The Impacts of Police Traffic Stops on Freeway Operations and Capacity

Material Information

The Impacts of Police Traffic Stops on Freeway Operations and Capacity
Carrick, Grady Thomas
Place of Publication:
[Gainesville, Fla.]
University of Florida
Publication Date:
Physical Description:
1 online resource (154 p.)

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Civil Engineering
Civil and Coastal Engineering
Committee Chair:
Washburn, Scott S
Committee Members:
Srinivasan, Sivaramakrishnan
Yin, Yafeng
Elefteriadou, Ageliki L
Bejleri, Ilir
Graduation Date:


Subjects / Keywords:
Free flow speeds ( jstor )
Freeways ( jstor )
Highway traffic ( jstor )
Motor vehicle traffic ( jstor )
Police ( jstor )
Roads ( jstor )
Speed ( jstor )
Traffic congestion ( jstor )
Traffic flow ( jstor )
Transportation ( jstor )
Civil and Coastal Engineering -- Dissertations, Academic -- UF
capacity -- emergency -- enforcement -- freeway -- lighting -- move -- police -- stop -- traffic
Electronic Thesis or Dissertation
born-digital ( sobekcm )
Civil Engineering thesis, Ph.D.


Some causes of non-recurring road congestion, like work zone activities, weather, traffic crashes, and vehicle disablements are well researched and quantifiable. Police traffic stops occur more frequently than crashes and vehicle disablements combined, though their impact on roadway capacity and operation has not been studied. Staged stop scenarios using a marked Florida Highway Patrol (FHP) vehicle and civilian research vehicle allowed for researchers to analyze the move over behavior of more than 9,000 outside lane vehicles approaching the police traffic stop. Additional examination Florida Department of Transportation traffic detector data at the location of over 13,000 historical FHP stops, combine to illuminate the subject. The speed of vehicles passing those enforcement stops was reduced in a statistically significant way, 1.3 and 4.6 miles per hour respectively. Move over laws create a mandate for motorists to vacate the lane adjacent to the stop. Overall, three out of four vehicles move over in compliance with the law. When red and blue lights are used exclusively, compliance is higher, along with earlier merges and greater reductions in speed among non-moving vehicles. Neither opposite direction rubbernecker effect, nor turning off forward-facing emergency lights were found to be statistically significant. From a highway capacity standpoint, the enforcement stop can be modeled as a theoretical lane blocking event, adjusted for motorists' compliance. Based on this research, police enforcement stops reduce available capacity between 54 and 58 percent on two-lane freeways, 35 and 41 percent on three lane facilities, 23 and 30 percent on four lane facilities, and 14 and 22 percent where five lanes are present. Regression analysis examined factors to explain the speed of vehicles passing stops and showed that the number of lanes, posted speed limit, and ambient lighting conditions were all statistically significant issues. From a policy standpoint, police should use their emergency lighting equipment throughout the duration of enforcement stops since they improve move over compliance and ostensibly safety. Though not modeled implicitly, breakdown is likely not a deterministic measure, so police activity during periods of high traffic flow and congestion should be reserved for compelling traffic safety needs. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (Ph.D.)--University of Florida, 2012.
Adviser: Washburn, Scott S.
Statement of Responsibility:
by Grady Thomas Carrick.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Carrick, Grady Thomas. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
864880505 ( OCLC )
LD1780 2012 ( lcc )


This item has the following downloads:

Full Text




2 2012 Grady T. Carrick


3 To Carla, Jacob, Joshua and Jonah


4 ACKNOWLEDGMENTS I would like to thank my graduate advisor, Dr. Scott Washburn for his support and friendship, along with the other members of the faculty who served on my committee; Dr. Siva Srinivasan, Dr. Lily Elefteriadou, Dr. Yafeng Yin, and Dr. Ilir Bejleri. The Flor ida Highway Patrol was instrumental in providing institutional support for this project. The men and women of the agency performed their routine duties in an admirable manner, providing the traffic stop data that was the basis for analysis. Colonel David Brierton and Trooper Roy Blanco were both helpful in their contribution of time and encouragement. The Florida Departm ent of Transportation and their Central Data Warehouse (CDW) proved that storage of historical traffic data is a worthwhile endeavor. Kate Norris from the University of Florida Geoplan Center, Pete V ega from FDOT District Two, as well as Jim Hilbert and Eric Gordon from the Florida Turnpike Enterprise all proved themselves more than accommodating in the collection and reduction of traffic data that was important to this project. Fellow students Vipul Modi Xiaoyu Zhu and Nagendra Dhakar were very helpful in accessi ng and analyzing that traffic data in the CDW. I thank my wife for her unselfish support and encouragement, and my mother for guiding me in my youth, supporting me throughout life, and being my inspiration As a single parent and breast cancer survivor, s he has my unending admiration


5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURE S ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 The Importance of Traffic Law Enforcement to Freeway Operation ........................ 17 Freeway Incidents as Causes of Congestion ................................ .......................... 18 Estimating Police Traffic Stops as a Cause of Congestion ................................ ..... 19 Frequency of Traffic Stops ................................ ................................ ............... 21 Duration of Traffic Stops ................................ ................................ ................... 21 Relative Intensity of Traffic Stops ................................ ................................ ..... 22 ................................ ................................ ....... 23 Research Objective ................................ ................................ ................................ 25 2 LITERATU RE REVIEW ................................ ................................ .......................... 27 Incidents ................................ ................................ ................................ ................. 27 Work Zones ................................ ................................ ................................ ............ 31 Weather ................................ ................................ ................................ .................. 33 Traffic Crashes ................................ ................................ ................................ ....... 35 Vehicle Disa blements ................................ ................................ ............................. 37 Effects of Traffic Enforcement ................................ ................................ ................. 39 3 .................... 46 Staging Traffic Stop Scenarios ................................ ................................ ............... 46 Experiment Vehicles and Positioning ................................ ............................... 46 Location Selection Criteria ................................ ................................ ................ 47 Surveillance and Data Recording ................................ ................................ ..... 48 Patrol Vehicle Lighting Considerations ................................ ............................. 50 Experimental Design ................................ ................................ ........................ 53 Analysis of Simulated Stops/Move Over Behavior ................................ .................. 53 Evaluate Video at Staged Stops ................................ ................................ ....... 54 Frequencies for outside lane volumes ................................ ....................... 55 Statistical significance of emergency vehicle lighting ................................ 55


6 Description of early/late merge (lane change) ................................ ............ 56 Assessment of available gaps and driver behavior ................................ .... 57 Introducing a new move over decision making model ............................... 58 Vehicle Speed Evaluation ................................ ................................ ................. 60 ................................ 60 Speeds from roadway detectors ................................ ................................ 61 Rubbernecking Impacts for Opposite Direction Traffic ................................ ..... 62 4 ESTIMATING TRAFFIC FLOW IMPACTS AT STOP LOCATIONS ........................ 77 Historical Stop Data Collection ................................ ................................ ............... 78 Collecting Police Traffic Stop Data ................................ ................................ ... 78 Collecting Florida Freeway Traffic Data ................................ ............................ 79 Spatially Combining Data Sets ................................ ................................ ......... 80 Traffic Stop Data Collection and Redaction Results ................................ ......... 81 Traffic Data Mining and Assembly ................................ ................................ .......... 82 Data Mining Objective and Methodology ................................ .......................... 83 Data Assembly ................................ ................................ ................................ 84 Statistical Analysis of Historical Stops ................................ ................................ .... 87 Stop Attributes ................................ ................................ ................................ .. 87 Speed at Stop Location s ................................ ................................ .................. 88 Plot Speed and Flow Rate at Stop Locations ................................ ................... 88 Regression Analysis and Analysis of Parameters ................................ ............ 89 Modeling Stops Using Van Aerde ................................ ................................ ........... 91 Model Parameters ................................ ................................ ............................ 93 Selection of Stop Sites for Modeling ................................ ................................ 94 Model Results ................................ ................................ ................................ ... 95 Summary of Capacity Impacts ................................ ................................ ................ 96 Existing HCM framework for incident capacity reductions ................................ 97 Measuring changes in traffic behavior at stop locations ................................ ... 98 HCM Alternative Speed Flow Calculations ................................ ....................... 99 5 RESULTS AND CONCLUSIONS ................................ ................................ ......... 134 Re search Results ................................ ................................ ................................ 134 Operational Impacts ................................ ................................ ....................... 134 Capacity Impacts ................................ ................................ ............................ 135 Policy Recommendations ................................ ................................ ..................... 137 Futu re Research Recommendations ................................ ................................ .... 137 APPENDIX A FLORIDA MOVE OVER CITATIONS BY COUNTY ................................ .............. 139 B SPOT SPEED STUDY TRACKING FORM ................................ ........................... 143 C STOP DETECTOR GIS DATA CONSOLIDATION ................................ ............... 144


7 LIST OF REFERENCES ................................ ................................ ............................. 145 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 154


8 LIST OF TABLES Table page 1 1 FHP Activities for 2009 Number of roadway incidents by type ......................... 26 1 2 FHP 2009 Mean duration of incidents by type in ................................ .............. 26 2 1 Reductions of capacity in literature ................................ ................................ ..... 43 2 2 Values form the HCM proportion of capacity available under incidents .............. 43 2 3 HCM Values for work zone flow rates wi th normal lane reductions noted .......... 44 2 4 Representative HCM capacity reductions values attributed to weather .............. 44 2 5 HCM capacities under varying environmental conditions ................................ ... 45 3 1 Volumes at simulated stop locations ................................ ................................ .. 73 3 2 Data collection even t summary ................................ ................................ ........... 73 3 3 Outside lane frequencies ................................ ................................ .................... 74 3 4 Move over compliance and lighting configuration paired sample test ................. 75 3 5 Observed percentage of late merge maneuvers. ................................ ................ 75 3 6 Gap analysis ................................ ................................ ................................ ....... 75 3 7 Average speed of vehicles that did not move over. ................................ ............ 75 3 8 Average lane speeds from detectors at staged stop locations ........................... 76 3 9 Traffic volumes at stop locations ................................ ................................ ........ 76 3 10 Opposite direction mean speeds under stop condi tions ................................ ..... 76 4 1 FHP CAD data format ................................ ................................ ....................... 127 4 2 STEWARD Data Content/Format ................................ ................................ ..... 127 4 3 Combined stops and detectors data elements table ................................ ......... 128 4 4 Sample sizes for past traffic incident studies ................................ .................... 128 4 5 Final data table for historical traffic stops ................................ ......................... 128 4 6 Segment relationship between stops and detectors ................................ ......... 129


9 4 7 Descriptive statistics for average lan e speeds ................................ .................. 129 4 8 Comparison of mean speeds for non stop and at stop conditions for all stops 129 4 9 Regression results for Speed_AtStop explanatory variables ............................ 129 4 10 Descriptive statistics and comparison of mean speeds for non stop (NS) and at stop (AS) for POST_SPEED variable ................................ ........................... 130 4 11 Descriptive statistics and comparison of mean speeds for non stop (NS) and at stop (AS) for NUM_LANES variable ................................ ............................. 130 4 12 Descriptive statistics and comparison of mean speeds for non stop (NS) and at stop (AS) for Ambient variable ................................ ................................ ...... 130 4 13 Locations Selected for Van Aerde Modeling ................................ ..................... 131 4 14 Consolidation of Traffic Stream Calibration Software output by detector. ......... 131 4 15 Change in free speed parameter ................................ ................................ ..... 131 4 16 Change in speed at capacity parameter ................................ ........................... 132 4 17 Change in capacity parameter ................................ ................................ .......... 132 4 18 Values from HCM for capacity remaining at incident ................................ ........ 133


10 LIST OF FIGURES Figure page 1 1 Sources of congestion (FHWA, 2010) ................................ ................................ 26 3 1 Photo of research vehicle and FHP vehicle in stop configuration ....................... 63 3 2 Photo of the researcher using a laser based speed measuring device .............. 64 3 3 Florida move over citations by year since inception ................................ ........... 64 3 4 Simulated stop site A I 95 NB @ 295mm, Flagler Co. ................................ ..... 65 3 5 Simulated stop site A ITS Camera Ima ge ................................ ........................ 65 3 6 Simulated stop site B I 95 SB S. of Old St. Augustine Rd, Duval Co. .............. 66 3 7 Simulated Stop Site B ITS Camera Image ................................ ...................... 66 3 8 Simulated Stop Site C I 95 SB N. of Pecan Park, Duval Co. ........................... 67 3 9 Simulated Stop Site C ITS Camera Image ................................ ...................... 67 3 10 Simulated Stop Site D ................ 68 3 11 Simulated Stop Site D ITS Camera Image ................................ ...................... 68 3 12 Simulated Stop Site E I 4 EB @ 124.4 mm, Volusia, Co. ................................ 69 3 13 Simulated Stop Site E ITS Camera Image ................................ ...................... 69 3 14 Image of emergency top lights only in operation configuration A. .................... 70 3 15 Image of emergency top lights plus amber directional arrow configuration B .. 70 3 16 Images of an amber directional arrow in right to left operation configuration C ................................ ................................ ................................ ......................... 71 3 17 ......................... 71 3 18 Move Over Lane Changing Model ................................ ................................ ...... 72 3 19 Average lane speed from roadway detectors ................................ ..................... 72 4 1 Florida counties (highlighted) and instrumented roadways used in project ...... 101 4 2 FHP mobile computer traffic stop input screen ................................ ................. 102


11 4 3 Duval County stops and detectors ................................ ................................ .... 103 4 4 Volusia County stops and detectors ................................ ................................ 104 4 5 Brevard County stops and detectors ................................ ................................ 105 4 6 Seminole County stops and detectors ................................ .............................. 106 4 7 Orange County stops and detectors ................................ ................................ 107 4 8 Osceola County stops and detectors ................................ ................................ 108 4 9 Hillsborough County stops and detectors ................................ ......................... 109 4 10 Broward County stops and detectors ................................ ................................ 110 4 11 Zoomed representation of stops and detectors ................................ ................ 111 4 12 Stop reduction process and resulting cases ................................ ..................... 112 4 13 Timeline representation of sampling methodology. ................................ .......... 112 4 14 Day of week distribution for stops ................................ ................................ ..... 112 4 15 Time of day distribution for stops ................................ ................................ ...... 113 4 16 Distribution of stops by daylight ................................ ................................ ........ 113 4 17 Three types of freeway flow similar to illustration in the HCM ........................... 114 4 18 Speed Flow plot for historical stops At Stop Condition ................................ .. 114 4 19 Speed Flow plot for historical stops Non Stop Condition ............................... 115 4 20 Speed Flow plot for historical stops POST_SPEED=70 mi/h ........................ 115 4 21 Speed Flow plot for historical stops POST_SPEED=65 mi/h ........................ 116 4 22 Speed Flow plot for historical stops POST_SPEED=55 mi/h ........................ 116 4 23 Speed Flow plot for h istorical stops POST_SPEED=50< mi/h ......................... 117 4 24 Speed Flow plot for historical stops NUM_LANES=2 ................................ ....... 117 4 25 Speed Flow plot for historical stops NUM_LANES=3 ................................ ....... 118 4 26 Speed Flow plot for historical stops NUM_LANES=4 ................................ ....... 118 4 27 Speed Flow plot for historical stops NUM_ LANES=5/6 ................................ .... 119


12 4 28 Speed Flow plot for historical stops Ambient=Day ................................ ........... 119 4 29 Speed Flow plot for historical stops Ambient=Night ................................ ......... 120 4 30 Speed Flow Plot for Location 410201 ................................ ............................... 120 4 31 Speed Flow Plot for Location 501351 ................................ ............................... 121 4 32 Speed Flow Plot for Location 510082 ................................ ............................... 121 4 33 Speed Flow Plot for Location 510311 ................................ ............................... 122 4 34 Speed Flow Plot for Location 510422 ................................ ............................... 122 4 35 Speed Flow Plot for Location 510511 ................................ ............................... 123 4 36 Speed Flow Plot for Location 510522 ................................ ............................... 123 4 37 Speed Flow Plot for Location 510711 ................................ ............................... 124 4 38 Speed Flow Plot for Location 510731 ................................ ............................... 124 4 39 Speed Flow Plot for Location 511322 ................................ ............................... 125 4 40 Adjusted speed flow curves ................................ ................................ .............. 126


13 LIST OF ABBREVIATION S A ADT Annual Average Daily Traffic AIMSUN A transportation simulation software program (Spain) CAD Computer Aided Dispatch CAF Capacity Adjustment Factor CDW Central Data Wareho u s e CORSIM CORridor SIMulation A microscopic traffic simulation software program (US) DHSMV Florida Department of Highway Safety and Motor Vehicles FBI Federal Bureau of Investigation FDOT Florida Department of Transportation FFS Free Flow Speed FHP Florida Highway Patrol FHWA Federal Highway Administration FTP File Transfer P rotocol GIS Geographic Information System GPS Global Positioning System HCM Highway Capacity Manual I 95 Interstate 95 I 75 Interstate 75 I 295 Interstate 295 I 275 Interstate 275 I 4 Interstate 4 I 595 Interstate 595 ITS Intelligent Transportation System LAT/LONG La titude/Longitude


14 LED Light Emitting Diode LOS Level of Service MDC Mobile Data Computer MUTCD Manual on Uniform Traffic Control Devices NCSC National Center for State Courts NHTSA National Highway Traffic Safety Administration p c / h /lane P assenger cars per hour per lane PHF Peak Hour Factor RTMS Remote Traffic Microwave Sensor STEWARD Statewide Transportation Engineering Warehouse for Archived Regional Data SQL Standard Query Language TIM Traffic Incident Management TMC Traffic Management C enter TRC University of Florida Transportation Research Center TSS Traffic Sensor Subsystem UF University of Florida VISSIM A multi modal microsimulation traffic flow modeling software program (Germany) VMT Vehicle Miles Traveled veh / h / l n vehicles per hour per lane WAF Weather Adjustment Factor


15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE IMPACT S OF POLICE TRAFFIC STOPS ON FREEWAY OPERATIONS AND CAPACITY By Grady T. Carrick May 2012 Chair: Scott S. Washburn Major: C ivil Engineering Some causes of non r ecurring road congestion like work zone activities, weather, traffic crashes, and vehicle disablements ar e well researched and quantifiable Police traffic stops occur more frequently than crashes and vehicle disablements combined though their impact on roadway capacity and operation has not been studied. Staged stop scenarios using a marked Florida Highway Patrol ( FHP ) vehicle and civilian research vehicle allowed for researchers to analyze the move over behavior of more than 9,000 outside lane vehicles approaching the police traffic stop. Additional e xamin ation Florida Department of Transportation traffic detector data at the location of over 1 3 ,000 historical FHP stops combine to illuminate the subject. The speed of vehicles passing those enforcement stops was reduced in a statistically significant way 1.3 and 4.6 miles per hour respectively Move over laws create a mandate for motorists to vacate the lane adjacent to the stop. Overall, three out of four vehicles move over in compliance with the la w Wh en red and blue lights are used exclusively compliance is higher, along with earlier merges and greater reductions in speed among non moving vehicles. Neither o pposite direction


16 rubbernecker effect nor turning off forward facing emergency lights w ere found to be statistically significant. From a highway capacity standpoint, the enforcement stop can be modeled as a theoretical lane blocking event adjusted for motorists compliance. Based on this research police enforcement stops reduce available capacity between 54 and 58 percent on two lane freeways, 35 and 41 percent on three lane facilities 23 and 30 percent on four lane facilities, and 14 and 22 percent where five lane s are present Regression analysis examined factors to explain the speed o f vehicles passing stops and showed that the number of lanes, posted speed limit, and ambient lighting conditions were all statistically significant issues. From a policy standpoint, police should use their emergency lighting equipment throughout the durat ion of enforcement stops since they improve move over compliance and ostensibly safety. Though not modeled implicitly, breakdown is likely not a deterministic measure, so police activity during periods of high traffic flow and congestion should be reserve d for compelling traffic safety needs.


17 CHAPTER 1 INTRODUCTION The Importance of Traffic Law Enforcement to Freeway Operation The mandate for traffic safety is evident in familiar national traffic crash statistics. While overall fatalities have fallen in recent years and the death rate based on VMT shows favorable trends, traffic crashes continue to be a crucial public health and safety issue. In 2008, 37,261 people were killed and another 2,346,000 injured in nearly 6 million police report ed tr affic crashes in the U.S. Someone dies on American roadways every 14 minutes, and in 2006, road deaths were the leading cause of death for every age from 3 to 34 (NHTSA 2008 ) Since driver behavior is a dominant cause of traffic crashes, the need for tra ffic law enforcement is apparent. The objective of enforcement at the individual level is to c hange the behavior of the person stopped and on a broader level, create general deterrence for all drivers. Evidence suggests that the threat of being caught is a strong motivation for most drivers to obey motor vehicle laws (Tay 2005 ). Orderly operation of vehicles on our roadways improves safety and capacity. Order is created by a consistent set of roadway rules accompanied by a perceived risk of sanction if they are not followed. A visible presence of traffic enforcement reinforces the perceived risk, and consequently encourages lawful behavior. A regular and continuous level of visible enforcement has been found to be more effective than sporadic intensifi ed deployments (Smith and Yoo k 2009 ) As a necessary part of freeway operations t here is likely a balance between optimum visible enforcement to maintain order and a level that adversely impacts the traffic stream. While one might intuitively suspect dr ivers reduce their speed upon observation of a police vehicle, l ittle


18 research has been undertaken to explore the relationship between visible traffic enforcement by police and the characteristics of traffic flow. Freeway Incidents as Causes of Congestion According to the 2010 Urban Mobility Report, congestion c auses U.S. travelers about $115 billion dollars in lost hours and fuel (Schrank and Lomax 20 10 ). Roadway congestion is becoming a common feature of highway travel, but it transcends the obvious conclusion that there are simply too many cars and too few lanes. The Federal Highway Administration (FHWA) estimates that 55% of delays are attributable to non recurring congestion such as crashes, breakdowns, work zones or weather (Cambridge Systematics 2004 ). nificant (Schrank a nd Lomax 2009 ). The FHWA (2002) speculates that about 25% of congestion can be linked to incidents specifically, such as vehicle disablements and crashes. If one considers that at least 1 in 4 congested conditions on urban freeways are incident related, developing ways to reduce the occurrence and/or impact of those incidents can greatly improve operations. While unexpected events like crashes and disabled vehicles obstructing travel nition is not quite as tangible. The Manual on Uniform Traffic Control Devices (MUTCD) defines an incident as, an emergency road user occurrence, a natural disaster, or other unplanned event (FHWA 200 9) The Highway Capacity Manual (HCM) a ny occurrence on a roadway that rgency or


19 The FHWA Traffic recurring event that causes a ( PB Farradyne 200 0 ) This definition is much more encompassing and likely more realistic in transportation management. The latest version of the Traffic Incident Handbook (2010) maintains the prior definition but adds commentary about post 9 / 11 / post Katrina views of inc ide nts incident response and the National Incident Management System (NIMS) as an overarching guide ( ATRI and SAIC 2010 ) Law enforcement traffic stops have been largely discounted in capacity research, or if mentioned, only tangentially considered If included they are simply lumped into a roadway debris vehicle breakdowns, and other occurre nces. The HCM definition would include law enforcement stops, as would the Traffic Incident Handbook definition, though neither publication is explicit in their inclusion. Neither publication contains any detail about the conduct of law enforcement traffic stops. Li terature is replete with research examining the impact of crashes, work zones, and weather, but there is ver y little specific to the topic of traffic enforcement. The Highway Capacity Manual (HCM) briefly mentions the need for research in this area. It is worthwhile to study the operational impacts of police stop s for the simple reason that unlike most inciden ts, they are completely discretionary. Estimating Police Traffic Stops as a Cause of Congestion While they occur thousands of time each day across the U.S., t he police traffic stop is rarely considered in U.S. traffic incident management (TIM) circles. Mos t TIM discussion and planning centers on lane blocking events and the mit igation of


20 congestion that comes from incident detection, response, and clearance activities. Current TIM practices, policies, planning, and training do not include or address traffi c stops as an incident typ ology The police traffic stop is also noticeabl y absent in the Highway Capacity Manual and in the discussion of capacity reduction s associated with incidents. There is a need and a value to examining the degree to which traffic enforcement operations contribute to congestion. The need stems from the absence of research on the topic. The value speaks to the fact that police enforcement action is largely discretionary, with individual officers determining the time, place, and du ration of the events. If enforcement choices were made with consideration to their impact on the operation of th e roadway, greater efficiency may be possible. Law enforcement agenc ies have informally suspected negative flow impacts associated with police activity on urban freeways for decades. Feedback from several urban agencies describe informal directives that encourage officers to minimize traffic enforcement activities, minimize the use of emergency lighting on the shoulder, or temper the freq uency at which enforcement is undertaken during peak periods in deference to traffic flow. When one considers the natural tendency of drivers to slow the shoulder, the suspicions of law enforcement are reinforced. If police consider potential impacts and motorist behavior apparently reinforces those ideas why i s research on the topic lacking? The answer may lie in perceptions about the frequency, duration, and relat ive intensity of the police traffic enforcement stop.


21 Frequency of Traffic Stops The National Center for State Courts (NCSC) estimates more than 56 million infraction citations were issued in the United States in 2007, a figure that does not include crimin al traffic violations or parking tickets ( LaFountain et al. 2009 ) Florida Department of Highway Safety and Motor Vehicle (DHSMV) records document over 4 million Florida traffic citations in 2009 ( Florida DHSMV 2011 ) Considering that police stops may re sult in warnings or no sanction at all, the number of stops may actually be much higher than these citation numbers. To frame the relative frequency of traffic stops, comparing their occurrence with crashes and disablements at an agency level is beneficial According to Florida Highway Patrol (FHP) traffic stop data records from 2010, and agency crash and disabled vehicle statistics from 2009, FHP traffic stops occur more frequently than crashes and disablements combined ( Florida DHSMV 2011 ) Since Florid a Department of Transportation (FDOT) Road Rangers operate on freeways patrolled by the FHP, adding their statistics for assisting motorists can create a realistic picture of vehicle breakdowns, though there may be some minimal duplication when both agencies are on t he scene. When adding 20 0 9 129,149 Road Ranger records as reported by the FDOT central office the number of di sabled vehicles increases to 417,518 Table 1 1 shows the frequency of enforcement stops compared with crashes and disablement s. Even with the most conservative figures, traffic stops still outpace both crashes and disablements. Duration of Traffic Stops Perhaps one reason that traffic stops are largely discounted in research is their typical short duration. Florida Highway Patr ol computer aided dispatch records and


22 records from the Florida Dep artment of Transportation Sunguide system for 2010 provide insight into the duration of various types of highway incidents. The typical FHP crash takes about 2.5 hours to complete and a tr affic stop lasts just above 20 minutes. The stop duration includes the time required for the trooper to initiate the pull over maneuver, safe l y approach the violator, secure the necessary documents, electronically create sanction ing documents, and return to the violator vehicle to conclude the stop. According to FDOT R o ad R anger records obtained from the S unguide software, the duration for the average disabled vehicle is more than 26 minutes. Table 1 2 shows the mean time to clear crashes, disablements, and traffic stops respectively. Relative Intensity of Traffic Stops While police enforcement rarely involves any blockage of lanes, the presence of a marked patrol vehicle on the shoulder of the roadway with emergency equipment activated likely distracts other motorists more than an abandoned or disabled vehicle. Considering that the process of stopping a motorist typically involves the vehicles slowing and changing lanes to arrive at the stopped shoulder position, additional turbulence in the traffic stream is likely present. Traffic impacts may even go beyond the direction of travel for the stop, as evidenced by Dutch researchers who estimate d that rubbernecking on the part of drivers on the opposite side of a highway incident can re du ce capacity by as much as 50% (Knoop Hoogendoorn, Van Zuylen 2008) O ther research found those effects to be less prominent reducing capacity around 30% (Knoop Hoogendoorn, and Adams 2009) A U.S. study found that gawker impacts on capacity near cras hes constitute a 12.7 percent reduction (Teng and Masinick 2004 ) Some effects are intuitive, but as indicated in the limited research available, difficult to quantify. While capacity reductions have been attributed to a disabled vehicle on the


23 should er and a crash on the shoulder, the traffic stop scenario with a marked police vehicle and emergency lights activated has not been determined. On a continuum the enforcement action likely falls between the simple disabled vehicle and multiple vehicles on t he shoulder as a result of a traffic mishap, but that measure of intensity has not yet been determined. It is certain however, that the intensity cannot be discounted. Working near roadways presents dangers for public safety transportation, and other professions. Police are confronted with dangers associated with traffic as well as those that accompany taking enforcement action. The Federal Bureau of Investigation (FBI) uniform crime reports track officer deaths nationally and of the 1,264 officers killed in the line of duty between 2000 and 2009, 415 were killed in automobile collisions, including 47 performing traffic stops and 73 directing traffic or assisting motorists (U.S. DOJ 2010 ) In addition, another 101 were k illed feloniously while conducting traffic stops (U.S. DOJ 2010 ) With nearly half of all police officer death s occurring in the roadway environment, the dangers are apparent. Workers from other disciplines such as emergency medical services (EMS), fire towing, and transportation are also in danger. To promote safety for police officers and other responders on our highways e very U S state except Hawaii has implemented some type of law requiring drivers to slow down or move over when approaching a stopped emergency vehicle. from state to state in their mandate for drivers. The National Committee on Uniform Traffic Laws and Ordinances published a model language for these laws, and most fit the general format. The laws are usually connected with existing laws governing yielding to emergency vehicles and contain definitions of emergency sc e nes, a responsibility to ch ange lanes when


24 available, a responsibility to slow down, and some state mandated education effort (Carson 2010 ) In Florida, the law requires drivers to vacate the lane adjacent to the emergency vehicle when gaps are available on multi lane facilities, or to slow to 20 miles per hour below the speed limit on single lane roadways or when an opportunity to change lanes is not present. The following excerpt is from Florida Statutes: Florida Statute 316.126(1)(b) When an authorized emergency vehicle making u se of any visual signals is park ed or a wrecker displaying amber rotating or flashing lights is performing a recovery or loading on the roadside, the driver of every other vehicle, as soon as it is safe: 1. Sh all vacate the lane closest to the emergency vehicle or wrecker when driving on an interstate highway or other highway with two or more lanes traveling in the direction of the emergency vehicle or wrecker, ex c e p t when otherwise directed by a law enforcemen t officer. If such movement cannot be safely accomplished, the driver shall reduce speed as provided in subparagraph 2. 2. Shall slow to a speed that is 20 miles per hour less than the posted speed limit when the posted speed limit is 25 miles per hour o r greater; or travel at 5 miles per hour when the posted speed limit is 20 miles per hour or less, when driving on a two lane road, except when otherwise directed by a law enforcement officer. ( State of Florida 2010 ) Most of these laws have been introduce d in just the last ten years, and their impact on traffic flow has not been studied. If the impact of law enforcement stops is unknown then additional impacts reinforce the compelling nature of this research. Florida is actua lly an excellent candidate state for examining move over behavior because it s law has been noted as having been highly publicized and well reinforced with enforcement (B i erling and Li, 2009 )


25 Research Objective The objective of this research is to understand the impact of enforcement stops on traffic by 1) examining the move over behavior of drivers passing police stops and 2 ) examining the characteristics of traffic flow at locations near police traffic stops on f reeways. This research objective w ill be accomplished through the conduct of the following tasks: A r eview of the relevant literature. Collect and analyze traffic flow data from marked FHP vehicle and civilian research vehicle. Collect and analyze archiv ed traffic flow data corresponding to historical stop locations. Individual chapters describe the methodological conduct of staged traffic stops and the mining of historical stop data. A separate chapter synthesizes the analytical results from both forms o f traffic stops. Finally, conclusions and recommendations are made to advance the state of the practice concerning police traffic stops and the analysis of traffic operations under the influence of law enforcement activities.


26 Figure 1 1. Sources of co ngestion (FHWA, 2010) Table 1 1. FHP Activities for 2009 Number of roadway incidents by type Crashes Disabled Vehicles Traffic Stops 217,033 417,518* 766,404 *Includes 129,149 FDOT Road Ranger Assists Table 1 2. FHP 2009 Mean duration of incidents by type in Crashes Disabled Vehicles Traffic Stops 02:29:21 00:26:24 00:20:32 Format HH:MM:SS Work Zones 10% Bad Weather 15% Traffic Incidents 25% P oor Signal Timing 5% Special Events/Other 5% Bottlenecks 40% Sources of Congestion


27 CHAPTER 2 LITERATURE REVIEW A systematic review of literature examines the body of research concerning highway incidents in general, work zones, weather, crashes, and vehicle breakdowns. Past research related to each of these non recurring events is presented, as well as a summary of how each are handled in the Highway Capacity Manual Although there is little research that directly lin ks police enforcement with freeway operations, there have been a number of studies that explore the efficacy of enforcement as a speed reduction tool. The final section of the literature review presents past research relating enforcement and speed, an impo rtant requisite for this product. Incidents ( Qin and Smith 2001 ). A wealth of research is available conc erning the topic of traffic incidents, but much of it centers on estimating and modeling incident duration characterizing incidents or in calculating delay attributed to those events. Still other efforts examine incident management strategies associated with detection, response, and clearance as parts of a comprehensive traffic incident management (TIM) program Predictive models complete most of the incident research typologies to date Since incidents impact freeway flow, there has been a great deal a ttention focused on determining how long they last and how they look. Research that relates incidents and capacity are less common, but their review is beneficial for the study at hand. Goolsby (1971) provided foundational research into the topic of incid His research


28 has been referenced for decades and the findings of capacity reductions attributed to crashes and disablements on a three lane highway in Houston, TX were among the first to quant ify impacts. He was among the first to note that the loss of capacity is actually greater than the percentage of roadway lanes blocked. The research found that capacity is reduced by 50% when one lane is blocked, 79% when two lanes are blocked, and 33% w hen the event is limited to the shoulder. Blumentritt Ross, Glazer, Pinnell, and McCasland (1981) examined capacity reductions due to freeway incidents and determined that the width of the highway was a factor in addition to the number of lanes blocked. Their effort determined that a should er incident reduced capacity on 2, 3, and four lane highways at a rate of 25%, 16%, and 11% respectively. This is significant, because it estimates shoulder events actually have about half of the impact that was previo usly thought. Qin and Smith (2001 ) revisited the issue of capacity reductions attributed to over 200 crashes in their study of data from the Hampton Roads Smart Traffic Center in Virginia. Most notable is their assertion that incidents a re better modeled as a random variable rather than deterministically. They surmised that capacity is reduced by 63% when one of three lanes is blocked, and 73% when two lanes are blocked. Their estimations concerning shoulder events were not conclusive, given the sample used and the methodology. Their estimate of a 21% reduction was lower than that of Goolsby. While their lane blocking measurements effectively shift the state of the practice, they concede that the methodology is not good at detecting small changes in ca pacity, which are common in shoulder events that do not block travel lanes (Qin and Smith 2001 ).


29 In an effort to explain a broad range of temporary capacity loses on a national level, Chin, Franzese, Greene, and Hawang (2002) examined how crashes, breakd owns, work zones, and weather impact freeways using estimates and probabilities for those events. C apacity loss and delay were their primary outcome measures Like other works, t he y categorize shoulder crashes according to the number of travel lanes and e stimate reduced capacity at 25% and 16% for two and three lane facilities respectively Research in Europe examined the flow discharge rate at 90 incidents where con gested conditions were present. Comparing incident conditions with non incident condition s, the research found that an incident on the shoulder lane reduces the efficiency of the roadway by 28%, and ironically, a drop of 31% was observed i n the ( Knoop et al. 2009 ). When congested conditions are present, t he impact of shoulder events is apparently more obvious. Lu and Elefteriadou (2010) examined incidents in five metro areas in North America, comparing non incident and incident conditions using logistic regression. Their research examined th e number of lanes present and/or blocked and found that shoulder events on 2 lane freeway segments reduce capacity by about 23%. Like Smith et al (2003), they found that a limited number of data points precluded making definitive assessments about should e r events on freeways particularly those with more than two directional lanes Prevedouros, Halkias and Papandreou (2008) examined freeway incidents in the U.S., U.K., and Greece to develop analytical models explaining duration and queue


30 length, based on i ncident variables like number of crash vehicles, response times, duration and number of lanes blocked. They found that shoulder incidents had a capacity reduction factor of .79 and .83 for two and three l ane freeways respectively which translate into 21% and 17% reductions Table 2 1 presents the reductions in capacity associated with crashes and other incidents, based on the findings of the above research. I t should also be noted that these values are for crashes and/or incidents and the number of vehic les on the shoulder and the number of emergency vehicles present may introduce a value of intensity that is not controlled. Sinha and Hadi (2007) examined the impacts of incidents on capacity in microsimulat i on. They noted that there is a need for calibr ation of the models used in capacity analysis software. Target capacit y values and the simulation of incident blockages were adjusted in three microscopic packages (CORSIM, VISSIM, and AIMSUN). Capacity reductions of 44%, 30%, and 37% were noted for each respective package, which is lower than the HCM value of 51%. The rubberneck parameter and location of upstream warning parameter in CORSIM were found to be most effective in modeling incident capacity reduction (Sinha and Hadi) Alvarez and Hadi (2010) recommended ways to calibrate those models using ITS data. The HCM (2010) freeway methodology applies reductions to capacity measures when a traffic crash or vehicle breakdown occur s The reductions consider the number of travel lanes and the number of la nes blocked. As evidenced by literature, the amount of capacity lost is not proportional to the ph ysical lane blockage. Table 2 2 represents values in the HCM for the proportion of capacity available under incident conditions.


31 Work Zones Polus and Swartz man (1999) studied the flow characteristics at suburban freeway work zones, to include one work zone where a marked patrol vehicle was parked on the shoulder. The limited number of sites precluded definitive capacity measures, but it was noted that the pr esence of a patrol vehicle on the shoulder reduced headways and the presence of stationary patrol vehicles as a speed management tool in work zones Twenty four Nor th Carolina freeway work zones were analyzed by Dixon, Hummer, and Lorscheider (1996) to determine where capacity is lowest within a work zone. Whether the site was urban or rural, as well as the intensity of the work were determined to be important varia bles. Heavy work in urban locations rendered capacity values of 1,500 vehicles per hour per lane, while rural sites were approximately 1,200 vehicles per hour per lane In rural Iowa work zones, lane closure volumes ranged from 1,400 to 1,600 and queu es s wiftly change. Maze, Schrock and Kamyab (2000) noted that historical measures of volumes and capacity provided by roadway detectors in a particular work zone shed light on the congestion problem and may provide insight into when traffic management counter measures might be needed. While the current state of the practice for work zone capacity measurement is contained in the Highway Capacity Manual, some have proposed enhancing currently used work zone capacity methodologies by applying a driver familiarity in the equation (Heaslip Louisell and Collura, 2008 ) The efficiency at which drivers navigate work zones is variable depending upon driver demographics a nd those unique to localities or regions. Stratification of driver populations considered familiarity, adaptability,


32 aggressiveness, and accommodation to render a driver population factor. Si milarly Al Kaisy and Ha l l (2001) looked at driver populations a t freeway work zones, applying adjustment factors based on trip purposed Differentiation between time of day and day of week were viewed as factors for populations and adjustments were estimated accordingly Avrenli, Benekohal, and Ramezani (2009) examin ed the speed flow relationship and capacity in their analysis of work zone locations that do not have lane closures. They used a three regimen model to determine that observed free flow rates of about 800 pc/h/lane are lower than the HCM and transition re gimen speeds actually decreased at a higher rate than the basic segment model in the HCM. When combined with the congested regimen, the three regimen work zone capacity was estimated at 1900 pc/h/lane Short Term Interstate lane closures were the focus of twenty two work zones in a South Carolina study by Sarasua, Davis, Clarke, Kottapally, and Mulukutla (2004) They sought to develop a methodology for improving lane closure policy by evaluating capacity given a variety of work zone characteristics. They found that the prevailing capacity measure of 800 pc/h/lane pc/h/lane After gathe ring field data, Heaslip, Kondyli, Arguea, Elefteriadou, and Sullivan (2009 ) used microsimulation to develop capacity models for work zones in Florida. They examined geometric, traffic, and work related parameters and found that their simulation models estimated capacity values well when compared to observations.


33 They also were close to those rendered by the methods in the Highway Capacity Manual. The Highway Capacity Manual is quite complete in its methodology for work zone activities. The methodology considers both short term and long term work zone settings, the difference m ainly relating to the duration, intensity, and type of traffic control features used. Lane shifts or closures may accompany either. For short term work zones, the intensity of the work, influence of heavy vehicles, and presence of ramps are all considered Because long term work zones are highly dependent upon si te specific variables, their operational capabilities are highly variable. If local capacity values are not available, the HCM provide s a table as a guide. Table 2 3 presents the values in the HC M for capacity values for long term work zones in veh/h/ln Weather Kyte, Zhatib, Shannon, and Kitchener (2001) examined the effects of weather on rural Interstates in Idaho to find that free flow speeds are affected by the conditions of the pavement, wind, and visibility. Vehicle speeds were examined under good conditions and compared with a variety of weather related conditions detected by advanced weather and visibility sensing equipment deployed at roadway locations. Rkha, Farzaneh, Arafeh, and Ste rizin (2008) sought to quantify the impact of precipitation and visibility on key traffic stream parameters by examining weather and traffic flows in Minneapolis St. Paul, Baltimore, and Seattle. The impacts of rain and snow were the focus of the research and it was found that both types of precipitation impact free flow speeds, speed at capacity, and capacity. The intensity of the precipitation was noted as an important factor in the study that examined aggregated


34 loop detector data at a five minute inter val. The larger study by Hranac et al. (2006) drilled into roadway capacity during rain and snow, and using a car following model developed by Van Aerde, they provided weather adjustment factors, WAF. Agarwal, Maze, and Souleyrette (2005) noted that weath er reduces operating speeds and capacities on roadways, and their research focused on rain, snow, and visibility. Freeway traffic flows from roadway detectors in the Twin Cities were matched with 8 weather stations located at freeway roadside to estimate impacts on flow. They noted capacity reductions of 10 17% for rain, 19 27% for snow, and 12% for visibility. Speeds were reduced by 4 7%, 11 15%, and 10 12% for each respective type of r than those specified in th The Highway Capacity Manual realizes that common weather events can have an adverse impact on travel. Rain and snow are all fairly well researched types of weather, and their impacts on capacity have been quantified, based on the heavy or light precipitation. Based in part on the findings of Agarwal, fog is grouped into a visibility category and along with wind, and now considered as part of the 2010 HCM methodology. The impact of flow and weather on Southern California accidents were analyzed by Golob and Recker (2003). They examined loop detector data near the crash location, immediately prior to the crash occurrence, and related environmental attributes like weather, road surface condition, and lighting to the incident. They found that crash severity was more strongly linked to volume than speed, and that fixed object and multi


35 vehicle collisions were likely on wet roads. Rear end type collisions were noted to occur more frequently on dry road s. The HCM examines wet and dry pavement, along with lighting conditions and day of the week as environmental factors. Daylight versus night are noted as relevant, since nighttime conditions are present during peak hours in some locations (HCM). Table 2 5 is represents the values from the HCM that describe capacities on German Autobahns under varying conditions. Traffic Crashes Expressways in three southern California counties provided inciden t data reviewed by Golob, Recker, and Leonard (198 7) Integrating crash report data and police dispatch logs, researchers compared differences in means for crash variables to identify relationships. The resulting distributions showed that the type of collision, lane blockage, and injury were significan t factors affecting duration. As was mentioned in the previous sec tion, a separate work from examined the relationship between speed, density, and crash attributes (Golob Recker, and Alvarez 2003) The work by Golob, et al ( 2003) additionally looked at c rashes and traffic detector information to define the traffic characteristics where freeway crashes occur. The influence of mean volume median speed, and temporal variations in volume and speed were examined. Th e y determined that the relationships potentially provide some foundation for a real time safety tool. Aljanahi Rhodes, and Mecalfe (1999) studied the relationship between traffic speed under free flow conditions in the U.K. and Bahrain. Their effort drew strong links between variability of traffic speeds and crashes and they conclude that enforcing lower speeds or reducing the spread of vehicle speeds would improve safety


36 Pande, Abdel Aty, and Hsai (2005) used archived traffic detector data to develop a p redictive crash model. Through various statistical applications, the spatiotemporal variation in speeds was used to identify black spots and consequently patterns of increased risk Crashes, as a type of incident clearly contribute to congestion, but they are also catalysts for secondary crashes on freeways. Vlahogianni Karlaftis, Golias, and Halkias (2010) examined spatiotemporal characteristics of incidents to determine that not only the traffic conditions at the time of the initial crash but also the ir contribution to secondary events Minimizing secondary crashes is an often espoused goal of TIM, but one continues to be an area of research need. One limitation of this study is that it only considers secondary crashes as a result of queues caused by c rashes, not other forms of incidents. Incidents on Houston, Texas freeways were examined by Ullman and Ogden (1996) using data collected by the Houston Police Department from 1986 to 1992. istics, the team the incidents at interchange locations, and noted the high representation of trucks as involved vehicles Jones, Janssen, and Mannering (1991) examined crash data and police dispatch times in Seattle, Washington, using mutivariate statistical models to analyze characteristics that affect incident frequency and duration. Later, Nam and Mannering examined incidents in Washington State using hazard based an alysis involving duration models for occurrence, notification, and responder arrival (Nam and Mannering ).


37 As mentioned ea r lier, Smith et al. (2003) studied traffic detector information near crash locations and confirmed that capacity reductions were grea ter than percentages corresponding to the physical lane blockage. Their work refined the decades old reductions established by Goolsby Their work was groundbreaking from the perspective that they modeled accidents as a random variable rather than a deter ministic one. They define the accident capacity as the minimum 15 minute oversaturated flow at the upstream of a bottleneck created by an accident. They did find difficulty however, in modeling shoulder accidents in this way due to limitations in their data set, most notably the majority of shoulder events involved only small capacity reductions. This is relevant because shoulder events, weather, dis ablements, and traffic stops do not typically block travel lanes and therefore may possess similar characteristics in terms of capacity impacts As was noted previously, the Highway Capacity Manual methodology provides a table that represents the proportio n of freeway capacity available under incident (crash and vehicle breakdown) conditions. Table 2 2 previously listed, provides capacity reductions associa ted with incidents in the HCM. Vehicle Disablements the associated capacity loss and (Chin et al. ) The number of vehicle breakdowns on U.S. freeways is difficult to estimate, but Chin, et al. (2002) pinned the number at over 7 million annually a t the turn of the century Disablements count representing 42 percent of all incidents versus 12 percent for crashes (Dowling, Skab ardonis Carrol and Wang 2004 ) Considering that 96 percent of breakdowns are confined to the shoulder, their


38 relative impact is obviously less than events that block lanes ( Skabardonis 1997 and Skabardonis 1999). Giuliano (1988) described incident duration as a function of incident characteristics, using descriptiv e statistics and statistical analysis Using a combination of state crash records and police dispatch logs, from Interstate 10 in Los Angeles, California, he concluded that duration was affected by the type of incident, time of day, and lane closure. The research noted that vehicle disablements and other incidents were shorter in duration that crashes. Wang, et al. (2005) analyzed the characteristics of vehicle breakdowns as a subset of incidents to examine the duration, vehicle type, location and report ing systems using fuzzy logic and artificial neural networks. The research found artificial neural network models to be more accurate than fuzzy logic, and made conclusions that there is a need for standardization of the way in which incident information is collected In their I 10 field data research, Skabardonis et al. ( 1999) determined that delay depends on duration and intensity of an incident and found that on 37% of incidents cause delay in the traffic stream They did examine shoulder breakdowns an d pointed to freeway service patrols as an asset in reducing the duration of those. The value of freeway service patrols in responding to vehicle disablements and mitigating congestion has prompted their deployment in many urban areas around the United St ates Khattak, et al. (2004) developed placement criteria in an effort to deploy service patrols using a decision support tool to maximize the ir effectiveness. Similarly Yin (2006) examined deployment methodologies as well as techniques for optimization of service patrol beats


39 Traffic Incident Management strategies aimed at disabled vehicles have proven to be effective countermeasures. Their benefit centers on reducing the duration of those incidents, whether they be lane blocking or shoulder disablements. In terms of capacity and operations, reduced duration can be extrapolated. Dougald and De m e tsky (2008) created a benefit cost estimate for service patrols in Hampton Roads, Virginia and found them to provide a benefit ratio of 5.4:1 The effectiveness of service patrols can also be measured by the favorable views of motorists and transportati on stakeholders like law enforcement, towing, and other incident responders as evidenced in the Florida DOT Statewide Road Ranger Survey for Incident Responders (FDOT ). Ta ble 2 2 previously listed, provides capacity reductions associated with incidents in the HCM. Vehicle crashes are considered a subset of incidents Effects of Traffic Enforcement There is a well founded need for visible traffic enforcement to support a safety culture among drivers. lations, but also by the risk of detection, which is in turn determined by the amount of (Elvik ) The seminal study for linking enforcement with motorist speeds was conducted by noted traffic safety expert J. Standard Baker more than a half c entury ago. Using stationary and roving patrol vehicles, the average speeds of passing motorists were examined. His research team also used increased patrols in a before/after experiment. The findings supported the belief that police presence has the ef fect of decreasing motorist speeds (Baker ) Little research has linked visible enforcement and capacity, but there have been studies similar to the early work of Baker that quantify simple changes in motorists speeds. These are beneficial to the research at hand, because of the relationship between speed, density, and flow.


40 Haas, Jones, and Kirk (2003) examined the relationship between police patrol and motorists s peeds for the Oregon DOT They found that small, but statistically significant changes in 8 5th percentile speeds were detected in conjunction with enhanced patrol operations at six enforcement sites Goldenbeld and Van Schagen (2005) examined Dutch enforcement in a rural providence over a five year period to conclude that there was a decrease in mean speeds and violators over time. It was noted that a public information campaign accompanied 1st year decreases and intensified enforcement accompanied 4th year decreases. One to two hours of selective enforcement per week were complimented by warni ng signs that remained continuously Australian research used stationary patrol vehicles on two study segments to measure the effect on vehicle speeds. Examining speed distributions using T test, one site showed reductions in speed while the other did not The site with reductions also showed some halo effect for two days in the morning commute, but not the afternoon. The author concluded that police presence may reduce urban speeds but discounts any general deterrent value (Amour 1986 ) Hauer and Ahlin (1982) conducted experiments to evaluate how speeds change after enforcement, when compared to a comparison roadway segment. The tags of passing vehicles were tracked to identify trends over several days, passing a stationary patrol vehicle, not engaged in actual enforcement. The average speed of vehicles was reduced at the site of enforcement, upstream, and downstream. The concept of a distance halo effect described vehicles slowing as they pass active enforcement and how far they travel before returni ng to original speeds, in this case degradation was cut


41 by half every 900 m. Another effect of enforcement is a time halo effect, wherein a motorist may alter their speed at the location of enforcement in subsequent days after observation, in this case be tween 3 and 6 days, depending upon repetition in the staged enforcement Sisiopiku and Patel (1999) used roving police patrols on Michigan roadway segments to validate a reduction in speeds where police are present, but they also found that a distance hal o is not supported since motorists resume travel speeds after passing police vehicles A series of spot speed studies in Saudi Arabia sought to evaluate posted speed limits, 85th percentile speeds, and the effect of police cars and automated enforcement si tes on speeds. A static police car was used to measure the effectiveness of potential enforcement. Although reductions in speed were noted at the location of the patrol car and camera installations, no downstream observations were made to determine if a halo effect was present (Al Gham di 2006 ) A series of studies in Norway evaluate the effects of police enforcement from a number of perspectives. Vaa (1997) enlisted officers to conduct enforcement on a 35km segment for a total of six weeks, bound by 2 and 8 week before and after periods respectively. Reductions in speeds of 1 to 4 km/hr were noted as well as a time halo of several weeks after the conclu sion of the enforcement. A separate effort varied the intensity of enforcement that concluded high i ntensity efforts were most effective at reducing speeds (De Waard and Rootjers ) A survey of Queensland drivers found that visible enforcement was a strong deterrent to speeding behavior. Interestingly, marked patrol cars in the traffic stream


42 tended to h ave more effect than a static patrol vehicle on the shoulder, though both were noted as effective (Soole Watson, and Lennon, 2009 ) Work zones present unique dangers, particularly where speed is concerned. Consequently, several studies unique to work zon e speed are wo rth reviewing. It has been found that stationary patrol cars, roving patrols, and officer flaggers were all effective means by which to reduce work zone speeding (Richards, Wunderlich and Dudek 1985 ) Similarly, Noel, et al. (1987) found the presence of a marked vehicle in the work zone was a good deterrent to speeding Medina, Benekohal, Hajbabaie, Wang, and Chitturi (2009) compared the downstream effectiveness of automated enforcement with speed trailers and patrol car presence with lights on and off. They conclude that camera systems were more effective at reducing mean speeds 1.5 miles downstream of the work zone. In a study of north Florida work zones, the FDOT found that their motorist awareness system, consisting of warning lights, spe ed signs, and motorist feedback, reduced speeds better than standard work zone signage, and an additional 2 to 3 m i/ h speed reduction was achieved when combine d with visible enforcement (Reddy and HNTB 2008 ). con junction with enforcement in Utah found reductions of 9 mph, versus 7 miles per hour without police presence (Saitom and Bowie 2003). The preponderance of research supports the value of visible law enforcement vehicles in work zones as a speed management strategy. Though there has been substantial research concerning the efficacy of enforcement as a speed control technique, little has been done to relate those effects to t he operation and capacity of freeway facilities The Highway Capacity Manual does not


43 consider the operational impacts of police traffic enforcement The HCM does mention enforcement tangentially as it relates to posted speed limits and the extent of enforcement as being a limitation of th e basic freeway methodology Table 2 1. Reductio ns of capacity in literature Author Shoulder of 3 Lanes Goolsby 0.67 Blumentritt, et al. 0.84 Smith, et al. N/A Prevedouros, et al. 0.83 Chin, et al 0.84 Knoop, et al. 0.72 Table 2 2. Values form the HCM proportion of capacity available under incidents Number of Directional Lanes) Shoulder Disablement Shoulder Accident One Lane Blocked Two Lanes Blocked Three Lanes Blocked 2 0.95 0.81 0 .035 0.00 N/A 3 0.99 0.83 0.49 0 0.17 0.00 4 0.99 0.5 0 0.58 0 0.25 0.13 5 0.99 0.87 0.65 0 0.40 0.20 6 0.99 0.89 0.71 0 0.50 0.26 7 0.99 0.91 0.75 0 0.57 0.36 8 0.99 0.93 0.78 0 0.63 0.41


44 Table 2 3. HCM Values for work zone flow rates with normal lane reductions noted State 2 to 1 3 to 2 3 to 1 4 to 3 4 to 2 4 to 1 TX 1,340 1,170 NC 1,690 1,640 CT 1,500 1,800 1,500 1,800 MO 1,240 1,430 960 1,480 1,420 NV 1,375 1,400 1,375 1,400 OR 1,400 1,600 1,400 1,600 SC 950 950 WA 1,350 1,450 WI 1,560 1,900 1,600 2,000 FL 1,800 1,800 VA 1,300 1,300 1,300 1,300 1,300 1,300 IA 1,400 1,600 1,400 1,600 1,400 1,600 1,400 1,600 1,400 1,600 1,400 1,600 MA 1,340 1,490 1,170 1,520 1,480 1,170 Def. 1,400 1,450 1,450 1,500 1,450 1,350 Table 2 4. Representative HCM capacity reductions values attributed to weather Type of Condition Intensity Reduction Average Reduction Range Rain >0 0.10 in./h 2.01 1.18 3.43 >0.10 0.25 in./h 7.24 5.67 10.10 >0.25 in./h 14.13 10.72 17.67 Snow >0 0.05 in./h 4.29 3.44 5.51 > 0.05 0.10 in./h 8.66 5.48 11.53 >0.10 0.50 in./h 11.04 7.45 13.35 >0.50 in./h 22.43 19.53 27.82 Temperature <50 F 34 F 1.07 1.06 1.08 <34 F 4 F 1.50 1.48 1.52 < 4 F 8.45 6.62 10.27 Wind >10 20 mi/h 1.07 0.73 1.41 >20 mi/h 1.47 0.74 2.19 Visibility <1 0.50 mi 9.67 One site <0.50 0.25 mi 11.67 One site <.25 mi 10.49 One site


45 Table 2 5. HCM capacities under varying environmental conditions Freeway Lanes Weekday or Weekend Daylight Dry Dark Dry Daylight Wet Dark Wet 6 Weekday 1,489 1,299 1,310 923 (% Change) (13%) (12%) (38%) 6 Weekday 1,380 1,084 1,014 (% Change) (21%) (27%) 4 Weekday 1,739 1,415 1,421 9131 (% Change) (19%) (18%) (47%) 4 Weekday 1,551 1,158 1,104 (% Change) (25%) (29%)


46 CHAPTER 3 Staging Traffic Stop Scenarios they be considered in any examination of potential impacts of enforcement on operations. An observational research component is well suited for this purpose. The objective is to observe a large sample of vehicles approaching a roadside traffic stop and measure compliance with the statutory requirements to 1) move over and/or 2) slow down. A staged roadside stop scenario, using an official police vehicle behind a civilian research vehicle, afford s the experimental control needed to min imize external factors. By using staged stop scenarios, a unique insight into the behavior of traffic approaching an emergency vehicle on the shoulder is made possible. The following sections describe the selection of roadway locations, surveillance and data recording techniques, and emergency lighting configurations associated with the designed st op experiment. Experiment Vehicles and Positioning Two vehicles are used in each simulated traffic stop, a civilian research vehicle and a fully marked Florida Highway Patrol cruiser. As research participants, the FHP has agreed to provide the research t eam with a marked cruiser and a uniformed member of the agency. The civilian research vehicle chosen is a sport utility vehicle that can comfortably accommodate a research assistant who will face oncoming traffic, observ ing passing vehicles and operate re cording (video and speed) equipment. A curtain or other shade is used to shield the research assistant from view, so that


47 passing motorists are not alarmed, mistake the researcher as pointing a weapon, or think that the officer in the simulated stop is in any danger. Location Selection Criteria The state of Florida is an excellent location for observational testing since it was among the first states to implement move over legislation The state has extensively publicized the law, has erected highway sign age, used promotional items, and has actively reinforced the law with targeted traffic enforcement. Since implementation in 2002, Florida officers have issued nearly 100,000 citations and countless warnings for vi olations. Figure 3 3 represents a chart t hat illustrates move over traffic citations in Florida and a complete list by county is included in A p pendix B. To select specific roadway segments and locations within the state, roadway geometry, the presence of intelligent transportation system (ITS) ar chitec ture, and geographic diversity are relevant. Freeway segments with two and three directional lanes in both rural and urban settings are desired. To minimize the impacts of interchanges, homogen e ous segments at least a mile from ramps are required. Since sight distance is relevant to perception reaction, no vertical or horizontal curves are present. The Florida Department of Transportation has extensively deployed Intelligent Transportation System (ITS) hardware on urban expressways in Florida. Hundreds of roadway cameras stream high quality images via a fiber optic network. The cameras are capable of being controlled remotely at traffic management centers operated jointly by the FDOT and FHP. Since most cameras are permanently mounted on polls high above the roadway, they offer a view of traffic far down the road. While the images afforded by these video cameras are not recorded by FDOT as a matter of practice,


48 they generously agreed to digitally record locations where staged stops would be co nducted as part of this research. ITS cameras and traffic detectors are needed to record passing traffic and monitor lane occupancy and speeds. Sites are free of work zone activity, incidents, and legitimate enforcement activity. Selecting s ites near Jac ksonvi lle and Orlando mitigates regional differences in driver composition that may be present. Figures 3 4 through 3 15 are aerial photographs and ITS camera images of the stop loc ations for Interstate 95 in Duval and Flagler Counties Interstate 4 in Volusia County y Each aerial photograph shows the roadway geometry, proposed location of the traffic stop vehicles, location of the ITS camera, and the location of the TSS detector for opposite travel direction veh icles. Note that the still images used to illustrate ITS camera views were captured from web site thumbnails a ctual video images are of significantly higher quality. Another feature of site selection involves traffic volumes. Traffic count information ob tained from the FDOT indicates an annual average daily traffic (AADT) v ehicle count ranging from 31,900 to 87,771 (FDOT, 2009 ) Traffic volumes are beneficial to estimate move over behavior in a variety of densities so that available gaps can be accommodat ed in the observations. Table 3 1 represents the simulated traffic stop sites and the accompanying vehicle counts in AADT and DDHV at the location Surveillance and Data Recording Traffic passing e ach staged stop location will be video recorded at the FDO T Traffic Management Center (TMC) for later analysis Additionally, a research te am member will use a laser device to measure and log the speeds of passing vehicles. Where possible, data from FDOT traffic detectors will be reviewed using the TMC


49 SunGuide software. The following sections describe the active surveillance techniques used The research team, through the partnership with the FDOT, is able to digitally record a live video stream from the camera nearest the stop. Ahead of the recording, the res earch team consulted with the TMC operators to arrange camera angles that maximize the upstream view of approaching traffic. A researcher in the rear of the research vehicle use s a Kustom Signals ProLaser III, a laser speed measuring device to m easure the lowest speed of each vehicle passing the stop in the outbound lane. The laser is tripod mounted for the comfort of the user, and s ince a laser cannot operate through automotive glass, the researc h vehicle must have a rear window that can be low ered or raised or the vehicle must be altered in such a way that the laser can operate without interference from the rear glass Concealment of the researcher by means of a curtain or other device is necessary to prevent alarming passing motorists who may perceive a threat to themselves or the officer in the simu lated stop. The researcher record s speeds using a log form that captures the date, time, location, and police vehicle emergency lighting configuration being used. The log contains periodic time stamps and /or vehicle descriptions for later matching to the ITS video if necessary. The log form is included in Appendix B Each TMC in Florida uses a custom software application called SunGuide to manage the ITS systems and data. T hrough the software, real time and historical traffic detector information is available via a variety of customizable reports. For purposes of this experiment, traffic volumes and speeds, by lane, are reported at 15 minute intervals


50 for the time period one hour before the de ployment through the end of the staged stop experiment. Three separate detector s tations are polled for volume and speed data, the one closest to the stop location, and the nearest upstream and downstream stations Patrol Vehicle Lighting Considerations P olice vehicles on the shoulder with red and blue lights activate d garner the attention of every passing motorist. Since much of the research on emergency vehicle lighting is well over thirty years old, t here has been debate recently concerning the types of lights, color of lights, and intensity during day and night conditions (Flannagan, Blower, and Devonshire, 2008) The basic intent of emergency vehicle lighting is conspicuity, warning motorists to clear a path while the vehicle is moving, or to be aw are when stationary. Those same red and blue lights can also be distracting to vehicles. For years, many have wondered if drivers are somehow attracted to emergency lights, causing them to unconsciously driv e their vehicle toward the light of a stopped patrol car resulting in a crash ot h effect has been debated, but no conclusive evidence exists to support the theory (FEMA 2009 ) T he mere may exist has made some re consider if more is truly better where emergency lights are concerned (Ashton 2006 ) T he involvement of rear end crashes involving stopped emergency The typical traffic stop involves the officer activating red and/or blue emergency lights located on the roof of the patrol have the unique agency color and markings, but use less conspicuous lights mounted in


51 the grill or interior of the patrol car. Unmarked patrol cars typically have these covert style emergency lighting devic es as well. The variety of light types, configurations, number, and colors used by police nationally are countless. A technical evaluation of emergency lighting is beyond the scope of this study, as is a behavioral analysis of driver perceptions of emerg ency lighting. With that said, observed differences in driver compliance with move over laws is possible by varying the color and configuration of emergency lighting devices on the stopped police vehicle. Since the Florida Highway Patrol is the primary tr affic law enforcement agency on Florida freeways, they are key partners in this research project. Coincidentally, the agency has also been a national leader in design, testing, and implementation of emergency lighting equipment (Wells 2004 ) The lighting devices used by the FHP provide an excellent foundation for evaluation of the application of various lighting techniques and their general effects on approaching traffic. All marked FHP vehicles use a combination of blue and red LED lights facing the fr ont rear and side of the vehicle. Most of these lights are roof mounted, while a part of the marked vehicle fleet is a slick top configuration with the colored LEDs along the top inside front and rear windows. Slick top vehicles also have side facing the vehicle B pil l ar or C pillar The intensity of the lights are altered, based on the available ambient light, by means of a photo cell. An alternating pattern of red and blue are used to maximize both colors and that patter n is different fo r vehicles in motion and vehicles in park. All marked vehicles have a separate bar of amber lights in the rear deck win dow that can sequence to flash on and off, move to the left, move to the right, or move from the center outward.


52 For this study, the lig hting patter n will be adjusted to determine if different light usage has an effect on passing vehicles. Three patterns will be used for vehicles approaching the rear of the vehicle, emergency top lights only (A) emergency top lights plus amber directiona l lights (B) and amber directional lights with no emergency top lights (C) During each deployment of the staged traffic stop scenario, each of these configurations will be used. Figures 3 14 through 3 16 are photographs of each emergency light configura tion in operation. During t he duration of each staged stop the tree lighting configurations were each used During the first portion of deployment, the emergency top lights only used then the amber directional arrow added, and finally, just the amber dir ectional arrow used. When emergency top lights are in use, those lights facing the front of the patrol vehicle are generally visible to motorists traveling in the opposite direction. Because of onlooker or rubbernecker effects, the impacts of traffic sto ps are potentially not limited to the same direction of travel as the stop. The effect of stops on opposite direction traffic is another aspect of research that can be explored with the simulated traffic stop scenarios. Since FHP emergency lights on all vehicles (marked, slick top, and unmarked) can be reduced to only rear facing lights, a comparison of opposite direction impacts is possible. A portion of each experiment will use front facing emergency lights There is no legal mandate for drivers approaching from the opposite direction to move over or slow approaching a stopped emergency vehicle, but a rubberneck effect is intuitive. Since simulated stop locations are chose n near FDOT detectors, the mean speed of opposite direction vehicles can be compared when front facing lights are used


53 and when they are not used. Changes in mean speeds can be used to estimate the impacts of stops on opposite direction operations. Forward facing emergency lights will be reduced during a portion of each simulated traffic stop. The times associated wit h front facing light use are logged by the research team so that opposite direction detector informat ion can be reviewed. Figure 3 17 shows a photograph of emergency lights activated, with the forward facing LEDs not used. Experimental Design Six separate deployments of the simulated or staged traffic stop were used for this research project. Freeway segments with two and thr ee directional trav el lanes were selected, and at each stop location, three separate lighting configurations tested. The first deployment was conducted on June 1, 2011 to ensure that the experimental design was sound and the remaining deployments conducted in the following 6 weeks All aspects of the data collection worked as planned. Using the AADT counts for each test locatio n, a suitable sample size was estimated, using a confidence level of 95%. Table 3 2 shows the number of data collection events number of lanes, l ighting configurations and estimated samples based on directional AADT. Analy sis of Simulated Stops/Move Over Behavior A variety of lighting configurations are used during each simulated stop deployment, to include emergency lighting, emergency plus directional amber, and directional amber only. During each phase, traffic in th e outside lane of travel can be observed from overhead ITS cameras to gauge early mer ge, late merge, or no merge, and gap acceptance. Additionally, each n on measured by


54 a rear facing observer to record the lowest vehicle speed before passing the stop location. Based on the data gathered in the observational stu dy, frequencies anal ysis a nd contingency tables provide insight into driver move over behavior. By varying emergency lighting configurations, inferential statistical techniques can determine which is most effective in supporting move over behavior. Op posi te direction traffic is monitored using vehicle detector equipment to compare mean speeds when drivers are faced with forward facing emergency lights versus when those lights are not used. This opposite direction of travel rubbernecker component of the st udy can help determine the extent of rubbernecking in stop scenarios. Evaluate Video at Staged Stops ITS cameras provide high quality video from a good vantage point to view the behavior of cars approaching stop locations. With the ability to zoom and adju st camera angles remotely, these devices can record traffic approaching a stop from several hundred feet upstream of the staged stop. The video obtained for each simulated stop was viewed several times. During the first viewing, accurate counts of vehicle s vacating the right lane upstream of the stop were made. During the second viewing, the proximity of the merge maneuver was classified, to determine the number of late merges. The final viewing of the video allowed the researcher to evaluate available g aps in instances where the right lane vehicle did not execute a lane change. Video observation of mo ve over behavior was tabulated and is presented in the following sections.


55 Frequencies for outside lane volumes During the staged roadside stop experiments 9,051 right lane vehicles were observed approaching. Just under 75 percent, 6,762 vehicles, vacated the lane upstream of the stop location. Of the remaining vehicles, 132 slowed the requisite 20 miles per hour below the posted speed limit. The moving and slowing vehicle combinations made for an overall move over compliance of 75.9%. When breaking down the frequencies by lighting configuration, there were 2,869 vehicles passing under lighting configuration A, 3,243 under lighting configuration B, and 2, 849 under lighting configuration C. Vehicles moving to an adjacent lane were noted as 2,239, 2,605, and 1,918 for each respective lighting configuration. Slowing vehicles were distributed fairly evenly. The overall compliance percentage (combined move o ver and slowing) for lighting A and B were both 79.7 percent, and lighting C dropped to 68.8 percent. While summary information in some respects, the lane frequencies are also important for statistical analysis that follows. A tabular presentation of outs ide lane volumes approaching staged stops, counts for move over and slow down behavior, and assessments concerning overall compliance are included in Table 3 3 Statistical significance of emergency vehicle lighting While overall move over compliance is t he essence of the research objective, considering the influence of different emergency vehicle lighting configurations is beneficial. Practitioners hold differing views on the appropriate use of emergency lighting while at roadside, with some maintaining that reducing the use of red and blue lights minimizes traffic impacts, and others asserting that red and blue lights are


56 essential to safety. Examining move over compliance across different lighting configurations may help answer the question. Frequenci e s for move over compliance can be tested statistically, given the variable of lighting type used on the emergency vehicle. Lights can be set to A red/blue, B red/blue with an amber directional arrow, or C an amber directional arrow with no red/blue lights The statistical significance of the different ligh t configurations is achieved by comparing the means in a paired t test There is no statistical significance between using red/blue and adding the amber directional arrow, however compliance drops off in a statistically significant way when just the amber arrow i s used, as depicted in Table 3 4 Using red and blue emergency lights increases move over compliance by 16% over using the amber directional light by itself. The increase in compliance is statist ically significant and should serve as guidance for practitioners. Description of early/late merge (lane change) Lange change modeling is predominantly used in microscopic simulation because of the impact that those movements have on traffic flow (Ben Ak iva, Choudhury, and Toledo, 2006). The models can classify changes in lane as discretionary or mandatory er perceives a lane offers better conditions, the change is a discretionary lane change (DLC) (Ben Akiva et somewhere between the two and therefore an interesting study. Thou gh legally required, like most laws its compliance hinges on a variety of factors such as knowledge of the law, level of enforcement, and driver behavior. For purposes of this research, an


57 imaginary line approximately 500 feet upstream of the staged stop was used to differentiate early and late lane change or merge. That point was calculated using the known distance between lane markings and transferred to the video monitor during playback Outside lane drivers were observed making lane changes well upst ream of the staged stops, as much as an estimated half mile. Drivers appeared to model the behavior of others and lane changes often appeared to cascade upstream. Red and blue lights were effective in communicating warning to drivers, as 93.6 and 92.9 pe rcent of merges were executed upstream of the 500 foot delineation for lighting configurations It is not known if the color of the lights differentiated the roadside event or if the distance of visibility in the light spectrum pl ayed a role in driver perception. The number of late moves was noticeabl y higher when red and blue emergency lighting was not in use A late merge was twice as likely to because of the pr oximity to the roadside stop, guidance to practitioners is again evident Table 3 5 shows the occurrence of late merge maneuvers. Assessment of available gaps and driver behavior Lane changes are predicated upon available space in the target lane, also called a gap. The smallest acceptable gap is also referred to as the critical gap (Ben Akiva et al., 2006). Most examination of gap acceptance relates to intersection and merge locations, but some research has applied the concepts to lane change scenario s. The space between the subject vehicle and the vehicle ahead in the target lane is called the lead gap and the space between the vehicle and the one behind in the target lane is


58 called the lag gap (Ben Akiva et al., 2006). tendant speeds of the vehicles, including their own, in relation to the space forms the basis for determining the smallest acceptable gap, or critical gap. In order to understand move over behavior, it is necessary to evaluate available gaps as a potential explanatory factor for non move over behavior. During four deployment s video was evaluated to determine if a suitable gap was available to the driver who did not move over Table 3 6 represent the available gaps, and it reflects that gap availability/acceptance was not a factor in move over behavior across the d ifferent locations In almost all cases there was an opportunity to execute a lane during the approach to the roadside stop. One might suspect that during more congested conditions, there may be fewer available gaps, i mpacting move over compliance, but for purposes of this research, it was not a factor. Conversely, and anecdotally, many successful merges were observed where the critical gaps appeared insufficient. Forced merges ap peared to occur quite often, but due to the angle of the ITS cameras, and upstream perspective, quantifying forced merges would prove to be quite subjective. The ITS video obtained from the Flagler and Volusia locations was suitable for obtaining move over counts, but not of sufficient quality for making objective assessments concerning gap availability, therefore those locations were not included in the gap analysis. Introducing a new mov e over decision making model ion to change lanes can be illustrated with a flowchart depicting the input, processes, decisions, and actions involved. Like a decision tree, branches of the flowchart represent alternative actions for the driver who


59 is approaching the roadside traffic s top, that culminate in the decision to move over, slow down, do some combination of the two, or neither. The observation of more than 9,000 right lane vehicles passing staged stops allows one to conceptualize the driver decision making process. The observ ation of a police vehicle on the shoulder of a roadway ahead provides the initial stimulus for the move over decision. How that stimulus is cognitively aw, is the foundation of the model. Two fundamental branches form upon this observation, an awareness that a move over situation exists, and a If the driver understands the legal obligation of a move over scenario, they immediately begin assessing available gaps in adjacent lanes. The essence of most move over laws is a requirement to vacate the lane adjacent to the emergency responder. For the right lane driver approaching a right shoulder s top on a mul t ilane facility, a move would be in order to comply with the law and an evaluation of available gaps ensues. As mentioned earlier, the critical gap is the smallest acceptable gap for executing a lane change maneuver. If the available gap is greater than the critical gap, then a free merge occurs and the lane change is executed. If the available gap is less than the critical gap, the driver can either squeeze into the uncomfortable space in a forced merge, or elect not to make a lane change. The driver who stays in the right lane typically slows as they pass the roadside stop, which is explored further in subsequent sections.


60 Upon initial observation of the police activity ahead, the driver may fail to process the stimulus as an obligation to act. This may stem from a lack of knowledge of the move over law, a lapse in recollection of the law, a distraction, or some other cognitive failure. If the driver fails to process the stimulus as an obligation to act, they still may process the roadsid e traffic stop as a general hazard. The secondary processing of the roadside stop involves the approaching driver making an assessment about the situation in general, in which case they may perceive a general hazard and slow, or not perceive any hazard an d continue status quo. The modeling of move over behavior cannot possibly include all driver environment interactions, but hopefully illustrates the process for the typical driver. Such a model can be used for future research and application to micro simu lation. Figure 3 18 is a graphic representation of a general move over model. Vehicle Speed Evaluation Speed is a fundamental operating parameter for freeway facilities. Changes in speed at stop locations can aid in the understand ing of how those events affect operations. The speed of vehicles that do not move over, the average lanes speeds at the stop location, and the average lane speeds of opposite direction lanes are all important to a complete understanding of stops affect speeds. Examine s peeds of v m ove o Since most move over laws require drivers to slow down when they cannot vacate the lane on multi lane facilities, the speed of passing vehicles in the outside lane is very important to gauge compliance and develop models. T he speed of each vehicle passing the staged traffic stop in the outside lane of travel is recorded using a laser speed measuring device by a researcher located inside one of the staged stop vehicles.


61 A distribution of vehicle speeds was created for all v ehicles passing the staged s top s to measure mile per hour speed reduction. Table 3 7 shows the average speeds of right lane vehicles that did not move over, as captured by laser speed measurement during staged stops It should be noted that of the 2,289 vehicles that did n ot move over in data collection, only 132 vehicles, 5.8 percent complied with the 20 mile per hour speed reduction requirement of the Florida law. The majority of those cases were encountered at the I 4 Volusia location, where a congested condition occurred for a short time during the data collection. Technically speaking, those vehicle s complied with the required speed reduction, though it was likely the product of the congested condition By a ttr ibuting driver compliance with the slow down portion of the Florida move over law in these samples provides an estimate that is decidedly conservative. Speeds from roadway detectors Examination of the SunGuide software for the roadway detectors at the st op location s can shed additional light on the impacts of the police traffic stop. When 15 minute intervals are charted i n a time se ries, the average lane speeds at the detector locations illustrate that speeds at the stop location are noticeably lower during the stop. This indicates that a perception of the blue and red lights causes a speed reduction at the stop location detecto r, as evidenced by the drop in average speeds coinciding with the onset of the staged stop s Such a finding reinforces tradit ional beliefs that active traffic enforcement has a deterrent effect. Across all types of emergency lighting, the mean reduction in traffic speeds is 4.6 miles per hour, when compared with the mean speeds during the time intervals before the staged stop was initiated. Average lanes speeds for all locations drop from 69.7


62 miles per hour pre stop, to about 65 miles per hour during staged stops, and then back to 67.0 miles per hour immediately after the staged stops ended. Across six different locations a nd multiple 15 minute time intervals at each, the evidence is conclusive that traffic passing police enforcement generally slows. Figure 3 19 and Table 3 8 depict how vehicles passing a traffic stop generally slow. To frame the changes in average lane s peeds at the stop location, a table depicting the lane volumes for the time period is beneficial. Table 3 9 shows lane volumes for the same time increments that were used in the preceeding discussion on average lane speed s According to data from detecto rs at the stop locations, 22,339 vehicles passed during the pre stop and post stop conditions and 29,449 passed during the time when the staged stop was present for a total of 51,788 estimated vehicles Rubbernecking Impacts for Opposite Direction Traff ic Little research has been done to understand how police activity on the shoulder distracts motorists The impact on those traveling in the opposite direction of travel is even sparser Again, by changing the emergency lighting configuration on the patro l vehicle, the opposite travel lanes can either view red/blue lights, or those forward facing lights can be dimmed. Staged stop locations w ere chose n that include an opposite direction roadway detector that samples the opposite direction travel lanes at a location approximately parallel to the staged stop. Opposite direction s peeds under non stop, stop with forward facing lights, and those with forward lights dimmed can be interpreted for statistical significance The phenomena of on looker or rubbernec ker delays for traffic traveling past an incident in the opposite direction of travel is believed, but not well researched. A traffic incident with multiple responder vehicles, lane blockage, and queue formation would


63 likely be a spectacle worthy of distr action. A police enforcement stop at roadside is probably more innocuous though noticeable D etector speeds shows average speeds for opposite direction vehicles were 72.3 miles per hour before the staged stops began. During stops with forward lights en gaged, average speeds dropped a slight 0 .2 miles per hour to 72.1 mile s per hour. When the front lights were dimmed during the stop, the average lanes speeds in the opposite direction were back to 72.4 miles per hour. These measures would tend to reinfor ce a slight impact on opposite direction travel, a s well as the efficacy of reducing forward facing lights as mitigation for t hat impact, though n either a rubbernecker effect, nor reducing forward lights was statis tically significant. Table 3 10 shows the mean speeds, obtained for roadway detectors opposite staged stops, based on the presence of forward facing emergency lights. Figure 3 1. Photo of research vehicle and FHP vehicle in stop configuration (Photo courtesy of Grady Carrick )


64 Figure 3 2. Photo of the researcher using a laser based speed measuring device (Photo s courtesy of Grady Carrick ) Figure 3 3. Florida move over citations by year since inception 0 5000 10000 15000 20000 25000 30000 2002 2003 2004 2005 2006 2007 2008 2009 Total Citations


65 Figure 3 4. Simulated stop site A I 95 NB @ 295 mm, Flagler Co. (Photo courtesy of Google Earth ) Figure 3 5. Simulated stop site A ITS Camera Image (Photo courtesy of FDOT.)


66 Figure 3 6. Simulated stop site B I 95 SB S. of Old St. Augustine Rd, Duval Co. (Photo courtesy of Google Earth.) Figure 3 7. Simulated Stop Site B ITS Camera Image (Photo courtesy of FDOT.)


67 Figure 3 8. Simulated Stop Site C I 95 SB N. of Pecan Park, Duval Co. (Photo courtesy of Google Earth.) Figure 3 9. Simulated Stop Site C ITS Camera Image (Photo courtesy of FDOT.)


68 Figure 3 10. Simulated Stop Site D Turnpike SB @ 2 84 mm Lake Co. (Photo courtesy of Google Earth.) Figure 3 11. Simulated Stop Site D ITS Camera Image (Photo courtesy of FDOT.)


69 Figure 3 12. Simulated Stop Site E I 4 EB @ 124.4 mm, Volusia, Co. (Photo courtesy of Google Earth.) Figure 3 13. Simulated Stop Site E ITS Camera Image (Photo courtesy of FDOT.)


70 Figure 3 14. Image of emergency top lights only in operation configuration A (Photo courtesy of Grady Carrick.) Figure 3 15. Image of emergency top lights plus amber directional arrow configuration B (Photo courtesy of Grady Carrick.)


71 Figure 3 16. Images of an amber directional arrow in right to left operation configuration C (Photo s courtesy of Grady Carrick.) Figure 3 (Photo courtesy of Grady Carrick.)


72 Figure 3 18. Move Over Lane Changing Model Figure 3 19. Average lane speed from roadway detectors 61.0 62.0 63.0 64.0 65.0 66.0 67.0 68.0 69.0 70.0 71.0 Pre-Stop Lighting A Lighting B Lighting C Post Stop AVERAGE LANE SPEED STOP LIGHTING CONDITION


73 Table 3 1. Volumes at simulated stop locations Location AADT Study Direction K Factor D Factor DDHV (AADT)(K)(D) Site A 45,500 23,500 10.12 54.81 2524 Site B 87,771 44,093 9.54 52.97 4435 Site C 61,160 30,711 10.78 53.58 3533 Site D 31,900 15,950 11.01 55.93 1963 Site E 56,000 28,000 8.65 53.65 2599 Table 3 2. Data collection event summary Number of Events No. Lanes Lighting Sample Size 3 3 A,B,C 390 3 2 A,B,C 390


74 Table 3 3. Outside lane frequencies I 95 OSA I 95 OSA Turnpike I 95 Flagler I 4 Volusia I 95 Pecan Park Total Overall (veh/h/ ln) 851 1120 443 852 799 687 Right Lane Vehicles 1705 2149 1414 1091 1816 876 9051 Move Over Vehicles 1256 1472 1104 974 1170 786 6762 Slowing Vehicles 8 13 10 1 99 1 132 Total Compliance 1264 1485 1114 975 1269 787 6794 Overall Compliance % 74.1 69.1 78.8 89.4 69.9 89.8 75.9 Lighting A (veh/h/ln) 768 883 437 599 704 623 Right Lane Vehicles 525 588 382 369 669 326 2859 Move Over Vehicles 407 439 305 343 446 299 2239 Slowing Vehicles 4 1 4 1 29 1 40 Total Compliance 411 440 309 344 475 300 2279 Overall Compliance % 78.3 74.8 80.9 93.2 71.0 92.0 79.7 Lighting B (veh/h/ln) 827 1096 443 780 799 637 Right Lane Vehicles 582 743 663 3 02 683 270 32 43 Move Over Vehicles 462 522 536 2 68 466 251 2605 Slowing Vehicles 4 11 1 0 45 0 61 Total Compliance 466 533 537 2 68 511 251 2666 Overall Compliance % 80.1 71.7 81.0 88.7 74.8 93.0 79.7 Lighting C (veh/h/ln) Right Lane Vehicles Move Over Vehicles Slowing Vehicles Total Compliance Overall Compliance % 851 598 387 0 387 64.7 1120 818 511 1 512 62.6 441 369 263 5 268 72.6 797 320 263 0 263 82.2 765 464 258 35 293 63.1 687 280 236 0 236 84.3 2849 1918 41 1959 68.8 Lighting configuration A Emergency blue and red, B Emergency blue and red plus amber directional arrow, C amber direction arrow only. Veh/h/ln measure was averaged over the entire deployment which lasted up to three hours.


75 Table 3 4. Move over compliance and lighting configuration paired sample test t df Sig. (2 tail) Mean Std. Dev Std. Error Mean 95% Confidence Interval of the Difference Lower Upper Pair A B .316666 2.4733 1.0097 2. 9123 2.2789 .314 5 .766 Pair A C 10.1166 2.5023 1.0215 7.4906 12.7427 9.903 5 .000 Table 3 5. Observed percentage of late merge maneuvers. Lighting I 95 OSA I 95 OSA Turnpike I 95 Flagler I 4 Volusia I 95 Pecan Park Average A 8.8 7.7 7.5 2.9 7.4 3.7 6.4 B 10.0 10.0 6.9 6.3 5.2 4.4 7.1 C 18.9 16.0 20.2 13.3 12.0 7.2 14.6 Average 12.6 11.3 11.5 7.5 8.2 5.1 Table 3 6. Gap analysis Non Move Over Vehicles I 95 OSA Turnpike I 95 OSA I 95 Pecan Park Row Total Gap Available 413 301 629 87 1430 Percentage 91.98% 97.10% 92.91% 96.67% 93.71% No Gap 36 9 48 3 96 Percentage 8.02% 2.90% 7.09% 3.33% 6.29% Total Non Moving 449 310 677 90 1526 Percentage 100% 100% 100% 100% 100% Table 3 7. Average speed of vehicles that did not move over. Lighting I 95 OSA I 95 OSA Turnpike I 95 Flagler I 4 Volusia I 95 Pecan Park Mean Speed Reduction A 62 63 62 63 59 63 8.2 B 61 61 62 67 61 65 7.2 C 62 62 62 68 59 67 6.6 Total 62.0 62.1 62.0 65.7 59.7 64.8


76 Table 3 8. Average lane speeds from detectors at staged stop locations I 95 OSA I 95 OSA I 4 Volusia I 95 Flagler I 95 Pecan Park Turnpike Avg Speed Pre Stop 70.3 70.9 69.7 74.6 71.7 61.0 69.7 Lighting A 69.3 68.8 62.3 71.2 68.0 55.3 65.8 Lighting B 69.6 67.3 61.4 70.3 66.0 53.1 64.6 Lighting C 69.8 67.4 62.4 71.5 66.8 51.0 64.8 Post Stop 71.9 68.4 63.2 73.6 69.7 55.3 67.0 Table 3 9. Traffic volumes at stop locations I 95 OSA I 95 OSA Turnpike I 4 Volusia I 95 Flagler I 95 Pecan Park Row Total Pre Stop 2099 2055 987 1240 1597 1144 9122 Lighting A 2167 2339 955 1138 1397 1178 9174 Lighting B 2444 2990 1345 1340 1744 992 10855 Lighting C 2411 3432 640 862 1234 841 9420 Post Stop 2502 3837 736 1460 2453 2229 13217 Table 3 10. Opposite direction mean speeds under stop conditions I 95 OSA I 95 OSA I 4 Volusia I 95 Flagler I 95 Pecan Park Turnpike Avg. Speed Pre Stop 69.8 74.8 74.4 73.8 73.6 67.7 72.3 Forward On 69.4 74.1 74.1 74.2 73.8 67.0 72.1 Forward OFF 69.8 73.7 73.7 76.3 72.8 68.0 72.4


77 CHAPTER 4 ESTIMATING TRAFFIC FLOW IMPACTS AT STOP LOCATIONS Roadways are characterized by a great deal of variability in their traffic stream. Traffic stops occur thousands of times each day on the roadways around the United States, and because the day, time, location, and circumstances of each stop are slightly different, a great deal of variability is also p resent where stops occur. To analyze the behavior of traffic at stop locations requires a data set that is large enough to allow the specific impacts due to the enforcement stop to be isolated from the impacts due to a variety of other factors Incident re search typically involves a few hundred event observations, at best, which was viewed as inadequate for the task at hand. For this project, the number of data points necessary would likely have to be in the thousands, to account for the measures of traffic impacts that were anticipated a priori. Modern transportation and law enforcement operations leverage automation and data collection, where those data were previously less accessible. Greater reliability is available in the temporal and spatial accuracy, timely collection, and long term storage of traffic flow and traffic stop activities. The data required for studying historical traffic stops are traffic flow data, provided by the Florida Department of Transportation (FDOT), and enforcement stop data, pr ovided by the Florida Highway Patrol (FHP). speed, volume, and density of traffic at specific points along the roadway. About two thousand detector locations obtain and store this information at intervals as frequent as every 20 seconds, 24 hours per day, 365 days per year. When one considers that FHP troopers make many stops near these detector locations, there is the potential to relate the two sets of data. Mapping stop location s to detector locations, and subsequently


78 mapping stop times with detector measurement times enables us to obtain traffic information before and during those stops. The following sections describe how the police stop data and traffic detector data are obt ained and filtered for analysis. The study area for this research centers on eight counties in northern, central, and south Florida. All of the roadways are classified as part of the U.S. Interstate Highway System, and there is a mixture of mainly two and three directional lane freeway segments. All of the highways are divided by barrier wall or guardrail in the urban areas, and by grass median or barrier in the rural areas. A total of approximately 324 miles of freeway are part of this study, encompassing 1,186 roadway traffic detectors. The Florida Highway Patrol is the law enforcement agency with primary jurisdiction for patrol, enforcement, and crash investigation on all of these roadways. Figure 4 1 is a map of the study area with counties and roadways highlighted. The subsequent sections describe the data collection, synthesis, and analysis used to examine historical traffic stop data. This chapter identifies how the two principle data sets, police traffic stop and roadway detector data, are collected related to each other, and finally analyzed. Historical Stop Data Collection Collection of historical stop data involves collecting police traffic stop data, collecting freeway traffic data, and spatially relating those sets. The following sections descr ibe the collection of data needed for this project. Collecting Police Traffic Stop Data The Florida Highway Patrol (FHP) computer aided dispatch (CAD) system is the principle records management system (RMS) for the agency. The system provides both live an d historical information for agency activities, trooper activities, and incidents. The


79 system facilitates dispatcher entry of telephone calls for service from the public and from other agencies. Troopers in the field communicate with the CAD system via m obile data computer (MDC) and via voice communication with dispatch personnel. Activities of troopers are systematically captured and stored in the CAD system as incidents. Incidents may be traffic crashes, arrests, vehicle tows, traffic stops, or a mult itude of other types of calls for service. Each incident record contains dozens of data fields and scores of attributes. Important date/time stamps are attached to com pleted. Each record also contains specific location data, indicating where the incident occurred, typically on a street or highway for FHP personnel. The location can be entered as a text string by the dispatcher or trooper, which is verified with the CA D system and longitude. When a trooper self initiates a CAD call via the MDC, the dispatche r can accept the GPS coordinates and they apply to the incident record. It is the policy of the FHP to log all interstate traffic stops via the CAD. Figure 4 2 depicts a sample input screen for a trooper creating a traffic stop incident and Table 4 1 show s the data format for the FHP CAD system for traffic stop data. Collecting Florida Freeway Traffic Data Many freeways in Florida are instrumented with vehicle detectors (typically inductance loops or remote traffic microwave sensors (RTMS)) that collect la ne specific traffic information (speed, occupancy, volume) on a continuous basis. Th ese data is stored in the respective FDOT district SunGuide software system.


80 The Statewide Transportation Engineering Warehouse for Archived Regional Data (STEWARD) at the University of Florida Transportation Research Center is a central data warehouse (CDW) for the traffic information that is collected by the thousands of roadway detectors around the state. The archived data provides an excellent resource for the study of traffic occupancy and speed. For this project, 1,191 individual traffic detector stations have been identified on study segments in the eight Florida counties. Given the study period, the rate of sampling, and the number of detector stations, millions of data points for traffic data are available. Since the data are stored in an Oracle database, it readily lends itself to query using a structured query language S QL. The station level traffic sensor subsystem (TSS) data format is listed in Table 4 2. Spatially Combining Data Sets The fundamental analysis for this project involves observing the behavior of the traffic stream at the time and place of police traffic s tops. To accomplish this, the location of the enforcement stops must be matched to the location of traffic detector equipment. This spatial referencing is best accomplished in a geographic information system (GIS) environment. The result will reduce the traffic stop data set to only those events that occurred within study segments, identify the detector nearest the stop, identify if the nearest detector is upstream or downstream, and identify the distance to the detector. Appendix C contains a detailed s et of steps used to join the data sets in the GIS environment.


81 Traffic Stop Data Collection and Redaction Results The traffic stop and traffic detector tables were combined spatially to produce a tabular and graphic representation of the data set. When ap plying the directional split to the study segments, separate road segments are created. The directional road segments contain individual stop records, referenced to the detector identifier for that same direction of travel. Combined records are stored in four separate directional (N/S/E/W) tables and shape files. Table 4 3 shows the combined data elements after the spatial join. An important characteristic of the combined table is the location of the stop in relation to the detector. The roadways were su b segmented every 200 meters, and the stops (ID and NAME ) and detectors (ID _1 and NAME_1 ) both plotted within those sub s egments. This allows for offset distance and direction (upstream or downstream of detector) to be deduced and captured as new data ele ments RDLOC_DIFF and STREAM. Equation 4 1 is a representation of the relationship produced. ID minus ID_1 = RDLOC_DIFF (4 1 ) Where: RDLOC_DIFF is ( ) = Downstream RDLOC_DIFF is (+) = Upstream RDLOC_DIFF is 0 = Same segment Figures 4 3 through 4 10 are graphic examples of the GIS fusion of stops and detectors for each of the eight Florida counties in the study area. It should be noted that mapped stop points often represent multiple stops at the same location. No symbology for frequency is neces sary since the maps are only intended to illustrate the relationship between stop and detector locations. Since some of the detail is lost because of scale, F igure 4 11 is provided to show the detail that is present in the GIS product.


82 Computer aided disp atch records for 252,956 FHP Interstate traffic stops occ urring between July 2010 and September 201 1 were obtained from the agency Since these records reflected the entire state, the initial filters were county, roadway segment description, and finally spatially with roadway segments. The red uction process left 35,101 candidate records, to be moved into the GIS processing. Initial visualization in GIS eliminat ed stops where there was an apparent incongruence between the GPS point of the stop and the ro adway description input by the officer. This occurs very frequently in interchange areas where the violation was observed on the freeway, but the stop occurred on an exit ramp or immediately off the facility. This was found to be very common in urban are as. Since the objective is to estimate the impacts of the net result of the stop data reduction was 21, 14 4 traffic stop events on instrumented facilities in the eigh t county study area. Because malfunctioning detectors, missing warehouse data, and other types of attrition were anticipated, the goal was to have an initial data set that could withstand reduction a nd still render a suitable size sample The 21,144 reco rds moved into the next phase of data collection, mining and assembly of CDW data. Traffic Data Mining and Assembly The product of the GIS fusion of the detector and traffic stop data is a spatiotemporally related data set. Because the stop data is spatia lly linked to detectors, the CDW can be queried to extract the traffic data at that detector, before, during, and after the traffic stop. The essential variables for query of the CDW are TSS Station ID, Date of Stop, and Time of Stop. From these paramete rs, the query is constructed to mine traffic information.


83 Due to the volume of data contained in the CDW, all data for the dates of interest Due to the size of the data set separate files were created for each month, by FDOT district. The result was 60 separate files that would ultimately expedite the search and retrieval of data. The methodology for mining traffic data and the assembly of the data set is set forth in the following sections. Data Mining Objective and Methodology Most incident impact analyses compare incident and non incident conditions at locations, to interpret changes as incident impacts. Similarly, comparing the traffic behavior before the stop and the traffic behavior during the stop potentially sheds light on the influence of the traffic stop on passing traffic. Enforcement stop conditions are those captured by the start and stop time stamps in the FHP computer aided dispatch system. Recall that the se time stamps are created by the officer at roadside, using a mobile data computer, and are therefore considered the most valid representations of the stop timing possible. Identifying non stop conditions must attempt to minimize influences of weather, i ncidents, and work zones, while attempting to model similar traffic volumes present. By obtaining the non stop samples from the same detector, immediately preceding the stop, the most reliable similarity is possible. Such a methodology minimizes the influ ence of weather, incidents, and work zones. The objective of traffic data is to obtain the s peed and volume at a one minute resolution for the time before during and after every stop event. To accomplish this, scripts were written to query the CDW data. While the GIS fusion rendered 21,14 4 candidate stops, it was anticipated that missing data, inoperative detectors, and other factors would cause significant attrition. Fortunately, d ata for just over 15,000 valid


84 records were found in the CDW, and moved to secondary processing and data assembly. Data Assembly With valid data for a portion of the stops available from the CDW, processing steps involved event de confliction, date de confliction, and final data scrubbing. Event de confliction requires that we ensure that there is not another stop occurring during the sampling period at the same detector location A script was constructed to identify where such conflicts were present and those 900 records were flagged and later rem oved from the data set. Date de confliction involves those stops that start on one day, immediately prior to midnight, and subsequently end on the following date. This conflict complicates data collection, since the CDW stores data by date. There were 4 99 cases where a date conflict existed, and those instances were flagged and also removed from the data set. A review of the remaining cases removed 26 that had missing speed data, and 472 that had volume data that were above the theoretical capacity of a facility. The net result o f all data processing was 13,416 valid stops, complete with the added variables of speed and volume from the CDW. In t erms of incident studies, 13,416 data points compares very favorably but likely necessary due to the inherent variability of traffic streams and the expectation that shoulder/stop events may impact traffic st reams in subtle ways. Table 4 4 shows a comparison of sample sizes in incident studies. The principle data mined from the CDW was a one minute resolution of speed and volume from the detector identified as nearest to the location of the stop. Using scripts, the process was automated to obtain the date, time, and CDW station identifier from the stop data, and subsequently retrieve the speed and volume data fr om the CDW. A


8 5 sample of these one minute time slices covered each stop record, 20 minutes before the start of the stop time stamp, t 1 to 30 minutes after the conclusion of the stop, t 2 From that isolation of data points, the necessary resolution could be collected. The duration of a police enforcement action varies greatly, depending on the officer, the nature of the stop, and the complexity of the situation. For this reason, the duration of at stop conditions are aggregated to the actual number of min utes, up to 15 minutes. Given this method of aggregation, the mean stop duration was 9.4 minut es. T he non stop condition was aggregated at a 5 minute resolution, commencing at 20 minutes before the time of the stop Some separation between the non stop a nd at stop periods is desirable, given the fact that t he exact stop time is in actuality somewhat uncertain The officer pull over maneuver and officer safety aspects at the initial moments of a stop can vary by officer and/or circumstance ( e g ., some officers may make a vehicle approach before initiating the stop on their mobile computer). F igure 4 13 represent s the sampling methodology. The key CDW variables are speed and volume at the location of the stop. Vehicle speeds were obtained at a 1 minute resolution from the CDW and aggregated to the required resolution by averaging the data points. The average volume for the sample period was presented as a 1 minute average From those data points, new data fields were aggregated for non stop and at sto p conditions The variables Speed Non Stop and Speed At Stop along with Non Stop Volume and At Stop Volume were created and added to the data set.


86 The calculated volume of the segment was derived by mul tiplying the volume average by 60 and dividi ng the result by the number of lanes at the location, to achieve an estimate of veh/ h /ln Because t he original data set provided the date, time, latitude, and longitude, a calculation of historical sunrise and sunset are possible with calculations. Once sunrise and sunset are calculated the period of twilight can be derived as the time thirty minutes before and after both sunrise and sunset The ambient lighting conditions of night and day were calculated for each record, with twilight being included i n the night category Figure 4 12 is a graphic representation of how the stops are distributed among those lighting conditions. Day of the week, year, month, calendar day, hour, and minute were all variables created from the date and time for the start of each enforcement stop. The remaining variables were carried forward from the GIS process discussed earlier. The result of the data assembly process is a data set capable of providing the information necessary for this research. T able 4 5 is a list of the final variables for each of the 13,416 records. Stops that are closer to the detector likely provide the most accurate representations of traffic impact. For this reason, only those stops occurring within four 656 foot ( 200 meter ) sub seg ments of a roadway detector were kept in the initial GIS join Of the more than 13,400 stops used 3,610 were in the same sub segment as a detector. Another 5, 0 78 stops were within one sub segment of the detector sub segment. When the 2,452 stops withi n 2 sub segments are added, 83 percent of the total data set are represented. Another 2,276 stops are within 3 or 4 200 meter sub


87 segments, still a valid distance for purposes of this project. Table 4 6 shows the frequency of stops occurring within 200 me ter sub segments of a detector, with 0 representing that the stop and detector are within the same 200 meter sub segment Statistical Analysis of Historical Stops Analysis of the more than 13 thousand freeway enforcement stops c an set the stage for underst anding their impacts on freeway operations and capacity. The attributes of those stops precedes examination of average lane speeds, speed flow relationships, and finally regression analysis. Stop Attributes The attributes of stops are easily presented in charts, depicting the distribution of days of week, time of day and ambient lighting conditions Figures 4 1 4 through 4 1 6 are graphic representations of those attributes. Observing police enforcement stops in their natural occurrence, there is not a lot of difference in the day of the week, though maximizes resources on Fridays, wh ich likely explains that situation. Fewer crashes on weekends likely means more unobligated patrol time for officers to engage in enforcement on Saturdays. Traffic stops occur at all hours of the day and night, as one might expect. Increases in staffing and traffic volume translate in to increased stops activity during the morning rush hours, and mid day spikes occur during shift overlaps when maximum staffing is present, at around 3:00 pm. Between 11:00 pm and 5:00 am, drops in stops coincide with reduc ed staffing levels. The ambient lighting conditions of a traffic stop may be important, given emergency vehicle lighting equipment has different qualities of visibility during dark,


88 daylight, and the transition period between the two, twilight. From this data set, 68 percent of stops occurred during periods of daylight, and 32 percent during darkness or twilight. Speed at Stop Locations The average speed of traffic passing traffic stops is fundamental to understanding differences between at stop and non st op conditions. Average lane speeds from detectors at each stop location, aggregated at a 5 minute resolution for non stop conditions, and up to 15 minutes during each stop can be tested descriptively and the means compared. Table 4 7 is a presentation o f the descriptive statistics for average lane speeds. A t test was used to compare the mean speeds of the non stop and at stop conditions The results of the comparison of means established that there is a difference in mean speeds that is statistically significant at th e 95 percent confidence level Table 4 8 shows the results of the paired comparison. Plot Speed and Flow Rate at Stop Locations The relationships between the parameters of speed, flow, and density are relevant to the uninterrupted flow me thodologies. In the HCM, capacity is defined as the maximum hourly rate at which persons or vehicles can reasonably be expected to traverse a point or uniform section of a lane or roadway during a given time period under prevailing roadway, traffic, and c ontr ol conditions (HCM). The array of traffic flow conditions for freeways are represented in the HCM in terms of undersaturated flow, oversaturated flow, and queue discharge flow. Figure 4 1 7 is a graphic representation of the different flow regimes depi cted in the HCM.


89 By plotting speed and flow rate at each stop location a similar visualization is possible for stop and non stop conditions The preceding section shows that the mean difference in speeds is just over 1 mile per hour ; however that difference is statistically significant. The speed flow plots for the stops are all quite similar. Figures 4 1 8 and 4 1 9 are the speed flow plots for the at stop and non stop conditions respectively. Regression Analysis and Analysis of Parameters A regression model w as used to evaluate parameters in the data set. Speed at stop locations was modeled to evaluate the significance of variables for speed limit ( POST _SPEED) number of travel lanes (NUM_LANES) ambient light (Ambient) and traffic volume ( Vol_AtStop_Calc) Equation 4. 2 rep r e se nts the model and Equation 4. 3 represents the model with the estimated coefficient values Table 4 9 presents the statistical result s of the model which had a R square d value of .227. While this measure would gene rally be considered low, it is not unexpected, given the great variability typically observed in speed flow plots on large data sets. The model shows that the speed limit of the roadway, number of travel lanes, ambient lighting and traffic volume all were statistically significant factors. Given the sample set is largely comprised of undersaturated conditions, they might be better described as significant for those conditions. (4. 2 ) Where:Speed _AtStop = speed at traffic stops Post speed = posted speed limit ( mi/h ) Num_ lanes = the number of directional lanes of travel Ambient = ambient light conditions (1 day, 0 night) Vol_Calc=volume at stop locations ( veh/h/ln ) (4. 3)


90 The simple linear relationship depicts a constant sl o pe from the free flow s peed to the speed at capacity, which serves the purpose of id entifying explanatory variables. T he functional form is different from HCM models that depict a constant speed that changes to a curvilinear section that points to capacity. Unlike typical regre ssion forms, the constant term is not a starting point for the function, but a component for a free flow speed calculation. The constant works in tandem with the posted speed variable to provide an estimate of free flow speed. Since the r egression model indicates that variables of POST _SPEED, NUM_LANES, Ambient and Volume_AtStop_Calc may be explanatory factors for the speed of vehicles passing during at stop conditions additional examination of their categorical speeds is illustrative To isolate the variables a series of t tests comparing the mean speeds in non stop and at stop conditions is useful. One might question if the two samples are best evaluated as paired or unpaired and certainly a case can be made both ways. Both pai red and unpaired statistical comparisons were performed and the results were nearly identical, so the paired option is presented herein. There were 3043 9022 1101 and 250 stops in the categories of POST SPEED for 70, 65, 55, and 50below mile per hour speed zones respectively. Between non stop and at stop conditions, the difference in mean speeds is 0 .9 1.4 1.5 and 1.4 mi/h respectively. While these differences are less than those observed in the staged stop scenarios discussed in Chapter 3 all pr oved statistically significant at the 95 percent confidence level. Table 4 9 provides a summary of the statistical comparison of non stop and at stop speeds for the categories of POST _SPEED.


91 There were 2142 7174 3407 and 693 stops in the categories of N UM_ LANES for 2, 3, 4, and 5/6 lane segments respectively. Between non stop and at stop conditions, the difference in mean speeds was 0 .8 1.3 1.5 and 1.9 mi/h respectively. Again, less than those observed in the staged stop scenarios, but all proved statistically significant at the 95 percent confidence level. Table 4 10 provides a summary of the statistical comparison of non stop and at stop speeds for the categories of NUM_LANES. There were 9086 and 4330 stops in the categories of day and night for the variable Ambient respectively. Between non stop and at stop conditions, the difference in mean speeds was 1.3 and 1.4 mi/h respectively. Both were statistically significant at the 95 percent confidence level. Table 4 11 provides a summary of the statistical comparison of non stop and at stop speeds for the categories of Ambient. The stop parameters of POST _SPEED, NUM_LANES, and Ambient all appear to be valid explanatory factors for speed at stop locations. Like the overall comparison of me ans, each categorical value had a reduction in mean speed between non stop and at stop conditions. To illustrate the relationship of these variables with volume, speed flow plots for each variable attribute are provided in Figures 4 20 through 4 29 Each figure overlays both the non stop and at stop data points. Modeling Stops Using Van Aerde Having collected and analyzed aggregated traffic data for non stop and at stop conditions at 13,416 events using these data to model the traffic stream is the next step in understanding how those events impact operations and capacity. Drilling down to the level of individual detectors, it is possible to examine the behavior of traffic for a given location, acro ss a variety of conditions. Th is insight is possible be cause of the


92 relationship between flow, speed, and density, the foundation of the fundamental diagram (Rakha and Arafeh, 2010) M acr oscopic simulation models are widely used to evaluate the behavior of traffic streams particularly those that relate speed and density Those that consider a microscopic car following behavior where headways are constant are steady state models. Models can also be categorized as single regime or multi regime. While multi regime models overcome deficiencies along some portion of the density range, they are difficult to calibrate (Hranac, 2006) The widely used Greenshields single regime linear speed density model represents the speed flow relationship as parabolic, where the speed at capacity is eq ual to half the free flow spee d (Rakha and Crowther, 2006) Greenshields model requires two parameters, free speed and jam density/c apacity to calibrate (Rakha and Crowther, 2006) The Pipes car following model is multi regime and assumes that in the uncongested regime speed is insen sitive to density (Rakha and Crowther, 2006). I t is used in several microscopic simulation models like and VISSIM Pipes requires free speed, jam density headway and a driver sensitivity factor to calibrate. As was mentioned, the model can be difficult to calibrate. Van Aerde (1995) and Van Aerde and Rakha (1995) created a model that essentially combines the Greenshields and Pipes models ( Rakha and Crowther, 2002 ) The Van Aerde model is a single regime, steady state model that has proven very effective in estimating the behavior of traffic streams. The Van Aerde model is implemented using the Traffic Stream Calibration Software, SPD_CAL.exe, which uses


93 a n iterative and heuristic hill climbing technique to compute output parameters (Rakha, 2007) The m odel, via the software, is able to evaluate speed and volume parameters at a location to model the speed flow relationship. From inputs of speed, flow, and density, the program produces free flow speed, speed at capacity, capacity, and jam density. This is highly desirable, given the objective of understanding the difference between non stop and stop conditions. The following sections discuss the parameters of the Van Aerde model, the selection of individual detector sites for evaluation, and an interpret ation of the results. Model Parameters The model created by Van Aerde ( 1995 ) and Van Aerde and Rakha (1995) assumes the functional form described by Equation 4. 4 The model estimates parameters c 1 c 2 c 3 and K as depicted in Equations 4. 5 through 4. 8 ( Rakha and Crowther, 2006) producing results u f u c q c and k j (4. 4 ) (4. 5 ) (4. 6 ) (4. 7 ) (4. 8 ) Where: c 1 =fixed distance headway constant (km), c 2 =first variable distance headway constant (km2/h), c 3 =second variable distance headway constant (h), u f free speed (km/h), u c =speed at capacity (km/h),


94 q c =flow at capacity (veh/h), k j =jam density (veh/km), and m=is a constant used to solve for the three headway constants (h/km) The essential inputs for the Traffic Stream Calibration Software come from the traffic stop data set, S peed_ Non Stop, Speed_AtStop, Vol_NonStop_Calc, and Vol_AtStop_Calc. Upon guidance from the software, t he density variable is set to default ( 1) in the program largely because the CDW stores occupancy in lieu of density. Selection of Stop Sites for Modeling The historical stop data set contains 13,416 individual stop records from a 15 mont h period in late 2010 through the summer of 2011. These stops occurred at 767 unique detector locations. A frequency distribution for all detector locations isolated those locations with larger numbers of stops. This is important to secure a sufficient number of data points for the model to work properly. It was not only important to draw on locations with a large number of stop data points, locations with many high flow rate and conges ted data points generally improve the Van Aerde model fitting process Because of the unique sampling in this study, t rial and error demonstrated that the combination of these three factors were practical requisites for successful runs. Consequently, not all stop locations were suitable for modeling with the Traffic Stream Calibration Software. At least t en different detector locations found success using the software. These locations represent a wide range of attributes including 2, 3, 4, and 5 lane se gments, segments with 55, 65, and 70 mi/h limits, and AADT ranging from 24,500 to 304,000. Eight hundred and twenty six records, 6 percen t of the total, were accounted for at the 10 locations. All locations had stops across the spectrum of days of the we ek and


95 hours of the day. Table 4 13 show s the detector id numbers general geographic locations, and number of lanes, posted speed, and AADT at these locations Model Results The Traffic Stream Calibration Software was used to develop speed flow plots, fo r both the non stop and at stop sample, for the 10 detector locations t he software develops the model based on the raw speed flow data, and then the fitted curve representing that model is plotted over the raw data plot. The fitted curve represent s the b ehavior of traffic through the undersaturated and oversaturated regimes Figures 4 29 through 4 39 are the speed flow diagrams for each of the locations. The software provides a tabular output of key parameters ( u f u c q c and k j ) for non stop and at stop conditions. Table 4 14 is a consolidation of those output s by detector location. Hranic, et al. (2006) used the Van Aerde model to compare traffic a t locations under adverse weather conditions and those under base conditions. Such a methodology rendered a weather adjustment factor WAF, for the output parameters of the model. Th e WAF was achieved by computing a ratio of the parameters under existing conditions relative to the base condition (Hranic, 2006). Such a methodology was thought possib le with the current effort Using each of the Van Aerde model parameters ( u f u c and q c ) a ratio of at stop and non stop conditions was evaluated to determine if it might create an enforcement adjustment factor or EAF. The change in each parameter bet ween the at stop and base condition are provided along with the ratio in Table s 4 15 through 4 17 It quickly bec o me s apparent that the prospect of replicating the WAF created by Hranic, et al. is not possible given the current data set The product of the Van Aerde modeling was somewhat inconsistent, and parameter ratios were fairly small.


96 Unlike continuous sampling by detector locations for days or even weeks, sampling stop events, occurring randomly during the typical day, renders fe w data points at the highest flow rates. This exacerbates the tendency for the Van Aerde model to under estimate actual capacity values (Washburn, et al., 2010). Selecting non stop samples from times immediately preceding enforcement stops helps control for extraneous effects like weather, incidents, and work zones, but unfortunately it does not fill the high flow rate gaps that are likely needed to form the base capacity condition. While the relative change in traffic parameters between non stop and at stop conditions show slight effects of stops, the actual capacity of the segments are very likely higher than those estimated by the model. Summary of Capacity Impacts A priori, this research sought to model the police traffic stop as a subset of traffic incidents, by emulating past research that has compared freeways under incident conditions and those where prevailing conditions were present. Such a methodology appear s well suited for events where the duration and/or intensity of the event alters the traffic stream in a statistically significant way. With a sample size that far exceeds most of these efforts, examining historical stops has revealed that freeway stops d o not rise to the level of influence seen in crashes and lane blocking incidents. Mean segment speeds are reduced by 1. 3 miles per hour. When plotted on speed flow graphs, differences in data points representing stop conditions and non stop conditions ar e not always obvious, as seen in the Van Aerde speed flow diagrams. Identifying capacity impacts are challenging, given the diverse roadway, user, and environmental circumstances that exist. Evaluating police enforcement stops and potential impacts on cap acity is challenging because changes in the traffic stream are


97 intuitive, but not readily apparent. Considering that there are not a large number of stops in heavy traffic flow conditions only makes the proposition more difficult. Modeling potential capa city impacts can be approached in three different ways, within the framework of existing incident based capacity reductions, through measuring changes in traffic behavior, and using HCM alternative speed flow calculations. Existing HCM framework for incide nt capacity reductions The values from the HCM chart that depicts the proportion of capacity available in incident conditions is represented in Figure 5 1. Intuitively, one might suspect that a police traffic stop on the shoulder of a freeway would be mor e distracting than a disabled vehicle, but perhaps not as noticeable as multiple vehicles on the shoulder with a traffic crash. Since police enforcement stops rarely physically block travel lanes, examining the boundaries of the shoulder disablement and t he shoulder accident would seem logical. This approach would, however, neglect the important impact of the move over law. Since move over laws require drivers to vacate the lane closest to emergency responders on multi lane facilities, that lane becomes theoretically blocked, at least to the degree which motorists comply with the law. It may therefore, be more appropriate to view the police traffic stop as an event blocking one lane, adjusted for motorists compliance. Such an approach would be more accu rate than simply modeling the traffic enforcement action as a shoulder incident. Move over laws require drivers vacate the lane adjacent to the emergency responder vehicle. Between 68.8 and 79.7 percent of drivers were observed to move over on Florida fre eway segments in this research, depending upon the patrol vehicle lighting configuration in use. The controlled experiment simulated what would be over event I t is possible that a higher level of intensity might


98 accompany huma n activity at the roadside stop like interviews or vehicle searches. In addition, the presence of multiple enforcement vehicles may also alter a measure of intensity. The same HCM table that specifies available remaining capacity for shoulder disablements and shoulder crashes also provides for remaining capacity when lanes are blocked. Applying that portion of the table to the stop scenario is evaluated further in the HCM alternative speed flow calculations section which follows. Measuring changes in tra ffic behavior at stop locations Comparing non stop and at stop conditions at individual detector locations make s it possible to control for roadway geometry and other factors. Identifying detectors where this was possible proved challenging because there was a need for a minimum number of stop data points, stops in both regime, and stops where higher flow rates were present. In addition, sampling small time slices over a period of 15 months deviates from the traditional application of the technique where continuous samples are retrieved for successive time periods. The Van Aerde model appears not to have been appropriate, given the data that was available in this study. Even when data satisfied conditions for the model to produce an output, the deficiency of data points under high flow rate conditions preclude d accurate capacity estimates. With that said, the speed flow plots and overlay of the model do help illustrate undersaturated conditions. Though the model did not illuminate the enforcement stop ca pacity issue, it did highlight a significant issue in modeling the events, the need for more data in higher flow periods.


99 HCM Alternative Speed Flow Calculations The HCM prescribes capacity values for freeways. Adjustment factors are a principle way in wh Geometric conditions, vehicle composition, and driver attributes are types of adjustments. The freeway facilities methodology also considers work zones, weather, and incidents as poten tially affecting freeway segments. Past research into these events has rendered tables in the HCM to guide modeling their impacts. Separate sections for each type of impacting event were thoroughly examined in C hapter 2 Beyond the HCM tables, the M anu al also provides for a way to calculate an alternative speed flow curve specifically where there is a reduction in capacity, which may be attributed to situations like work zones, weather or incidents. The formula for the alternative speed calculation is seen in Equation 4.10 ( 4.10 ) Where: S=segment speed (mi/h), FFS=segment free flow speed (mi/h), C=original segment capacity (pc/h/ln), CAF=capacity adjustment factor (CAF=1.0, use speed estimation procedures), v p =segment flow rate (pc/h/ln) Such a formula proves useful in the case at hand, since the capacity impacts associated with a police enforcement stop have never been quantified. This research has established that free flow speeds are 1.3 mi/h less at stop locations. The HCM provides values for capacity when one lane is blocked. Applying the observed range of compliance between 68.8 and 79.7 percent as a weighted reduction, the capacity

PAGE 100

100 impacts of the law enforcement stop can be m odeled. Equation 4.9 is used for applying the compliance weight to the HCM table, replicated as Table 5.1 herein, would be: (4.9) Where: c i j = HCM value for available capacity when one lane bl ocked for total lanes i through j a = percent driver compliance with move over law c a = move over compliance weighted capacity reduction Using the alternative speed flow formula, enforcement stops have a capacity reduction range of 54 to 58 percent for a two lane freeway, 35 to 41 percent for a three lane freeway, 23 to 30 percent for a four lane freeway, and 14 to 22 percent in the case of a five lane facility. Figure 4 40 shows the speed flow curves created using the alternative formula with capacity adjustment factors. Note that the lines representing free flow speed of 70 and th at adjusted for the 1. 3 mi/h reduction in speed are very similar, and c onverge near the point of capacity. Ranges for the number of directional travel lanes are identified within the speed flow curves Naturally, freeways with fewer lanes are impacted more severly.

PAGE 101

101 Figure 4 1. Florida counties (highlight ed) and instrumented roadways used in project

PAGE 102

102 Figure 4 2. FHP mobile computer traffic stop input screen

PAGE 103

103 Figure 4 3. Duval County stops and detectors

PAGE 104

104 Figure 4 4. Volusia County stops and detectors

PAGE 105

105 Figure 4 5. Brevard County stops and detectors

PAGE 106

106 Figure 4 6. Seminole County stops and detectors

PAGE 107

107 Figure 4 7. Orange County stops and detectors

PAGE 108

108 Figure 4 8. Osceola County stops and detectors

PAGE 109

109 Figure 4 9. Hillsborough County stops and detectors

PAGE 110

110 Figure 4 10. Broward County stops and detectors

PAGE 111

111 Figure 4 11. Zoomed representation of stops and detectors

PAGE 112

112 Figure 4 12. Stop reduction process and resulting cases Figure 4 13. Timeline representation of sampling methodology. Figure 4 1 4 Day of week distribution for stops Total FHP Stops 252,956 Filtered by County 35,101 Post GIS Spatial Join 21,144 CDW Search and Filter 13,416 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 Sunday Monday Tuesday Wednesday Thursday Friday Saturday Percent of Total Stops Day of Week

PAGE 113

113 Figure 4 1 5 Time of day distribution for stops Figure 4 1 6 Distribution of stops by daylight 0 200 400 600 800 1000 1200 1400 Midnight 1am 2am 3am 4am 5am 6am 7am 8am 9am 10am 11am Noon 1pm 2pm 3pm 4pm 5pm 6pm 7pm 8pm 9pm 10pm 11pm Number o Traffic Stops Traffic Stops by Hour of Day Night 32% Day 68% Ambient Lighting at Stops

PAGE 114

114 Figure 4 17. Three types of freeway flow similar to illustration in the HCM Figure 4 18. Speed Flow plot for historical stops At Stop Condition

PAGE 115

115 Figure 4 19. Speed Flow plot for historical stops Non Stop Condition Figure 4 20. Speed Flow plot for historical stops POST _SPEED=70 mi/h

PAGE 116

116 Figure 4 21. Speed Flow plot for historical stops POST _SPEED=65 mi/h Figure 4 22. Speed Flow plot for historical sto ps POST _SPEED=55 mi/h

PAGE 117

117 Figure 4 23. Speed Flow plot for historical stops POST _SPEED=50< mi/h Figure 4 24. Speed Flow plot for historical stops NUM_LANES=2

PAGE 118

118 Figure 4 25. Speed Flow plot for historical stops NUM_LANES=3 Figure 4 26. Speed Flow plot for historical stops NUM_LANES=4

PAGE 119

119 Figure 4 27. Speed Flow plot for historical stops NUM_LANES=5/6 Figure 4 28. Speed Flow plot for historical stops Ambient=Day

PAGE 120

120 Figure 4 29. Speed Flow plot for historical stops Ambient=Night Figure 4 3 0. Speed Flow Plot for Location 410201

PAGE 121

121 Figure 4 3 1. Speed Flow Plot for Location 501351 Figure 4 3 2. Speed Flow Plot for Location 510082

PAGE 122

122 Figure 4 3 3. Speed Flow Plot for Location 510311 Figure 4 3 4. Speed Flow Plot for Location 510422

PAGE 123

123 Figure 4 3 5. Speed Flow Plot for Location 510511 Figure 4 3 6. Speed Flow Plot for Location 510522

PAGE 124

124 Figure 4 3 7. Speed Flow Plot for Location 510711 Figure 4 3 8. Speed Flow Plot for Location 510731

PAGE 125

125 Figure 4 3 9. Speed Flow Plot for Location 511322

PAGE 126

126 Figure 4 40. Adjusted speed flow curves

PAGE 127

127 Table 4 1. FHP CAD data format Database Field Name Description of Data INCIDENTNO Incident Number CSTREET On Street DIR Offset Direction CXSTREET1 Cross Street CCOUNTY County CMAP_X Latitude CMAP_Y Longitude COMPLAINT Type of Incident PUNITA Primary Unit ID DATE RECEIVED Date Received TIME RECEIVED Time Received DATE 1098 Date Cleared TIME 1098 Time Cleared Table 4 2. STEWARD Data Content/Format Database Field Name Description of Data STAT_NDX Station Index STAT_CDW Station CDW Descrip Description Road Roadway DET_UNIT Detector Unit LATITUDE Latitude LONGITUDE Longitude STATE_MP State Mile Post ROADWAY_ID Roadway Identifier ROAD_MP Roadway Mile Post MAX_SPEED Maximum Speed NUM_LANES Number of Lanes UPNODE Upstream Node LANE_CAP Lane Capacity DET_TYPE Detector Type COUNTY County CNT_STAT Count Station

PAGE 128

128 Table 4 3. Combined stops and detectors data elements table Data Element Name FID_ INCIDENTNO CSTREET DIR CXSTREET1 CCOUNTY CMAP_X CMAP_Y COMPLAINT PUNITA DATE_RECEI DATE_1098 TIME_RC TIME_98 ST_NAME FIRST_DIR_ ID NAME STAT_NDX STAT_CDW DESCRIP STATUS ROAD DET_UNIT LATITUDE LONGITUDE STATE_MP ROADWAY_ID ROAD_MP MAX_SPEED NUM_LANES UPNODE LANE_CAP DET_TYPE COUNTY CNT_STAT DISTRICT DIRECT_NS ROAD_NAME ST_NAME_1 FIRST_DI_1 ID_1 NAME_1 Distance_2 RDLOC_DIFF STREAM Table 4 4. Sample sizes for past traffic incident studies Author Year Country Location Sample Size Goosby 1971 US Houston 2,271 Giuliano 1988 US Los Angeles 776 Jones, et al 1991 US Seattle 2,156 Khattak, et al 1994 US Chicago 109 Nam and Mannering 1995 US Washington State 681 Ullman and Ogden 1996 US Houston 612 Garib, et al 1997 US California 2,181 Skabardonis, et al 1997 US San Francisco 2,181 Skabardonis, et al 1999 US Los Angeles 1,239 Qin and Smith 2001 US Virginia 258 Ozbay and Noyan 2002 US Virginia 577 Wang, et al 2005 UK UK 1,080 Knoop, et al. 2009 Neth Netherlands 90 Lu and Elefteriadou 2010 US MN, OR, Toronto 98 Table 4 5. Final data table for historical traffic stops Data Element Names StopId Speed_NonStop Day COUNTY Vol ume NonStop Month ROAD Volume_NonStop_Calc Year POST _SPEED Speed_AtStop StartHour NUM_LANES Volume_AtStop StartMin Ambient Volume_AtStop_Calc LONGITUDE LANE_CAP STATION_CDW LATITUDE Absolute Distance

PAGE 129

129 Table 4 6. Segment relationship between stops and detectors 200 meter segments Frequency Percent Cumulative Percent 0 3610 26.9 2 6.9 1 5078 37.9 64.8 2 2452 18.3 83.0 3 1338 10.0 93.0 4 938 7.0 100.0 Total 13416 100.0 Table 4 7. Descriptive statistics for average lane speeds N Min Max Mean Std. Deviation Non Stop Speed 13 4 16 7.2 95.4 66.7 5 4 10.0286 At stop Speed 13 4 16 7.9 96.6 65.466 10.3934 Valid N ( listwise) 13 4 16 Table 4 8. Comparison of mean speeds for non stop and at stop conditions for all stops Diff. Diff Diff All Stops N Mean Std. Deviation Std. Error Mean Mean Std. Dev iation Std. Error Mean P Value Non Stop 13416 66.75 10.0286 .0866 1.2876 4.2008 .0363 .000 At Stop 13416 65.47 10.3934 .0997 95 percent CI Table 4 9. Regression results for Speed_AtStop explanatory variables Unstandard Coefficients Std. Coeff. Model B Std. Error Beta T Sig. Constant 34.673 1.449 .392 23.924 .000 Vol_AtStop_Calc .009 .000 .392 43.253 .000 POST _SPEED .583 .020 .238 29.421 .000 NUM_LANES .779 .114 .057 6.838 .000 Ambient 3.847 .198 .173 19.432 .000 Dependent Variable: Speed_AtStop

PAGE 130

130 Table 4 10 Descriptive statistics and c omparison of mean speeds for non s top (NS) and at stop (AS) for POST _SPEED variable Diff. Diff Diff Speed mi/h N Mean Std. Deviation Std. Error Mean Mean Std. Dev iation Std. Error Mean P Value NS 70 mi/h 3043 71.81 9.936353 0.1801259 0.9791 3.971908 0.072003 0.000 AS 70 mi/h 3043 70.83 10.21364 0.1851525 NS 65 mi/h 9022 66.09 9.093008 0.0956265 1.3905 4.025631 0.042382 0.000 AS 65 mi/h 9022 64.70 9.425967 0.0992476 NS 55 mi/h 1101 60.10 10.84394 0.0326809 1.4682 5.202227 0.156782 0.000 AS 55 mi/h 1101 58.95 11.00599 0.3316924 NS <=50 250 58.33 10.41729 0.6588412 1.9484 7.019149 0.44393 0.000 AS <=50 250 56.38 10.93771 0.7677966 95 percent CI Table 4 11 Descriptive statistics and comparison of mean speeds for non stop (NS) and at stop (AS) for NUM_LANES variable Diff. Diff Diff Lanes N Mean Std. Deviation Std. Error Mean Mean Std. Dev iation Std. Error Mean P Value NS 2 Lane 2142 74.32 6.100463 0.1318114 0.8379 4.064558 0.097822 0.000 AS 2 Lane 2676 73.48 6.362524 0.1374737 NS 3 Lane 7174 65.56 10.37228 0.1224598 1.2497 3.918534 0.045264 0.000 AS 3 Lane 7174 64.31 10.58202 0.1249362 NS 4 Lane 3407 64.05 9.378024 0.1606665 1.5229 4.453732 0.076302 0.000 AS 4 Lane 3407 62.53 9.878489 0.1692406 NS 5/6 Lane 3043 71.81 9.935343 0.1801259 0.9790 3.971908 0.072003 0.000 AS 5/6 Lane 3043 70.83 10.21364 0.1851525 95 percent CI Table 4 12 Descriptive statistics and comparison of mean speeds for non stop (NS) and at stop (AS) for Ambient variable Diff. Diff Diff Day/Night N Mean Std. Deviation Std. Error Mean Mean Std. Dev iation Std. Error Mean P Value NS Night 4330 66.69 9.231564 0.1402914 1.3691 3.976184 0.060426 0.000 AS Night 4330 65.32 9.486525 0.1441661 NS Daylight 9086 66.79 10.38725 0.1089716 1.2486 4.303454 0.045147 0.000 AS Daylight 9086 65.54 10.79866 0.1132879 95 percent CI

PAGE 131

131 Table 4 13 Locations Selected for Van Aerde Modeling STATION_CDW Location No. Lanes Speed AADT 410201 I 95 NB @ Broward Blvd 5 65 304000 501351 I 95 NB @ MM 239.9 2 70 24500 510082 I 4 WB West of SR 545 3 65 110000 510311 I 4 EB West of SR 535 4 65 113500 510422 I 4 WB West of C FL Parkway 4 65 153974 510511 I 4 EB West of Sand Lake 4 65 153974 510522 I 4 WB West of Sand Lake 4 65 153974 510711 I 4 EB West of John Young Pkwy 3 55 161000 510731 I 4 EB @ John Young Parkway 3 65 134000 511322 I 4 WB @ SR 434 3 65 133908 Table 4 14 Consolidation of Traffic Stream Calibration Software output by detector. u f u f u c u c q c q c Detector Non Stop At Stop Non Stop At Stop Non Stop At Stop 510711 61.0 58.5 40.3 37.6 1512 1542 510351 55.9 54.0 45.2 43.7 1940 1869 510511 55.4 53.1 43.9 44.3 1776 1706 510731 71.5 69.5 57.6 58.5 1713 1721 410201 75.8 73.2 51.3 52.3 1786 1834 510522 62.3 63.0 52.0 45.3 1647 1617 510422 64.1 63.6 51.9 49.8 1611 1559 510082 68.0 69.7 58.9 52.5 1865 1695 510311 62.2 62.4 52.9 48.2 1533 1457 511322 64.1 63.4 50.6 50.6 1534 1549 Table 4 15 Change in free speed parameter u f u f Detector Non Stop At Stop Diff Ratio 510711 61.0 58.5 2.5 0.96 510351 55.9 54.0 1.9 0.97 510511 55.4 53.1 2.4 0.96 510731 71.5 69.5 2.0 0.97 410201 75.8 73.2 2.6 0.97 510522 62.3 63.0 0.7 1.01 510422 64.1 63.6 0.6 0.99 510082 68.0 69.7 1.6 1.02 510311 62.2 62.4 0.2 1.00 511322 64.1 63.4 0.7 0.99 AVG 1.0 0.98 Ratio 1.61

PAGE 132

132 Table 4 16 Change in speed at capacity parameter u c u c Detector Non Stop At Stop Diff Ratio 510711 40.3 37.6 2.7 0.93 510351 45.2 43.7 1.5 0.97 510511 43.9 44.3 0.4 1.01 510731 57.6 58.5 0.9 1.02 410201 51.3 52.3 1.0 1.02 510522 52.0 45.3 6.7 0.87 510422 51.9 49.8 2.1 0.96 510082 58.9 52.5 6.4 0.89 510311 52.9 48.2 4.7 0.91 511322 50.6 50.6 0.1 1.00 AVG 2.2 0.96 Ratio 4.23 Table 4 17 Change in capacity parameter q c q c Detector Non Stop At Stop Diff Ratio 510711 1512.0 1542.0 30.0 1.02 510351 1939.7 1869.4 70.3 0.96 510511 1776.1 1705.8 70.3 0.96 510731 1712.7 1720.8 8.1 1.00 410201 1786.0 1834.3 48.3 1.03 510522 1646.9 1616.6 30.3 0.98 510422 1611.0 1559.0 52.0 0.97 510082 1865.0 1694.7 170.3 0.91 510311 1533.1 1457.2 75.9 0.95 511322 1534.1 1548.5 14.4 1.01 AVG 36.8 0.98 Ratio 2.06

PAGE 133

133 Table 4 18. Values from HCM for capacity remaining at incident Number of Directional Lanes) Shoulder Disablement Shoulder Accident One Lane Blocked Two Lanes Blocked Three Lanes Blocked 2 0.95 0.81 0 .35 0.00 N/A 3 0.99 0.83 0.49 0.17 0.00 4 0.99 0.5 0.58 0.25 0.13 5 0.99 0.87 0.65 0.40 0.20 6 0.99 0.89 0.71 0.50 0.26 7 0.99 0.91 0.75 0.57 0.36 8 0.99 0.93 0.78 0.63 0.41

PAGE 134

134 CHAPTER 5 RESULTS AND CONCLUSIONS potentially impacts the traffic stream. Stops are typically fairly short in duration, and since the y do not physically block lanes, their intensity is less than most other incident types. With that said, this research has produced results that show that stops do impact the traffic stream. A summary of the results are presented along with policy implications and future research recommendations in the fol lowing sections Research Results Operational Impacts The speed of vehicles passing traffic l aw enforcement stops is reduced in a statistically significant way. The mean speed reduction was 4.6 mi /h during st aged stop experiments, and about 1. 3 mi /h during historical stops. This difference might be explained by controlled experiment site selection that allowed for adequate driver sight distances, absence of horizontal/vertical curves, no influence of work zones and interchanges, and ideal weather. The mean speed of vehicles traveling in the o pposite direction was reduced by a scant 0 .2 mi /h which was determined to not be statistically significant. Move over laws create a mandate for motorists to vacate the lane adjacent to the stop. Overall, three out of four vehicles move over in compliance with the law, and that figure climbs to nearly 80 percent when red and blue lights are used exclusively. Conversely, compliance drops to 68.8 percent when amber deck lights are used in lieu of red and blue emer gency lighting.

PAGE 135

135 Driver compliance with the requirement to slow when not moving over was quite low. Less than 6 percent of right lane vehicles achieved the legally mandated reduction of 20 mi /h The mean speed reduction of vehicles passing staged stops in the right lane was 7.4 mi /h During staged experiments, drivers who failed to move over had available gaps 93.7 percent of the time. This would indicate that a segment of drivers were unaware of their obligation to move over, had a cognitive lapse, or so me other reason for not executing a lane change. In terms of evaluating critical gaps, t he opposite was observed, but decidedly difficult to measure, many drivers appeared to make forced merges, ostensibly to comply with the law. The emergency lighting co mponent of the study revealed that the use of blue and red emergency lights was important for move over compliance. The addition of an amber directional arrow did not improve driver move over compliance. Reducing forward facing lights did minimize the im pact on traffic passing the stop in the opposite direction, however that change was not statistically significant. Regression analysis examined factors to explain the speed of vehicles passing stops and showed that the number of lanes, posted speed limit, and ambient lighting conditions and traffic volume were all statistically significant issues. The extent of influence and nature of attributes associated with these variables were beyond the scope of this effort. Capacity Impacts This research has explore d the behavior of traffic around police traffic stops to quantify those effects. Even with extensive data collection in both the designed experiment and historical analysis of actual stops, capacity quantification remains

PAGE 136

136 elusive. With that said, because of the absence of any research in this arena, conclusions about capacity impact s do serve as a credible starting point for inquiry into the subject If the police traffic stop is modeled strictly as a shoulder, incident type event, reducing available capac ity somewhere between a shoulder disablement and a shoulder crash would seem reasonable Based upon related incident work, these HCM values are logical applications for the police enforcement stop, but they would prove highly suspect considering the influence of move over laws. Because most state move over laws were implemented in since 2000, the majority of incident studies have no reference or consideration of their influence. To fully appreciate the move over law impacts, police t raffic stop s can be modeled as a theoretical lane blocking event where move over compliance fractionally applies HCM values for available capacity under a one lane blocked incident condition. Since it is known that the average speed passing enforcement st ops is reduced by 1.3 mi/h, that reduction, and the weighted capacity adjustment factor can be applied to the HCM method for calculating an alternative speed flow curve. With the use of alternative speed flow curves, t he law enforcement stop might have a capacity reduction range of 54 to 58 percent for a two lane freeway, 35 to 41 percent for a three lane freeway, 23 to 30 percent for a four lane freeway, and 14 to 22 percent in the case of a five lane facility. Again, there is logic in this approach, but because the move over maneuver is not a mandatory lane change the intensity of the theoretical lane blockage may be an uncontrolled factor.

PAGE 137

137 Policy Recommendations Law enforcement should use emergency lighting throughout the duration of traffic stops, sin ce it increases compliance with move over laws. Coincidentally, the use of emergency lighting may be a condition required by many state move over laws Reducing emergency lights during stops does not appreciably reduce adverse operational impacts, but it does cause late merges and lower speed reductions among passing vehicles, potentially impacting safety. Forward facing emergency lights should be reduced to minimize the distraction to motorists traveling in the opposite direction. While the rubbernecker impact was not statistically significant for the stop scenario, the slight reduction in speed was minimized when emergency lighting was reduced. State move over laws should be standardized to promote a more consistent requirement for drivers. Standardizat ion should address the types of responder vehicles at roadside, to include any vehicle displaying warning lights. This would create an inclusion for law enforcement, fire, EMS, towing, transportation, and e in jeopardy at roadside. The laws should also be standardized concerning the obligation of approaching drivers. A consistent requirement to vacate the adjacent lane or slow if unable to do so should be included. Since this research has evidenced diffi culty in messaging an absolute speed reduction of 20 miles per hour, different language should be considered. Driver education and public information efforts should be used to amplify the move over law requirements. Specific emphasis should be placed on s afely executing the lane change, when possible. Additionally, the slow in lieu of moving part of public compliance/awareness appears lacking. While the relationship between traffic volume and stop impacts remain imprecise law enforcement should consider that move over mandates may impact congested conditions disp roportionally when available ga p s are reduced. The HCM should include this research to indentify a new category of capacity adjustment, or potentially fold it into existing commentary on incidents Future Research Recommendations The staged stop methodology should be replicated or expanded to other states, to regionally validate levels of move over compliance. The staged stop methodology should be expanded to other classifications of roadways, to determine if impacts differ from the freeway study herein.

PAGE 138

138 The staged stop methodology should be replicated with other types of responder vehicles to determine compliance differences among vehicles types and lighting configurations. The staged stop methodo logy might be enhanced to better observe upstream gap acceptance and the measurement of forced merge behavior. Short duration staged stops should be conducted in congested conditions to model the impact of stops at or near breakdown. Where breakdown occur s as a result of the traffic enforcement stop, queue formation and discharge should be evaluated. Q ualitative research should be conducted with individuals who have been cited for move over violations to determine their attribution for the violations, i.e. did not know about the law, driver distraction, lapse in attention, lapse in judgment, etc. The emergency vehicle lighting and conspicuity should be evaluated further to identify the effects of light color, intensity, and pattern. Similarly, vehicle conspicuity in the form of vehicle color and/or markings, might be examined to determine their ef fectiveness. In terms of modeling and simulation, a lane change algorithm for the police stop situation needs to be developed so it can be implemented in simulation tools. This will allow for additional experimentation and evaluation.

PAGE 139

139 APPENDIX A F LORIDA MOVE OVER CITATIONS BY COUNTY Failure to Yield/Move Over Convictions by County by Year Charge Code 511 is used for both Failure to Move Over and Failure to Yield for Emergency Vehicles County 2009 2008 2007 2006 2005 2004 2003 ALACHUA 382 966 330 383 91 100 73 BAKER 21 48 15 2 3 7 1 BAY 169 163 171 152 54 52 51 BRADFORD 27 63 27 31 20 26 21 BREVARD 355 479 399 420 327 405 147 BROWARD 2,392 1,710 702 1,043 360 375 194 CALHOUN 3 5 3 1 CHARLOTTE 63 239 379 113 71 25 35 CITRUS 92 135 115 65 19 10 16 CLAY 47 42 27 16 15 18 17 COLLIER 1,017 474 344 292 403 292 204 COLUMBIA 20 19 23 16 8 7 13 DE SOTO 2,650 2,394 2 10 2 1 DIXIE 7 18 11 8 1 5 3 DUVAL 10 31 571 531 306 227 138 ESCAMBIA 424 939 42 55 41 26 31 FLAGLER 76 58 68 128 32 15 13 FRANKLIN 125 276 5 2 2 GADSDEN 2 35 37 20 12 4 GILCHRIST 144 226 4 1 3 3 9

PAGE 140

140 County 2009 2008 2007 2006 2005 2004 2003 GLADES 3 8 34 65 14 44 55 GULF 39 51 3 4 3 1 1 HAMILTON 2 1 4 5 2 HARDEE 4 10 6 21 3 2 2 HENDRY 5 15 60 45 58 41 23 HERNANDO 29 30 183 98 86 54 46 HIGHLAND 138 155 155 146 123 19 16 HILLSBOROUGH 144 174 494 462 406 261 117 HOLMES 758 674 2 4 3 4 3 INDIAN RIVER 28 24 52 31 17 11 15 JACKSON 145 174 7 8 10 10 4 JEFFERSON 42 73 6 1 5 9 LAFAYETTE 29 127 1 3 1 1 LAKE 3 3 231 188 112 79 49 LEE 459 310 458 263 185 182 110 LEON 408 696 22 46 29 15 7 LEVY 129 212 70 31 24 20 13 LIBERTY 33 49 4 2 2 MADISON 1 1 13 9 3 9 4 MANATEE 40 45 106 75 22 16 31 MARION 156 68 118 95 54 35 34 MARTIN 234 139 85 68 76 76 38 MIAMI DADE 334 123 1,379 1,192 190 441 382 MONROE 97 127 79 181 101 81 108

PAGE 141

141 County 2009 2008 2007 2006 2005 2004 2003 NASSAU 18 12 31 34 14 33 12 OKALOOSA 58 85 74 72 30 18 8 OKEECHOBEE 25 16 25 20 19 7 4 ORANGE 843 1,018 828 571 207 176 139 OSCEOLA 384 389 261 215 65 63 49 PALM BEACH 1,637 1,585 519 508 283 123 132 PASCO 154 169 58 89 93 103 116 PINELLAS 391 543 337 361 266 227 156 POLK 530 622 422 578 271 159 119 PUTNAM 15 16 27 45 8 5 19 SANTA ROSA 43 50 78 40 48 24 10 SARASOTA 352 309 284 118 86 67 42 SEMINOLE 342 627 323 204 110 72 40 ST JOHNS 123 275 148 113 64 33 22 ST LUCIE 564 270 183 205 140 67 48 SUMTER 127 140 115 36 27 11 13 SUWANNEE 11 23 11 4 1 4 2 TAYLOR 26 21 10 2 6 4 10 UNION 1 1 VOLUSIA 429 499 409 314 194 205 130 WAKULLA 1 12 3 5 4 WALTON 13 20 17 18 10 14 5 WASHINGTON 21 10 13 8 6 4 Unknown 326 402 116 232 243 229 199

PAGE 142

142 County 2009 2008 2007 2006 2005 2004 2003 TOTAL 17,717 18,690 11,133 10,136 5,500 4,659 3,321

PAGE 143


PAGE 144

144 APPENDIX C STOP DETECTOR GIS DATA CONSOLIDATI ON The following bullets outline the steps of the fusion of the stop and detector data in a GIS. Segment Preparation Locate road segment in NavTeq Streets layer Export directional road segments and use Xtools pro to snip boundaries Clean up roadway selection set by deleting unnecessary roads Populate the DIRECTION field for each N/S/E/W direction. Dissolve based on DIRECTION field Use Xtools pro to sub segment roadways every 200m, creating a RDID field to fix direction (ascending/descending) and assign numeric ID (starting with 0) Break collective roadway segments into separa te directional layers N/S/E/W Detector Preparation Create a detectors layer from the tables containing detector ID and lat/long information Break detector locations into directional layers N/S/E/W Spatial join to the proper directional segment to get the closest RDID. Create new field for the I D _1 using the RDID Traffic Stop Data Preparation Create a stops layer from the tables containing stop incidents and lat/long provided Break stop locations into directional layers N/S/E/W Spatial join to the proper directional segment to get the closest RDID. Create new field for ID, containing the RDID value Spatially Relate Detectors and Stops Join the detector information to the traffic stops via a spatial join. Create the field RD LOC_DIFF and calculate ID minus the ID_ 1 For RD LOC_DIFF that is negative, the detector is downs tream of the stop, for RDLOC_DIFF that is positive, the detector is upstrea m of the stop, and for RDLOC_DIFF segment. Th e Dist _2 field provides the distance in meters between the stops and detectors.

PAGE 145

145 LIST OF REFERENCES Agarwal, M., Maze, T.H., Souleyrette, R., 2005 Impact of Weather on urban Freeway Traffic Flow Characteristics and Facility Capacity. Ames, IA. Center for Transportation Studies. Available: Last accessed January 12, 2012. Al Ghamdi, A.S. 2006 Analy sis of speeding in Saudi Arabia and e ffective ness of enforcement methods. Transportation Research Record 1969 1 9. Aljanahi, A.A.M., Rhodes, A.H., Metcalfe, A.V., 1999 Speed, speed limits and road traffic accidents under free flow conditions. Accident Analysis and Prevention 31 (1) 161 168. Al Kaisy A., Hall, F., 2001 Examination of effect of driver population at freeway reconstruction zones. Transportation Research Record 1776, 35 42. A lvarez, P., Mohammed H., 2010 Use of ITS data to calibrate micr oscopic simula t i on models for incident conditions. Submitted to AHB45 for Publication in Transportation Research Part C. American Transportation Research Institute (ATRI), and Science Applications I nternational Corporation (SAIC), 2010 2010 Traffic incide nt management handbook update Washington, D.C., U.S. Federal Highway Administration. Armour, M. 1986 The e ffect of p olice p re sence on urban driving speeds. ITE Journal 56 (2), 40 45. Ashton, R., 2006. Practical v ehicle e quipment. The Police Chief 73 (1) 40 Avrenli, K., Benekohal, R.F., Ramezani H., 2009. Traffic Flow Characteristics and Capacity in Intelligent Work Zones. West Lafayette, IN, NESTRANS Available: http://ntl.b Last accessed January 12, 2012. Baird, M., Cove, L., Horne, F., Jacobs B., 2003 service patrol (HELP) Program. Transportation Research Record 1856 87 95. Baker J S. 1 954. Effects of enforcement on vehicle s peeds. Highway Research Board Bulletin Number 91, 33 38. Ben Akiva, M.E., Choudhury, C.F., Toledo T., 2006. Lane changing Models. Available: df Last accessed January 12, 2012.

PAGE 146

146 Bierling, D. H. Li Y., 2009 The Texas move over act: Driver knowledge, understanding, and compliance College Station, TX, Texas Transportation Institute. Available: 2009 4.pdf Last accessed January 12, 2012. B lumentritt C.W., Ross D.W., Glazer J., Pinnell C., McCasland W.R., 1981. Guidelines for selection of ramp control systems N ational C ooperative H ighway R esearch P rogram Report No. 232 Cambridge Systematics & Texas Transportation Institute 2004. Traffic congestion and reliability: linking solutions to problems. Washing ton, D.C., Federal Highway Administration. Available: Last accessed January 12, 2012. Carson, J. 2010. Best practices in traffic incident Management Washington, D.C., Federal Highway Administration. Last accessed January 12, 2012. Chin, S.M., Fra nzese, O., Greene, D.L., Hawang H.L., 2002. Temporary losses of highway capacity and impacts on performance. Oak Ridge, TN, Oak Ridge National Laboratory Courage, K. Lee S., 2009. Development of a central data warehouse for statewide ITS and transporta tion data in Florida: phase III final report. Tallahassee, FL, Florida Department of Transportation. Available: 93_rpt.pdf Last accessed January 12, 2012. De Waard, D. Rootjers T 1994. An experimental study to evaluate the e ffectiv eness of different methods and i ntensitie s of law e nforcement on driving speed on motorways. Accident Analysis and Prevention 26 (6) 751 765. D ixon, K.K.., Hummer, J.E., Lorscheider A.R., 1996. Capacity for North Carolina freeway work zones. Transportation Research Record 1529, 27 34. Dougald, L.E. Demetsky M.J., 2008. Assessing return on investment of freeway safety service patrol programs Transportation Research Record 2047, 19 27. Dowling, R., Skabardonis, A., Carroll, M., Wang Z 2004. Methodology for measuring recurrent and nonrecurrent traffic congestion. Transportation Research Record 1867, 60 68. Elvik, R. Christensen P., 2007. T he deterrent effect of increasing fixed p enalties for traffic offences: the Norwegian experience. Journal of Safety Research 38 (6) 689 695.

PAGE 147

147 Federal Emergency Management Administration 2009. Emergency vehicle visibility and conspicuity study. (FA 323). Available: Last accessed January 12, 2012. Federal Highway Administration 2010. Focus on congestion relief: desc ribing the congestion problem. Available: Last accessed January 12, 2012. Federal Highway Administration: O ffice of Operations 2010 Reducing non recurring congestion November 18, 2010, Retrieved from Federal Highway Administration website: non cong.h tm Last accessed January 12, 2012. F ederal Highway Administration, 2009. Manual on uniform traffic control devices (MUTCD) Washington, D.C.: U.S. Department of Transportation, Federal Highway Administration. Fenno, D.W., Ogden, M.A., 1998. Freeway servi ce patrols: a state of the practice. Transportation Research Record 1634 28 38. Flannag an, M.J., Blower, D.F., Devonshire J.M., 2008. Effects of warning lamp color and intensity on driver vision. Ann Arbor, Michigan. The University of Michigan Transport ation Research In stitute Florida Department of Highway Safety and Motor Vehicles 2011. Facts and figures 2009 2011 Available: Last accesse d January 12, 2012. Florida Department of Transportation 2009. FDOT Traffic Viewer Online 2009 Available: Last accessed January 12, 2012. Florida Department of Transportation. 2010. Statewide Road Ranger Survey for Incident Responders : FY 2009 2010. Available: Final_2010.pdf Last accessed January 12, 2012. Garib, A., Radwan, A.E., Al Deek. H., 1997. Estimating magnitude and duration of incident delays Journal of Transportation Engineering 123 (6), 459 466. Gipps, P.G., 1986 A model for the structure of lane changing decisions. Transportation Research Part B 20 (5) 403 414. Giuliano, G. 1988. Incident characteristics, frequency, and duration on a high volume ur ban freeway, Irvine, CA: Institute of Transportation Studies, University of California, Irvine.

PAGE 148

148 Go ldenbeld, C. V an Schagen I 2005. The effects of speed e nforcement with mobile radar on speed and a ccidents. Accident Analysis and Prevention 37 (6) 1135 1144. Golob, T.F., Recker W.W., Leonard J.D., 1987. An analysis of the severity and incident d ura tion of truck involved freeway a ccidents. Accident Analysis and Prevention 19 (5) 375 395. Golob, T.F., Recker W.W. Alvarez V.M 2004. Freeway safe ty as a function of traffic flow. Accident Analysis and Prevention 36 (6) 933 946. Golob, T F., Recker W.W 2003. Relationships among urban freeway accidents, traffic flow, weather, and lighting conditions. Journal of Transportation Engineering 35 (4) 342 353. Goolsby, M.E., 1971. Influence of incidents on freeway quality of service. Highway Res earch Record No.349, 41 46. Haas, K.J., Jones, B., Kirk A.R., 2003. The effect of law enforcement deployment FHWA OR DF 04 04) U.S. Department of Transportation H adi, M., Xiao Y. Zhan C. Gan A 2010 Use of HCM freeway procedures combined with ITS data archives to assess the benefits of incident management systems. In Transportation Research Board of the National Academies, 90 th Annual Meeting CD ROM. Hauer, E. Ahlin F.J., 1982. Spee d enforcement and speed choice. Accident Analysis and Prevention 14 (4), 267 278. Heaslip, K., Kondyli, A., Arguea, D., Elefteriadou, L., Sullivan, F., 2009. Estimation of freeway work zone cap acity through simulation and field data. Transportation Research Record 2130, 16 24. H easlip, K., Louisell, C., Collura J., 2008. Driver population adjustment factors for the highway capacity manual work zone capacity equation. In Tran s portation Research Board of the National Academies, 88 th Annual Meeting CD ROM. Hidas, P. 2002. Modeling lane changing and merging in microscopic traffic simulation, Transportation Research Part C 10 (5) 351 371. Hranac, R., Sterzin, E., Krechmer, D., Rakha, H., Farzaneh M., 2006. Empirical studies on traffic flow in inclement weather. (FHWA HOP 07 073). U.S. Department of Transportation. Jones, B., Janssen, L., Mannering F., 1991. Analysis of the frequency and duration of freeway accidents in Seattle." Accident Analysis and Prevention 23 (4) 239 255.

PAGE 149

149 Khattak, A., Rouphail, N., Monast, K., Havel J., 2004. Method for priority ranking and expanding freeway service patrols. Transportation Research Record 1867 1 10. Khattak, A.J., Schofer, J.L., Wang M 1994 A simple ti me sequential procedure for predicting freeway incident duration. ( PATH UCB ITS PRR 94 26). Institute of Transportation Studies, University of California at Berkeley. Knoop, V L Hoogendoorn, S.P., Van Zuylen H., 2008. Capacity r ed uction at incidents: Empirical data collected from a helicopter Transportation Research Record 2071, 19 25. Knoop, V ., Hoogendoorn, S., Adams K., 2009. Capacity reductions at incident sites on motorways. Euopean Journal of Transport and Infrastructure Research 9 (4), 363 379. Kyte, M., Khatib, Z., Shannon, P., Kitchener F., 2001. Effects of weather on free flow speed. Transportation Research Record 1776 60 68. LaFountain, R., Schauffler, R., Strickland, S., Bromage, C., Gibson, S., Mason, A., Raftery W 2009 Exami ning the work of state courts: A national perspective from the court statistics project. National Center for State Courts, Available: sp/2008_files/EWSC_2008_cvr.jpg Last accessed January 12, 2012. Li, J., Lan, C., Gu, X., 2006. Estimation of incident delay and its uncertainty on freeway networks. Transportation Research Record 1959, 37 45. Lu, C. Elefteriadou, L. 2010. An investigation of freeway capacity before and during incidents. In Transportation Research Board of the National Academies 90 th Annual Meeting CD ROM. May, A.D. 1990 Traffic Flow Theory Prentice Hall. Maze, T.H., Schrock, S.D., Kamyab A 2000. Capacity of freeway work zone lane closures Mid Continent Transportation Symposium 2000 Proceedings Ames, Iowa Medina, J. C., Benekohal, R.F., Hajbabaie, A., Wang, M., Chitturi M.V., 2009. downstream effects of speed p ho to radar enforcement and other speed redu ction treatments on work zones. Transportation Research Record 2107 24 33. Nam, D., Mannering F., 2000. An exploratory hazard based analysis of highway incident duration. Transportation Research Part A 34 ( 2 ) 85 102. National Highway Traffic Safety Administration 2009. Traffic safety facts 2008: A compilation of motor vehicle crash data from the fatality analysis reporting system and the general estimates system. Available: http://www nrd. Last accessed January 12, 2012.

PAGE 150

150 N ational Research Council 2010. Highway Capacity Manual, 2010 Edition Transportation Research Board N oel, E. C., Dudek, C.L., Pendleton, O.J., McGee, H.W., Sabra Z.A 1987. Speed control through work zones: techniques evaluation and implementation guidelines. ( FHWA IP 87 4) U.S. Department of Transportation. Olmstead, T., 1999. Pitfall to avoid when estimating incident induced delay by using deterministic queuing models. Transportation R esearch Record 1683, 38 46. Ozbay, K., Noyan N 2006. Estimation of incident clearance times using bayesian networks approach. Accident Analysis and Prevention 38 (3), 542 555. Pal, R., Latoski, S.P., Sinha K.C., 1998 Investigation of freeway incident characteristics and their influence in planning an incident management program. Transportation Research Record 1634 46 55. Pande, A Abdel Aty M. Hsia L., 2005. Spatiotemporal variation of risk Preceding crashes on freeways. Transportation Research Re cord 1908, 26 36. PB Farradyne Inc. 2000 Traffic incident management h andbook Washington, D.C.; U.S. Department of Transportation, ITS Joint Program Office. Polus, A., Shwartzman Y., 1999. Flow characteristics at freeway work zones and increased deterr ent zones. Transportation Research Record 1657 18 23. Prevedouros, P., Halkias, B., Papandreou, K., Kopelias P 2008. Freeway incidents in the United States, United Kingdom, and Attica Tollway, Greece: characteristics, available capacity, and models. Tr ansportation Research Record 2047 57 65. Qi, Y., Smith B.L., 2004. Identifying nearest neighbors in a large scale incident data archive. Transportation Research Record 1879 89 98. Qi, Y. Teng H 2004. Integrated modeling approach to total incident delay. Transportation Research Record 1895 46 54. Q in, L., Smith B.L., 2001. Characterization of accident capacity reduction (UVACTS 15 0 48). Charlottesville, VA: University of Virginia Smart Travel Laboratory. Rakha, H., and B. Crowther., 2002. Compar ison of Greenshields, Pipes, and Van Aerde Car Following and Traffic Stream Models. Transportation Research Record 1802, 248 262. Rakha, H. 2007. Traffic Stream Calibration Software. Available: Last accessed March 26, 2012.

PAGE 151

151 Rakha, H ., Arafeh, M., 2010. Calibrating Steady State Traffic Stream and Car Following Models Using Loop Detector Data. Transportation Science, 44 (2), 151 168. Reddy, V, HNTB Corporation 2008. Evaluation of innovative safety treatments Volume 4, a study of the effectiveness of motorist awareness system in construction work zones. T allahassee, FL: Florida Department of Transportation Richards, S., Wunderlich, C., Dudek C ., 1999. Field evaluation of work z o ne speed control techniques. Transportation Research Record 1693 66 78. Rkha, H., Farzaneh, M., Arafeh, M., Sterizin E., 2008. Inclement weather impacts on freeway traffic stream behavior. Transportation Research Record 2071, 8 18. Saito, M. Bowie J 2003. Efficacy of speed monitoring displays in increasing speed limit compliance in highway work zones. Utah Dept. of Transportation, Research Division, Salt Lake City Sarasua, W.A., Davis, W.J., Clarke, D.B., Kottapally, J., Mulukutla P., 2004. Evaluation of interstate highway capacity for short term work zone lane closures. Transportation Research Record 1877 85 94. S chrank, D.L. Lomax T.J., 2010. 2010 Urban mobility report College Park, TX: Texas Transp ortation Institute, Texas A&M University. S chrank, D.L., Lomax T.J., 2009. 2009 U rban mobility report College Park, TX: Texas Transportation Institute, Texas A&M University. S inha, P., Hadi, M., Wang A 2007. Modeling reductions in freeway capacity du e to incidents in microscopic simulation models. In Tran s portation Research Board of the National Academies, 85 th Annual Meeting CD ROM. Sisiopiku, V.P. Patel H 1999. Study of the impact of police enfor cement on motorists' speeds. Transportation Research Record 1693 31 36. Skabardonis, A., Petty, K.F., Varaiya P.P 1999. Los Angeles I 10 field experiment: incident patterns. Transportation Research Record 1683, 22 30. Skabardonis, A., Petty, K.F., Noeimi, H., R ydzewski, D., Varaiya P.P., 1997. I 880 Field Experiment: Analysis of Incident Data. Transportation Research Record 1603 72 79. Skabardonis, A., Petty, K.F., Varaiya, P., Bertini R 1998. Evaluation of freeway service patrol at Los Angeles freeway site (PATH UCB ITS PRR 98 13). Univer sity of California, Berkeley.

PAGE 152

152 Smith, B.L., Qin L., Venkatanarayana R 2003 Characterization of freeway capacity reduction resulting from traffic accidents. Journal of Transportation Engineering 129 (4), 362 268. S mith, B.L., Yook, D. McGhee, C.C., 200 9. Investigation of enforcement techniques and technologies to support high occupancy vehicle and high occupancy toll operations. Charlottesville, VA: Virginia Transportation Research Council. Soole, D., Watson, B., A. Lennon A 2009. The impact of police speed enforcement practices on self reported speeding: An exploration of the effects of visibility and mobility. Proceedings of the 2009 Australasian Road Safety Research, Policing and Education Conference : Smarter, Safer Directions. November 10 12 2009, Sydney Conventio n and Exhibition Centre, Sydney. State of Florida. Florida State Statute s, 2010. Operation of vehicles and actions of pedestrians on approach of authorized emergency vehicle Available: Last accessed January 12, 2012. Sullivan, E.C. 1997 New m odel for p redicting f reeway i ncidents and i ncident d elays. Journal of Transportation Engineering 123 (4), 267 275. T ay, R. 2005. General and specific deterrent effects of traffic enforcement: do we have to catch offenders to reduce crashes? Journal of Transport Economics and Policy 39 ( 2 ) 209 223. Teng, H. Masinick J.P. 2004. An analysis of the impact of rubbernecking on urban freeway traffic Charlottesville, Va Center for Transportation Studies. Turner F airbank Highway Research Center, Daniel Cons ultants, Dudek & Associates, 1987. Speed Control Through Work Zones: Techniques, Evaluation, and Implementat ion Guidelines. ( FHWA IP 87 4 ) U .S. Depa rtment of Transportation U.S. Department of Justice Federal Bureau of Investigation 2010. Law Enforcement Officers Killed & Assaulted. Available: Last a ccessed January 12, 2012. U.S. Department of Justice Federal Bureau of Investigation 2010. Law Enforcement Officers Killed & Assaulted. Available: Last accessed January 12, 2012. Ullman, G.L, Ogden M.A., 1996. Analysis of major freeway incidents in Houston, Texas. Transportation Research Record 1554 221 227. Vaa, T. 1997. Increased Police Enforcement: Effects on Speeds. Accident Analysis and Prevention 29 (3), 373 385.

PAGE 153

153 Van Aerde, M., 1995. Single Regime Speed Flow Density Relationship for Congested and Uncongested Highways. Presented at 74th Annual Meeting o f the Transportation Research Board, Washington, D.C. Van Aerde M. and Rakha H., 1995. Multivariate Calibration of Single Regime Speed Flow Density Relationships, Vehicle Navigation and Information Conference (VNIS). Seattle, WA. Vlahogianni; E.I., Karlaft is, M.G., Golias, J.C., Halkias B.M., 2010. Freeway o pera tions, spatiotemporal incident c harac teristics, and secondary crash o ccurrence Transportation Research Record 2178 1 9 Wang, W., Chen, H., Bell. M.C., 2005. Vehicle breakdown duration modeling. Journal of Transportation and Statistics 8 (1) 75 84. Washburn, S.S., Yin, Y., Modi, V., Kulshrestha A 2010. Investigation of Freeway Capacity ( TRC FDOT 73157 2010 ) University of Florida Transportation Research Center. Well s J. D. 2004. Florida Highway Patrol Emergency Lighting R esearch & Prototype Evaluation. Wu, N., Rakha, H., 2009. Derivation of Van Aerde Traffic Stream Model from Tandem Queuing Theory. Tra nsportation Research Record 2124 18 27. Yin, Y ., 2006. Optimal fleet allocation of freeway service patrols. Networks and Spatial Economics 6, 221 234. Yin, Y., Washburn, S.S., Wu, D., Kulshrestha, A., Modi, V., Michalaka, D., Lu, J. 2012. Managed Lane Operations Adjusted Time of Day Pricing vs. Near Real T ime Dynamic Pricing: Volume I: Dynamic Pricing and Operation of Managed Lanes. (UF Project 00081551) University of Florida Transportation Research Center.

PAGE 154

154 BIOGRAPHICAL SKETCH Grady Carrick is an accomplished manager with the Florida Highway Patrol and a lifelong learner. He has served on numerous local, state, and national committees and panels related to transportation safety and traffic operations including assignments with the Federal Highway Administration (FHWA) National Highway Traffic Safet y Administration (NHTSA) and the International Association of Chiefs of Police (IACP) He is a graduate of the FBI National Academy and he possesses graduate degrees in Criminal Justice, Public Administration, and Civil Engineering (Transportation).