengths
9	136.05	9.01
10	198.98	59.44	significant increase in queue lengths	decrease in NB queue lengths
11	168.95	35.38	significant increase in queue lengths	significant EB reduction in queue lengths
Wet weather with Dry Optimal	124.8	0

Table 28. Average Queue Length Percent Change, Benefits, and Costs

The maximum queue lengths showed some fairly large changes as a result of re-timing the signal for optimal dry conditions. In dry conditions, the maximum queue length (average over all approaches) decreases from 482.9 to 358.6 vehicles. In wet conditions, the maximum queue length decreases from 544.8 to 403 vehicles. The effects of wet weather on the maximum queue length stay the same, at an approximately 12 percent increase on both timing plans.

One of the wet weather timing plans reduces the maximum queue length. Plan 8 shows an overall decrease in maximum queue lengths, but it also has a significant increase in the southbound queue length. Most of the plans which have only a slight overall increase in queue length also do not have any large increases on any of the approaches. The results are shown in Figure 31 and Table 29.

Figure 31. Maximum Queue Length vs. Signal Timing Plan

Timing Plan Index	Maximum Queue Length	% Change from Dry Optimal Plan & wet conditions	Cost	Benefits
1	410.78	1.94
2	500.65	24.24	significant increase NB and SB
3	410.78	1.94
4	430.53	6.84	significant increase SB	significant decrease EB
5	407.53	1.13	increase NB	decrease SB and EB
6	541.58	34.39	significant increase
7	411.43	2.10		significant decrease SB
8	401.30	-0.42	significant increase SB	decrease in maximum queue length
9	407.35	1.09		large decrease SB
10	456.83	13.36	significant increase	NB decrease
11	448.08	11.19	significant increase	significant decrease EB
Wet weather with Dry Optimal	402.98	0.00

Table 29. Maximum Queue Length Percent Change, Benefits, and Costs

6.4.6 Findings

When developing new timing plans in Synchro, many of the plans were similar to each other. In future research, it may be desirable to calibrate additional parameters in Synchro in order to better reflect changes due to weather conditions. Such parameters could include the lane utilization factors or the headway factors. Additionally, validation should occur to ensure that Synchro accurately reflects the conditions seen in VISSIM during wet weather. This could include comparing the delay values given by Synchro with the delay value calculated from VISSIM output results.

Under both wet and dry conditions, optimizing the signal showed significant improvement in most performance measures compared to the original timing plan under the same weather conditions. In many cases, the percent difference in the performance measures between dry and wet conditions decreased under the dry-weather optimized timing plan. It would therefore appear that there are benefits to updating the signal timing plan during both dry and wet weather conditions. It is likely that the current timing plan at Emmet Street and Barracks Road was the optimal timing plan sometime in the past, however growth in traffic can render old timing plans obsolete.

While optimizing the signal improved conditions as compared to the original timing plan, deterioration in performance measure results was still seen when comparing dry weather to wet weather, even with the optimized plan in effect. When special wet weather timing plans were created and implemented, little if any improvement was noted in the performance measures examined when compared to the dry weather optimized plan under wet conditions. Some timing plans showed improvements in performance measures on one or two approaches, while showing a significant worsening of performance measures on the remaining approaches. Therefore it seems unlikely that there would be significant system-wide benefits to special wet-weather signal timing plans at isolated intersections. However specific negative effects of wet weather at isolated intersections, such as increased queue lengths on one particular approach, can be mitigated at the expense of performance on other approaches. Additionally, there may be benefits to developing special wet-weather signal timing plans on arterial networks with multiple intersections. When dealing with multiple intersections, offsets can be adjusted in addition to splits and cycle lengths, which may yield more promising results in future research projects.

6.5 Summary

From the steps performed in this chapter, one can see that there are noticeable effects of rain on traffic flow at an isolated signalized intersection. The performance measures examined included travel time, delay, stop delay, vehicle volumes, number of stops, average queue length, and the maximum queue length. Each showed some deterioration when the saturation flow rate and desired speeds were adjusted to simulate wet weather conditions. At high and low traffic volumes, the effects of wet weather on the performance measures were less than when the traffic volume was near the intersection’s capacity. Therefore, when the intersection is near capacity, the effects will be most observable.

When the signal was optimized, improvements in performance measures were seen. The percent difference in the performance measures between dry and wet weather changed favorably with the new optimized timing plans, meaning wet weather caused less deterioration in traffic flow. The special wet weather timing plans examined in this project did not improve performance measures at the isolated intersection as a whole. Therefore, special wet weather timing plans at an isolated intersection would not likely be a cost-effective solution to mitigating the effects of wet weather, unless one specific negative effect is mitigated at the expense of performance on other approaches. However these special wet-weather timing plans may be an effective mitigation strategy for arterial networks where offsets can be altered in addition to the cycle length and splits.

CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS

The final results from this project are summarized below, and recommendations for future research presented.

7.1 Conclusions

From the literature review:

• The effect of weather on traffic operations is a topic that is gaining interest among researchers and transportation officials
• Weather can impact traffic flow from both the supply-side and the demand-side
• The supply of capacity is affected by changes in physical conditions and driver behavior characteristics

From the data collection:

• Collecting data during inclement weather is difficult due to the adverse conditions presented
• Data quality and concerns with keeping equipment dry and safe are two problems that may arise during data collection in inclement weather conditions
• Saturation headways increased by an average of 4.92 percent during wet weather
• Saturation flow rate was reduced by 4.68 percent during wet weather
• Free flow speeds decreased by 9.44 percent during wet weather

Based on the literature review, it would appear that these driver behavior parameters might vary significantly between one intersection and another, or even from one approach to another at the same intersection, based on the exact conditions at each location. Additionally, these values may have been influenced by the intensity of the rainfall, which varied throughout the data collection process. More research should be conducted with a larger data sample in order to draw additional conclusions on the effects of variations in rainfall on traffic flow parameters.

The VISSIM microscopic simulation model was chosen to simulate an isolated intersection during dry and wet conditions. The car-following model parameters and desired speed parameters were adjusted in order to simulate the differing weather conditions. While changing these parameters showed differences in system performance during dry and wet conditions, other parameters could have been changed to replicate driver behavior during inclement weather conditions. These parameters could include:

• Lane Changing Behavior
• Reaction Time to the Amber Signal Phase
• Look Ahead Distance

Calibrating these additional parameters may yield a simulation model which more accurately replicates different weather conditions.

The simulation study showed that wet weather had noticeable effects on traffic flow at the isolated intersection, including:

· Increased Travel Time

· Increased Delay

· Increased Stop Delay

· Increased Number of Stops

· Increased Average Queue Length

· Increased Maximum Queue Length

· Decrease in Total Volume through the Intersection (Reduction in Capacity)

· The effects of weather were greatest when the intersection was nearest to its capacity.

Though all traffic flow parameters deteriorated when the intersection reached capacity, it appeared that the wet weather caused the deterioration in the performance measures to occur at a lower traffic volume than under dry conditions. This is indicative of a lower intersection capacity due to wet weather.

New signal timing plans were developed using Synchro, and it appeared that:

• Optimizing the timing plan had a positive effect during wet and dry conditions.
• A reduction in the percent difference between wet and dry conditions in the traffic flow parameters was observed when the timing plan was optimized.

This reduction in the percent difference equates to a lessening of the effect of rainy weather on the traffic flow. Special wet weather timing plans were devised by attempting to replicate weather conditions in Synchro. These special wet weather timing plans did not produce a significant improvement in traffic flow when compared to the optimal dry-weather timing plan. This was likely because the timing plans at an isolated intersection are mainly a function of vehicle input volumes, and would therefore be similar under wet and dry conditions when the only input changes were a uniform decrease in speeds and saturation flows on all approaches.

7.2 Recommendations for Future Research

In order to better understand and utilize the results of this research project and to learn more about the effects of rainy weather at signalized intersections, more research should be conducted in several areas.

First, better data collection methods and sources should be developed. Any data source should be able to accurately obtain data during wet weather conditions, and should have the ability to automatically archive large amounts of data. Data collection should have a resolution fine enough to automatically measure the saturation flow rate. ITS devices can also be used to measure traffic flow parameters directly, such as delay, travel time, and queue length. AVI technology, such as tracking electronic toll transponders or cell phone signals could be used for these purposes. A modified version of the Smart Travel Van might also be of use for collecting data in locations that do not have a plethora of ITS devices. Large amounts of data should be collected and analyzed to better determine trends caused by weather.

A more realistic simulation model should be developed. This simulation model should be microscopic, and should examine the effects of calibrating a wide variety of parameters that may be influenced by weather in addition to car-following and speed, such as lane changing behavior, acceleration and deceleration rates, braking distance, look ahead distance, and reaction time. This calibration should take place using data collected in the field during dry and wet weather. The desired end result would be a simulation software package which would accurately portray the differences in the field between dry and wet weather. A detailed analysis on the variance of the results within the simulation software package should be conducted to determine the minimum number of runs required to accurately portray the field conditions.

Research should be conducted to determine how parameters should be calibrated between two simulation software packages, such as VISSIM and Synchro, if one package is used for simulating field conditions and another for producing signal timings. The default parameters should be examined to ensure that the conditions portrayed in one simulation model are accurately represented in the other.

The effect of rainfall rate on speed, saturation flow, and other parameters should be examined. This would allow researchers to determine if the rainfall rate would affect intersection performance or mitigation strategies.

Additional research may wish to examine the effects of adjusting signal timing along arterial networks. Reductions in speed can affect the signal timing offsets along a network, and as a result different effects of weather on performance may be seen. Additionally, mitigation strategies may better be applied to a larger network than at a three-signal network or at an isolated intersection. Research specific to which mitigation strategies work along networks would be useful to transportation engineers for improving traffic conditions in the real world. If further research shows that there are reasonable strategies for mitigating the effects of weather at signalized intersections or arterial networks, the results should be validated in the field.

REFERENCES

1. FHWA Report: Economic Impact of Highway Snow and Ice Control, Final Report. FHWA-RD-77-95, 1977.

2. Roosevelt, D. S., Hanson, R. A. Availability and Accuracy of the Virginia Department of Transportation's Road Weather Information System. Virginia Transportation Research Council (VTRC 98-R21RB), December 1977.

3. Foretell. www.foretell.com Accessed: May 21, 2004.

4. Highway Capacity Manual 2000. Transportation Research Board: Washington, 2000.

5. Smith, Brian L., et. al. An Investigation Into the Impact of Rainfall on Freeway Traffic Flow. Transportation Research Board 2004 Annual Meeting: Washington, 2004.

6. Weather Responsive Traffic Management. Federal Highway Administration. http://ops.fhwa.dot.gov/weather/best_practices/WxRspTfcMgmtTRB2004.pdf Accessed: May 21, 2004.

7. Lieu, Henry and Lin, Shiow-Min. Benefit Assessment of Implementing Weather-Specific Signal Timing Plans Using CORSIM. Transportation Research Board 2004 Annual Meeting: Washington, 2004.

8. Perin, J., Martin, P.T., and Hansen, B.G. “Modifying Signal Timing During Inclement Weather.” Transportation Research Record 1748, Transportation Research Board: Washington, 2001.

9. Maki, P. J. Adverse Weather Traffic Signal Timing. 69^th Annual Meeting of the Institute of Transportation Engineers: Las Vegas, 1999.

10. Bernardin Lochmueller and Associates, Inc. Anchorage Signal System Update – Final Report, 1995.

11. Agbolosu-Amison, Seli J., Sadek, Adel W., and ElDessouki, Wael. Inclement Weather and Traffic Flow at Signalized Intersections: A Case Study from Northern New England. Transportation Research Board 2004 Annual Meeting: Washington, 2004.

12. Smart Travel Lab at the University of Virginia. http://smarttravellab.virginia.edu/ Accessed: May 21, 2004.

13. National Oceanic and Atmospheric Administration. http://www.noaa.gov/ Accessed: May 21, 2004.

14. Garber, Nicholas J. and Hoel, Lester A. Traffic and Highway Engineering. Brooks/Cole: Pacific Grove, 2002.

15. Robertson, H. Douglas, editor. Manual of Transportation Engineering Studies. Institute of Transportation Engineers, Washington: 2000.

16. Wilbur Smith Associates. Traffic Signal Study – Barracks Road and Emmet Street, Charlottesville, Virginia, 2002.

17. CORSIM. http://www.fhwa-tsis.com/corsim_page.htm Accessed May 21, 2004.

18. VISSIM 3.70 – User Manual. Innovative Transportation Concepts, Inc.: Corvallis, 2003.

19. Park, B. Brian. Traffic Signal Optimization Project handout, 2002.

APPENDIX A:

Field Obtained Values for Saturation Headway and Free-Flow Speed

Saturation Headway in seconds – Wet Conditions

27-Oct	19-Nov	16-Mar
1.60	1.65	1.64
1.65	1.66	1.67
1.68	1.69	1.72
1.70	1.76	1.75
1.72	1.81	1.77
1.73	1.84	1.78
1.76	1.89	1.80
1.77	1.89	1.83
1.80	1.90	1.83
1.81	1.90	1.85
1.81	1.91	1.87
2.00	1.94	1.90
2.04	1.95	1.91
2.10	1.96	1.92
2.19	1.98	1.96
2.36	2.00	1.96
2.41	2.01	1.97
2.43	2.02	1.97
2.46	2.03	1.99
2.51	2.04	2.02
2.53	2.07	2.02
2.56	2.07	2.03
3.10	2.10	2.05
	2.12	2.06
	2.16	2.13
	2.21	2.14
	2.22	2.16
	2.25	2.19
	2.33	2.29
	2.33	2.46
	2.33	2.49
	2.33	2.60
	2.34	2.65
	2.38	2.67
	2.44	2.73
	2.50	2.76
	2.56	2.79
	2.69	2.84
	3.03	2.86
		2.91
		2.95
		2.97

Saturation Headway in Seconds – Dry Conditions

17-Nov	20-Nov	18-Feb	19-Feb
1.59	1.62	1.60	1.72
1.73	1.65	1.66	1.77
1.81	1.69	1.67	1.77
1.83	1.69	1.68	1.81
1.86	1.70	1.70	1.83
1.92	1.71	1.72	1.87
1.93	1.72	1.75	1.89
1.94	1.75	1.79	1.90
1.98	1.76	1.80	1.94
2.01	1.78	1.81	1.97
2.08	1.78	1.83	1.98
2.08	1.80	1.86	1.99
2.10	1.82	1.87	2.03
2.10	1.82	1.87	2.06
2.12	1.82	1.88	2.08
2.30	1.83	1.90	2.09
2.32	1.85	1.91	2.15
2.71	1.89	1.92	2.15
	1.89	1.94	2.37
	1.91	1.94	2.55
	1.92	1.97	2.58
	1.93	1.98	2.63
	1.93	1.98	2.77
	1.95	2.00
	2.00	2.01
	2.01	2.03
	2.04	2.06
	2.06	2.07
	2.07	2.08
	2.11	2.11
	2.13	2.14
	2.17	2.14
	2.17	2.16
	2.19	2.20
	2.24	2.23
	2.30	2.25
	2.35	2.32
	2.41	2.34
	2.44	2.36
	2.45	2.41
	2.54	2.43
	2.57	2.46
	2.85	2.49
		2.60

Free Flow Speeds in Miles per Hour

Dry		Wet
21-May		18-May
39.9	47.8	35.3	45.0
40.4	48.0	35.4	45.0
41.3	48.1	37.9	46.3
41.5	48.1	38.3	46.8
42.5	48.2	38.4	47.2
42.5	48.3	38.6	47.5
42.8	48.3	39.3	47.6
43.1	48.3	39.4	47.6
43.4	48.5	39.7	48.0
44.1	48.8	39.9	48.2
44.3	49.0	40.2	49.4
44.4	49.2	40.3	49.5
44.5	49.2	40.4	50.2
44.8	49.4	40.4	50.3
44.8	49.7	40.7	53.6
45.0	49.9	40.9
45.9	50.1	41.5
45.9	50.3	41.8
46.5	50.6	41.9
46.8	50.6	42.0
46.9	50.9	42.2
47.0	51.2	42.7
47.1	51.2	42.9
47.1	51.4	42.9
47.2	52.1	43.0
47.4	52.8	43.1
47.5	52.8	43.3
47.6	53.8	43.5
47.6	55.4	44.0
47.6	57.6	44.3
47.7	61.2	44.5
47.8		44.8

APPENDIX B:

Effects of Rain on Individual Intersection Approaches (from VISSIM)

DRY	Travel time (s)
	NBL	NBT	NBR	SBL	SBT	SBR
Volumes	trav time	trav time	trav time	trav time	trav time	trav time
25%	52.94	39.95	31	44.93	37.54	30.01
50%	60.79	46.85	37.24	53.1	40.5	30.77
75%	76.85	55.78	49.07	65.17	46.93	31.58
100%	95.38	66.14	58.68	145.18	59.79	36.91
125%	130.49	94.19	87.08	308.06	90.41	52.22
150%	151.91	117.82	110.35	305.02	94.59	54.17
175%	156.45	118.96	110.49	308.54	93.46	53.45
200%	155.21	118.93	111.16	317.7	94.31	54.19
225%	157.2	118.43	111.23	299.71	93.83	54.33
250%	157.65	119.06	110.37	311.67	94.68	55.62
DRY	Travel time (s)
	EBL	EBT	EBR	WBL	WBT	WBR	AVERAGE
Volumes	trav time	trav time	trav time	trav time	trav time	trav time	trav time
25%	48.2	45.69	37.09	43.01	47.74	30.63	40.93831
50%	54.73	55.68	44.57	50.67	55.25	32.4	46.73109
75%	65.77	67.2	56.77	62.62	65.9	38.25	55.78395
100%	92.36	120.97	110.3	77.55	83.21	53.89	78.20135
125%	212.49	252.16	233.18	107.87	113.91	83.87	133.4102
150%	219.15	279.18	264.02	113.73	116.13	84.07	144.4605
175%	227.11	269.94	255.33	113.33	117.11	85.6	145.0485
200%	224.59	280.94	272.35	113.49	117.94	87.19	146.7004
225%	227.03	280.91	263.12	114.13	117.1	86.02	144.9938
250%	229.07	276.24	268.93	113.1	116.93	84.51	146.4156

WET	Travel time (s)
	NBL	NBT	NBR	SBL	SBT	SBR
Volumes	trav time	trav time	trav time	trav time	trav time	trav time
25%	54.91	43.22	33.85	46.73	39.47	32.64
50%	66.83	49.24	37.81	55.21	43.15	33
75%	79.75	60.12	49.67	70.45	49.73	34.13
100%	98.78	71.92	63.32	190.94	71.97	43.82
125%	147.48	113.77	104.43	312.11	92.98	54.55
150%	152.55	122.59	113.85	321.78	96.72	56.75
175%	153.01	121.98	112.34	318.57	95.73	56.68
200%	158.33	123.21	113.79	311.59	97.06	57.57
225%	156.39	122.97	115.57	325.45	96.71	56.85
250%	154.5	123.71	113.74	338.64	96.95	55.72
WET	Travel time (s)
	EBL	EBT	EBR	WBL	WBT	WBR	AVERAGE
Volumes	trav time	trav time	trav time	trav time	trav time	trav time	trav time
25%	50.41	48.58	39.12	46.14	52.13	32.67	43.37834
50%	58.04	59.28	46.06	53.73	58.99	34.28	49.40376
75%	70.45	73.27	59.97	66.59	70.36	41.72	59.67388
100%	123.02	176.2	164.75	87.66	92.46	62.96	95.93856
125%	229.91	282.71	274.33	113.06	118.43	87.43	144.2941
150%	239.6	289.98	279.97	114.6	120.39	87.33	150.4108
175%	240.82	299.38	289.69	116.83	121.32	89.58	150.8838
200%	238.09	305.91	286.38	116.17	121.55	88.44	151.374
225%	237.83	301.16	286.01	115.34	121.68	88.25	152.3265
250%	237.98	302.72	280.05	114.07	120.12	86.22	152.418

% change	Travel time
	NBL	NBT	NBR	SBL	SBT	SBR
Volumes	trav time	trav time	trav time	trav time	trav time	trav time
25%	1.97	3.27	2.85	1.8	1.93	2.63
50%	6.04	2.39	0.57	2.11	2.65	2.23
75%	2.9	4.34	0.6	5.28	2.8	2.55
100%	3.4	5.78	4.64	45.76	12.18	6.91
125%	16.99	19.58	17.35	4.05	2.57	2.33
150%	0.64	4.77	3.5	16.76	2.13	2.58
175%	-3.44	3.02	1.85	10.03	2.27	3.23
200%	3.12	4.28	2.63	-6.11	2.75	3.38
225%	-0.81	4.54	4.34	25.74	2.88	2.52
250%	-3.15	4.65	3.37	26.97	2.27	0.1
% change	Travel time
	EBL	EBT	EBR	WBL	WBT	WBR	AVERAGE
Volumes	trav time	trav time	trav time	trav time	trav time	trav time	trav time
25%	2.21	2.89	2.03	3.13	4.39	2.04	2.440027
50%	3.31	3.6	1.49	3.06	3.74	1.88	2.67267
75%	4.68	6.07	3.2	3.97	4.46	3.47	3.889925
100%	30.66	55.23	54.45	10.11	9.25	9.07	17.73721
125%	17.42	30.55	41.15	5.19	4.52	3.56	10.88395
150%	20.45	10.8	15.95	0.87	4.26	3.26	5.950273
175%	13.71	29.44	34.36	3.5	4.21	3.98	5.835323
200%	13.5	24.97	14.03	2.68	3.61	1.25	4.673601
225%	10.8	20.25	22.89	1.21	4.58	2.23	7.332687
250%	8.91	26.48	11.12	0.97	3.19	1.71	6.002382

DRY	Volume (veh)
	NBL	NBT	NBR	SBL	SBT	SBR
Volumes	volume	volume	volume	volume	volume	volume
25%	14.7	198.7	22.7	65.3	208.3	34.7
50%	30.8	393.7	48.9	135.5	426.5	66.6
75%	49.5	589	69.9	199.9	637.6	106.1
100%	64.2	792.2	94.9	255.2	850.5	142.4
125%	76.9	968.6	118	222.4	788.6	140.1
150%	81	990.7	120	227.2	798	133
175%	81	996.3	120.6	228.4	823.3	133.3
200%	77.7	995	122.4	223.5	781.3	136.9
225%	79.6	1001.6	116.1	230.6	830.3	141.4
250%	77.2	993.4	123.4	224.9	799.5	135.5
DRY	Volume  (veh)	DRY
	EBL	EBT	EBR	WBL	WBT	WBR	TOTAL
Volumes	volume	volume	volume	volume	volume	volume	volume
25%	126.2	61.1	12.5	58.2	49.4	52.7	904.5
50%	258.3	121.5	23.1	120.9	94.4	107.6	1827.8
75%	387.1	188.9	39.7	188.1	130.9	165.7	2752.4
100%	515.7	242.1	53.7	242.9	182.5	221.5	3657.8
125%	518	237.7	49.7	288.4	216.7	251	3876.1
150%	499.8	240.9	45.7	295.1	215.6	246.1	3893.1
175%	509.2	234.1	49	291.6	209.2	248.2	3924.2
200%	496.7	241.5	48.1	285	215.5	248.5	3872.1
225%	499.4	236.9	49.2	284.9	210.2	256.8	3937
250%	510.3	232.7	51.3	289.9	210.8	251.1	3900

WET	Volume (veh)
	NBL	NBT	NBR	SBL	SBT	SBR
Volumes	volume	volume	volume	volume	volume	volume
25%	17.7	194.7	23.2	62.1	208.2	38.1
50%	33.3	391.9	48	133.9	422.2	71.4
75%	45.8	589.3	71.1	195.9	636.5	110.1
100%	70.2	790.2	94.9	241.1	817.7	139.4
125%	68.4	949.5	111.4	216.6	776.9	131.1
150%	72.3	948.1	115	214.3	772.9	126.7
175%	73.9	960.4	115.7	214.8	777.5	130.6
200%	80.4	954	108.6	218.6	774.3	128.3
225%	76.6	948.2	113.8	216.9	746.8	125.6
250%	76.9	947.3	115.2	214	737.8	129.1
WET	Volume  (veh)
	EBL	EBT	EBR	WBL	WBT	WBR	TOTAL
Volumes	volume	volume	volume	volume	volume	volume	volume
25%	123.6	58.9	11.9	54.4	45.7	57.2	895.7
50%	256.3	122.5	24.6	119.3	93	110.7	1827.1
75%	399.4	177.6	38.4	192.2	132.5	162.4	2751.2
100%	506.8	234.1	48.2	241.6	178.7	223.6	3586.5
125%	469.7	220.3	42	275.7	209.4	240.1	3711.1
150%	475.6	219.8	45.1	279.5	201.9	250.5	3721.7
175%	465.8	218	49.4	281.5	199.4	235.1	3722.1
200%	468.1	219.7	45.8	278.5	200.6	246	3722.9
225%	468	224.5	47.2	270.3	202.9	247.2	3688
250%	465.7	219.2	47.8	280.3	197.4	252.6	3683.3

% change	Volume
	NBL	NBT	NBR	SBL	SBT	SBR
Volumes	volume	volume	volume	volume	volume	volume
25%	20.41	-2.01	2.2	-4.9	-0.05	9.8
50%	8.12	-0.46	-1.84	-1.18	-1.01	7.21
75%	-7.47	0.05	1.72	-2	-0.17	3.77
100%	9.35	-0.25	0	-5.53	-3.86	-2.11
125%	-11.05	-1.97	-5.59	-2.61	-1.48	-6.42
150%	-10.74	-4.3	-4.17	-5.68	-3.15	-4.74
175%	-8.77	-3.6	-4.06	-5.95	-5.56	-2.03
200%	3.47	-4.12	-11.27	-2.19	-0.9	-6.28
225%	-3.77	-5.33	-1.98	-5.94	-10.06	-11.17
250%	-0.39	-4.64	-6.65	-4.85	-7.72	-4.72
% change	Volume
	EBL	EBT	EBR	WBL	WBT	WBR	TOTAL
Volumes	volume	volume	volume	volume	volume	volume	volume
25%	-2.06	-3.6	-4.8	-6.53	-7.49	8.54	-0.97
50%	-0.77	0.82	6.49	-1.32	-1.48	2.88	-0.04
75%	3.18	-5.98	-3.27	2.18	1.22	-1.99	-0.04
100%	-1.73	-3.3	-10.24	-0.54	-2.08	0.95	-1.95
125%	-9.32	-7.32	-15.49	-4.4	-3.37	-4.34	-4.26
150%	-4.84	-8.76	-1.31	-5.29	-6.35	1.79	-4.4
175%	-8.52	-6.88	0.82	-3.46	-4.68	-5.28	-5.15
200%	-5.76	-9.03	-4.78	-2.28	-6.91	-1.01	-3.85
225%	-6.29	-5.23	-4.07	-5.12	-3.47	-3.74	-6.32
250%	-8.74	-5.8	-6.82	-3.31	-6.36	0.6	-5.56

DRY	Delay (s)
	delay	delay	delay	delay	delay	delay
Volumes	NBL	NBT	NBR	SBL	SBT	SBR
25%	27.2	16.24	12.22	23.72	13.93	6.71
50%	34.72	22.87	18.32	31.66	16.91	7.46
75%	50.05	31.77	30.1	43.7	23.17	8.16
100%	68.83	41.5	39.46	122.97	35.72	13.55
125%	103.31	70.08	68.03	284.58	66.52	28.76
150%	123.6	92.46	89.4	278.67	69.28	30.39
175%	126.92	92.64	89.48	278.81	67.73	29.55
200%	124.05	92.48	89.65	288.32	68.46	30.01
225%	123.1	92.24	89.61	267.81	68.25	30.44
250%	125.87	92.82	89.01	280.26	68.85	31.58
DRY	Delay (s)	DRY
	delay	delay	delay	delay	delay	delay	delay
Volumes	EBL	EBT	EBR	WBL	WBT	WBR	AVERAGE
25%	18.61	16.98	8.4	19.23	18.82	7.29	13.8415
50%	25.16	26.65	16.31	26.96	26.48	9.12	19.09469
75%	35.94	37.97	27.83	38.51	36.55	14.89	26.96274
100%	61.78	90.01	79	52.82	52.54	29.23	46.54541
125%	180.55	221.1	202.54	82.61	83.95	59.91	97.82157
150%	185.91	246.11	229.88	89.41	86.26	60.38	107.3667
175%	193.31	236.48	224.46	89.55	88	62.19	107.3359
200%	190.73	246.21	237.77	89.7	88.58	63.85	109.1476
225%	193.46	244.08	229.42	90.44	87.97	62.59	106.7931
250%	194.63	240.83	233.52	89.33	87.53	61.11	108.3937

WET	Delay (s)
	delay	delay	delay	delay	delay	delay
Volumes	NBL	NBT	NBR	SBL	SBT	SBR
25%	25.62	16.86	12.77	22.84	13.41	6.26
50%	36.99	22.57	16.34	31.14	17.03	6.78
75%	49.97	33.53	28.17	45.89	23.45	7.64
100%	68.27	44.77	41.7	165.75	45.45	17.46
125%	116.58	86.5	82.32	285.17	66.02	27.84
150%	121.92	94.1	90.21	293.07	68.84	29.8
175%	120.63	93.32	89.48	287.13	67.12	29.79
200%	125.71	94.17	90.18	280.81	68.72	30.44
225%	121.37	93.72	90.94	293.19	67.69	29.9
250%	118.68	94.75	90.1	304.43	68.24	29.09
WET	Delay (s)
	delay	delay	delay	delay	delay	delay	delay
Volumes	EBL	EBT	EBR	WBL	WBT	WBR	AVERAGE
25%	18.03	16.67	8.09	20.48	20.37	7.27	13.72057
50%	25.51	27.27	14.94	27.9	27.25	8.96	19.09924
75%	37.65	41.01	28.69	40.26	38.19	16.3	28.11259
100%	89.99	143.46	133.43	61.36	60.27	37.42	60.52295
125%	194.71	247.99	240.26	86.46	85.54	61.78	105.563
150%	202.82	253.45	244.99	88.45	87.73	61.67	110.3903
175%	204.17	261.73	253.02	90.65	88.83	63.98	110.6382
200%	200.47	267.34	248.96	90.18	89.3	62.93	110.8109
225%	201.18	259.77	246	89.45	89.72	62.58	111.8426
250%	200.04	264.04	244.38	88.32	88.09	60.78	112.1237

% change	Delay
	delay	delay	delay	delay	delay	delay
Volumes	NBL	NBT	NBR	SBL	SBT	SBR
25%	-5.81	3.82	4.5	-3.71	-3.73	-6.71
50%	6.54	-1.31	-10.81	-1.64	0.71	-9.12
75%	-0.16	5.54	-6.41	5.01	1.21	-6.37
100%	-0.81	7.88	5.68	34.79	27.24	28.86
125%	12.84	23.43	21.01	0.21	-0.75	-3.2
150%	-1.36	1.77	0.91	5.17	-0.64	-1.94
175%	-4.96	0.73	0	2.98	-0.9	0.81
200%	1.34	1.83	0.59	-2.6	0.38	1.43
225%	-1.41	1.6	1.48	9.48	-0.82	-1.77
250%	-5.71	2.08	1.22	8.62	-0.89	-7.88
% change	Delay
	delay	delay	delay	delay	delay	delay	delay
Volumes	EBL	EBT	EBR	WBL	WBT	WBR	AVERAGE
25%	-3.12	-1.83	-3.69	6.5	8.24	-0.27	-0.87
50%	1.39	2.33	-8.4	3.49	2.91	-1.75	0.02
75%	4.76	8.01	3.09	4.54	4.49	9.47	4.26
100%	45.66	59.38	68.9	16.17	14.71	28.02	30.03
125%	7.84	12.16	18.62	4.66	1.89	3.12	7.91
150%	9.1	2.98	6.57	-1.07	1.7	2.14	2.82
175%	5.62	10.68	12.72	1.23	0.94	2.88	3.08
200%	5.11	8.58	4.71	0.54	0.81	-1.44	1.52
225%	3.99	6.43	7.23	-1.09	1.99	-0.02	4.73
250%	2.78	9.64	4.65	-1.13	0.64	-0.54	3.44

DRY

Stop

Delay (s)