Pavement Marking Specifications Compliance and Modeling of Retroreflectivity Performance by Luana Clara de Sena Monteiro Ozelim A thesis submitted to the Graduate Faculty of Auburn University in partial fulfillment of the requirements for the Degree of Master of Science Auburn, Alabama May 6, 2012 Keywords: pavement markings, thermoplastic, retroreflectivity, service life, modeling Copyright 2012 by Luana Clara de Sena Monteiro Ozelim Approved by Rod E. Turochy, Chair, Associate Professor of Civil Engineering Robert L. Vecellio, Associate Professor of Civil Engineering Jeffrey J. LaMondia, Assistant Professor of Civil Engineering ii Abstract This study provided an analysis of a database containing 7,840 observations of 40 highway pavement marking projects constructed in Alabama in 2007 to determine the extent to which marking properties complied with ALDOT specifications. A statistical analysis was also performed on these observations to determine mean, standard deviation, and coefficient of variation by color and type of markings for marking properties. Results showed that retroreflectivity had significant variation. This research also developed models of retroreflectivity performance over time for thermoplastic markings, which was executed for 15 projects that had measurements of retroreflectivity for the same locations in 2007, 2008, 2009, and 2010. A linear model considering age, initial retroreflectivity, and AADT as independent variables was the one which best represented data, with R2 values of 0.398 for white markings and 0.479 for yellow markings. iii Acknowledgements The preparation of this thesis during these years was possible because of the incredible people I have in my life. I am deeply grateful to Dr Rod E. Turochy for giving me the confidence to keep going with my research and the guidance to avoid getting lost when obstacles appeared. I also thank him for the patience when helping me with my English, as I am Brazilian, I know it was tough. Rod Turochy was a wonderful advisor and I just thank him and his wife, Kathy Gregory, for being the most supportive people here in the United States. I am very grateful to all staff of the Department of Civil Engineering of Auburn University for the help and kindness during hard times. Dr Robert L. Vecellio and Dr Jeffrey J. LaMondia, members of my committee, thank you for the advices and availability to always help and clarify any doubts I had related to the research. Gabriel Maciel Leite, an inspiration for me and for everybody that knows him, thank you for being such this strong person that helped me since a long time ago and changed my life. Derong Mai, my friend, I am so grateful for all your help when I needed, all the conversations, and all the times you made me laugh. Despite the geographical distance, my parents were a constant source of support. Unfortunately, their English is not very good, so I have to thank them in Portuguese. Juvenal e Lenilza, devo tudo de bom que acontece na minha vida a voc?s. Obrigada por sempre confiarem em mim e me deixarem viver meus sonhos. M?e, sei que n?o foi f?cil ficar tanto tempo nos Estados Unidos sem falar ingl?s, sempre ouvindo meus desabafos, sempre paciente com meu iv mau humor e estresse em v?rios momentos, eu nunca teria conseguido sem voc?. I also would like to tell my brother, Luan, that he is the best person I know, thank you for just being you. I know I can do anything because my family is the best, none of this could have happened without them, so this thesis is dedicated to them. v Table of Contents Abstract ........................................................................................................................................... ii Acknowledgements ........................................................................................................................ iii List of Tables ................................................................................................................................. ix List of Figures ................................................................................................................................ xi Chapter One .................................................................................................................................... 1 Introduction ..................................................................................................................................... 1 1.1. Background .......................................................................................................................... 2 1.2. Objectives ............................................................................................................................ 3 1.3. Scope .................................................................................................................................... 4 1.4. Outline.................................................................................................................................. 5 Chapter Two.................................................................................................................................... 6 Literature Review............................................................................................................................ 6 2.1. Characterization of Pavement Markings .............................................................................. 7 2.2. Retroreflectivity Threshold .................................................................................................. 9 2.3. Pavement Marking Retroreflectivity Modeling ................................................................. 14 vi 2.4. Pavement Marking Materials Specifications and Requirements ....................................... 21 2.5. Economic Evaluation of Thermoplastic Materials ............................................................ 23 2.6. Service Life of Thermoplastic Pavement Markings .......................................................... 29 2.7. Summary of findings.......................................................................................................... 31 Chapter Three................................................................................................................................ 34 Methodology ................................................................................................................................. 34 3.1. ALDOT specifications study ............................................................................................. 34 3.2. Data Mining and Analysis ................................................................................................. 35 3.3. Retroreflectivity Modeling................................................................................................. 40 3.3.1. Preparation of Data for Modeling ............................................................................... 44 3.3.2. Model Fitting .............................................................................................................. 46 3.4. Benefit/Cost Calculation .................................................................................................... 49 3.5. Modeling of Retroreflectivity Over Time .......................................................................... 49 3.6. Summary of Chapter Three ................................................................................................ 52 Chapter Four ................................................................................................................................. 53 Results ........................................................................................................................................... 53 4.1. ALDOT Data Mining and Analysis ................................................................................... 53 4.2. Retroreflectivity Modeling................................................................................................. 55 4.2.1. Retroreflectivity Curve ............................................................................................... 55 4.2.2. Model Fitting Analysis ............................................................................................... 56 vii 4.3. Benefit/Cost Calculation .................................................................................................... 59 4.4. Modeling of Retroreflectivity Over Time .......................................................................... 61 4.4.1. Approach One: Initial Retroreflectivity and Age of Marking as Independent Variables????????????????????????????????63 4.4.2. Approach Two: Initial Retroreflectivity, Age of Marking, and Traffic Volume as Independent Variables........................................................................................................... 69 4.4.3. Service Life ................................................................................................................. 71 4.5. Summary of Chapter Four ................................................................................................. 74 Chapter Five .................................................................................................................................. 76 Conclusions and Recommendations ............................................................................................. 76 5.1. Conclusions ........................................................................................................................ 76 5.2. Recommendations .............................................................................................................. 78 5.3. Recommendations for Subsequent Studies ........................................................................ 79 References ..................................................................................................................................... 81 Appendix A ................................................................................................................................... 85 Project STPSA-0185(500): Pictures ............................................................................................. 85 Appendix B ................................................................................................................................... 89 List of Mileposts by Project .......................................................................................................... 89 Appendix C ................................................................................................................................... 93 Retroreflectivity Modeling for Yellow Markings: Initial Retroreflectivity and Age as Independent Variables .................................................................................................................. 93 Appendix D ................................................................................................................................... 98 viii Retroreflectivity Modeling for Yellow Markings: Initial Retroreflectivity, Age, and Traffic Volume as Independent Variables ................................................................................................ 98 ix List of Tables Table 1 ? Threshold dry retroreflectivity values suggested by FHWA to define end of pavement marking service life....................................................................................................................... 11 Table 2 ? Minimum Maintained Retroreflectivity Levels for Longitudinal Pavement Markings 12 Table 3 ? Recommended minimum retroreflectivity values (mcd/m2/lux) .................................. 14 Table 4 ? Class of Traffic Stripe ................................................................................................... 21 Table 5 ? Initial Daytime Chromaticity for yellow materials ....................................................... 23 Table 6 ? Initial Daytime Chromaticity for yellow materials ....................................................... 23 Table 7 ? Summary of models to predict retroreflectivity ............................................................ 32 Table 8 ? Summary of Predicted Service Life .............................................................................. 33 Table 9 ? List of the 40 Pavement Marking Projects in the 2007 ALDOT Database .................. 35 Table 10 ? Meaning of ALDOT Database Entries ....................................................................... 37 Table 11 ? Minimum Required Values for Pavement Markings .................................................. 38 Table 12 ? Acceptable Values for Pavement Markings................................................................ 38 Table 13 ? Initial Daytime Chromaticity Coordinates (Corner Points) ........................................ 39 Table 14 ? Projects Included in the Retroreflectivity Modeling ................................................... 41 Table 15 ? Total Number of Retroreflectivity Curves .................................................................. 44 Table 16 ? ALDOT Historical Traffic Data ................................................................................. 46 Table 17 ? Material Property Compliance with Specifications .................................................... 54 Table 18 ? Color Check Comparison............................................................................................ 54 x Table 19 ? Statistical Calculations ................................................................................................ 55 Table 20 ? Retroreflectivity Values of Project NHF-STPSAF-0053(525), for MP36 ................. 56 Table 21 ? Models in the Literature that Best Represented Actual Data ...................................... 58 Table 22 ? Example of Benefit/Cost and Service Life calculations ............................................. 59 Table 23 ? Developed Models considering initial retroreflectivity and age as independent variables ........................................................................................................................................ 62 Table 24 ? Developed Models considering initial retroreflectivity, age, and AADT as independent variables.................................................................................................................... 62 Table 25 ? Predicted Service Life: Initial Retroreflectivity and Age as Independent Variables .. 72 Table 26 ? Predicted Service Life: Initial Retroreflectivity, Age, and AADT as Independent Variables ....................................................................................................................................... 73 xi List of Figures Figure 1 ? Glass Bead Retroreflection ............................................................................................ 9 Figure 2 ? Relative crash probability versus low retroreflectivity on freeways: white edge lines 13 Figure 3 ? Relative crash probability versus low retroreflectivity on freeways: yellow edge lines ............................................................................................................................................... 13 Figure 4 ? Pattern representative of newly placed pavement markings ....................................... 16 Figure 5 ? Pattern for established sites ? markings older than about 300 days ............................ 17 Figure 6 ? Measuring user benefit: Kansas and Virginia ............................................................. 27 Figure 7 ? Alternative measurement of benefit ............................................................................ 28 Figure 8 ? Information from 2007 Database ................................................................................. 36 Figure 9 ? Initial Daytime Chromaticity: White ........................................................................... 39 Figure 10 ? Initial Daytime Chromaticity: Yellow ....................................................................... 40 Figure 11 ? Location of the 15 Projects with Retroreflectivity Data from 2007 to 2010 ............. 42 Figure 12 ? Section of Project STPSA-0185(500), on SR-185 .................................................... 43 Figure 13 ? Example of Correspondence from Stations to Mileposts .......................................... 44 Figure 14 ? Four Methods of Riemann Summation ..................................................................... 47 Figure 15 ? Comparison Between Areas Under Curves ............................................................... 48 Figure 16 ? Plot of Retroreflectivity Values of Project NHF-STPSAF-0053(525), for MP36 .... 56 Figure 17 ? Comparison between Models for Project STPSA-0185(500), MP10, Solid White .. 57 Figure 18 ? Thamizharasan et al. Linear Extrapolation of Actual Data ....................................... 60 xii Figure 19 ? Sitzabee et al. Extrapolation of Actual Data ............................................................. 60 Figure 20 ? Linear Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables .................................................................................................................. 63 Figure 21 ? White Markings Linear Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables .......................................................... 64 Figure 22 ? Power Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables .................................................................................................................. 65 Figure 23 ? White Markings Power Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables .......................................................... 66 Figure 24 ? Quadratic Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables ...................................................................................................... 67 Figure 25 ? White Markings Quadratic Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables .......................................................... 67 Figure 26 ? Exponential Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables ...................................................................................................... 68 Figure 27 ? White Markings Exponential Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables ............................................ 69 Figure 28 ? Linear Regression estimates for White Markings: Initial Retroreflectivity, Age, and AADT as Independent Variables .................................................................................................. 70 Figure 29 ? White Markings Linear Regression for Different Initial Retroreflectivity and AADT Values: Initial Retroreflectivity, Age, and AADT as Independent Variables .............................. 71 1 Chapter One Introduction Highway systems are comprised of a number of elements, each with own role. Pavement markings are elements which can convey regulations, guidance, warnings, and may also be used as supplements to other traffic control devices. As pavement markings guide road users, providing them information to understand what happens in the roadway, they can improve safety by reducing the risk of accidents. There are many different materials and colors used for pavement markings (e.g. primarily white and yellow in the U.S.). Many factors can influence the performance of pavement markings over time, and consequently, their useful life. Higher traffic volumes, for example, may cause more rapid marking degradation, a centerline marking might have greater wear than an edge line, and higher percent of trucks may cause greater deterioration. Also, geographic location and climate can influence pavement marking durability. Pavement markings need to be visible, during the day as well as at night, to support the driver?s understanding of the roadway, supplementing other traffic control devices or used alone to convey regulations, guidance, or warnings. A sufficient level of visibility, or serviceability, can be determined through measurement of a number of properties. Minimum thickness requirements support visibility of the marking and also avoid premature maintenance. Chromaticity quantifies the color of the marking, and it needs to meet color specifications to distinguish a yellow marking from a white marking very clearly. For each chromaticity, there is a 2 unique optimal color having its maximum luminance, which represents the amount of light emitted from a particular area. Retroreflectivity measures the incident light from a vehicle?s headlights reflected back toward the general area of the light source, particularly the eye of the driver of the vehicle; this property is critical to nighttime visibility. The rate of degradation of pavement marking retroreflectivity can be influenced by many factors, such as type of material, geographic location and climate, traffic volume, and percent of heavy vehicles. Determination of service life is essential to maintenance of markings, and it is closely related to retroreflectivity. This topic is becoming increasingly important, which can be attested to by the significant development of research on establishing minimum retroreflectivity levels, including a proposed amendment to the Manual on Uniform Traffic Control Devices (MUTCD) that would create a national standard (FHWA, 2010). 1.1. Background Over the past decade, much advancement has been made with respect to quality and durability of pavement marking materials. When quantifying the performance of pavement markings, properties such as retroreflectivity, thickness, luminance, and chromaticity are typically measured. Agencies responsible for construction and maintenance of highways typically develop criteria for pavement markings to determine whether a new installation is acceptable. The Standard Specifications for Highway Construction, produced by the Alabama Department of Transportation (ALDOT), is the document which provides the requirements for all projects performed in the State of Alabama (ALDOT, 2008). The edition applicable to the data analyzed in this study was published in 2006 and, since then, several changes on traffic 3 stripe, markings, and legends specifications have been made and included in a series of five Special Provisions during 2007. The most recent publication is the draft of the 2012 edition of the Standard Specifications for Highway Construction, but only the 2007 Special Provisions apply to projects studied in this research. There has not been a detailed study related to these changes in ALDOT specifications and the quality of pavement marking projects in Alabama have not recently been evaluated. As a result, there are some unanswered questions, such as the effects of ALDOT specification changes, the cost-effectiveness of materials used by the ALDOT for pavement markings, and durability of retroreflectivity. A database containing observations of pavement marking properties on 40 projects in Alabama in 2007 was obtained from ALDOT. Additionally, data from 2008 to 2010 for 15 of these projects were available for analysis. The 2007 database includes properties such as thickness, retroreflectivity, luminance, and chromaticity; the 2008, 2009, and 2010 data has only retroreflectivity measurements. This thesis utilized this information to answer the important questions mentioned above related to pavement markings in Alabama. 1.2. Objectives To guide the analysis of data applied to this study, and also understand how pavement markings in Alabama have performed, the main objectives of this research project are: 1. Document recent changes in ALDOT specifications (2007 Special Provisions); 2. Mine the database to: a. Determine the extent to which observations from the 40 projects from 2007 meet specifications; 4 b. Execute a statistical analysis of the observations from the 40 projects from 2007 (mean, standard deviation, and coefficient of variation by color and type, for each property measured); c. Perform a benefit/cost analysis of pavement markings used in Alabama; d. Model change in retroreflectivity over time, utilizing data from 2007 to 2010; e. Develop a framework for future research, stating limitations of this study and how the results of this research can be extended in long-term studies. 1.3. Scope A general analysis of 7,840 observations from all 40 projects in the 2007 database was performed. Properties of the markings of the projects in Alabama were compared to specifications to examine the percentage of observations that met ALDOT requirements. A statistical analysis, including mean, variance, standard deviation, and coefficient of variation is also provided for properties such as chromaticity, thickness, luminance, and retroreflectivity. This study also examined the feasibility of establishing a methodology to determine benefits based on retroreflectivity. Cost data were obtained from the ALDOT Tabulation of Bids database. Retroreflectivity modeling is also presented. Mathematical models were developed associating available variables, within the limitations of the database. This thesis can also serve as a starting point for more detailed studies in the future. Finally, the data included in these analyses were collected by ALDOT, and these data represent only a small portion of all ALDOT- sponsored projects that were constructed in 2007. 5 1.4. Outline Chapter Two includes a detailed literature review which consists of a summary of pavement markings characterization, including materials, colors, and measurable properties related to them, such as thickness, retroreflectivity, luminance, and chromaticity. This chapter also includes an overview of pavement marking retroreflectivity models in the literature. In this chapter, a summary of pavement marking materials specifications and requirements in Alabama is presented, according to the 2006 Edition of the ALDOT Standard Specifications for Highway Construction and subsequent Special Provisions. Methods regarding economic evaluation, including benefit/cost estimation examples, and details about pavement markings service life and related retroreflectivity thresholds conclude the literature review. Chapter Three contains the methods used to achieve the objectives of this research project. It provides an overview of changes in ALDOT specifications during 2007, as well as the approach to performing data mining and analysis of the 2007 database. Chapter Three also provides a framework for the retroreflectivity modeling performed in this thesis. This process includes the preparation of data for modeling and application of existing models in the literature to the current dataset. The feasibility of a benefit/cost analysis, given the available data, is also provided in this chapter. Finally, the development of several new models of retroreflectivity over time is explained in Chapter Three. Chapter Four includes the results related to the application of the methodology explained in Chapter Three. Chapter Five presents the conclusions of this research and gives recommendations for subsequent studies, considering limitations of the existing dataset and approaches to overcome them. 6 Chapter Two Literature Review Pavement markings have an essential role in the highway system (Fu and Wilmot, 2008). In some cases, markings are used to supplement other traffic control devices such as signs and signals. In other instances, markings are used alone to effectively convey regulations, guidance, or warnings in ways not obtainable by the use of other devices (FHWA, 2009). The risk of accidents in roadways is reduced due to pavement markings, as they provide the driver?s understanding of the roadway and his or her ability to stay on course (Montebello et al., 2000). Characteristics of pavement markings can vary among the different materials and colors available, these variables include luminance, retroreflectivity, and cost for each type of pavement marking. An important issue with pavement markings is durability; the development of mathematical models that can predict the service life of pavement markings is a common approach to quantifying durability. These models typically refer to retroreflectivity as the main variable of analysis and typical independent variables are initial retroreflectivity, age of marking, and traffic. A highway agency typically develops requirements that have to be met. In Alabama, acceptable values for pavement marking parameters are presented in ALDOT?s ?Standard Specifications for Highway Construction? (ALDOT, 2008). Economic evaluation is also an important analysis for the project. This chapter describes general characteristics of pavement markings, indicating the most commonly used types. It also provides an overview of existing models to predict life service of 7 pavement markings. Finally, it presents a summary of different methods of economic evaluation analyses. 2.1. Characterization of Pavement Markings There are a wide variety of characteristics, costs, and benefits among the different pavement marking materials which can make it a challenging task deciding which factors should be considered when selecting the type of pavement marking material for a particular road. There are several manufacturers of pavement marking materials competing for business and they distribute information on their products and on the competition?s products, making comparisons based on this information difficult to understand and somewhat unreliable. (Montebello et al., 2000). There are numerous types of materials used for pavement markings in the field today, including paint, epoxy, tape, and thermoplastic (Thomas et al., 2001). Each material has its own set of unique characteristics related to durability, retroreflectivity, installation cost, and life-cycle cost. Marking types are also wide-ranging and they include pavement and curb markings, delineators, colored pavements, channelizing devices, and islands (FHWA, 2009). Most used materials for pavement and curb markings placement are paints or thermoplastics (FHWA, 2009). The materials used for markings should provide the specified color throughout their useful life. The Manual on Uniform Traffic Control Devices (MUTCD) determines that markings shall be yellow, white, red, blue, or purple. Black, in conjunction with one of the aforementioned colors, is also a usable color (FHWA, 2009). Pavement markings include longitudinal lines, transverse lines, words, and symbols. Longitudinal markings include centerlines, lane lines, and 8 edge lines on paved streets and highways (Fu and Wilmot, 2008). The general functions of longitudinal lines are: a double line indicates maximum or special restrictions, a solid line discourages or prohibits crossing (depending on the specific application), a broken line indicates a permissive condition, and a dotted line provides guidance or warning of a downstream change in lane function (FHWA, 2009). White and yellow are the two most commonly used colors for pavement markings (Fu and Wilmot, 2008). White markings, when used for longitudinal lines, delineate the separation of traffic flows in the same direction or the right-hand edge of the roadway. Yellow markings, when used for longitudinal lines, delineate the separation of traffic traveling in opposite directions, the left-hand edge of the roadways of divided highways and one-way streets or ramps, or the separation of two-way left-turn lanes and reversible lanes from other lanes (FHWA, 2009). Marking systems should offer the best possible performance at the lowest possible cost. Regarding performance, the purpose of the markings is to facilitate safe and efficient traffic flow on highways (Cuelho et al., 2003). Configurations and visibility requirements of pavement markings are generally well-defined by publications such as the Manual on Uniform Traffic Control Devices (MUTCD) (Cuelho et al., 2003). Each standard marking shall be used only to transmit the meaning prescribed for that marking in the MUTCD (FHWA, 2009). All necessary markings should be in place before any new highway, private road open to public travel, paved detour, or temporary route is opened to public travel (FHWA, 2009). Service life is one of the first factors to be considered when choosing a pavement marking material. Many factors can influence the service life of a pavement marking material including weather conditions, winter maintenance activities, installation conditions and quality, retroreflective optics used, binder type, binder thickness, binder color, traffic volume, and the 9 minimum selected retroreflectivity level. Among these, the major factors that are known to have significant impacts on marking service life are traffic volume and minimum required pavement marking retroreflectivity level (Songchitruksa et al., 2011). 2.2. Retroreflectivity Threshold The determination of when a pavement marking material is no longer serviceable is rather complex. Given a quantified terminal condition of a pavement marking material, it is not a simple task to forecast the remaining service life. The performance of a pavement marking material has been judged based primarily on its retroreflectivity. Retroreflectivity has been used extensively in past studies as an important factor in analyzing the performance and cost- effectiveness of a pavement marking material (Zhang et al., 2006). Retroreflectivity can be defined as the portion of incident light from a vehicle?s headlights reflected back toward the eye of the driver of the vehicle. Retroreflectivity is provided in pavement marking materials by glass or ceramic beads that are partially embedded in the surface of the material (Thomas et al., 2001). Figure 1 shows how retroreflection occurs. Figure 1 ? Glass Bead Retroreflection SOURCE: Thomas et al., 2001 10 Pavement markings are typically retroreflective. This retroreflective property of the pavement markings is essential for nighttime visibility. Retroreflectivity is typically measured in units of millicandelas per square meter per lux (mcd/m2/lux) using retroreflectometers. The candela is the International System (SI) unit for luminous intensity, in a given direction, of a source that emits monochromatic radiation of frequency 540?1012 hertz and that has a radiant intensity in that direction of 1?683 watt per steradian. A common candle emits light with roughly 1 candela luminous intensity. The SI unit for illumination is defined as lux, and it represents the illumination produced by a luminous flux of 1 lumen distributed uniformly over an area of 1 square meter, or the illumination produced at a surface all points of which are at a distance of one meter from a uniform point source of one candela (NIST, 1979). According to the MUTCD, markings that have to be visible at night shall be retroreflective except for the cases when ambient illumination assures that the markings are adequately visible (FHWA, 2009). In 2006, The National Cooperative Highway Research Program sponsored a study focused on non-intersection, non-daylight crashes in California during 1992-1994 and 1997-2002 and related them to the retroreflectivity of the longitudinal pavement markings on the road at the time of the crashes. Over 118,000 crashes were considered in the study, which covered over 5,000 miles of state maintained freeways and highways in California. A main finding of the NCHRP study is that the amount of retroreflectivity is not important to driver safety as long as the marking is present and visible to drivers (Bahar et al., 2006). Therefore, there is a need to specify a minimum retroreflectivity value to determine if a marking is ?visible?. Many researchers adopt FHWA candidate criteria for minimum pavement marking retroreflectivity in their studies (FHWA, 2000). These recommended guidelines are impacted by three factors: speed, roadway type, and the presence/absence of raised retroreflective pavement 11 markers (RRPM) or lighting (Migletz and Graham, 2002). For freeways, minimum guideline retroreflectivity values of 150 mcd/m2/lux and 100 mcd/m2/lux are recommended for white and yellow pavement markings, respectively, when there is no RRPM or lighting; while 70 mcd/m2/lux is used for both white and yellow pavement markings when there is RRPM or lighting (FHWA, 2000). Table 1 shows these values. Table 1 ? Threshold dry retroreflectivity values suggested by FHWA to define end of pavement marking service life Material Roadway type/speed classification Non-freeway ? 40 mph Non-freeway ? 45 mph Freeway ? 55 mph White 85 100 150 White with lighting or RRPM 30 35 70 Yellow 55 65 100 Yellow with lighting or RRPM 30 35 70 SOURCE: FHWA, 2000 When installed, the retroreflectivity of yellow markings is typically about 35% lower than that of white markings. As white and yellow markings are usually replaced at the same time, there is a lower minimum value for yellow markings, as shown in Table 1 (Fu and Wilmot, 2008). In 2010, the Federal Highway Administration (FHWA) had drafted a revision to the 2009 Edition of the Manual on Uniform Traffic Control Devices (MUTCD) to specify minimum retroreflectivity values for the pavement marking standard (FHWA, 2010). The FHWA is currently reviewing the docket comments received and the proposed revisions regarding maintaining minimum retroreflectivity of longitudinal pavement markings are tentatively designated as Revision 1 to the 2009 edition of the MUTCD. The proposed revision establishes that public agencies or officials having jurisdiction shall use a method designed to maintain 12 retroreflectivity of white and yellow longitudinal pavement markings, at or above the minimum levels in Table 2. (FHWA, 2010). Table 2 ? Minimum Maintained Retroreflectivity Levels for Longitudinal Pavement Markings Roadway Type Posted Speed (mph) ? 30 35 - 50 ? 55 Two-lane roads with centerline markings only n/a 100 250 All other roads n/a 50 100 SOURCE: Based on FHWA, 2010 Parker and Meja (2003) studied the relationship between retroreflectivity levels and user perception, depending on user?s age. In addition to retroreflectivity, subjective ratings from a survey conducted with the participation of the New Jersey driving public along a 32-mi circuit were measured. Multiple regression techniques were used to correlate the average scores reported by the study participants for each specific roadway section with the corresponding measured retroreflectivity. The threshold value of acceptable versus unacceptable retroreflectivity, for both yellow and white markings, was between 80 and 130 mcd/m2/lux for New Jersey drivers younger than 55 and between 120 and 165 mcd/m2/lux for drivers over 55. Smadi et al.?s (2008) analysis of safety effectiveness related to marking retroreflectivity showed that low retroreflectivity, less than 200 mcd/m2/lux, is correlated to a higher crash probability. Values higher than 200 mcd/m2/lux did not show significant increase in crash probability (Smadi et al., 2008). Figure 2 and Figure 3 show the behavior of crash probability for retroreflectivity lower than 200 mcd/m2/lux for freeways. 13 Figure 2 ? Relative crash probability versus low retroreflectivity on freeways: white edge lines SOURCE: Smadi et al., 2008 Figure 3 ? Relative crash probability versus low retroreflectivity on freeways: yellow edge lines SOURCE: Smadi et al., 2008 14 In 2008, Debaillon et al. (2008) used a computer model called the Target Visibility Predictor (TarVIP) to study pavement marking retroreflectivity needs. Key factors affecting pavement marking visibility included pavement marking configuration, pavement surface type, vehicle speed, vehicle type, and presence of raised retroreflective pavement markers. The recommended values of minimum retroreflectivity can be observed in Table 3. Table 3 ? Recommended minimum retroreflectivity values (mcd/m2/lux) Roadway Marking Configuration Without RRPMs With RRPMs ? 50 mph 55 - 65 mph ? 70 mph Fully marked roadways (with centerline, lane lines, and edge lines, as needed) 40 60 90 40 Roadways with centerlines only 90 250 575 50 SOURCE: Debaillon et al., 2008 2.3. Pavement Marking Retroreflectivity Modeling In the mid-1990s Michigan State University (MSU) evaluated the performance of several pavement marking materials for the Michigan DOT (Lee et al., 1999). Lee et al. (1999) developed the following linear regression model for both white and yellow thermoplastic markings. In order to establish if a model is good to represent the behavior of actual data, goodness-of-fit measures can be considered. One of these measures is the coefficient of determination, R2, which represents the fraction of total variation in the dependent variable that is explained by the independent variables, and its value ranges from 0 to 1, with 1 indicating that the regression line perfectly fits actual data. The R? value for Lee et al (1999) model is 0.14: g1851=?0.3622g1850+254.82 (Eq. 2.1) 15 where Y = retroreflectivity of pavement markings (mcd/m2/lux); X = age of marking in days. In 2002, Abboud and Bowman (2002 [2]) determined pavement marking retroreflectivity using field retroreflectivity readings for 520 mi of longitudinal pavement markings in 9 Alabama counties. The minimum retroreflectivity threshold was determined to be 150 mcd/m2/lux. Logarithmic regression analysis was used to establish the following relationship between pavement marking retroreflectivity and prolonged traffic exposure, for white thermoplastic markings: g1844g3013 =?70.806lng4666g1848g1831g4667+639.66 (Eq. 2.2) where RL=pavement marking retroreflectivity (mcd/m2/lux); ln = natural logarithm; VE = vehicle exposure. Vehicle exposure is the total number of vehicles that have traversed a point on a highway up to the time frame of interest and represents the prolonged effect of traffic over time, combining the effects of marking age and traffic volume on the deterioration rate, and it is a function of time and traffic volume per lane and was expressed by Abboud and Bowman as: g1848g1831 =g1827g1830g1846g3013g3041 ?g1842g1839g1853g1859g1857?30.4?10g2879g2871 (Eq. 2.3) where: g1827g1830g1846g3013g3041 = average daily traffic per lane (thousands of vehicles/day/lane); g1842g1839g1853g1859g1857 = age in months (using a month-to-day conversion factor of 30.4). 16 Abboud and Bowman (2002 [1]) developed another exponential regression model to depict the relationship between pavement-marking retroreflectivity and vehicle exposure (VE). The value of R? for this method, related to white thermoplastic markings is 0.58: g1844g3013 =?70.806lng4666g1848g1831g4667+150.55 (Eq. 2.4) Thamizharasan et al. (2003) identified patterns of retroreflectivity change over time in South Carolina, as it can be observed in Figures 4 and 5. Figure 4 shows the first pattern, where retroreflectivity increases for a short period of time, usually before 300 days after pavement marking application, then gradually decreases thereafter, for newly placed markings. The initial increase in retroreflectivity is because of more glass beads becoming exposed after some amount of wear. Figure 4 ? Pattern representative of newly placed pavement markings SOURCE: Thamizharasan et al., 2003 The second pattern, which can be observed in Figure 5, is where retroreflectivity decreases gradually with time. This is for the well-established markings that have passed the initial increase period, defined by Thamizharasan et al. as 300 days. 17 Figure 5 ? Pattern for established sites ? markings older than about 300 days SOURCE: Thamizharasan et al., 2003 Thamizharasan et al. (2003) developed two models to predict marking retroreflectivity, including a non-linear model for the time the retroreflectivity increases when markings are newly applied and a linear model for the time retroreflectivity decreases to a minimum value. The models were stratified by marking color (white or yellow), surface type (AC or PCC), and marking material (thermoplastic or epoxy). Traffic volume was not found to be significant in the analysis. The non-linear model, for white thermoplastic materials, has a R? value of 0.22: g1830g1861g1858g1858g1857g1870g1857g1866g1855g1857uni0009g1861g1866uni0009g1844g1857g1872g1870g1867g1870g1857g1858g1864g1857g1855g1872g1861g1874g1861g1872g1877=?0.0005?g1830g1853g1877g1871g2870+0.18?g1830g1853g1877g1871+1.10 (Eq. 2.5) where: g1830g1853g1877g1871 = age of pavement marking (days). For yellow thermoplastic materials, the R? value is 0.20: g1830g1861g1858g1858g1857g1870g1857g1866g1855g1857uni0009g1861g1866uni0009g1844g1857g1872g1870g1867g1870g1857g1858g1864g1857g1855g1872g1861g1874g1861g1872g1877=?0.0001?g1830g1853g1877g1871g2870+0.04?g1830g1853g1877g1871+1.23 (Eq. 2.6) The linear model, for white thermoplastic materials, has R? value of 0.47: g1830g1861g1858g1858g1857g1870g1857g1866g1855g1857uni0009g1861g1866uni0009g1844g1857g1872g1870g1867g1870g1857g1858g1864g1857g1855g1872g1861g1874g1861g1872g1877=?0.06?g1830g1853g1877g1871?6.80 (Eq. 2.7) 18 For yellow thermoplastic materials, R? value is 0.21: g1830g1861g1858g1858g1857g1870g1857g1866g1855g1857uni0009g1861g1866uni0009g1844g1857g1872g1870g1867g1870g1857g1858g1864g1857g1855g1872g1861g1874g1861g1872g1877=?0.03?g1830g1853g1877g1871?3.63 (Eq. 2.8) The National Transportation Product Evaluation Program (NTPEP) has collected retroreflectivity data from various sites located in different regions of California. The variables contained within the NTPEP data set include age of marking, color, material type, traffic volume, pavement surface, climate region, and snow removal (Bahar et al., 2006). The polynomial model developed to predict the retroreflectivity was: g1844 = 1g2010 g2868+g2010g2869?g1827g1859g1857+g2010g2870?g1827g1859g1857g2870 (Eq. 2.9) where: g1844 = retroreflectivity of pavement marking (mcd/m2/lux); g2010g2868,g2010g2869,g2010g2870uni0009 = model parameters to be estimated; g1827g1859g1857 = age of pavement marking (months). For white thermoplastic markings in a hot humid climate and no usual snow removal, the values found in the NTPEP study were g2010g2868 = 2.42 x 10-3, g2010g2869 = 1.32 x 10-4, and g2010g2870 = -1.18 x 10-6. For yellow markings in the same conditions, g2010g2868 = 4.89 x 10-3, g2010g2869 = 1.85 x 10-4, and g2010g2870 = -8.00 x 10-8 (Bahar et al., 2006). A general linear model was developed by Sitzabee et al. (2009) in North Carolina for thermoplastics on asphalt based on the variables that were validated by the effects test (time, initial retroreflective value, AADT, color, and lateral location). The thermoplastic model produced an R? value of 0.60 which is greater than those found in previous studies reviewed in the literature: 19 g1844g3013 =190+0.39?g1844g3013g3284g3289g3284g3295g3284g3276g3287 ?2.09?g1872g1861g1865g1857?0.0011?g1827g1827g1830g1846+20.7?g1850g2869?20.7 ?g1850g2870+19?g1850g2871?19?g1850g2872 (Eq. 2.10) where: RL = retroreflectivity in mcd/m2/lux; RL initial = initial retroreflectivity in mcd/m2/lux; time = time since installation in months; AADT = annual average daily traffic in vehicles per day; X1=1 if edge line, 0 otherwise; X2=1 if middle line, 0 otherwise; X3=1 if white line, 0 otherwise; X4=1 if yellow line, 0 otherwise. In 2009, Clarke and Yan (2009) performed a retroreflectivity investigation in 85 sites with 90-mil thermoplastic longitudinal markings located in 14 counties across the state of Tennessee. Three models were developed by color, a linear, a logarithmic, and a quadratic. The models for white 90-mil thermoplastic had R2 values of 0.015 (linear), 0.008 (logarithmic), and 0.017 (quadratic) and are expressed by Equations 2.11, 2.12, and 2.13, respectively: g1851 =298.004?0.053g1872 (Eq. 2.11) g1851 =331.366?9.510lnuni0009g4666g1872g4667 (Eq. 2.12) g1851 =287.019+0.022g1872?0.000092g1872g2870 (Eq. 2.13) where Y = retroreflectivity (mcd/m2/lux); t = age of marking (days). 20 Yellow 90-mil thermoplastic models, which had R2 values of 0.003 (linear), 0.007 (logarithmic), and 0.017 (quadratic) are represented by Equations 2.14, 2.15, and 2.16, respectively. g1851 =150.233+0.016g1872 (Eq. 2.14) g1851 =120.845+6.181lnuni0009g4666g1872g4667 (Eq. 2.15) g1851 =133.718+0.124g1872?0.000g1872g2870 (Eq. 2.16) Traffic volume, typically expressed as AADT (Annual Average Daily Traffic), is a continuous parameter that measures the volume of traffic on the roadway in vehicles per day. Traffic volumes vary from day to day and over the course of a year, and are another source of variation that contributes to the complexity of the analysis (Songchitruksa et al., 2011). Thamizharasan et al. (2003) argued that AADT was not significant and was accounted for as a function of time. However, some reports indicate that AADT has a significant impact on pavement marking degradation apart from time (Sitzabee et al., 2009). Different sources are considered to analyze retroreflectivity behavior over time. Karwa and Donnell (2011) used Artificial Neural Networks (ANN) to model the degradation pattern of pavement marking retroreflectivity (PMR) in North Carolina as a function of several input variables, including the initial PMR, age of the markings, and traffic flow characteristics. It was found that the degradation of thermoplastic pavement markings occurs generally at a nonlinear rate, and the rate of decay appears to differ among different pavement marking types. It was also found that there may be a different degradation process according to the geographic location of the markings as well as by the color of the marking. The initial PMR also appears to be an important service life predictor. The variability in traffic volume, however, does not appear to 21 have a strong association with retroreflectivity degradation for most of the pavement marking types (Karwa and Donnell, 2011). 2.4. Pavement Marking Materials Specifications and Requirements On contracts for the Alabama Department of Transportation (ALDOT), requirements for all projects are based on the Alabama Standard Specifications for Highway Construction (ALDOT, 2008). The 2007 Special Provisions, an interim update to the 2006 Edition of ALDOT?s Standard Specifications for Highway Construction, will be the reference for this thesis as the projects on which data were collected are governed by the 2006 Edition and its subsequent Special Provisions. These specifications define classes of traffic stripe. The required type of material is designated by "Class" in accordance with Table 4 (ALDOT, 2007). Table 4 ? Class of Traffic Stripe Class of Traffic Stripe Class Material 1 Paint 1H High Build Paint 2 Standard Thermoplastic Material 2T Thin Film Spray Applied Thermoplastic Material 3 Tape W Warranted Traffic Marking Material SOURCE: ALDOT, 2007 Thermoplastics are generally composed of four ingredients: binder, glass beads, titanium dioxide and calcium carbonate. The binder is used to hold the mixture together as a rigid mass, the glass beads are used to provide reflectivity, the titanium dioxide is used for reflectivity enhancement, and calcium carbonate or sand is used as an inert filler material. Typical thermoplastic markings are 15 to 33 percent binder, 14 to 33 percent glass beads, 8 to 12 percent titanium dioxide and 48 to 50 percent filler (Cuelho et al., 2003). 22 Thermoplastic is a blend of solid ingredients that becomes liquid when heated and returns to a solid state on cooling. Thermoplastics are classified into two types: hydrocarbon-based plastics derived from petroleum, and alkyd, which is a naturally occurring resin (Migletz and Graham, 2002). Alkyd-based binders are more widely used because they are resistant to chemical decomposition from motor oil and other hydrocarbon contaminants. Thermoplastics can be applied to the roadway surface by spraying or extrusion. Extruded thermoplastics are thicker than sprayed thermoplastics (Fu and Wilmot, 2008). According to Lindly and Marci, (2006), thermoplastic markings are all-weather pavement markings. These markings should be visible at night during a rainfall of up to 0.25 inch per hour. Thermoplastic has been used successfully in warmer climates for a number of years. Class 2 thermoplastic is the marking type for which data will be analyzed in this research. Class 2 thermoplastic may be applied to asphalt and concrete surfaces, and according to ALDOT specifications, they have to be placed to produce a minimum uniform thickness of 0.100 inch for all stripes. Thickness less than 95% of this required value may be deemed unacceptable. The retroreflectivity is required to be a minimum of 450 mcd/m2/lux for white stripe and 350 mcd/m2/lux for yellow stripe. The Contractor may be compelled to replace stripe that is 90% or less of this retroreflectivity requirement (ALDOT, 2007). Class 2T thermoplastic markings are also analyzed in this study. Class 2T thermoplastic may be applied to asphalt and concrete surfaces, and ALDOT specifications require that they need to be placed to produce a minimum uniform thickness of 0.040 inch. Thickness less than 95% of this required value is not acceptable. The retroreflectivity has to be a minimum of 300 mcd/m2/lux for white stripe and 250 mcd/m2/lux for yellow stripe. The Contractor is required to replace stripe that is 90% or less of this retroreflectivity requirement (ALDOT, 2007). 23 ALDOT specifications require initial daytime chromaticity for yellow materials to fall within the box created by the coordinates presented in Table 5 (ALDOT, 2007). Table 5 ? Initial Daytime Chromaticity for yellow materials Initial Daytime Chromaticity Coordinates (Corner Points) 1 2 3 4 X 0.530 0.510 0.455 0.472 Y 0.456 0.485 0.444 0.400 SOURCE: ALDOT, 2007 The initial daytime chromaticity for white materials must fall within the box created by the coordinates showed on Table 6 (ALDOT, 2007). Table 6 ? Initial Daytime Chromaticity for yellow materials Initial Daytime Chromaticity Coordinates (Corner Points) 1 2 3 4 X 0.355 0.305 0.285 0.335 Y 0.355 0.305 0.325 0.375 SOURCE: ALDOT, 2007 According to ALDOT specifications, luminance factor requirements for white markings are daylight luminance factor at 45 degrees / 0 degrees - 50% minimum, and daylight luminance factor at 45 degrees / 0 degrees ? 35% minimum for yellow markings. 2.5. Economic Evaluation of Thermoplastic Materials Another critical factor to consider when choosing the best material for pavement markings is the associated cost. But it is not only the initial cost that should be analyzed, as a material with a higher initial cost could also have a longer lifetime, possibly resulting in a more cost-effective material (Thomas et al., 2001). The total cost of pavement markings includes not only the cost of the material, but also the cost of the crew and the application equipment, as well as manufacturer guarantees over a 24 specified time, in which case manufacturers replace deteriorating materials free of charge if their product does not achieve certain guidelines (Thomas et al., 2001). Basic costs of the materials and the equipment, and time required for their installation, fundamentally determine marking system costs, but secondary issues also can have a noticeable impact. The volume of markings to be installed and whether or not markings are installed by private firms or public agencies are some secondary issues that can influence the cost of pavement markings. Usually, the greater volume of markings to be installed, the lower the unit cost of their installation. The reasons for the differential in costs between private and public agency installation are uncertain, although they may be related, in part, to volume of work (Cuelho et al., 2003). Even if there are many factors to be considered when choosing the type of pavement marking material to be used, the decision is often dictated by the initial cost. There are limited guidelines leading to a more cost-effective selection and this approach to pavement marking can result in lack of durability, poor retroreflectivity, increased long-term costs and increased exposure to traffic for staff (Montebello et al., 2000). Service life is an important parameter in selecting a marking system and it is mainly determined from the level of retroreflectivity provided by the pavement marking. Other conditions that may end the service life of a marking include detachment from the pavement, extensive loss of pigment, and obliteration by pavement maintenance activities. Some of the major factors that affect the performance and service life of a particular type of pavement marking include type of road surface, volume of traffic, orientation with respect to traffic, and schedule of pavement maintenance activities (Cuelho et al., 2003). Thermoplastic materials have been used in the United States since the 1950s, and they are one of the most widely used pavement marking materials (Jiang, 2008). Thermoplastic markings 25 provide excellent performance when applied properly, being the most durable of the commonly used pavement marking systems. The life of thermoplastic markings varies widely because of its dependence on installation procedures, volume of traffic, atmospheric conditions when placed, and snowplow activity. The range of life expectancy is typically from four to seven years (KDOT, 2002). This relatively long service life can sometimes exceed the interval between pavement maintenance activities (Cuelho et al., 2003). There are some issues to be considered when applying thermoplastic to pavement markings, in addition to service life. One of the advantages of using thermoplastic is that the material can be re-applied over older thermoplastic markings, thereby refurbishing the older marking as well as saving on the costs of removing old pavement markings. However, thermoplastic color and appearance are disadvantages. Thermoplastic is grayish, making it less visible by day and it has a tendency to crack. (Jiang, 2008). Thermoplastics are expensive in comparison to conventional paints, with installed costs ranging between $0.19 and $0.26 per linear foot based on a four-inch wide longitudinal strip (KDOT, 2002). Thermoplastics, on the other hand, are the most durable of the commonly used pavement marking systems, which can result in a more cost-effective use of thermoplastics in the long-term (KDOT, 2002). In Alabama, the averages of contract bid prices are $0.26 per linear foot for Class 2T thermoplastic materials and $0.65 per linear foot for Class 2 thermoplastic materials (ALDOT, 2010[1]). The objective of a pavement marking economic evaluation is to identify the most economical pavement marking materials. Some methods have been employed. For example, Kansas DOT (KDOT, 2002) developed a sophisticated methodology to determine the most economical type of pavement marking to be used under different conditions. Materials are 26 selected based on the remaining pavement service life, traffic volume level, and a Brightness Benefit Factor (BBF). In the Kansas DOT document, there is one table for each remaining service life, between 1 and 7 years; columns of these tables represent AADT levels (<5,000, 5,000-50,000, and >50,000 veh/day) and rows represent BBF for each type of material; the material with the highest BBF represents the best combination of durability, retroreflectivity, and cost for the considered remaining service life and ADT. The BBF is a benefit/cost ratio representing the combined effects of a material?s retroreflectivity, durability, and installed cost. The BBF is defined as: g1828g1828g1832 =g1844g3028g1846g3046g1845 (Eq. 2.17) where: Ra = average useful retroreflectivity over the anticipated service life of the project in mcd/m2/lux; Ts = pavement marking service life in years; S = average cost per unit length in dollars per meter. In the calculation of BBF, additional retroreflectivity over the minimum required level is considered a benefit to the user. As a result, the higher the retroreflectivity obtained during the material lifetime, the higher the BBF (KDOT, 2002). Cottrell and Hanson (2001) used cost-effectiveness analysis to select marking materials for Virginia DOT (VDOT). They found that there is not much benefit in using a marking with a retroreflectivity value greater than 600 mcd/m2/lux compared to one with a value of 300 mcd/m2/lux. As a result, their study did not use retroreflectivity as a benefit, only service life. Figure 6 illustrates the two methods. The horizontal axis is time measured by month; the vertical axis is retroreflectivity; Rmin is the minimum retroreflectivity threshold (Fu and Wilmot, 2008). 27 Figure 6 ? Measuring user benefit: Kansas and Virginia SOURCE: Fu and Wilmot, 2008 Lindly and Wijesundera (2003) used life cycle cost analysis (LCCA) to compare different marking materials for Alabama DOT. LCCA assumes the alternatives yield the same level of service, provided the retroreflectivity of the pavement markings meets the minimum value requirement. LCCA requires identification of pavement marking service life and total cost and then calculates the net present worth (NPW) or the equivalent uniform annual cost (EUAC). The material with the lowest NPW or EUAC is selected. Loetterle et al.?s study of public perception of pavement marking retroreflectivity leads to an intuitively appealing way to measure user benefit. Based on their data, user benefit can be considered to increase approximately in a linear fashion when retroreflectivity is under 200 mcd/m2/lux, while above 200 mcd/m2/lux, there is little additional benefit. Thus, provided the retroreflectivity is at least 200 mcd/m2/lux, full benefit of the pavement marking is received (Loetterle et al., 2001). 28 On the other hand, below a certain minimum threshold value of retroreflectivity, pavement markings are considered to be unacceptable and have no value to the driving public. This minimum value indicates the end of service life of the pavement marking. Calling these two values Rmax and Rmin, respectively, then Figure 7 demonstrates an alternative measurement of benefit. Figure 7 ? Alternative measurement of benefit Fu and Wilmot (2008) shows that the suggested user benefit is measured by the area between Rmin and Rmax. Benefit is measured in units of month*vehicle*mcd/m2/lux. The benefit/cost ratio can be simply calculated by: g1854g1857g1866g1857g1858g1861g1872 g1855g1867g1871g1872 = g1854g1857g1866g1857g1858g1861g1872?g1827g1827g1830g1846 g1872g1867g1872g1853g1864uni0009g1855g1867g1871g1872 (Eq. 2.18) 29 2.6. Service Life of Thermoplastic Pavement Markings Migletz and Graham (2002) established a retroreflectivity of 150 mcd/m2/lux as a threshold to predict the service life of white thermoplastic pavement markings on freeways. The obtained mean estimated service life was 22.6 months, in a range of 7.4 to 49.7 months. The mean service life of yellow markings on freeways in the Migletz and Graham (2002) study was 24.7 months, in a range of 11.0 to 41.6 months. The minimum retroreflectivity threshold for yellow thermoplastic was considered as 100 mcd/m2/lux. Minimum retroreflectivity thresholds were based on the FHWA suggested values (FWHA, 2000). Abboud and Bowman (2002 [1]) considered in their study the useful lifetime of white edge lines. For low ADT (<2500 veh/day), the service life for white edge thermoplastic markings was found to be 53 months; for intermediate ADT (2500 to 5000 veh/day), useful lifetime dropped to 18 months; for high ADT (>5000 veh/day), the service life was 10.5 months. According to the FHWA (2000), the threshold for minimum acceptable retroreflectivity was considered equal to 150 mcd/m2/lux and the midpoint of each ADT range was used as representative (Abboud and Bowman, 2002 [1]). Thamizharasan et al. (2003) developed a model to predict retroreflectivity and established the threshold of 100 mcd/m2/lux, for both yellow and white markings, as the minimum acceptable retroreflectivity; this threshold was chosen based on an NCHRP study (Andrady, 1997). Average service life for white thermoplastic marking on asphalt pavement was found to be 65 months, for retroreflectivity variation in a range from 30 to 690 mcd/m2/lux, with an average of 203 mcd/m2/lux. For yellow thermoplastic marking on asphalt pavement, the average useful lifetime was 103 months, for retroreflectivity variation in a range from 26 to 429 mcd/m2/lux, with an average of 135 mcd/m2/lux. 30 Sitzabee et al. (2009) used an AADT of 10,000 veh/day to estimate the service lives of thermoplastic markings. For white edge lines, average service life was 102 months; for white broken lines, average useful lifetime was 84 months; for yellow edge lines, average service life was 85 months; and for yellow broken lines, average useful lifetime was 65 months. For white markings, the minimum retroreflectivity value was 150 mcd/m2/lux; for yellow markings, the considered threshold was 100 mcd/m2/lux, based on most common research recommendations. Initial retroreflectivity was 375 mcd/m2/lux for white markings and 250 mcd/m2/lux for yellow markings. Clarke and Yan (2009) performed a study which investigated changes in retroreflectivity over time for 90-mil thermoplastic markings of 85 sites in 14 counties of Tennessee. The average service life of the 90-mil thermoplastic white markings is 1,100 days (36 months) while that of the yellow markings is 900 days (30 months). Karwa and Donnell (2011) predicted service life of pavement markings using Artificial Neural Networks. Service life was estimated for white and yellow line types, considering centerlines and edgelines separately. Two different threshold values for minimum retroreflectivity were considered: 150 and 200 mcd/m2/lux. Also, three different ADT levels were considered, along with two different percent truck and two different initial PMR levels. White edgeline markings generally had longer predicted service lives than all other pavement marking types. The shortest mean predicted service life was most often computed for the yellow centerline markings. There was also considerable variability in the predicted service life across the engineering divisions, which shows the importance of geographic location when developing models to predict service life. In addition, increasing the ADT from 5,000 to 25,000 reduced the mean predicted service life by 7 months or less. The impact of different vehicle types was also 31 analyzed, and it was found that roadways with higher truck traffic volumes decrease pavement markings service life due to abrasion between the tires and the markings (Karwa and Donnel, 2011). 2.7. Summary of findings Pavement markings have an essential role in the highway system. It is a complex decision, balancing many competing objectives, to decide which factors should be considered when selecting the type of pavement marking material for a particular road. This is essentially an agency?s policy decision, which can be supported by research. There is a wide variety of characteristics, costs, and benefits of the different pavement marking materials. There are numerous types of materials used for pavement markings in the field today, including paint, epoxy, tape, and thermoplastic. Thermoplastic is the most widely used. AADT was sometimes not significant for service life analysis. However, most reports indicate that AADT has a significant impact on pavement marking degradation apart from time. The performance of a pavement marking material has been judged based primarily on its retroreflectivity. There are many models to predict marking retroreflectivity. The best prediction found in the literature is a general linear model that was developed by Sitzabee et al. (2009), with R? equal to 0.60 for thermoplastics. Table 7 shows a summary of all models cited in Section 2.3 of this thesis. 32 Table 7 ? Summary of models to predict retroreflectivity Study Location Type of Marking Independent Variables R2 Lee et al. (1999) Michigan White and Yellow Thermoplastic Age of marking 0.14 Abboud and Bowman (2002 [1]) Alabama White Thermoplastic ADT per lane 0.58 Age of marking Abboud and Bowman (2002 [2]) White Thermoplastic ADT per lane N/A Age of marking Thamizharazan et al. (2003) Non-Linear South Carolina White Thermoplastic in Asphalt Age of marking 0.22 Initial Retroreflectivity Yellow Thermoplastic in Asphalt Age of marking 0.2 Initial Retroreflectivity Thamizharazan et al. (2003) Linear White Thermoplastic in Asphalt Age of marking 0.47 Initial Retroreflectivity Yellow Thermoplastic in Asphalt Age of marking 0.21 Initial Retroreflectivity Bahar et al. (2006) California White Thermoplastic, Hot Humid, No snow removal Age of marking N/A Yellow Thermoplastic, Hot Humid, No snow removal Age of marking N/A Sitzabee et al. (2009) North Carolina White and Yellow Thermoplastic Initial Retroreflectivity 0.6 Age of marking AADT Marking location (edge, middle) Marking color (white, yellow) Clarke and Yan (2009) Linear Tennessee White 90-mil Thermoplastic Age of Marking 0.015 Yellow 90-mil Thermoplastic Age of marking 0.008 Clarke and Yan (2009) Logarithmic White 90-mil Thermoplastic Age of Marking 0.017 Yellow 90-mil Thermoplastic Age of marking 0.003 Clarke and Yan (2009) Quadratic White 90-mil Thermoplastic Age of Marking 0.007 Yellow 90-mil Thermoplastic Age of marking 0.017 33 Economic evaluation is also important when deciding which material will be used for pavement markings. Basic costs of the materials and the equipment and service life are the most characteristics analyzed to perform a cost-effective evaluation. The objective of a pavement marking economic evaluation is to identify the most economical pavement marking materials. Some methods have been employed. Prediction of service life, usually based on retroreflectivity levels, varies from one model to another. A summary of predicted service life for pavement markings analyzed in Section 2.6 can be observed in Table 8. Adopted retroreflectivity thresholds are also in Table 8. Table 8 ? Summary of Predicted Service Life Study Marking Retroreflectivity (mcd/m2/lux) Service Life (months) Threshold Average Initial Minimum Maximum Average Migletz and Graham (2002) White Thermoplastic 150 - 7.4 49.7 22.6 Yellow Thermoplastic 100 - 11 41.6 24.7 Abboud and Bowman (2002) White Edge Thermoplastic 150 - 10.5 53 - Thamizharasan et al. (2003) White Thermoplastic on Asphalt 100 203 - - 65 Yellow Thermoplastic on Asphalt 100 135 - - 103 Sitzabee et al. (2009) White Edge Thermoplastic 150 375 - - 102 White Middle Thermoplastic 150 375 - - 84 Yellow Edge Thermoplastic 100 250 - - 85 Yellow Middle Thermoplastic 100 250 - - 65 Clarke and Yan (2009) White 90-mil Thermoplastic - 296.1 - - 92 Yellow 90-mil Thermoplastic - 154.5 - - 75 Karwa and Donnell (2011) Yellow Edgelines 150, 200 300, 400 (values differ according to ADT and percent trucks) Yellow Centerlines 150, 200 300, 400 White Edgelines 150, 200 300, 400 White Skip Lines 150, 200 300, 400 34 Chapter Three Methodology Chapter Three provides details of the methods applied to this thesis with the purpose of achieving the objectives stated in Chapter One. These methods include the data mining and analysis of a database of pavement markings applied in 2007 highway projects for ALDOT. This process determined the extent to which observations from this database meet the ALDOT Standard Specifications for Highway Construction for thickness, retroreflectivity, luminance, and color of pavement markings. A statistical analysis was also executed to evaluate trends in mean, standard deviation, and coefficient of variation for each class and color of markings. Retroreflectivity modeling based on models found in the literature was developed in this research project. A methodology for development of new models related to preparation of data and model fitting, as well as considerations regarding benefit/cost analysis are also discussed in this chapter. 3.1. ALDOT specifications study The database analyzed in this thesis was developed by ALDOT in 2007 and includes measurements of pavement markings observed at 7,840 locations from 40 projects in Alabama. In this year, five different General Application Special Provisions for Traffic Stripe, which constitute updates to ALDOT?s Standard Specifications for Highway Construction, were approved. A documentation of the recent changes in these special provisions was prepared. 35 3.2. Data Mining and Analysis The database studied in this thesis contains multiple pavement marking measurements from 40 projects in Alabama. The list including the number of these projects and their locations can be observed in Table 9. Table 9 ? List of the 40 Pavement Marking Projects in the 2007 ALDOT Database Project Number County 99-307-203-100-701 Covington County 99-307-346-010-701 Henry County STPNU-3128(200) Geneva County STPNU-3140(200) Geneva County 99-307-164-167-701 Geneva County STPNU-2014(200) Covington County EB-0042(507) Mobile County EB-0074(513) Cullman County 99-302-391-101-701 Lauderdale County 99-307-162-088-703 Coffee County STPNU-4816(201) Marshall County 99-304-154-009-701 Cleburne County STPAA-0052(506) Geneva County 99-305-632-069-703 Tuscaloosa County STPSA-0021(515) & 99-306-434-021-701 Lowndes County STPNU-CN07(203) (1Y) OLD McGEHEE Montgomery County STPNU-CN07(203) (1W) OLD McGEHEE Montgomery County STPNU-CN07(203) (2Y) MARLER Montgomery County STPNU-CNO7(203) (2W) MARLER Montgomery County 99-308-663-089-706 & STPSA-0089(500) Wilcox County EB-0016(505) Baldwin County STPSA-0185(500) & 99-306-074-185-701 Butler County STPAA-0079(506) Blount County STPNU-3423(201) Henry County 99-302-473-013-704 Marion County 36 Table 9 (Continuation) ? List of the 40 Pavement Marking Projects in the 2007 ALDOT Database Project Number County EB-0004(509) Jefferson County STPSA-0001(529) & 99-301-285-001-705 Etowah County 99-303-595-003-709 Shelby County NHF-STPSAF-0053(525) & 99-307-234-053- 701 Dale County NHF-0056(500) & BRF-0102(527) Montgomery County STPNU-1713(201) Colbert County 99-302-473-171-706 Marion County STPNU-2221(201) Cullman County 99-305-632-069-702 Tuscaloosa County EBF-0012-(522) Houston County EBF-0012-(522)B Houston County EB-0035(506) DeKalb County 99-303-582-004-708 St. Clair County STPSA-0079(505) & 99-301-484-079-708 Marshall County STPSA-0079(505) & 99-301-484-079-708(2) Marshall County All projects in the 2007 database contain measurements of Class 2 (Standard Thermoplastic) or Class 2T (Thin Film Spray Applied Thermoplastic), and colors are white or yellow. After identifying changes in 2007 General Application Special Provisions, a data mining plan was developed. The original database obtained from ALDOT was a Microsoft Access file that was then imported into Microsoft Excel. All data were assembled into one table. A screen capture of Microsoft Excel, in Figure 8, shows an example of the available information for all 40 projects. Table 10 presents the meaning of each entry represented on Figure 8. Figure 8 ? Information from 2007 Database 37 Table 10 ? Meaning of ALDOT Database Entries Entry Meaning PROJECT_NO Number used to identify the project CONTRACTOR Contractor of project COUNTY County where project is located BEAD_MAN Bead Manufacturer THERMO_MAN Thermoplastic Manufacturer PROJECT_MAN Project Manager INSPECTOR Inspector of the Project APP_METHOD Method used on the application of marking CLASS Class of marking LOT_ID Identification of marking in ALDOT records PLACED_DATE Date when marking was placed TEST_DATE Date when marking properties were measured RETRO_ID Identification of the equipment used to measure retroreflectivity COLOR_ID Identification of the equipment used to measure color STRIPE_TYPE Type of marking (solid or broken) STRIPE_COLOR Color of marking (white or yellow) DIRECTION Direction of traffic where marking was measured (e.g., "West" means "Westbound") Station Location of the marking Thickness Initial Thickness of the marking Retro Initial Retroreflectivity of the marking Luminance Initial Luminance of the marking Color x Coordinates of chromaticity measures Color y Color P/F Determination as to whether color passes or fails to meet specifications Analyses were performed using data available from the 2007 ALDOT database. The evaluation of measurements of retroreflectivity, thickness, luminance, and chromaticity in regard to specifications was the first executed process, followed by a statistical analysis. Pavement marking measurements were compared with values given in the Standard Specifications for Highway Construction Special Provisions (ALDOT, 2007) to determine whether the values in the database complied with ALDOT requirements. As it was observed in Section 3.1, there were 38 no changes in quantitative requirements from one special provision to another. Therefore, all Special Provisions presented the same required values shown on Table 11. Table 11 ? Minimum Required Values for Pavement Markings Class Color Thickness (in) Retroreflectivity (mcd/m2/lux) Luminance (%) 2 White 0.10 450 50 2 Yellow 0.10 350 35 2T White 0.04 300 50 2T Yellow 0.04 250 35 Although the values shown in Table 11 are the minimum required, the special provisions inform that a thickness greater than 95% of the minimum value is acceptable and a retroreflectivity greater than 90% of the minimum value meets requirements (ALDOT, 2007). Therefore, the values used to verify if the measured thickness, retroreflectivity, and luminance in Alabama were acceptable are shown in Table 12. Table 12 ? Acceptable Values for Pavement Markings Class Color Thickness (in) Retroreflectivity (mcd/m2/lux) Luminance (%) 2 White 0.095 405 50 2 Yellow 0.095 315 35 2T White 0.038 270 50 2T Yellow 0.038 225 35 The analysis of color was also performed according to the 2007 General Application Special Provisions (ALDOT, 2007). The initial daytime chromaticity for white and yellow materials must fall within the box created by the coordinates shown in Table 13 for the chromaticity to be in compliance with specifications. 39 Table 13 ? Initial Daytime Chromaticity Coordinates (Corner Points) White Yellow x y x y 0.335 0.375 0.472 0.400 0.355 0.355 0.530 0.456 0.305 0.305 0.510 0.485 0.285 0.325 0.455 0.444 SOURCE: ALDOT, 2007 The box created by the plot of the coordinates from Table 13, for white materials, can be observed in Figure 9 and the plot of the coordinates from Table 13, for yellow materials, is shown in Figure 10. If measured chromaticity values were located within constrained space in Figures 9 and 10, then color observations complied with specifications. Figure 9 ? Initial Daytime Chromaticity: White 0.280 0.290 0.300 0.310 0.320 0.330 0.340 0.350 0.360 0.370 0.380 0.250 0.270 0.290 0.310 0.330 0.350 0.370 Color -White 40 Figure 10 ? Initial Daytime Chromaticity: Yellow After analyzing all field measurements from the 40 projects in Alabama, some conclusions could be made. The percentage of projects that were according to specifications requirements can be observed in Chapter Four: Results. Finally, statistical measures were calculated, for each marking color and type. Mean, standard deviation and coefficient of variation were computed to demonstrate the variation of the measured observations of the 40 projects in Alabama. This analysis can provide the information on the central tendency for each marking property and the dispersion from the average. The detailed statistical analysis for each class and color of markings can be seen in Chapter Four: Results. 3.3. Retroreflectivity Modeling For all 40 projects in the initial ALDOT database, from 2007, only 15 have retroreflectivity data from 2008, 2009 and 2010. Table 14 and Figure 11 describe the location of 0.350 0.370 0.390 0.410 0.430 0.450 0.470 0.490 0.510 0.440 0.460 0.480 0.500 0.520 0.540 Color -Yellow 41 these 15 projects. For these 15 projects, cost information was obtained from the Tabulation of Bids, from ALDOT (ALDOT, 2010 [1]). Table 14 ? Projects Included in the Retroreflectivity Modeling Project ID County Division Location Map Key EB-0016(505) Baldwin County 9 SR-16 (US-90) from Robertsdale to Florida State line 9A STPSA- 0185(500) Butler County 6 SR-185 from SR-263 in Greenville to the Lowndes county line 6A EB-0074(513) Cullman County 1 SR-74 (US-278) from east of CR-420 to I-65 in Cullman 1A NHF- STPSAF- 0053(525) Dale County 7 SR-53 (US-231) from near CR-63 in Pinckard to near CR-18 7A EB-0035(506) DeKalb County 1 SR-35 from SR-7 (US-11) to MP 25.40 in Fort Payne 1B STPSA- 0001(529) Etowah County 1 SR-1 (US-431) from west of I-59 in Attalla to 0.2 miles south of CR-137 1C 99-307-164- 167-701 Geneva County 7 SR-167 from the Florida state line to near the south city limit of Hartford 7B 99-307-346- 010-701 Henry County 7 SR-10 from SR-1 (US-431) to SR-95 in Abbeville 7C 99-302-391- 101-701 Lauderdale County 2 SR-101 from the north end of Wheeler Dam to SR-2 (US-72) in Elgin 2A STPSA- 0021(515) Lowndes County 6 SR-21 from the Wilcox county line to south of CR-45 near Mount Willing 6B 99-302-473- 171-706 Marion County 2 SR-171 from the Fayette county line in Winfield to SR-118 2B STPSA- 0079(505) Marshall County 1 SR-79 from SR-1 (US-431) through Columbus City to the Jackson county line 1D 99-303-595- 003-709 Shelby County 3 SR-3 (US-31) from Seventh Avenue in Calera to I-65 3A 99-305-632- 069-702 Tuscaloosa County 5 SR-69 from the north end of the Lake Tuscaloosa bridge to south of Windham Springs 5A STPSA- 0089(500) Wilcox County 8 SR-89 from SR-21 west of Snow Hill to the Dallas county line 8A 42 Figure 11 ? Location of the 15 Projects with Retroreflectivity Data from 2007 to 2010 SOURCE: Based on University of Alabama, 2011 43 All projects are located on State Routes, U.S. Routes, or Interstates, with 2 lanes (one in each direction) or four lanes (separated, two lanes on one side, two lanes on the other side). Figure 12 shows an example of a two-lane state route, a section of project STPSA-0185(500), in Butler County, on SR-185. More pictures showing details of these 4-year old pavement markings of project STPSA-0185(500) can be observed in Appendix A. Figure 12 ? Section of Project STPSA-0185(500), on SR-185 Retroreflectivity for 2007 data was measured at specific points, within stations. Retroreflectivity for 2008, 2009 and 2010 was measured in different locations, within mileposts. To build the retroreflectivity curves, it was necessary to correlate locations from the 2007 44 database and the 2008 to 2010 data. From ALDOT project records, the ?Begin Project? location, in station, and the corresponding milepost were identified. Figure 13 shows an example for project NHF-STPSAF-0053(525). The project begins at Milepost 32.1, which is the same location as Station 251+67. Figure 13 ? Example of Correspondence from Stations to Mileposts For each milepost, there was a range of corresponding stations. The average retroreflectivity for this range of stations was calculated and set as the milepost retroreflectivity value for 2007. One retroreflectivity curve was built for each milepost, according to color and type of marking. Table 15 shows the total number of mileposts, across all projects, for each color and type; therefore, the number of retroreflectivity curves available for analysis. The complete list of mileposts by projects can be observed in Appendix B. Table 15 ? Total Number of Retroreflectivity Curves TOTAL CURVES Solid White Broken White Solid Yellow Broken Yellow 76 17 67 32 3.3.1. Preparation of Data for Modeling The end of a pavement marking service life can be represented in a retroreflectivity curve as the time when the curve reaches a minimum retroreflectivity threshold. The adopted minimum 45 values are 150 mcd/m2/lux for white markings and 100 mcd/m2/lux for yellow markings, as these were the most common values found in the literature. The retroreflectivity curves that were built for each milepost, based on 4 years of data, do not reach these minimum retroreflectivity values; therefore, it is necessary to extrapolate curve points. Many existing models in the literature were applied to the available data to determine which one would be the best option to represent actual data. Among these models, those developed by Abboud and Bowman, and Sitzabee et al., required traffic information as input data. The Average Annual Daily Traffic (AADT) for the 15 project locations was obtained from the ALDOT website (ALDOT, 2010 [2]). As there was not an exact correspondence between locations in the ALDOT traffic database and the location of the projects? measurements, some assumptions were made when determining AADT for each milepost. Three locations from the ALDOT traffic database, similar to locations of projects, were chosen, and the average AADT for these three locations was considered as the AADT for each milepost of the considered project. To illustrate this process, Project 99-302-391-101-701, on SR-101, is shown as an example. It has retroreflectivity data at milepost 25.7, milepost 26.2, milepost 27.0, and milepost 28.0. The ALDOT traffic database contains traffic data at milepost 24.99, milepost 26.78, and milepost 28.24, as historical data, as shown in Table 16. The average of these three AADT historical data was calculated. From that, a linear regression analysis was applied in order to obtain an equation to estimate AADT as a function of time to relate AADT values to the times when retroreflectivity measurements were taken. All existing mileposts for this project (MP 25.7, MP 26.2, MP 27.0, and MP 28.0) will have the same AADT equation to calculate vehicles as a function of time. 46 Table 16 ? ALDOT Historical Traffic Data Traffic Counters 2010 Traffic Counters 2010 Traffic Counters 2010 Counter ID AL-39-522 Counter ID AL-39-521 Counter ID AL-39-520 Station 522 Station 521 Station 520 County 39 County 39 County 39 City N/A City N/A City N/A Route 101 Route 101 Route 101 Milepoint 24.99 Milepoint 26.78 Milepoint 28.24 AADT 2010 5990 AADT 2010 7390 AADT 2010 5640 AADT 2009 6580 AADT 2009 7370 AADT 2009 5360 AADT 2008 6380 AADT 2008 7140 AADT 2008 5150 AADT 2007 7050 AADT 2007 7480 AADT 2007 5400 AADT 2006 6980 AADT 2006 7300 AADT 2006 5270 AADT 2005 6890 AADT 2005 7590 AADT 2005 5540 AADT 2004 6930 AADT 2004 7760 AADT 2004 5300 AADT 2003 6540 AADT 2003 7460 AADT 2003 5100 AADT 2002 6370 AADT 2002 7040 AADT 2002 4860 K 11 K 11 K 11 D 55 D 75 D 55 TDHV 8 TDHV 7 TDHV 7 TADT 11 TADT 9 TADT 9 Heavy 50 Heavy 50 Heavy 50 Functional Class 6 Functional Class 6 Functional Class 7 SOURCE: ALDOT, 2010 [2] This calculated AADT was used to estimate models in the literature. For Abboud and Bowman model, AADT per lane was considered; for Sitzabee et al. model, AADT for all lanes was used to calculate retroreflectivity. 3.3.2. Model Fitting The determination of the best-fitting model from the literature when applied to ALDOT data was based on three performance measures: Area Under Curve Ratio, Average Model Error, and Average Percent Error. 47 The behavior of a curve can be represented by the area under it. The definite integral gives the area between the graph of the input and the x-axis. The technical definition of the definite integral is the limit of a sum of areas of rectangles, called a Riemann sum. There are four methods of Riemann summation: left sum, right sum, middle sum, and trapezoidal rule. The left Riemann sum approximates the function by its value at the left-end point; the right Riemann sum approximates the value at the right endpoint; the middle Riemann sum approximates the function at the midpoint of each interval; and the trapezoidal rule considers that the values of the function on an interval are approximated by the average of the values at the left and right endpoints (Thomas and Finney, 1996). Figure 14 shows how these four summation methods can be visualized. Figure 14 ? Four Methods of Riemann Summation SOURCE: Adapted from Thomas and Finney, 1996 48 The trapezoidal rule is applied to calculate the area under the curve resulted from actual data and the area under each curve representing a model in literature. Figure 15 illustrates the comparison between two curves considering this area. Figure 15 ? Comparison Between Areas Under Curves To have a numerical parameter of comparison between all models and a method to identify the ?best-fitting? model, a ratio is calculated for each model, for each milepost, as shown in Equation 3.1. g1844g1853g1872g1861g1867= g1827g1870g1857g1853uni0009g4666g1839g1867g1856g1857g1864g4667g1827g1870g1857g1853uni0009g4666g1827g1855g1872g1873g1853g1864uni0009g1830g1853g1872g1853g4667 (Eq. 3.1) The closer this ratio is to 1, the better fitting is the model. Another way to compare models is to calculate the average model error. This relationship is simply the difference between the model and actual data and it is given by Equation 3.2. g1827g1874g1857g1870g1853g1859g1857uni0009g1839g1867g1856g1857g1864uni0009g1831g1870g1870g1867g1870=g1839g1867g1856g1857g1864?g1827g1855g1872g1873g1853g1864uni0009g1830g1853g1872g1853 (Eq. 3.2) 49 The last method used to evaluate the suitability of models found in the literature is average percent error. It is calculated by Equation 3.3. g1827g1874g1857g1870g1853g1859g1857uni0009g1842g1857g1870g1855g1857g1866g1872uni0009g1831g1870g1870g1867g1870=g1839g1867g1856g1857g1864?g1827g1855g1872g1873g1853g1864uni0009g1830g1853g1872g1853g1827g1855g1872g1873g1853g1864uni0009g1830g1853g1872g1853 ?100 (Eq. 3.3) 3.4. Benefit/Cost Calculation Retroreflectivity curves developed for each milepost of the 15 projects in Alabama were extrapolated based on the best-fitting models from the literature, established according to the methodology described in Section 3.3.2. As shown in Chapter Two, there is a wide range of methods used to determine benefit/cost relationship of each available marking material. The method developed by Fu and Wilmot, considering the area between Rmin and Rmax as the benefit of the marking, was the most relevant to this study. According to what was analyzed in Chapter Two, it is possible to set Rmax = 200 mcd/m2/lux, Rmin = 150 mcd/m2/lux for white markings, and Rmin = 100 mcd/m2/lux for yellow markings. Cost data were available from ALDOT Tabulation of Bids (ALDOT, 2010[1]). AADT was obtained from ALDOT Traffic Data (ALDOT, 2010 [2]). The benefit/cost ratio can be calculated by Equation 3.4. g1854g1857g1866g1857g1858g1861g1872 g1855g1867g1871g1872 = g1854g1857g1866g1857g1858g1861g1872?g1827g1827g1830g1846 g1872g1867g1872g1853g1864uni0009g1855g1867g1871g1872 (Eq. 3.4) 3.5. Modeling of Retroreflectivity Over Time Using the data from the 15 projects with several years of observations, new models will be developed to represent the behavior of retroreflectivity over time in Alabama. First, locations are grouped by color and type: solid white, broken white, solid yellow, and broken yellow. The 50 Standard Specifications for Highway Construction Special Provisions (ALDOT, 2007), which were in effect at the time the projects included in this study were constructed, state that initial retroreflectivity shall be a minimum of 450 mcd/m2/lux for white markings and 350 mcd/m2/lux for yellow markings. Therefore, locations with initial retroreflectivity less than the required values are not considered in the modeling. In the literature, separate models for solid and broken markings are not presented. Therefore, two groups of data will be considered in this study: white and yellow; within each color, measurements from solid and broken lines will be combined. It can also be inferred from the literature that most models consider only age as the independent variable, some consider the initial retroreflectivity, and a few consider traffic volume. Regressions that consider only one independent variable are simple regressions; if 2 or more independent variables are included, they are defined as multiple regressions. For all developed models in this thesis, 2 or more independent variables are used; therefore, they are multiple regressions. When analyzing available data from ALDOT databases, two independent variables can be considered for the models: Initial Retroreflectivity and Age. AADT can be another independent variable as this data was obtained by the linear regression analysis based on ALDOT traffic data demonstrated in Section 3.3.1. Some research considers that AADT does not influence retroreflectivity degradation significantly; other studies affirm it is an important variable when estimating retroreflectivity over time. Therefore, this study will consider two approaches when predicting retroreflectivity models: the first one considers initial retroreflectivity and age of markings as independent variables, the second approach also considers AADT as an independent variable. 51 For the retroreflectivity modeling based on initial retroreflectivity and age of markings, four models will be evaluated according to the tendencies of data: linear, power, quadratic, and exponential. Models will be estimated using the software IBM Statistical Package for the Social Sciences (SPSS). The linear model general expression for two independent variables is represented by Equation 3.5. g1877=g1865g2869g1876g2869+g1865g2870g1876g2870+g1854 (Eq. 3.5) where: g1877 = Dependent Variable; g1876g2869,g1876g2870 = Independent Variables; g1865g2869,g1865g2870= Coefficients; g1854 = Constant. The power model can be represented by Equation 3.6. g1877=g1854?g1865g2869g3051g3117 ?g1865g2870g3051g3118 (Eq. 3.6) The general expression for a quadratic model is observed in Equation 3.7. g1877=g1865g2869g1876+g1865g2870g1876g2870 +g1854 (Eq. 3.7) The exponential model can be represented by Equation 3.8. g1877=g1876g2869g1857g2879g3029g3051g3118 (Eq. 3.8) Considering initial retroreflectivity, age of markings, and traffic volume as independent variables to model retroreflectivity, a linear model will be developed based on the tendencies of data and existing model forms in the literature with these three variables. The model will also be estimated using the software IBM Statistical Package for the Social Sciences (SPSS) and its general expression is observed in Equation 3.9. g1877=g1865g2869g1876g2869+g1865g2870g1876g2870 +g1865g2871g1876g2871+g1854 (Eq. 3.9) 52 Comparisons between all models, details related to each one and conclusions about the results obtained with all four regressions can be analyzed in Chapter Four: Results. 3.6. Summary of Chapter Three This chapter presented the methods used to analyze the 2007 database, which included the determination whether observations met specifications and a statistical analysis to evaluate trends in mean, standard deviation, and coefficient of variation. Also provided herein are details related to the methodology applied to retroreflectivity modeling based on models found in the literature and preparation of data for development of new models. Model forms under consideration include linear, power, quadratic, and exponential when considering initial retroreflectivity and age of markings as independent variables to predict retroreflectivity; a linear model will be developed when AADT is also considered. The results obtained with the use of methods from this chapter will be presented in Chapter Four: Results. 53 Chapter Four Results This chapter presents the results for the procedures adopted in this thesis, described in Chapter Three. Chapter Four provides a summary illustrating the proportion of all 7840 observations among the 40 projects in Alabama that met ALDOT specifications, considering the properties of retroreflectivity, thickness, luminance, and chromaticity. Statistical calculations of mean, standard deviation, and coefficient of variation are also presented in this chapter. The retroreflectivity modeling process provided several results. The development of retroreflectivity curves for each milepost of the 15 projects with data from 2007 to 2010 is described in this Chapter, as well as application and determination of best-fitting models from the literature. Considerations for the benefit/cost analysis are presented in this chapter. Finally, different proposed models of retroreflectivity over time are suggested to represent projects in Alabama. 4.1. ALDOT Data Mining and Analysis After analyzing all measures from the 2007 database of 40 projects in Alabama, some conclusions can be made pertaining to compliance with specifications and statistical attributes. The first two columns of Table 17 indicate how many observations met specifications and how many did not; how many of the observation points did not have a property measured is also represented in the table. The final two columns show the percentage of observations which 54 passed specifications and the percentage of total observations which was measured for the considered property of thickness, retroreflectivity, luminance, or chromaticity. Table 17 ? Material Property Compliance with Specifications Property Pass Do not pass Measured Not measured Total % Passing % Measured Thickness 6025 1551 7576 264 7840 79.53 96.63 Retroreflectivity 5344 2356 7700 140 7840 69.40 98.21 Luminance 5831 1865 7696 144 7840 75.77 98.16 Color 7227 459 7686 154 7840 94.03 98.04 In addition, the 2007 ALDOT database provided data on whether chromaticity values met specifications. In this analysis a color check was also performed, and results of the present analysis and ALDOT analysis were compared. As can be observed in Table 18, some conclusions of the color measurements being in accordance with the Special Provisions were different, which means the ALDOT analysis found some color measurements passed the criteria and this data analysis did not, or vice-versa. In spite of these differences, 98.70% of the total compared showed equal conclusions. Table 18 ? Color Check Comparison Different Equal Total % Compared %Equal 99 7536 7635 97.39 98.70 To describe the variation of the measured values from the 40 projects in Alabama in 2007, a statistical analysis was performed. Table 19 shows these measures for each analyzed group, by class and color of the material. 55 Table 19 ? Statistical Calculations Property Class Mean Standard Deviation Coefficient of Variation (%) Number of Observations (n) Thickness (in) Class 2 - white 0.117 0.034 29.06 3971 Class 2 - yellow 0.115 0.035 30.43 3031 Class 2T - white 0.078 0.043 55.13 306 Class 2T - yellow 0.065 0.033 50.77 268 Retroreflectivity (mcd/m2/lux) Class 2 - white 496 141.383 28.50 4000 Class 2 - yellow 330 72.020 21.82 3126 Class 2T - white 326 113.324 34.76 306 Class 2T - yellow 202 87.783 43.46 268 Luminance (%) Class 2 - white 55 10.189 18.53 3996 Class 2 - yellow 37 5.241 14.16 3126 Class 2T - white 65 5.437 8.36 306 Class 2T - yellow 40 2.910 7.28 268 Table 19 shows that the most consistent property was luminance, with the lowest coefficients of variation. Thickness appears to be the less consistent one, as coefficient of variation can be as high as 55.13%. Retroreflectivity also has very dispersed observations, with a high coefficient of variation of 43.44% among the four color/class categories, and also very different means when contrasting all classes and types. 4.2. Retroreflectivity Modeling 4.2.1. Retroreflectivity Curve Table 20 and Figure 16 show an example of the data and plot developed for all mileposts of the 15 projects that had data for four years. The total observations allowed for development of 76 curves for solid white markings, 17 for broken white, 67 for solid yellow and 32 for broken yellow markings. 56 Table 20 ? Retroreflectivity Values of Project NHF-STPSAF-0053(525), for MP36 Curve 3: MP 36 Solid Yellow Date Tested (months) Retroreflectivity (mcd/m2/lux) 0.00 249.90 8.40 291.00 22.77 224.00 34.67 220.50 Figure 16 ? Plot of Retroreflectivity Values of Project NHF-STPSAF-0053(525), for MP36 4.2.2. Model Fitting Analysis The end of pavement marking service life can be determined when retroreflectivity reaches minimum values. The plots developed in Section 4.2.1 had retroreflectivity data for three years after markings? application, which did not represent the end of service lives. Retroreflectivity data had to be extrapolated until it could reach minimum threshold values, and to accomplish that, several models existing in the literature were tested to determine which one best represented actual data. For solid white and broken white pavement markings, 10 models were compared to actual data, Lee et al., Abboud and Bowman [1], Thamizharasan et al. linear, Sitzabee et al, Abboud and Bowman [2], Thamizharasan et al. non-linear, Bahar et al., Clarke 57 and Yan linear, Clarke and Yan logarithmic, and Clarke and Yan quadratic. For solid yellow and broken yellow markings, the same models were tested, except for Abboud and Bowman models, which were developed specifically for white markings. An example of application of the models found in the literature to the observed data, using data from Project STPSA-0185(500), Milepost 10, Solid White, can be observed in Figure 17. Figure 17 ? Comparison between Models for Project STPSA-0185(500), MP10, Solid White -200.00 0.00 200.00 400.00 600.00 800.00 1000.00 0.00 10.00 20.00 30.00 40.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Date Tested (months) Project STPSA-0185(500) MP10 Solid White Actual Data Lee et al. Abboud and Bowman [1] Thamizharasan et al. Linear Sitzabee et al. Abboud and Bowman [2] Thamizharasan et al. Non- Linear Bahar et al. Clarke and Yan Linear Clarke and Yan Logarithmic Clarke and Yan Quadratic 58 The comparison between models, using the methods area under curve ratio, average model error, and average percent error, as described in Section 3.3.2, was performed. The three best-performing models for each color and type can be observed in Table 21. Number of locations indicate the number of curves for each marking color and type (the number of existing mileposts with data). Table 21 ? Models in the Literature that Best Represented Actual Data Marking Color and Type Number of Locations Best-Performing Models Area Under Curve Ratio Average Model Error Average Percent Error (%) Solid White 76 Thamizharasan Linear 1.02 Thamizharasan Linear 10.54 Sitzabee -7.48 Sitzabee 0.86 Sitzabee -55.84 Thamizharasan Linear 8.20 Thamizharasan Non-Linear 0.85 Thamizharasan Non-Linear -89.76 Thamizharasan Non-Linear -20.27 Solid Yellow 67 Sitzabee 0.91 Thamizharasan Non-Linear 24.79 Sitzabee -1.33 Thamizharasan Linear 1.14 Thamizharasan Linear 29.84 Clarke Quadratic -19.37 Thamizharasan Non-Linear 1.14 Sitzabee -30.44 Thamizharasan Non-Linear 22.49 Broken White 17 Sitzabee 0.97 Thamizharasan Non-Linear -13.68 Thamizharasan Non-Linear 0.78 Thamizharasan Non-Linear 1.09 Sitzabee -19.76 Sitzabee 4.91 Clarke Logarithmic 0.83 Thamizharasan Linear 76.96 Clarke Logarithmic -11.45 Broken Yellow 32 Thamizharasan Non-Linear 1.06 Thamizharasan Non-Linear 7.92 Thamizharasan Non-Linear 8.61 Thamizharasan Linear 1.06 Thamizharasan Linear 13.63 Sitzabee -9.58 Sitzabee 0.86 Sitzabee -40.49 Thamizharasan Linear 11.55 59 4.3. Benefit/Cost Calculation The determination of a benefit/cost ratio for each observation from the 15 projects with retroreflectivity data from 2007 to 2010 is described in this section. This analysis can identify which projects were best executed, by color and type, and make them models for future pavement marking construction in Alabama. The most adequate benefit/cost method to be applied in this study was the Fu and Wilmot method, and for that retroreflectivity extrapolation was required. From Table 21, it can be seen that Thamizharasan et al. Linear and Sitzabee et al. models were found to be the most appropriate to represent actual data. Therefore, an extrapolation of actual data, beyond the 4 years for which data are available, was performed, based on these two models. However, the service life, determined by when retroreflectivity reaches the minimum threshold value of 150 mcd/m2/lux for white markings and 100 mcd/m2/lux for yellow markings, is much higher than observed values for markings found in the literature. This fact can be illustrated by Figure 18, which shows estimated service life equal to 250 months; Figure 19, where estimated service life is 150 months; and Table 22, from where it can be observed that all service life values are greater than the maximum found in the literature (102 months for white markings). Table 22 ? Example of Benefit/Cost and Service Life calculations SOLID WHITE Project MP Benefit/Cost Thamizharasan et al. Model (month*vehicle*mcd/m2/lux) Benefit/Cost Sitzabeeet al. Model (month*vehicle*mcd/m2/lux) Service Life Thamizharasan et al. Model (months) Service Life Sitzabee et al. Model (months) 99-307- 164-167- 701 1 12379 7156 256 152 2 10587 6555 221 140 3 10946 6670 228 142 4 9883 6311 207 135 5 11381 6823 236 145 6 9742 6273 204 135 60 Figure 18 ? Thamizharasan et al. Linear Extrapolation of Actual Data Figure 19 ? Sitzabee et al. Extrapolation of Actual Data 61 Table 22 also shows that the Benefit/Cost values do not provide a useful idea of benefit related to cost; the meaning of the unit ?month*vehicle*mcd/m2/lux? is not tangible. In the literature, this method was valuable when comparing different materials, such as markings of Class 1 compared to markings of Class 2, and compared to markings of Class 2T, and determining the most cost-effective material among them. As the observations of the 15 projects in Alabama use only one type of material, thermoplastic Class 2, this benefit/cost analysis approach is not suitable in the case of the available data. 4.4. Modeling of Retroreflectivity Over Time The models found in the literature, when applied to the Alabama data and extrapolated to acceptable minimum retroreflectivity values, do not result in reasonable service life projections. Therefore, new models were developed to try to represent better the behavior of retroreflectivity over time. First, locations were grouped by color and type: Solid White (76 locations), Broken White (17 locations), Solid Yellow (67 locations), and Broken Yellow (32 locations). For model development, observations from all 15 projects were grouped only by color, one model for yellow and another for white markings; additionally, those below minimum initial retroreflectivity thresholds required by ALDOT were not considered. This approach to data organization for model development is consistent with those found in the literature. After applying this procedure, there are 63 observations for white markings and 42 for yellow markings. In the literature, it is possible to observe that most models consider initial retroreflectivity and age of marking as the independent variables to estimate retroreflectivity; some models also consider traffic volume as an independent variable. For this reason, two different approaches 62 were adopted in this study; first, four models were developed, by color, for data in Alabama: linear, power, quadratic, and exponential, as mentioned in Section 3.5. Table 23 shows all developed models for this approach and the corresponding R2 for white markings in 63 locations and yellow markings in 42 locations. The dependent variable is retroreflectivity (g1844), measured in mcd/m2/lux; the two independent variables are initial retroreflectivity (g1844g2868g4667, measured in mcd/m2/lux, and age of marking (g1872), measured in months. Table 23 ? Developed Models considering initial retroreflectivity and age as independent variables Color Model Equation R2 White Linear R = 0.3242Ro - 4.745t + 384.4 0.325 Power R = 366.8 * 1.001Ro *0.9900t 0.335 Quadratic R = Ro - 2.865t - 0.05300t2 + 1.231 0.181 Exponential R = Ro * e-0.009747t 0.261 Yellow Linear R = 0.1222R0 - 4.778t + 338.3 0.465 Power R = 276.8 * 1.001Ro *0.9839t 0.480 Quadratic R = Ro - 1.926t - 0.05119t2 - 23.67 0.320 Exponential R = Ro * e-0.01682t 0.440 The second approach to the retroreflectivity estimate considered initial retroreflectivity, age of marking, and AADT as independent variables. Only linear models were developed, by color, which was consistent with the existing models in the literature. Table 24 shows the models for this approach and the corresponding R2 for white markings in 63 locations and yellow markings in 42 locations. Table 24 ? Developed Models considering initial retroreflectivity, age, and AADT as independent variables Color Model Equation R2 White Linear R = 0.2296Ro - 4.967t - 0.004665AADT + 470.7 0.398 Yellow Linear R = 0.03928Ro - 4.932t - 0.002030AADT + 381.4 0.479 63 4.4.1. Approach One: Initial Retroreflectivity and Age of Marking as Independent Variables The linear model developed in this section was applied to estimate retroreflectivity over time for age and initial retroreflectivity values of actual data. The comparison for white markings between actual data points and linear model data points can be observed in Figure 20. It can be inferred from the figure that the range of data points representing the linear regression is not very similar to the range representing actual observations. The R2 value for this relationship, as shown in Table 23, was 0.325. The data set could not be divided into more ranges because the number of locations in each would not be significant. Figure 20 ? Linear Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables 0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Linear Regression) Linear Estimates Actual Data 64 In order to ascertain how different initial retroreflectivity values influence where the model intercepts the Y axis, and also to compare the range of points in the model to the range formed by actual data points, different initial retroreflectivity values were tested from age zero until 40 months. For this analysis, with white markings, initial retroreflectivity values were set equal to 450 mcd/m2/lux, 550 mcd/m2/lux, 650 mcd/m2/lux, and 750 mcd/m2/lux. It can be noticed in Figure 21 that the regression lines occupy the middle region of the area formed by actual data points. The initial points as obtained from the model are not actual initial retroreflectivity values. This happens because the general linear model has a constant and a coefficient multiplying initial retroreflectivity causing that when age is equal to zero, retroreflectivity is not equal to initial retroreflectivity. Figure 21 ? White Markings Linear Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables 0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Linear Regression) Actual Data Ro=450mcd/m2/lux Ro=550mcd/m2/lux Ro=650mcd/m2/lux Ro=750mcd/m2/lux 65 In trying to develop non-linear relationships between data and determine which models better represent retroreflectivity behavior over time in Alabama, a power regression model was developed. The comparison for white markings between actual data points and power model data points can be observed in Figure 22. In this case, as with the linear model, the range of data resulting from the power regression is narrower than the range of actual observations. The R2 for this relationship, as shown in Table 23, was 0.335. Figure 23 shows that the power regression curves for initial retroreflectivity values of 450, 550, 650, and 750 mcd/m2/lux also occupy the middle region of the area formed by actual data points; a similar pattern as was seen in Figure 23 for the linear model is apparent. Figure 22 ? Power Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables 0 100 200 300 400 500 600 700 800 900 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Power Regression) Power Estimates Actual Data 66 Figure 23 ? White Markings Power Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables A quadratic model was also developed in this section and applied to calculate retroreflectivity over time for age and initial retroreflectivity values of actual data. The comparison for white markings between actual data points and quadratic model data points can be observed in Figure 24. It is possible to observe that the range of data points from the quadratic regression is very similar to the range of actual observations, more so than was seen with the linear and power functions. The R2 value for this relationship, as shown in Table 23, was 0.181. It can be observed in Figure 25 that points projected by the quadratic regression curves are in almost the entire region formed by the actual data points. There is high accuracy on where regression lines start, their initial points are very close to the observed initial retroreflectivity values. This happens because the constant in the quadratic model is not very high (1.231) and there is not a coefficient multiplying initial retroreflectivity. 0 100 200 300 400 500 600 700 800 900 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Power Regression) Actual Data Ro=450mcd/m2/lux Ro=550mcd/m2/lux Ro=650mcd/m2/lux Ro=750mcd/m2/lux 67 Figure 24 ? Quadratic Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables Figure 25 ? White Markings Quadratic Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables 0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Quadratic Regression) Quadratic Estimates Actual Data 0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Quadratic Regression) Actual Data Ro=450mcd/m2/lux Ro=550mcd/m2/lux Ro=650mcd/m2/lux Ro=750mcd/m2/lux 68 This study included an exponential model as well. The comparison for white markings between actual data points and exponential model data points can be observed in Figure 26. Figure shows that the range of data points estimated by the exponential regression is very similar to actual observations. The R2 value for this relationship, as shown in Table 23, was 0.261. Figure 26 ? Exponential Regression estimates for White Markings: Initial Retroreflectivity and Age as Independent Variables Figure 27 shows that the exponential regression curves occupy almost all the area formed by actual data points. Unlike the other model forms developed, the initial points of the exponential model are exactly initial retroreflectivity values. This happens because when age is equal to zero, the exponential function of age has a value of one. 0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Exponential Regression) Exponential Estimates Actual Data 69 Figure 27 ? White Markings Exponential Regression for Different Initial Retroreflectivity Values: Initial Retroreflectivity and Age as Independent Variables The same process was conducted for yellow markings. Adopted initial retroreflectivity values, to verify the behavior of the developed models in the area occupied by actual observations, were equal to 350 mcd/m2/lux, 400 mcd/m2/lux, 450 mcd/m2/lux, and 500 mcd/m2/lux. Conclusions were the same as the observed for white markings. Detailed charts are presented in Appendix C. 4.4.2. Approach Two: Initial Retroreflectivity, Age of Marking, and Traffic Volume as Independent Variables With three independent variables, only a linear model was developed by color, as it is usually seen in the literature. Figure 28 shows the data estimated by the linear regression compared to actual data. Figure 29 illustrates estimated data for initial retroreflectivity values 0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings Exponential Regression Actual Data Ro=450mcd/m2/lux Ro=550mcd/m2/lux Ro=650mcd/m2/lux Ro=750mcd/m2/lux 70 equal to 450, 600, and 750 mcd/m2/lux and AADT equal to 350, 7000, and 18000 veh/day. These values were observed to be the minimum, average, and maximum values of actual initial retroreflectivity for white markings and traffic volume data, respectively, for the 15 study sites. Both figures show that the initial points of the model are not actual initial retroreflectivity values, because the linear model has a constant and a coefficient multiplying initial retroreflectivity and AADT causing that when age and AADT are equal to zero, retroreflectivity is not equal to initial retroreflectivity. However, this linear regression estimated values for retroreflectivity closer to actual values than the linear model that did not consider AADT as an independent variable; this can be verified by comparing Figures 28 and 29 to Figures 20 and 21, and also Table 24 to Table 23, which shows a higher R2 value (0.398) for the regression considering AADT. The same procedure was applied to yellow markings for initial retroreflectivity values equal to 350, 425, and 500 mcd/m2/lux, and the conclusions were similar; details can be seen in Appendix D. Figure 28 ? Linear Regression estimates for White Markings: Initial Retroreflectivity, Age, and AADT as Independent Variables 0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Linear Regression) Linear Estimates Actual Data 71 Figure 29 ? White Markings Linear Regression for Different Initial Retroreflectivity and AADT Values: Initial Retroreflectivity, Age, and AADT as Independent Variables 4.4.3. Service Life Another important issue to be considered is service life. Considering as thresholds for minimum retroreflectivity 150 mcd/m2/lux for white markings and 100 mcd/m2/lux for yellow markings, the predicted service life using each developed model in this section was calculated. Table 25 shows estimated service life for different values of initial retroreflectivity, considering initial retroreflectivity and age of marking as independent variables (approach one). It can be observed in Table 26 estimated service life also considering traffic volume as an independent variable (approach two). 0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) White Markings (Linear Regression) Actual Data Ro = 450 mcd/m2/lux and AADT = 350 veh/day Ro = 450 mcd/m2/lux and AADT = 7000 veh/day Ro = 450 mcd/m2/lux and AADT = 18000 veh/day Ro = 600 mcd/m2/lux and AADT = 350 veh/day Ro = 600 mcd/m2/lux and AADT = 7000 veh/day Ro = 600 mcd/m2/lux and AADT = 18000 veh/day Ro = 750 mcd/m2/lux and AADT = 350 veh/day Ro = 750 mcd/m2/lux and AADT = 7000 veh/day 72 Table 25 ? Predicted Service Life: Initial Retroreflectivity and Age as Independent Variables Model Color Predicted Service Life (months) Considered Initial Retroreflectivity (mcd/m2/lux) Minimum Retroreflectivity Threshold (mcd/m2/lux) White Linear 81 450 150 Power 134 Quadratic 54 Exponential 110 Linear 87 550 Power 144 Quadratic 65 Exponential 130 Linear 94 650 Power 154 Quadratic 74 Exponential 147 Linear 101 750 Power 164 Quadratic 83 Exponential 161 Yellow Linear 59 350 100 Power 85 Quadratic 51 Exponential 74 Linear 61 400 Power 88 Quadratic 58 Exponential 82 Linear 62 450 Power 92 Quadratic 64 Exponential 89 Linear 63 500 Power 95 Quadratic 70 Exponential 95 73 Table 26 ? Predicted Service Life: Initial Retroreflectivity, Age, and AADT as Independent Variables Color Model Predicted Service Life (months) Considered AADT (veh/day) Considered Initial Retroreflectivity (mcd/m2/lux) Minimum Retroreflectivity Threshold (mcd/m2/lux) White Linear 86 350 450 150 79 7000 68 18000 93 350 600 86 7000 75 18000 99 350 750 93 7000 82 18000 Yellow Linear 60 350 350 100 57 7000 53 18000 61 350 425 58 7000 54 18000 61 350 500 59 7000 54 18000 It could be observed in the literature that the service life for white thermoplastic markings was in a range between 22.6 and 102 months. It is important to notice that the maximum service life was a function of an initial retroreflectivity equal to 375 mcd/m2/lux. Table 25 shows that the maximum service life for white markings is 166 months, but it considers an initial retroreflectivity of 750 mcd/m2/lux, higher than the values found in the literature. For yellow thermoplastic markings, service life in the literature was in a range between 24.7 and 103 months. The maximum service life considered an initial retroreflectivity equal to 135 mcd/m2/lux. Table 25 shows that the maximum service life for yellow markings is 96 months, but it considers an initial retroreflectivity of 500 mcd/m2/lux, higher than the values 74 found in the literature. Therefore, values shown on Table 25 are considered consistent with what was found in the literature. Some high estimation for service life may be due to the high initial retroreflectivity values from actual observations. It can be observed in Table 26 the estimated service life values for different initial retroreflectivity and AADT values, considering three independent variables (approach two). Similar conclusions from service life estimated by approach one were found. 4.5. Summary of Chapter Four Chapter Four presented the results regarding the application of methodology described in Chapter Three. The analysis to verify if measures of markings met specifications showed that the most consistent property was chromaticity, with 94.04% of total measurements passing minimum required values. Thickness and luminance measurements presented 79.53% and 75.77%, respectively, of total observations meeting specifications. The property which presented fewer measurements passing minimum required values was retroreflectivity, with 69.40% observations meeting specifications. The statistical analysis of observations in the 40 projects of the 2007 database showed that the most consistent property was luminance, with the lowest coefficients of variation. Thickness was found to be the less consistent one, because some standard deviations are almost half of, or more than the mean value, and coefficient of variation could be as high as 55.13%. Retroreflectivity also had very dispersed observations, with a high coefficient of variation of 43.44% among the four color/class categories, and also very different means when contrasting all classes and types. 75 Modeling of retroreflectivity over time was performed and all observations of the 15 projects that had data for four years allowed the development of 76 curves for solid white markings, 17 for broken white, 67 for solid yellow and 32 for broken yellow markings. Four models were developed, by color, based on the observations of these 15 projects. Considering initial retroreflectivity and age of marking as independent variables, for white markings the R2 values were 0.325 for linear, 0.335 for power, 0.181 for quadratic, and 0.261 for exponential models; for yellow markings, the R2 values were 0.465 for linear, 0.480 for power, 0.320 for quadratic, and 0.440 for exponential models. Considering also AADT as an independent variable, the R2 values were 0.398 for white and 0.479 for yellow models, both linear. The benefit/cost method analyzed in this Chapter did not provide good results, especially because the service life, determined by when retroreflectivity reaches the minimum threshold value of 150 mcd/m2/lux for white markings and 100 mcd/m2/lux for yellow markings, was much higher than the observed values for markings in literature and, therefore, not reliable. 76 Chapter Five Conclusions and Recommendations 5.1. Conclusions The documentation of 2007 ALDOT Special Provisions changes showed that no significant modifications related to specifications were observed; text modifications constituted the changes among different Special Provisions. An examination of the 2007 database containing 7,840 observations for 40 projects in Alabama to determine whether thickness, retroreflectivity, luminance, and color complied with ALDOT specifications found that 79.53% of the observations complied with thickness specifications, 69.40% were according to retroreflectivity standards, 75.77% conformed to luminance specifications, and 94.03% were according to color standards. A statistical analysis was performed for all observations among the 40 projects, based on mean, standard deviation, and coefficient of variation, by color and type (class) of markings for thickness, retroreflectivity, and luminance. Results showed that luminance measures were consistent, with highest coefficient of variation of 18.33%. Thickness values were not very consistent within each color/type group, with a coefficient of variation as high as 55.13%, but means between groups were similar, varying from 0.065 to 0.117 in. Retroreflectivity, however, presented huge variations within groups, with standard deviation as great as 141.383 mcd/m2/lux, and between groups, with mean varying from 202 to 496 mcd/m2/lux. 77 Retroreflectivity modeling was also executed in this thesis. Joining retroreflectivity data from the 2007 database and subsequent ALDOT data on retroreflectivity from 2008 to 2010, it was possible to create models of retroreflectivity over time. Only 15 projects had measurements of retroreflectivity for the same locations from 2007, 2008, 2009, and 2010 databases, which represented 76 observations for solid white markings, 17 for broken white, 67 for solid yellow, and 32 for broken yellow markings. Since only four years of data exist, the created curves had to be extrapolated to reach minimum retroreflectivity values (until the end of service life). In order to accomplish this, the most relevant models in the literature were applied to the data to evaluate model fit. The Thamizharasan et al. and Sitzabee et al. were the models that best represented actual data from ALDOT projects. To determine best-fitting models, comparisons of area under curve ratio, average model error and average percent error between model curve and actual curve were performed. When extrapolating actual data based on the best-fitting models in the literature, it was found that service life was about twice the common values for thermoplastic materials found in the literature, and sometimes higher than typical pavement service life, being as high as 250 months (almost 21 years). Cost data was obtained from and organized based on the Tabulation of Bids, from ALDOT. However, without retroreflectivity curves that yielded realistic service lives, it was not possible to calculate benefits based on this model. In addition, benefit/cost relationships given by this method did not provide meaningful and applicable results for the observations of the 15 projects in Alabama, since there was only one type of material, thermoplastic Class 2, and this benefit/cost analysis is most appropriate when comparing different materials. 78 New models needed to be developed to represent actual points from projects in Alabama more realistically. The modeling of retroreflectivity over time considered retroreflectivity as the dependent variable; age and initial retroreflectivity were considered as the independent variables for the first modeling approach and traffic volume was also considered for a second modeling approach. Models developed when considering age and initial retroreflectivity as independent variables were linear, power, quadratic, and exponential, for each color. For white markings, there were 63 different locations and for yellow markings, 42. White markings had R2 values equal to: 0.325 (linear), 0.335 (power), 0.181 (quadratic), and 0.261 (exponential); yellow markings had R2 values equal to: 0.465 (linear), 0.480 (power), 0.320 (quadratic), and 0.440 (exponential). For the consideration of AADT as an additional independent variable, linear models were developed for the same locations by color and R2 values were 0.398 for white markings and 0.479 for yellow markings. In the literature, R2 values from models for white markings vary from 0.007 to 0.600 and from 0.003 to 0.600 to yellow markings; the R2 values of the models developed in this study are higher than most models in the literature. 5.2. Recommendations The linear model considering age of marking, initial retroreflectivity, and traffic volume as independent variables had the highest R2 value among all predicted models for white markings. For yellow markings, the power model considering age of marking and initial retroreflectivity as independent variables and the linear model considering traffic volume as an additional independent variable, yielded R2 values equal to 0.480. The linear model considering the three independent variables to estimate retroreflectivity over time was found to be the most adequate when representing Alabama data because of the simplicity in its equation, the capacity 79 to well-represent retroreflectivity for a time equal to zero, and the highest R2 values among predicted models. 5.3. Recommendations for Subsequent Studies Limitations to the present study can be noticed throughout this thesis. From all 40 projects applied in 2007 in Alabama, only 15 had retroreflectivity data, which were used to develop models. In addition, historical retroreflectivity measurements were made only from 2007 to 2010. This means that models can be improved if data are provided for new locations and measurements continue for the existing sites. New sites can also be added to increase sample size and provide more accurate results. The wide range of initial retroreflectivity values for different observations of the 15 projects was something that also deserves further investigation. A detailed analysis of the equipment used to measure each observation and development of adjustment factors between measures from one retroreflectometer to another can be ways to explain potential source of variability. Another point to be considered is which variables influence retroreflectivity behavior. In this thesis, age of marking, initial retroreflectivity, and traffic volume were the considered independent variables. However, additional variables might be analyzed to improve models. It was observed in the literature that white edgeline markings generally had longer predicted service lives than all other pavement marking types and the shortest mean predicted service life was most often computed for the yellow centerline markings, showing that position of marking on pavement may be considered when predicting retroreflectivity. In addition, it was observed in the literature that there was considerable variability in the predicted service life across different 80 engineering divisions, which shows the importance of geographic location when developing models to predict service life. The impact of different vehicle types was also mentioned in the literature, and it was found that roadways with higher truck traffic volumes decrease pavement markings service life due to abrasion between the tires and the markings; therefore, vehicle type might also be an independent variable to predict retroreflectivity. A benefit/cost analysis is useful, especially if information of other types of markings is available; comparison between different materials to determine the most appropriate to Alabama conditions can be interesting. Pay adjustment factors procedure in the 2008 ALDOT Standard Specifications for Highway Construction does not provide consistent justification on how the percentage that the stripe will be paid for was established. The stripe is accepted without a price adjustment for retroreflectivity 85% or greater than the minimum required value; for retroreflectivity less than 85% and greater than 50% the target value, the stripe is paid for at a percentage equal to the percentage determined from the measurements; for retroreflectivity less than 50%, the stripe needs to be removed. This procedure could be refined; therefore, the development of pay adjustment factors may also be another suggestion for future studies. This analysis can be based on the data that did not meet minimum requirements as well as those that did in the 2007 ALDOT database, in addition to a study considering benefit/cost models. 81 References Abboud, N. and Bowman, B. L. (2002) [1]. Cost and longevity-based scheduling of paint and thermoplastic striping. Transportation Research Record. Journal of the Transportation Research Board, No. 1794. Transportation Research Board of the National Academies, Washington, D.C., pp. 55?62. Abboud, N. and Bowman, B. L. (2002) [2]. Establishing Crash-Based Retroreflectivity Threshold. Presented at 81st Annual Meeting of the Transportation Research Board of the National Academies, Washington, D.C. Alabama Department of Transportation (ALDOT). (2006). Standard Specifications for Highway Construction: 2006 Edition. Montgomery, AL. Alabama Department of Transportation (ALDOT). (2008). Standard Specifications for Highway Construction: 2008 Edition. Montgomery, AL. Alabama Department of Transportation (ALDOT). (2007). Standard Specifications for Highway Construction, Special Provisions No. 06-0102, 06-0102[2], 06-0102[3], 06-0102[4], and 06-0102[5]. Montgomery, AL. Alabama Department of Transportation (ALDOT). (2010) [1]. Tabulation of BIDS. Web. 4 June, 2010. . Montgomery, AL. Alabama Department of Transportation (ALDOT). (2010) [2]. Traffic Data. Web. 15 October, 2010. . Montgomery, AL. Andrady A.L. (1997). Pavement Marking Materials: Assessing Environment-Friendly Performance. National Cooperative Highway Research Program Report 392. Transportation Research Board of the National Academies, Washington, D.C. Bahar, G., Masliah, M., Erwin, T., Tan, E. (2006). Web-Only Document 92: Pavement Markings Materials and Markers: Real-World Relationship Between Retroreflectivity and Safety Over Time. National Cooperative Highway Research Program. Transportation Research Board of the National Academies. Web. 16 September, 2010. . Washington, D.C. 82 Clarke, D. B. and Yan, X. (2009). Retroreflectivity Performance of 90-mil Thermoplastic Longitudinal Pavement Markings During the Early Application Period. Presented at 88th Annual Meeting of the Transportation Research Board of the National Academies, Washington, D.C. Cottrell, B. and Hanson, R. (2001). Determining the Effectiveness of Pavement Marking Materials. Virginia Transportation Research Council, Report No. VTRC 01-R9, Charlottesville, VA. Cuelho, E., Stephens, J., McDonald, C. (2003). A Review of the Performance and Costs of Contemporary Pavement Marking Systems. U.S. Department of Transportation, Federal Highway Administration. Report No. FHWA/MT-03-001/8117-17. Helena, MT. Debaillon, C., Carlson, P. J., Hawkins, H. G., He, Y., Schnell, T., Aktan, F. (2008). Review and Development of Recommended Minimum Pavement Marking Retroreflectivity Levels. Transportation Research Record. Journal of the Transportation Research Board, No. 2055. Transportation Research Board of the National Academies, Washington, D.C., pp. 71?77. Federal Highway Administration. (2000). Evaluation of All Weather Pavement Markings. Washington, D.C. Federal Highway Administration. (2009). Manual on Uniform Traffic Control Devices for streets and Highways (MUTCD): 2009 Edition. Washington, D.C. Federal Highway Administration. (2010). Proposed Pavement Marking Retroreflectivity MUTCD Text: Revision 1. Web. 5 November, 2010. Washington, D.C. Fu, H., Wilmot, C. G. (2008). Assessing Performance of Alternative Pavement Marking Materials. Louisiana Department of Transportation and Development. Louisiana Transportation Research Center. Report No. 08-4TA. Baton Rouge, LA. Jiang, Y. (2008). Durability and Retro-Reflectivity of Pavement Markings. U.S. Department of Transportation, Federal Highway Administration. Report No.: FHWA/IN/JTRP-2007/11, SPR-300. West Lafayette, IN. Kansas Department of Transportation (KDOT). (2002). Kansas Department of Transportation Pavement Marking Policy. Topeka, KS. Karwa, V., Donnell, E. T. (2011). Predicting Pavement Marking Retroreflectivity Using Artificial Neural Networks: Explanatory Analysis. American Society of Civil Engineers. Journal of Transportation Engineering, Vol. 137, No. 2, pp. 91-103. Reston, VA 83 Lee, J., Maleck, T., Taylor, W. (1999). Pavement Marking Material Evaluation Study in Michigan. Institute of Transportation Engineers Journal, Vol. 69, No. 7, pp. 44-51. Washington, D.C. Lindly, J. K. and Marci, A. (2006). Evaluation of Rumble Stripe Markings. The University of Alabama, Report No. 04405. Tuscaloosa, AL. Lindly, J. K. and Wijesundera, R. (2003). Evaluation of Profiled Pavement Markings. The University of Alabama, Report No. 01465. Tuscaloosa, AL. Loetterle, F. E., Beck, R. A., Carlson, J. (2001). Public Perception of Pavement-Marking Brightness. Transportation Research Record. Journal of the Transportation Research Board, No. 1715. Transportation Research Board of the National Academies, Washington, D.C., pp. 51?59. Migletz, J. and Graham, J. (2002). Long-Term Pavement Marking Practices: A Synthesis of Highway Practice. National Cooperative Highway Research Program Synthesis 306. Transportation Research Board of the National Academies, Washington, D.C. Montebello, D., Schroeder, J. (2000). Cost of Pavement Marking Materials. Minnesota Department of Transportation, Report No, 2000-11. St. Paul, MN. National Institute of Standards and Technology (NIST). (1979). International System of Units (SI). Web. 11 November, 2011. Parker, N. A. and Meja, M. S. J. (2003). Evaluation of Performance of Pavement Markings. Transportation Research Record. Journal of the Transportation Research Board, No. 1824. Transportation Research Board of the National Academies, Washington, D.C., pp. 123?132. Sitzabee, W. E., Hummer, J. E., Rasdorf, W. (2009). Pavement Marking Degradation Modeling and Analysis. American Society of Civil Engineers. Journal of Infrastructure Systems, Vol. 15, No. 03, pp. 190-199. Reston, VA. Smadi, O., Souleyrette, R.R., Ormand, D. J., Hawkins, N. (2008). Pavement Marking Retroreflectivity: Analysis of Safety Effectiveness. Transportation Research Record. Journal of the Transportation Research Board, No. 2056. Transportation Research Board of the National Academies, Washington, D.C., pp. 17?24. Songchitruksa, P., Ullman, G. L., Pike, A. M. (2011). Guidance for Cost-Effective Selection of Pavement Marking Materials for Work Zones. American Society of Civil Engineers. Journal of Infrastructure Systems, Vol. 17, No. 02, pp. 55-65. Reston, VA. 84 Thamizharasan, A., Sarasua, W., Clarke, D., Davis, W. (2003). A Methodology for Estimating the Lifecycle of Interstate Highway Pavement Marking Retroreflectivity. Presented at 82nd Annual Meeting of the Transportation Research Board of the National Academies, Washington, D.C. Thomas, G. B., Schloz, C. (2001). Durable, Cost-Effective Pavement Markings. Phase I: Synthesis of Current Research. Center for Transportation Research and Education, Iowa State University. Ames, IA. Thomas, G. B. Jr., Finney, R. L. (1996), Calculus and Analytic Geometry. 9th ed. University of Alabama. (2011). Alabama Maps. Web. 30 September, 2011. Zhang, Y., Wu, D. (2006). Methodologies to Predict Service Lives of Pavement Marking Materials. Transportation Research Forum. Journal of the Transportation Research Forum, Vol. 45, No. 3, pp. 5-18. Fargo, ND. 85 Appendix A Project STPSA-0185(500): Pictures 86 87 88 89 Appendix B List of Mileposts by Project 90 Project MP 99-302-391-101-701 Solid White Broken White Solid Yellow Broken Yellow 25.7 - 25.7 26.2 - 26.2 27 - 27 28 - 28 99-302-437-171-706 53 53 53 99-303-595-003-709 244 - 244 244 245 - 245 245 99-305-632-069-702 157.1 - - 157.1 158 - 158 - 159 - 159 159 160 - 160 - 161 - 161 161 162.1 - 162.1 - 163.1 - 163.1 - 164 - 164 - 165 - 165 - 166 - 166 - 167 - 167 - 99-307-164-167-701 1 - - 1 2 - 2 - 3 - - 3 4 - - 4 5 - - 5 6 - 6 - 99-307-346-010-701 216 216 216 - 217 - 217 217 91 EB-0016(505) 60 - 60 - 61 - 61 61 62 - 62 - 63 - 63 63 64 - 64 - 65 - 65 65 66 - 66 66 67 - 67 67 68 - - 68 69 - 69 69 70 - 70 70 71 - 71 - 72 - 72 - 73 - 73 - 74 - 74 - 75 - - 75 76 - - 76 EB-0035(506) - 23 23 23 - 24 24 24 - 25 25 - EB-0074(513) 76 76 76 - 77 77 77 77 NHF-STPSAF- 0053(525) 32 32 32 - 33 33 33 - 34 34 34 - 35 35 35 - 36 36 36 - STPSA-0001(529) 267 267 267 - 268 268 268 - 269 269 269 - 270 270 270 - 271 271 271 - 92 STPSA-0021(515) 98 - 98 - 99 - 99 99 100 - 100 - 101 - - 101 102 - 102 102 103 - 103 - 104 - 104 - 105 - 105 - 106 - - - 107 - 107 - 108 - 108 108 STPSA-0079(505) - - 72 - - - 73 - - - - 74 - - - 75 - - 76 - - - 77 - - - - 78 - - 79 - - - 80 - - - - 81 STPSA-0089(500) 1 - 1 - 2 - 2 - 3 - - 3 STPSA-0185(500) 8 - - - 9 - - - 10 - - - 11 - - - 12 - - - 13 - - - 14 - - - 93 Appendix C Retroreflectivity Modeling for Yellow Markings: Initial Retroreflectivity and Age as Independent Variables 94 0 100 200 300 400 500 600 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings (Linear Regression) Linear Estimates Actual Data 0 100 200 300 400 500 600 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings (Power Regression) Power Estimates Actual Data 95 0 100 200 300 400 500 600 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings (Quadratic Regression) Quadratic Estimates Actual Data 0 100 200 300 400 500 600 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings (Exponential Regression) Exponential Estimates Actual Data 96 0.00 100.00 200.00 300.00 400.00 500.00 600.00 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings Linear Regression Actual Data Ro=350mcd/m2/lux Ro=400mcd/m2/lux Ro=450mcd/m2/lux Ro=500mcd/m2/lux 0.00 100.00 200.00 300.00 400.00 500.00 600.00 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings Power Regression Actual Data Ro=350mcd/m2/lux Ro=400mcd/m2/lux Ro=450mcd/m2/lux Ro=500mcd/m2/lux 97 0.00 100.00 200.00 300.00 400.00 500.00 600.00 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings Quadratic Regression Actual Data Ro=350mcd/m2/lux Ro=400mcd/m2/lux Ro=450mcd/m2/lux Ro=500mcd/m2/lux 0.00 100.00 200.00 300.00 400.00 500.00 600.00 0.00 10.00 20.00 30.00 40.00 50.00 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings Exponential Regression Actual Data Ro=350mcd/m2/lux Ro=400mcd/m2/lux Ro=450mcd/m2/lux Ro=500mcd/m2/lux 98 Appendix D Retroreflectivity Modeling for Yellow Markings: Initial Retroreflectivity, Age, and Traffic Volume as Independent Variables 99 0.00 100.00 200.00 300.00 400.00 500.00 600.00 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings (Linear Regression) Linear Estimates Actual Data 0 100 200 300 400 500 600 0 10 20 30 40 50 Re tro ref lec tiv ity (m cd /m 2/ lux ) Age (months) Yellow Markings (Linear Regression) Actual Data Ro = 450 mcd/m2/lux and AADT = 350 veh/day Ro = 450 mcd/m2/lux and AADT = 7000 veh/day Ro = 450 mcd/m2/lux and AADT = 18000 veh/day Ro = 600 mcd/m2/lux and AADT = 350 veh/day Ro = 600 mcd/m2/lux and AADT = 7000 veh/day Ro = 600 mcd/m2/lux and AADT = 18000 veh/day Ro = 750 mcd/m2/lux and AADT = 350 veh/day