Quantifying the Effects of Preservation Treatments on Pavement Performance (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

A-1 A T T A C H M E N T Guide for Pavement Preservation Performance Measures

A-2 C O N T E N T S A-3 Chapter A-1 Introduction A-3 A-1.1 Background A-3 A-1.2 Objective A-4 A-1.3 Using this Guide A-4 A-1.4 Implementation Overview A-7 A-1.5 Organization of Guide A-8 Chapter A-2 Performance Measures A-8 A-2.1 Background A-10 A-2.2 Selection Criteria A-12 A-2.3 Recommended Performance Measures A-13 A-2.4 Alternate Performance Measures A-15 Chapter A-3 Data Requirements A-15 A-3.1 Overview of Data Requirements A-18 A-3.2 Minimum Data Required to Implement Performance Measures A-21 A-3.3 Desirable Data Elements for Implementing Performance Measures A-22 A-3.4 Obtaining Data for Developing Models in Support of Performance Measures A-23 A-3.5 Data Availability and Quality A-25 Chapter A-4 Implementation Process A-25 A-4.1 Calculating Initial Change in Condition A-26 A-4.2 Calculating Changes in Performance A-28 A-4.3 Calculating Performance, Service Life, and LCC A-31 Chapter A-5 Summary A-31 A-5.1 Performance Measures Implementation A-31 A-5.2 Other Considerations A-31 References A-35 Appendix A Case Studies A-35 AA-1 Initial Condition Jump and Long-Term Performance Changes A-46 AA-2 Calculating Changes in Performance, Service Life, and LCC A-53 AA-3 Implementation of Performance Measures A-59 AA-4 References

A-3 A-1.1 Background Pavement preservation provides a means for maintaining and improving the functional con- dition of an existing pavement segment through application of a preventative and responsive set of treatments that slow deterioration or correct isolated pavement defects (i.e., it defers costly pavement rehabilitation or reconstruction to a later time). While pavement preservation is not expected to increase structural capacity of the pavement, it can lead to improved performance, longer service lives, and reduced life cycle costs (LCC). Preservation is defined as work that is planned and performed to improve or sustain the condition of the transportation facility in a state of good repair, and which does not add capacity or structural value but does restore the overall condition of the transportation facility (FHWA 2016). A survey of U.S. state and Canadian province highway agencies (Rada et al. 2017) revealed that the most common preservation treatments for asphalt concrete (AC; hereafter referred to as “asphalt”) pavements are thin asphalt overlays (hereafter referred to as “thin overlays”), chip seals, microsurfacing, crack sealing, fog seals, and slurry seals. These treatments are intended to prevent the intrusion of moisture into the pavement structure and/or to protect the asphalt surface layer (Applied Pavement Technology, Inc. 2015). For portland cement concrete (PCC; hereafter referred to as “concrete”) pavements, the most common treatments are diamond grinding, partial and/or full depth patching, joint sealing, dowel bar retrofit (DBR), and spall repairs (Rada et al. 2017). These treatments are intended to address faulting, roughness, localized failures, and/or prevent intrusion of moisture into the pavement structure (Applied Pavement Technology 2015). Pavement preservation technology has been around for decades and it has been demonstrated, albeit often anecdotally, to be an effective approach to extend a pavement’s effective service life, improve safety and service condition, and is cost-efficient. However, there is a need to identify and use performance measures to quantify the effects of preservation on pavement performance, service life, and LCC. A-1.2 Objective The objective of this guide is to facilitate use, by highway agencies, of performance measures for assessing the effects of preservation treatments on pavement performance, service life, and LCC. Such use of performance measures will support the pavement management decision- making process, allowing highway agencies to address critical questions such as: • Does application of preservation treatments contribute to improved pavement performance of segments in a given network and/or of the overall network? C H A P T E R A - 1 Introduction

A-4 Quantifying the Effects of Preservation Treatments on Pavement Performance • Does application of preservation treatments affect the life of a given pavement segment, and how does this contribute to the overall serviceability of a pavement network? Do these treat- ments help maintain the level of service above a specific threshold and delay application of more costly rehabilitation treatments? • Is application of preservation treatments at the given point in time and/or given pavement condition cost effective, and contribute to lowering LCC for the network? Responses to these questions will help highway agencies consider the contributions of pres- ervation treatments to pavement performance and service life, and better assess the role of preservation in maintaining the level of service of a pavement network. Incorporating these measures in pavement and/or asset management systems would help selecting the right treat- ment at the right time, thereby improving the use of resources. The overarching consideration in meeting the objective of this guide is the ability of the per- formance measure to capture the effects of preservation treatments on pavement performance, and hence service life and LCC, recognizing that the application of a preservation treatment does not necessarily improve performance, extend service life, and/or reduce LCC. A-1.3 Using this Guide This guide contains recommended performance measures and the processes for using them to assess the effects of preservation treatments on pavement performance, service life, and LCC. The guide also provides guidance to identify and use alternate pavement performance measures for assessing the effects of using preservation treatments. Performance, service life, and LCC each have unique definitions that are discussed throughout this guide. However, quantifying each requires that the effects of preservation be modeled; modeling these effects is addressed in detail in the guide. Use of performance measures to assess the effect of preservation requires three key steps: (1) selecting the measures, (2) obtaining data to support use of the measures and evaluating the changes in the performance measures (immediate and long-term) resulting from the application of preservation; and (3) implementing the measures in a decision support context. This guide details each of those steps and discusses many of the issues that agencies may encounter when implementing performance measures. This guide provides a step-by-step process for imple- menting these measures and an appendix is included to provide examples of key elements or steps in the process and for modeling. The methods presented in this guide could be enhanced as measurement technologies and data sources continue to improve and allow better modeling and understanding of the effects of preservation on pavements. In addition, agencies may collect additional data (if not already available) to help select a different set of performance measures that commensurate with agency objectives. A-1.4 Implementation Overview The implementation of performance measures by highway agencies requires a systematic approach consisting of the following three steps: (1) selecting performance measures; (2) assess- ing the effects of preservation using the selected measures; and (3) incorporating the measures into highway agency practices. A flowchart summarizing this approach is shown in Figure A1.1; the main steps are described in the following.

Attachment A-5 Figure A-1.1. General approach to implementation of pavement preservation performance measures.

A-6 Quantifying the Effects of Preservation Treatments on Pavement Performance A-1.4.1 Selection of Performance Measures The highway agency selects the pavement preservation performance measures to use from those recommended in this guide, those identified by the agency, or a combination of the two. This guide recommends criteria for identifying and selecting performance measures as well as a process for implementing these measures. A-1.4.2 Assessing Effects of Preservation Treatments on Pavement Performance This step establishes the models for defining the effects of a given preservation treatment on pavement performance using the selected performance measures. Figure A-1.2 illustrates the effect of preservation treatments on pavement condition using a performance measure that is scaled between 0 (worst condition) and 100 (best condition). The initial effect, which is rep- resented in Figure A-1.2 by the term CJ, can be an improvement (as illustrated in the figure), a degradation in pavement condition, or no condition change depending on a number of factors (e.g., treatment type and condition of pavement when treatment was applied). The second effect shown in Figure A-1.2 is the relative change in deterioration between the treated pavement and a control pavement. At a given point in time (tt), the relative difference in condition between the treatment and control section (Ct′ − Ct) gives an indication of the change in performance resulting from preservation. This guide addresses the minimum and desired data requirements needed to develop the pavement preservation effect models, as well as data priorities and handling data issues (quality and others). Case studies are also presented to illustrate data considerations and the modeling of pavement preservation effects. Following the steps provided in the guide will help achieve a clear understanding of what the data are indicating and develop appropriate models for capturing the effects of preservation treatments on pavement performance. A-1.4.3 Incorporating Performance Measures into Highway Agency Processes The next step after establishing models describing the initial and long-term effects entails integrating those models into the agency’s tools (e.g., pavement and/or asset management sys- tems). This integration will vary from agency to agency, but as a minimum it will enable agencies to compare the effects of applying preservation to those of other treatment options such as do nothing or rehabilitation, in terms of pavement performance, service life, and LCC. 0 20 40 60 80 100 0 3 6 9 12 15 18 21 Pa ve m en t C on di ti on Percentage, year Original Deterioration Deterioration Post Preservation tt C't - Ct Treatment Performance CJ Figure A-1.2. Illustration of pavement preservation effect on pavement condition.

Attachment A-7 A-1.5 Organization of Guide This guide is organized into five chapters. This chapter provides background information, guide objective, overview of guide implementation process, and organization of the guide. Chapter A-2 use of performance measures to assess the effects of preservation treatments and support of pavement management decision-making processes, and includes criteria for select- ing performance measures, recommended performance measures, and steps for implementing alternate performance measures. Chapter A-3 summarizes the data requirements for implemen- tation of the performance measures, and provides guidance on obtaining data, data priorities, and resolving data issues (e.g., quality and completeness). Chapter A-4 provides a step-by-step procedure for using performance measures to assess the effects of preservation treatments on pavement performance, service life, and LCC, and to integrate those effects into the agencies’ existing processes to support pavement decision-making. Chapter A-5 summarizes the guide and discusses means for improving those models. The references cited throughout the guide also are presented. The guide includes an appendix that contains example case studies to illustrate how to deal with data issues and define the effects of preservation treatments on the initial and long-term condition of pavements, and how to implement performance measures within highway agencies’ processes.

A-8 This chapter discusses the selection of performance measures, presents a set of recommended measures, and discusses the use of agency specific measures. Figure A-2.1 presents a flow chart that illustrates the process for selecting a set of performance measures that can be used to develop models to assess the effectiveness of their preservation practices. A set of data requirements is also outlined along with recommendations on where to obtain these data. A-2.1 Background Much research has been conducted in recent years in an effort to clearly define performance measures, their attributes, and their uses. For example, NCHRP Report 551 (Cambridge System- atics 2006) defines performance measures as indicators that can be used to measure progress towards a goal and NCHRP Report 708 (Zietsman et al. 2011) defines performance measures as quantifiable criteria used to measure progress towards goals. Also, performance measures have also been defined as indicators of system effectiveness and efficiency, or, more generally, as tools to support performance management (Amekudzi and Meyer 2011). Specific to pavements, per- formance measures have been defined as ways to consistently assess their functional condition (Simpson et al. 2013). In addition to defining performance measures, literature defined the attributes that can be used to evaluate performance measures. For example, NCHRP Report 666 (Cambridge System- atics et al. 2010) identifies four criteria, Grant et al. (2013) identify seven criteria and the U.S. Army Corps of Engineers (2015) identifies 12 criteria against which a performance measure could be evaluated. Many of the criteria in these references are similar, and their differences relate to how the criteria are combined (e.g., a single criteria in one reference may be expressed as multiple criteria in another reference). The following are the main criteria that were considered for selecting recommended performance measures: • Is the measure feasible? Can the measure be obtained or objectively calculated from data col- lected by agencies at a reasonable level of accuracy and repeatability? Can the measure be used for a broad set of pavement types? • Is the measure understandable? Many measures can potentially communicate abstruse aspects of pavements, which is not desirable. In order to facilitate widespread implementation, con- sideration was given to whether the recommended measures can be used across a wide group of practitioners. • Is the measure useful in decision-making? In order for the measure to be useful in decision- making, it must be, or can be, directly linked to agency practices, and reflects the effect of preservation. For example, structural measures may be linked to agency decisions, but do not capture the effect of preservation (or functional performance). C H A P T E R A - 2 Performance Measures

Attachment A-9 Thus, pavement performance measures (as defined in this guide) are metrics that quantify the degree of achieving specific goals pertaining to performance. Although these measures are often used to assess the current condition of the pavement, they provide a basis for the deci- sions needed to achieve specific goals for the pavement asset. For a pavement performance measure to provide the basis for decision-making, it must have the ability to be forecasted over time given that many agencies do not make treatment decisions based on single year evaluations. Therefore, for selecting the recommended performance measures, consideration was given to those indicators for which adequate data could be obtained to develop models as a function of time. For purposes of this guide, pavement performance measures have been grouped into the fol- lowing four categories: • Individual distress/condition measures. Examples include the International Roughness Index (IRI), cracking, rutting, and faulting. • Composite pavement condition indexes/measures. Examples include the pavement con- dition index (PCI) developed by the U.S. Army Corps of Engineers (ASTM International 2009), the composite score (CS) developed by Texas Department of Transportation (DOT) (Texas DOT 2009), and the pavement condition rating (PCR) developed by Florida DOT (Florida DOT 2015). • Cost measures. Examples include the historical discounted equivalent uniform annual cost (EUAC) and equivalent single axle load (ESAL) efficiency used by Washington DOT (Luhr et al. 2010; Rydholm and Luhr 2014), the Vehicle operating cost index proposed for use by the New Zealand Ministry of Transport (Costello et al. 2013), and the asset sustainability index (ASI) proposed by Proctor et al. (2012). • Other measures. Examples include the pavement remaining service interval proposed by the FHWA (Rada et al. 2016), and pavement surface friction measurements (e.g., skid number). Many other performance measures exist today and many of those were documented (Rada et al. 2017). A set of performance measures recommended for use by highway agencies is pre- sented later in this chapter. However, highway agencies can select alternate measures or use a combination of the recommended and alternate measures. The criteria for identifying and deciding on those performance measures are presented. Figure A-2.1. Contents of Chapter A-2: Performance Measures.

A-10 Quantifying the Effects of Preservation Treatments on Pavement Performance A-2.2 Selection Criteria The selection of performance measures should be directly linked to the agency’s pavement management objectives. Some of these objectives will be common among many agencies, such as those related to maintaining an acceptable level of ride quality or maintaining an acceptable level of service for the lowest LCC. On the other hand, many agencies have objectives that are not necessarily widespread, such as certain safety or environmental objectives. With this in mind, the following questions can be used to drive the selection of performance measures: • What are the objectives of pavement management within the agency? The purpose of this question is to assess whether a given performance measure can be linked to the overall pavement management objectives for the agency. Inherent in the ques- tion is that, if treatment effects are related to one or more overall objectives of pavement management, then a measure should be selected to capture these effects. For example, if, along with ride quality, maintaining pavement friction above a certain threshold value is an objec- tive of pavement management decisions, then pavement friction (or some proxy of pavement friction) should be chosen along with IRI as a performance measure. • Does the set of performance measures capture the aspects of pavement performance that are considered important by the agency? Based on the complex nature of pavements, it is not expected that pavement performance can be defined by a single performance measure. A group of measures should be selected to adequately encompass the objectives of the agency. Important to selecting performance measures that capture the effect of preservation is under- standing the extent to which preservation contributes to pavement performance. Selection of the performance measures will be iterative and agency specific. Each agency will need to assess how preservation influences the performance of pavements, and whether the set of selected performance measures accurately reflect this influence. This guide proposes performance measures and illustrates how pavement preservation affects these measures. The process for determining whether or not a given performance measure is capable of capturing the initial and long-term effects of preservation treatments on pavement performance is detailed in Chapter A-3. This section addresses the recommended criteria for identifying and selecting pavement performance measures that merit assessment under the pro- cess described in Chapter A-3. For purposes of this guide, the criteria have been separated into three categories: general considerations, measure characteristics, and measure implementation. General Considerations 1. Number of measures – agencies should select the appropriate number of measures that will adequately reflect their objectives. In general, the fewer the performance measures the better because they will eliminate measures that duplicate the same outcomes, ease implementation and maintenance of measures within agency’s pavement and/or asset management tools, and reduce data demands. At the same time, however, the selected performance measures must address the full range of pavement types and preservation treatments used. 2. Data availability – performance measures must encompass necessary data elements to sup- port development (as well as periodic updates) of the performance prediction models. With- out IRI and faulting data, for example, the effect that diamond grinding will likely have on the performance of jointed concrete pavements cannot be demonstrated. Accordingly, the required data should provide measurements of: – Pavement condition prior to and after application of the treatment, as close as possible to the date of application of the treatment, for each combination of pavement type and preservation treatment.

Attachment A-11 – Long-term (preferably more than five years) pavement condition data for each combina- tion of pavement type and preservation treatment. – Reference or control (without application of preservation treatment) long-term (prefer- ably more than five years) pavement condition data for each pavement type to enable thorough comparisons and more accurate assessments of the long-term effects of preser- vation treatments on pavement performance. The availability of the required data within the highway agencies is vitally important in order to account for local conditions, including climate and traffic characteristics (agency personnel generally have a great deal of familiarity and experience with those data). However, highway agencies may choose to use alternate data sources (discussed further in Chapter A-3 and the case studies are presented in Appendix A). Measure Characteristics 3. Data quality – in the context of this guide, data quality is defined by how well it informs the steps required to implement the performance measures. Several key criteria that can be used to define data quality are detailed in Chapter A-3. Two important criteria for data quality are accuracy and precision. Accuracy is the condition or quality of being “true.” When applied to pavement condition measurements, true measurement refers to the actual condition of the pavement measured without errors. Precision, on the other hand, is the condition or quality of being able to repeatedly measure the same quantity with little random error; it is used in this guide to address the repeatability of the data measurements. Data quality is a significant factor in the outcomes of performance measures model devel- opment. Recognizing that the models will invariably contain errors, accounting for these errors is important for successful implementation of the selected performance measures. Chapter A-3 addresses the topic of data quality in detail, and four case studies are provided in Appendix A to illustrate the effect of data quality and the number of observations on the confidence of results. 4. Risk impacts – the ability of highway agencies to capture risk associated with application of preservation treatments is a key consideration in selecting performance measures. The World Road Association defines risk as the combination of the probability of a hazard and its consequences (World Road Association 2010). In the context of this guide, risk is associ- ated with the probability that a given treatment will have a worse than expected performance and therefore increased cost. If, for example, a treatment is applied to extend pavement life, but the service life extension was shorter than anticipated, there are cost and other effects associated with this under-performance. The performance prediction models also have errors associated with them that must be accounted for in the implementation of each performance measure. The smaller the model errors, the lower the degree of uncertainty and hence the lower the risks associated with the model predictions; this supports the need for high quality data for the selected performance measures. 5. Quantifiable – if performance measures are to accurately reflect the performance of the pres- ervation treatments, they must be able to generate its effect on the performance of the pave- ment. As such, it is important that the measures be quantifiable and, more specifically, be capable of determining initial and long-term effects in terms of performance metrics. Measure Implementation 6. Ease of use and implementation – this consideration must be carefully balanced against the five selection criterions. If there are alternate performance measure with better data availability and/or quality, serious consideration should be given to those measures. Alter-

A-12 Quantifying the Effects of Preservation Treatments on Pavement Performance natively, the agency could decide to use the performance measures they are familiar and comfortable with, if they are supported by available data. A-2.3 Recommended Performance Measures In the development of this guide, a number of performance measures were evaluated in terms of their ability to capture the effects of preservation treatments on pavement performance (Rada et al. 2017). Based on this evaluation, the following performance measures are recommended for consideration. • Ride quality (IRI) for both asphalt and concrete pavements. • Cracking for both asphalt and concrete pavements. • Rutting for asphalt pavements. • Faulting for concrete pavements. These measures are indicators of pavement condition at a given point in time, and their use as performance measures is derived from measuring their values over time to model the expected behavior of the pavement. Details on regarding how to model each of these models as a function of time are provided in this guide. It is recognized that there are many types of cracking that an agency may measure (e.g., fatigue cracking, transverse cracking, etc.), and that, even for the same cracking type, agencies may have varying standards as to what should be included in that type. For example, different results were obtained for the immediate change in longitudinal cracking following a chip seal application provided by some agencies, because some agencies included construction joints as part of longi- tudinal cracking. This guide does not recommend a specific cracking type to use; it only defines data quality measures that should be considered when selecting the cracking type. Similarly, both rutting and faulting may be measured using different techniques (e.g., point measurements such as three or five sensors or for rutting). This guide does not recommend a specific measurement methodology for rutting and faulting, but specifies a level of quality at which the data should be collected. These performance measures are recommended because they are used by most state highway agencies in their day-to-day pavement related activities, and the agencies are familiar with them. In addition, state highway agencies are required to submit these same measures as part of their Highway Performance Monitoring System (HPMS) data submittals (FHWA 2014). The final rules for assessing pavement condition, first outlined under the MAP-21 legislation, require the same measures for monitoring, tracking, and reporting on the condition of the state’s Interstate and National Highway Systems (Federal Register 2017). Also, these measures include many (but not all) of the distresses that typically trigger preservation treatments. Table A-2.1 summarizes how the most common preservation treatments affect the recom- mended performance measures for asphalt and concrete pavements, using data from the state highway agency and long-term pavement performance (LTPP) program and information found in the literature (Rada et al. 2018). This table also shows whether the effect of the treatment is reflected in the initial pavement condition jump and/or long-term pavement condition, noting that cases where no effect was demonstrated do not necessarily rule out effect. More or better quality data may show a different trend for some cases. In the absence of data and/or when the available data are of insufficient or questionable qual- ity, the use of local highway agency data is preferred, but other sources of data (e.g., LTPP data- base) can be used. This topic is discussed further in Chapter A-3 and the case studies provided in Appendix A illustrate key issues and considerations.

Attachment A-13 A-2.4 Alternate Performance Measures The recommended performance measures are considered satisfactory for quantitatively incor- porating the effects of preservation treatments into the pavement management decision-making process for many cases. Nonetheless, highway agencies may choose to use other measures (e.g., friction) because of availability and/or quality of data to support these measures, preference for these measures (e.g., composite pavement condition indexes), ease of incorporating these alter- nate measures within the agency’s pavement and/or asset management tools, or other reasons. This guide provides the criteria for identifying and selecting performance measures and the means to implement alternate performance measures. A flowchart to help guide the assessment of candidate measures against the criteria recom- mended by this guide is provided in Figure A-2.2. The six questions listed in this figure corre- spond to the identified selection criterion; suggestions are provided in case the specific criterion is not met. The measure was assumed to be able to assess the effects of preservation treatments on pavement performance. More detailed guidance concerning the data availability, data qual- ity, model development, and model implementation elements is provided in Chapter A-3. In addition, the case studies provided in Appendix A illustrate key issues. Examples of alternate performance measures that highway agencies may consider include: • Surface friction measurements for both asphalt and concrete pavements. A number of agen- cies collect and use these data (Speir et al. 2009; de Leon Izeppi et al. 2016). Several preserva- tion treatments (e.g., diamond grinding of concrete pavements, chip seals, microsurfacing, and slurry seals) have a significant effect on pavement surface friction; skid resistance is the only measure that captures these effects. • Composite indexes, such as the critical condition index (CCI) (McGhee 2002) used by Virginia DOT (VDOT). Other examples of composite indexes include the functional cracking index (FCI) and structural cracking index (SCI) used by Maryland State Highway Administration (SHA) (Rada et al. 2016), and the grinding index, faulting index, and dowel-bar retrofit (DBR) index (Jackson 2008) used by Washington DOT (WSDOT). Although specific composite pavement index measures are not recommended in this guide, these composite measures Pavement Type and Preservation Treatment Effect on Roughness (IRI) Effect on Cracking (at least one cracking Type) Effect on Rutting Effect on Faulting Initial Condition Change Long- Term Condition Change Initial Condition Change Long-Term Condition Change Initial Condition Change Long-Term Condition Change Initial Condition Change Long-Term Condition Change Asphalt Pavements Thin Overlay Yes Yes Yes Yes Yes Yes Chip Seal None Yes Yes Yes None None Microsurfacing None - Yes - - - Concrete Pavements Diamond Grinding – No DBR Yes - None - Yes - Diamond Grinding – With DBR Yes - None - Yes - - – Not assessed because of unavailability of data None No effect demonstrated Yes – Effect demonstrated using data from state highway agencies, LTPP, and/or literature Shaded cells Not applicable for pavement type and distress combination – – Table A-2.1. Preservation treatments and their effects on performance measures.

A-14 Quantifying the Effects of Preservation Treatments on Pavement Performance have shown to capture the effects of preservation treatments on pavement condition (Rada et al. 2018) After selecting performance measures, the next step in the performance measure implemen- tation process addresses the data requirements and development of performance models for both the immediate and long-term performance following treatment application. This step is discussed in detail in Chapter A-3 and illustrated with case studies provided in Appendix A. 1. Does alternate performance measure capture performance indicators not addressed by these measures being considered for replacement (e.g., skid resistance may be only measure that captures friction of pavement surface)? No Do not consider alternate performance measure unless there are other considerations (e.g., agency plans on transitioning to another measure that captures same performance indicator). Yes 2. Are data available to support development of performance models for alternate measure? No Yes Do not consider performance measure; take steps to address quality issues or identify alternate data source(s). 3. Is data quality adequate to support development of performance models for the alternate measure? No Yes Do not consider performance measure; take steps to generate needed data, or identify alternate data sources. 4. Can risk impacts be assessed using measure models? No Do not consider performance measure or identify alternate data source(s) to estimate model errors. Yes 5. Are effects of preservation treatments on pavement performance quantified by the alternate measure? No Do not consider performance measure. Yes 6. Is agency familiar with measure and is it easy to implement within agency's pavement management system (PMS)? No Consider other alternate performance measure. Yes Performance measure may be used by agency Figure A-2.2. Approach for assessing alternate performance measures.

A-15 The previous chapter detailed the selection of performance measures, presented a set of rec- ommended measures, and discussed several agency specific performance measures. The next step is to gather data to support the implementing of the measures (Figure A-3.1). This chapter provides guidance on the data required to implement the pavement preservation performance measures. First, a set of the minimum and desirable data required is defined. Then, a set of rec- ommendations on how an agency can obtain these data are outlined. Finally, a set of attributes for assessing the quality of data is presented and discussed. Information presented in this chapter is supplemented by the case studies presented in Appendix A. An important concept that is highlighted throughout this guide is that conclusions should not be based on the performance evaluation of a single or few pavement segments. An approach to esti- mating the changes in performance caused by preservation is presented in Section A-4.2. An exam- ple that shows practically no difference in performance existed between a control pavement and a preserved pavement when evaluating an individual pavement segment but when looking at many pavement segments, difference was observed. The next section outlines a simplified approach for determining the number of pavement segments needed to draw appropriate conclusions. A-3.1 Overview of Data Requirements Implementing the performance measures within an agency requires that data be available to support the development of models to assess the effectiveness of preservation. The data includes: • Construction history data that defines the type, timing, and location of preservation treat- ments on the pavement network. • Pavement condition measurements immediately prior to and following the application of preservation treatments. Measurements over several years following the preservation applica- tion are required in order to assess the effect of preservation on pavement performance. • Additional information that can be used to describe the sources of change in the performance measures. This information may include data such as pavement thickness, traffic informa- tion, or other variables that affect pavement performance. These data will be used to develop condition deterioration models for the selected performance measures. The main objective of implementing the performance measures is to assess the effective- ness of preservation. Therefore the data must capture changes in pavement condition resulting from preservation, and can be used to calculate the performance measures that are selected. Figure A-3.2 illustrates pavement condition measurements for a single pavement over 14 years, and relays several important concepts related to data requirements. Figure A-3.2 illustrates a performance measure that decreases over time with worsening pavement condition (the same concept applies to a measure that increases with worsening pavement condition, such as IRI). C H A P T E R A - 3 Data Requirements

A-16 Quantifying the Effects of Preservation Treatments on Pavement Performance The data in Figure A-3.2 show that pavement condition data are expected to have variability that would influence the conclusions about the effectiveness of preservation if evaluated on a segment-by-segment basis. As shown the calculated condition jump using data from a single segment is less than the condition jump calculated using data from several segments. This result demonstrates the importance of using data from several segments in developing the models. There are many ways to determine the minimum amount of data required to develop reli- able conclusions, and this amount is dependent on the variability/error in the data. However, a simple approach for determining the minimum number of segments needed is to define a toler- able margin of error, a desired confidence, and calculate the minimum number of preservation segments using Equation 3-1 (Gelman and Hill 2017). (3-1)2 2 2 n z c≥ σ δ    α where: n = number of segments required to develop statistically significant conclusions; zα = z-statistic from a standard normal distribution for a given confidence level (1 − α); σ = standard deviation of the measurement errors; 30 40 50 60 70 80 90 100 0 5 10 15 20 25 Pe rf or m an ce M ea su re V al ue Pavement age, year Pavement Condition Measurements Expected Pavement Deterioration Deterioration After Preservation Calculated condition jump from this site Condition jump calculated using data from several sites Differences in deterioration following preservation Figure A-3.2. Use of condition data to estimate condition jump and performance following application of preservation for a single segment. Figure A-3.1. Gathering data.

Attachment A-17 δ = tolerable margin of error; and c = number of categories in which the data are grouped (e.g., discretizing the pavements into good, fair, or poor condition, each with either high or low traffic results in c = 6). An example to estimate the number of segments required to draw with supported conclusions follows. Example of Calculating Number of Segments Required to Develop with Supported Conclusions Estimating the minimum number of sites required to estimate the condition jump using the IRI as a performance measure and given the following criteria: • Tolerable error(s) of ± 5 in/mile (range the agency is willing to accept for the calculation). • Standard deviation of error(s) 5 in./mile (assumed to be normally distributed). • Ninety-five percent confidence in the conclusions. First, find the z value associated with the 95 percent confidence interval. In this case, z = 1.96 (Gelman and Hill 2017). The condition jump is defined as the difference in two IRI measurements and, therefore, the standard deviation of the difference between two measurements with normally distributed errors is calculated as: σ = σ + σ12 22diff measurement measurement Assuming the measurements have the same standard deviation of errors, the standard deviation can be calculated as:  σ = σ + σ = σ = =2 2 5 7.072 2diff m m m Thus, assuming one category, the minimum number of sites required to draw statistically significant conclusions about the effectiveness of preservation is: ≥ σ δ     ≥    ≥α 1.96 7.07 5 82 2 2 2 2 2 n z c Thus to assess the effectiveness of preservation when it is placed on pavements in Good, Fair and Poor initial conditions, at least eight sites in each condition (i.e., minimum of 24 sites) are required. Furthermore, to discretize the pavements based on thickness (thick and thin pavements), 24 sites for each (i.e., a minimum of 48 sites) are required. Using this simplified approach, it can be seen that the minimum number of sites is related to the extent of the desired detail. If the effect of preservation is expected to be the same on all pavements regardless of the initial condition of the pavement, then only eight pavement segments are needed to draw. If the effect of preservation varies based on other factors (e.g., initial condition), a larger number of sites would be required.

A-18 Quantifying the Effects of Preservation Treatments on Pavement Performance The tolerable margin of error is defined by the agency and should be selected based on desired precision and resolution for the models. However, in absence of a known precision or resolu- tion, an estimate for the tolerable margin of error is the measurement error. There are other sources of error than measurement errors (e.g., variability between contractor’s workmanship). However, using the measurement error as an estimate of the overall error will provide a good estimate when calculating the number of required segments using Equation 3-1. The data in Figure A-3.2 all illustrate the need to know the timing of the condition mea- surements relative to the timing of application of preservation. For example, if a preservation treatment is applied in the year 2014, it is important to know whether the 2014 condition mea- surements occur before or after the application of preservation. Given the magnitude of the errors/variability in condition measurement, and that preservation may only slightly affect cer- tain performance measures, it may not always be obvious from condition data whether the condition measurement occurred before or after preservation condition data. Finally, when comparing performance over time for several pavement segments, differences in performance can be expected between the different segments. A potentially significant amount of these differences may be due to differences in the characteristics of the pavement segments (e.g., thickness or traffic loading) but some of these differences will be due to measurement errors. Therefore, it is desirable to collect as much information about each pavement segment as possible as described (see Section 1). The data requirements, along with recommendations about where to obtain these data, is discussed in this chapter. A-3.2 Minimum Data Required to Implement Performance Measures The minimum required information needed from the data are the immediate change in con- dition and change in performance over time following preservation. Besides defining the per- formance of the preserved pavement, this information can be used to calculate the service life extension and cost or other measures of effectiveness associated with the treatment. Generally it is preferred, but not required, to obtain the data for the condition jump and change in perfor- mance from the same source. For example, the data required for calculating the condition jump may be obtained from the pavement management system, but data from other sources (e.g., the LTP) may be used to estimate the changes in performance. Although, the agency may implement the measures using these two data sources, and the models should be revised later using only data from the pavement management system. A fundamental assumption in the development of performance models is that the pavements described by the model have similar characteristics. The models are based on using a sample of pavements, assumed to be representative of the entire population of pavements in the network. For example, models developed based on a sample comprising only interstate data will be appli- cable only to interstate pavements. Therefore, it is important to ensure that data are collected for pavement segments that are representative of the pavement network. This is why, at a minimum, the functional class of the pavement is a required data element needed to categorize pavements. However, other descriptive elements for the segments should be collected and included in the implementation discussed (see Section A-3.3). A-3.2.1 Data Required for Calculating Condition Jump The condition jump is defined as the change in condition that can be measured immediately following the application of preservation. Although the condition jump is calculated using a simple formula (i.e., the condition following preservation minus the condition prior to preservation),

Attachment A-19 complications may arise if the proper data are not available. The basic steps needed to calculate the condition jump and associated data are the following: 1. Group pavements into families having similar characteristics (to develop models for each grouping). At a minimum the grouping should be based on pavement type and pavement functional class. This requires gathering the following data: – Pavement type (e.g., rigid, flexible, or composite) – Functional class of the pavement segment 2. Identify location and timing of preservation treatments application. This requires gathering the following data: – Beginning and ending mileposts of the treated segment – Type of preservation treatment (e.g., microsurfacing, etc.) – Date preservation application – Additional work conducted on the segment (e.g., rut leveling) 3. Obtain condition measurements required to calculate performance measures. This includes: – Beginning and ending mileposts for the pavement segments for each year of condition measurement – Date of condition measurement – All pavement condition metrics needed to calculate the performance measures 4. Compare condition measurement just prior to application of preservation to that following preservation (on same pavement segment). – A complication would occur if different beginning and ending mileposts for the pavement segments stored in the pavement management system change over time. 5. Calculate change in condition resulting from application of preservation (see Chapter 4). An important consideration is the pavement segment length, as illustrated by the following considerations: • Potential errors related to defining beginning and ending points for condition measurement are introduced if the segment length is too short • As segment lengths increase, errors related to averaging condition measurements (e.g., crack- ing, rutting, etc.) and other properties (e.g., subgrade resilient modulus, etc.) are likely to increase. Therefore, segment lengths should be selected to minimize these errors. • If the defined beginning and end locations of pavement segments changed over time, select the pavement section having consistent characteristics between the changing beginning and ending points. Identifying where and when the preservation took place can generally be obtained from a construction history or contract history database when using agency data. One complication may arise if the information in the construction history database cannot be directly matched to the condition data (i.e., project level data stored in different databases are different). This issue is discussed in this chapter and in Appendix A. Another common issue that occurs is the change in the beginning and ending mileposts of condition measurements over time. For example, new beginning and ending mileposts are sometimes defined following the application of preservation treatments (or rehabilitation or reconstruction). As a result, different pavement segments are used for comparing the condition before and after preservation, that may invalidate the results of determining the condition jump associated with preservation. The following options address this issue: • Limiting the segments used in the analysis to those segments where the post-treatment seg- ment mileposts reasonably overlap with the pre-treatment segment mileposts by establishing a specified percentage of overlap recognizing that a relatively low percentage would introduce significant errors into the models.

A-20 Quantifying the Effects of Preservation Treatments on Pavement Performance • Reviewing the pavement condition database to determine if data stored for smaller increments is available. For example, many agencies collect and input data for small increments (e.g., 0.1 mile) into their pavement management system, and then aggregate the data over longer segments for pavement management purposes. If data for the smaller increments is available, the condition can be assessed using only the segments that received a preservation treatment (not the longer segments that may include segments that did not receive preservation). A-3.2.2 Data Required for Calculating Changes in Performance Estimating the change in performance following a preservation treatment requires compar- ing the condition of the pavement following the application of the preservation treatment to that of an equivalent control segment (i.e., a similar pavement with no preservation applied). The calculation of the rate of deterioration of pavement segments requires more data, much of which can be found in the same sources as those for this data required to calculate the immedi- ate change in condition. In addition, pavement deterioration is not only a function of time, but is also related to many potential explanatory variables; it is desirable that these variables be collected and the effec- tiveness of preservation assessed using these variables. The minimum data elements required for assessing the effectiveness of preservation using performance measures are discussed in Section A-3.3. The data required for assessing changes in performance resulting from preservation are those required for each of the basic steps for estimating these changes. 1. Group pavements into families. Outlined in Section A-3.2.1; having similar characteristics at a minimum based on pavement type and pavement functional class and gathering relevant data. 2. Identify location and timing of preservation treatments and gather relevant data. 3. Obtain condition measurements required to calculate performance measures. 4. Compare condition measurement on same pavement segment just prior to application of preservation to the several years of condition measurement following preservation. Because of the variability in pavement condition data and the relatively slow change in most perfor- mance measures over time, segments with less than 5 years of condition data should not be included in the analysis. 5. Identify control group and collect the condition data group required to calculate performance measures. The control group for a pavement family should not be selected from pavement segments that have no required work, but from the following options: – Pavement segments identified in the past as candidates for preservation, but were not included into the work program (e.g., pavement segments that have been deferred and allowed to deteriorate further). However, the control segments should come from the same population as the treatment segments. – The condition of preserved segment can be compared to that of similar segments that have not been treated. However, data from pavement segments that do not require treatment should not be used as controls. – Control segments should be selected from the same source as that for the preservation data. For example, if data from pavement management system is used to develop the models for preservation segments, control segment data should be obtained from the same pavement management system measurement as the preservation segments and the control segments are drawn from populations with similar characteristics. 6. Calculate change in performance resulting from application of preservation (see Sec- tion A-4.2).

Attachment A-21 When developing performance models using multiple pavement segments, it’s generally assumed that this group of pavement segments is representative of the population of pave- ment (i.e., pavements are of the same type, in the same functional class, and belong to the same population). To minimize the effects of differences in the pavement, additional variables that influence performance over time (e.g., thickness and traffic) could be collected. By adding these variables to the models, several sources of error can be accounted for, and the effect of assuming that all segments are sampled from the same population is reduced. A-3.3 Desirable Data Elements for Implementing Performance Measures This section outlines several desirable data elements that inform the initial condition jump and long-term performance models. However, agencies may collect other informative data elements (e.g., measures relating to construction and material quality). Several case studies of developing the models to describe the change in condition and change in performance following preservation are presented in Appendix A. One finding from devel- oping these case studies was that the immediate change in condition resulting from the appli- cation of preservation treatments was not generally related to any variables other than the condition prior to preservation and the type of preservation treatment. Therefore, the desirable elements for assessing the immediate change in condition are the same as the required elements. However, when calculating the changes in performance following application of preservation, several variables were found to be significant for informing the deterioration models (e.g., cumulative traffic loading and pavement thickness) and implementation of the performance measures. The development of the models can be completed without these data but their use is encouraged. The data elements identified as relevant for this report are: • Pavement type. • Condition prior to the application of preservation. • Immediate change in condition following the application of preservation. • Structural capacity of the pavement (e.g., Structural Number). • Traffic loading. This may be expressed in terms of average annual daily traffic (AADT), aver- age annual truck traffic or ESALs. • Thickness of the pavement surface and base layers (may be important in some cases; less important than structural capacity for assessing roughness and rutting performance). • Average annual precipitation (important to assessing rutting and roughness growth). • Average annual temperature (important for roughing and cracking performance models). • Number of freeze thaw cycles (important in several models, including the rutting perfor- mance model). An example that illustrates the effect of including several additional variables in model devel- opment is provided in Appendix A. However, each variable should be carefully evaluated in a systematic process (e.g., stepwise linear regression) in order to identify those variables that should be included in the model. The data elements required for the development of models that support implementation of the performance measures have been described; the minimum required data and the inclusion of additional data elements were outlined. The next section describes the sources of data that can be used to develop the models associated with the performance measures.

A-22 Quantifying the Effects of Preservation Treatments on Pavement Performance A-3.4 Obtaining Data for Developing Models in Support of Performance Measures Many sources exist for gathering the data required for developing models in support of the performance measures. However, it is important to recognize that certain data should be priori- tized above other data, and the data used to develop the models are assumed to be representative of the pavement network. Thus, data collected and stored by the agency should be given the highest priority. Several potential sources of data were identified, evaluated, and ranked in terms of their potential use for implementing the measures. Figure A-3.3 shows these data sources in terms of their relative priority for use in implementing the measures. In general, the use of agency specific data is prioritized above the use of other data because it reflects specific climatic conditions, construction practices and segment characteristics (e.g., subgrade soil types), each of which has considerable influence on pavement performance. However, in the absence of adequate data to support implementation of the performance measures, other sources of data are available, as is shown in Figure A-3.3. The priorities shown in Figure A-3.3 should be used together with an assessment of data quality to determine the appropriate source of data to support implementa- tion. If it is determined that Priority 1 data are not of adequate quality, Priority 2 data should be considered, and so forth (see Figure A-3.4). Data availability and quality are discussed in the next section. Priority 1 Agency Database Priority 4 Data from Neighboring Agencies Priority 5 Data from Similar Treatments It is highly desirable that the data come from the highway agency that is implementing the pavement performance measures because local conditions, climate and traffic characteristics can be accounted for. These data should come from the agency’s PMS and other databases. Priority 3 LTPP Database Priority 2 State/National Level Studies Use of data from other national/state level studies reported in the literature, and clearly documents how data were collected, quality of data, etc. The use of these data is acceptable for initial implementation until data for calibrating initial condition jump and long term performance models become available. Although at the national level, the LTPP database has high quality data that can be used to implement a large number of pavement performance measures. Use of data from other highway agencies with similar conditions (construction practices, traffic, climate, etc.). Use of these data should be considered as preliminary, until data specific to the pavement network become available. Use of data from similar treatments. For example, chip seal data could be used for microsurfacing. These data should be considered as preliminary until data specific to the preservation treatments become available. Figure A-3.3. Priorities for gathering data for implementing performance measures.

Attachment A-23 A-3.5 Data Availability and Quality A primary consideration in selecting data source to support implementation of the perfor- mance measures is the availability of the data. Ideally, the agency has available complete sets of data spanning many years of condition metrics that can be used to calculate the performance measures for the desired preservation treatments. However, this may not be the case for many agencies, and the agencies may need to rely on other data sources (see Figures A-3.3 and A-3.4). Data quality should also be a primary consideration in selecting the source of data to support implementation of the measures. Bergdahl et al. (2007) developed a comprehensive guide that can be used for assessing data quality. However, the data must be capable of identifying critical features of pavement performance and supporting the assessment of preservation effectiveness. This section of the guide outlines some important attributes related to data quality, and it dis- cusses how agencies can assess these attributes. The attributes are accuracy and reliability, preci- sion, source completeness, temporal completeness, and relevance. Appendix A also includes case studies that describe how to address data quality issues. Many attributes can be used to define data quality (Bergdahl et al. 2007 and Pipino et al. 2002). In this section, a few important elements pertaining to data assessment are discussed. Assessing the quality of data depends on the specific performance measures chosen because condition measurements can influence different performance measures in distinct ways. However, a few attributes need to be considered regardless of the performance measure to help determine if the quality of data from a specific source is good enough to use. These attributes include: • Accuracy and reliability. Pertains to how the measured condition accurately and reliably reflects the actual condition of the pavement. For example, if safety is a concern, and a preservation N o Yes Use agency data. Does the LTPP contain data for many sections that: • Are specific to the preservation treatments employed by the agency, • Have condition measurements that can inform the performance measures being implemented by the agency, and • Are representative of the climatic and site conditions of the agency? N o Yes Use LTPP data. • Models resulting from the analysis of pavement condition data • Limitations on the use of the results? N o N o Yes Use data from other agencies Priority 1 Data Priority 3 Data Priority 2 Data Priority 4 Data Do the agency’s various asset management systems contain the required data that are specific to the preservation treatments used by the agency, and is this data of adequate quality? Are there studies that have assessed the preservation treatments specific to the agency, and have clearly documented: Identify preservation treatments that are similar to those being considered, and proceed with identifying data (beginning again with priority 1 data) Are there data available from other agencies that meet the criteria for priority 1 data? Yes Use documented study data and models. Figure A-3.4. Flowchart for selecting the source of data.

A-24 Quantifying the Effects of Preservation Treatments on Pavement Performance treatment is adopted to improve safety (e.g., by increasing friction or reducing splash and spray), then the condition metric should accurately reflect changes in safety. • Precision. Pertains to measurement errors apparent in the data such that trends in the condi- tion as a function of time are masked by variability. The minimum level of precision required depends on the performance measures being modeled and the number of segments in the analysis (e.g., increasing the number of segments can compensate for poor precision). An estimate of the required precision of the data can be obtained by using the same approach for estimating the required number of pavement segments described in Section A-3.1 by input- ting the number of available segments for each pavement family and estimating the minimum required standard deviation of measurements (a proxy for precision). • Source Completeness. Pertains to the inclusion of all required data elements (and required metadata) to support implementation of the measures and is an essential part of the data source selection process. Metadata is the set of data that provides information about the remainder of the data (e.g., for example, the types of distresses and units of measurement). • Temporal Completeness. Pertains to the adequacy of the number of condition measurements over time to discern performance trends. Pavement condition data contains significant vari- ability that can have similar or larger magnitudes than the yearly growth rates of the condition. For example, Figure A-3.5 shows an example of roughness growth following the application of a thin overlay (for a single LTPP SPS-3 segment). As shown, five measurements were taken each time the roughness was assessed; the values in the first year varied by as much as 6.7 in./mile in this controlled experiment. Therefore, larger variations can be expected for roughness measure- ment on a larger network of pavements. • Relevance. Pertains to the quality of conclusions that can be developed using the selected data. Determining how well the selected data meets this attribute should be done following the development of the performance models, by evaluating the results that provide the expected information about the effectiveness of preservation (as discussed in Chapter A-4). The next chapter discusses implementation of the performance measures, and details the steps that should be taken for implementation. In addition, Appendix A provides supporting examples. Figure A-3.5. Example of roughness growth data.

A-25 This chapter details the process for implementing the performance measures and provides guidance on how to use these measures to assess the effectiveness of preservation. The approach to calculating the change in condition and changes in performance following preservation and the use of the measures to assess the effects of preservation is presented in Figure A-4.1, and is discussed in detail in the following sections. A-4.1 Calculating Initial Change in Condition The initial change in condition due to preservation is defined by the change in the pavement condition immediately following application of preservation. In many cases, condition measure- ments immediately prior to (e.g., within the month before) and immediately following (e.g., within the month after) application of preservation are not available. Therefore, to support implementa- tion of the measures, two condition measurements closest in time to the preservation treatment are generally used. As long as the condition data are collected relatively frequently (e.g., on an annual or biannual basis), and the rate of deterioration following preservation is not very high, using these condition measurements is expected to be adequate. However, if data is collected on a relatively infrequent basis, the approaches to data selection outlined in Chapter A-3 should be considered. It was found that the initial change in condition resulting from preservation for a given agency is only related to the specific treatment type and the condition prior to applying preservation (NCHRP Research Report 858). This was because the condition data used to calculate the imme- diate change in condition were generally obtained in close enough succession such that minimal pavement deterioration occurred between the measurements. The steps for calculating the initial change in condition resulting from the application of a preservation treatment are to: 1. Identify the condition measurements just prior to and immediately following the applica- tion of preservation for several segments. 2. Calculate the difference between the two data values for each segment; designate an improve- ment in pavement condition as a positive value. – If the selected performance measure decreases with decreasing pavement condition (e.g., most composite indexes), calculate the difference as the measurement after preservation minus the measurement before preservation. – If the selected performance measure increases with decreasing pavement condition (e.g., IRI), calculate the difference as the measurement before preservation minus the measure- ment after preservation. – It is likely that measurement variability will lead to differences that are both positive and negative, and the sign of the difference should not be changed. C H A P T E R A - 4 Implementation Process

A-26 Quantifying the Effects of Preservation Treatments on Pavement Performance 3. Perform a regression where the independent variable is the condition before the application of preservation, and the dependent variable is the calculated difference. Given that the inde- pendent variable will contain errors, a technique such as Deming regression should be used. Appendix A illustrates the calculation of the initial change in condition following the applica- tion of preservation using state and other data for several cases. A-4.2 Calculating Changes in Performance Changes in performance resulting from the application of preservation are estimated by eval- uating differences in the rate of change of the pavement condition after preservation relative to some control to assess the relative change caused by applying preservation (as is shown by examples in Appendix A). Approaches to selecting a control segment are discussed in Chap- ter A-3, along with the data required for the treatment and control segments. The changes in performance can be evaluated by considering the relative differences in rate of change or the difference in performances. • Relative differences in rate of change of performances. This approach involves defining a rela- tive rate of change in performance for both the preserved and control segments and express- ing the difference as a percent of the rate of change for the control segment. This approach requires less data elements without the need for developing detailed deterioration models. This is not the preferred approach because it will lead to overly general conclusions; pave- ments within the treatment group having different performances cannot be identified. For example, in developing performance models for control and treatment segments as illustrated in Appendix A, the treatment segment tended to perform better on average, but the control segment performed better in many cases. An example of evaluating and using this approach is given in Appendix A. • Differences in performance. This approach involves developing performance models for both control and treated segments. An example of developing such models for rut growth Performance Measures Incorporated into Agency Processes Figure A-4.1. The implementation process.

Attachment A-27 following a thin overlay is given in Appendix A. This approach is preferred, but it requires much more data to complete. Therefore, the relative differences in rate of change approach can be used if limited data is available and the absolute differences approach is used as more data becomes available. The following steps are required to evaluate changes in performance resulting from the appli- cation of a preservation treatment: 1. Obtain the condition measurements over time needed for developing models for the perfor- mance measures for both the control and treatment segments. – For the treated segments, the time of interest begins with the application of preservation, and condition measurements following the application of preservation or used in this assessment. – For control segments, the time of interest begins when the condition of the control seg- ment is approximately equal to that of the treatment segment after the application of preservation. – Figure A-4.2 illustrates the deterioration of treated and control segments, and the time period that should be evaluated for both segments. This analysis assumes that the control segment and treatment segment are not paired (i.e., not at the same site such that many independent variables are not the same). However, if the control and treatment segments are paired, many independent variables are the same for both, and the analysis provided in the appendix should be followed to develop deterioration models for each segment. As shown in Figure A-4.2 the preservation treatment was applied in year 7, and it caused an immediate change in the value of the performance measure. Based on the condition mea- surements shown in the figure, the time period of interest for the preserved segment begins at year 7 with the first condition measurement at year 8, and the condition after preservation is approximately 90, which corresponds to the condition of the control seg- ment at year 4. Thus, the time period of interest for the control segment begins at year 4. 2. Calculate the rate of deterioration or expected performance for the treated and control seg- ments using one of the following (or other) procedures: – Perform regression on the control and treated segments data to model the rates of deterio- ration and determine if the deterioration of the control and treated segments are different. 30 40 50 60 70 80 90 100 0 2 4 6 8 10 12 14 Pe rf or m an ce M ea su re V al ue Pavement age, year Deterioration of Control Segment Deterioration of Treated Segment Treated Pavement Condition Measurements Control Pavement Condition Measurements Start time period of interest Figure A-4.2. Pavement deterioration for preserved and control segments

A-28 Quantifying the Effects of Preservation Treatments on Pavement Performance This approach was used for the examples presented in Appendix A. For example, if changes in the performance measure versus time are assumed to be linear for the time period of condition measurements, then a linear regression model can be fit to the data, and the slope of the regression can be used for estimating differences in performance. – Calculate the area beneath the curve defined by the condition measurements over time. Carvalho et al. (2011) used this approach and Equation 4-1 to compare the performance of treated and control segments for analyzing the data. This method is sensitive to outliers, but if the data do not contain a significant amount of variability, it provides a good means for estimating performance. 2 (4-1) 1 0 1 1 1 0 1WeightedDistresses D D P P i i i n i i i n ∑ ∑ ( ) = +   += − + + = −  where Di = the distress in year i; Di+1 = the distress in year i+1; Pi+1 = the period of time between the subsequent distress measurements; and n = the number of years in the time period. 3. Compare the performance of the treated segment to the control segment. A-4.3 Calculating Performance, Service Life, and LCC After the immediate change in condition and change in performance resulting from preser- vation is calculated, the performance measures can be used to assess the effectiveness of pave- ment preservation (i.e., if the application of preservation treatments contributes to improved performance, increased service life, and decreased life cycle costs of a given pavement or for the overall pavement network). The models for the immediate condition jump and change in performance may be imple- mented into the pavement management system, or just used to estimate the effectiveness of pres- ervation. This section outlines the approaches for calculating improvements in performance, service life, and LCC. A-4.3.1 Estimating Effects of Preservation on Performance and Service Life The performance of a pavement is defined by its condition as a function of time, and is gener- ally calculated by integrating the function that defines pavement condition over time (shaded area in Figure A-4.3). The service life of a pavement is generally defined as the time it takes for the pavement to reach a predefined threshold (IRI of 150 in./mile in this example). The change in service life is the difference in time taken for the control pavement to reach the threshold and for the preserved pavement to reach the threshold (approximately a 6 year service life exten- sion in Figure A-4.3). In order to estimate the effect of preservation on the performance and service life of a pave- ment, both the immediate and subsequent changes in condition resulting from preservation should be known, as well as a condition deterioration model specific to the performance measure.

Attachment A-29 This model can be developed as part of implementation of the measures (presented in Appen- dix A for estimating rut growth following a thin overlay), or otherwise an existing deterioration model may be used. Given a deterioration model, immediate change in condition and changes in performance resulting from preservation, the approach to estimating the effect of preservation on the performance of a pavement as illustrated in Figure A-4.3. The approach consists of three steps: • Define threshold condition beyond which preservation is no longer a treatment option (T). • Define the time when preservation treatment is applied (t1), the time when the control seg- ment reaches the threshold (t2), and the time when the preserved pavement segment reaches the threshold (t3). • Calculate the area bound by the threshold and two deterioration models. An example illustrating this approach is provided in Appendix A. A-4.3.2 Estimating Effects of Preservation on LCC LCC is a fundamental part of assessing the effectiveness of preservation. Once the effects of pavement preservation on the condition and performance of a pavement are known, these effects are combined with direct costs to estimate LCC for two scenarios: (1) preservation is applied in the specified year, and (2) preservation is applied in a different year (or not applied). The scenario providing the lower cost or the higher benefits (normalized by costs) is the preferred option. Two approaches for assessing the effectiveness of preservation are outlined in this guide: (1) EUAC and (2) cost-benefit analysis. Equivalent Uniform Annual Cost Estimating the EUAC approach to support decision-making has been implemented by the Washington DOT using historic costs (Luhr and Rydholm 2015). However, the marginal EUAC 0 20 40 60 80 100 120 140 160 180 200 0 5 10 15 20 25P er fo rm an ce M ea su re V al ue - I R I (i n/ m ile ) in T hi s E xa m pl e Time (Years) Assumed Threshold (T) Improved Performance Resulting from Preservation t1 t2 t3 Deterioration of Control Segment p(t) Deterioration of Treatment Segment r(t) Step 1 Given T, p(t) and r(t), calculate t1, t2 and t3 Step 3 Estimate the area between the curves Change in Service Life Step 2 Calculate the change in service life as t3 – t2 Figure A-4.3. Estimating effect of preservation on pavement performance.

A-30 Quantifying the Effects of Preservation Treatments on Pavement Performance approach that describes the cost per year of life extension while accounting for the discounted worth of money over time is recommended. Equation 4-2 is used to calculate the EUAC.   1 1 1 (4-2)EUAC Cost i i i n n ( ) ( ) = + + − where i = the discount rate (typically 3 to 4 percent is assumed); n = the service life or the analysis period in years; and Cost = present worth treatment cost. Cost-Benefit Analysis Cost benefit analysis is an approach to normalizing costs by the expected benefits resulting from incurring the cost, as is detailed in Appendix A. However, because approaches to monetiz- ing the benefits associated with improvements in pavement performance are complex, proxy measures are generally used for benefits (e.g., service life extension). The EUAC is a form of cost benefit analysis that considers the initial costs and expected life extension values. The use of benefit (area under the curve) is expected to provide more information than only using the life extension because it accounts for both the life extension and the condition trend of the pave- ment. Figure A-4.4 illustrates the potential difference in those approaches. It shows two cases having different potential deterioration curves following the application of preservation, but the same life extension. Case 1 illustrates a pavement that exhibits higher values of the performance measure following treatment application than Case 2. This difference is captured when the per- formance is used as the benefit instead of relying only on the life extension. 0 20 40 60 80 100 0 5 10 15 20 25 30 35 Pe rf or m an ce M ea su re V al ue Percentage age, year Deterioration of Control Case 1 Case 2 Condition Threshold Life Extension Figure A-4.4. Performance after preservation for two example cases

A-31 A-5.1 Performance Measures Implementation This guide detailed a 3-step process for the implementation of performance measures to assess the effects of preservation treatments on pavement performance, service life, and LCC. The first step detailed how to select performance measures of the suggested four measures, and described a method to identify alternative measures. The second step related to required data elements, identification of the sources of the data, and gathering of the data. The third step addressed the development of models that capture the immediate change in condition and the long-term change in performance resulting from the application of preservation, for implementation in pavement management practices, and presented examples for the calculation of costs and ben- efits resulting from the application of preservation. A-5.2 Other Considerations In addition to the approaches in the guide, there are several other consideration such as the need for periodic training of personnel as it pertains to the analysis of pavement data. Also, the data may have particular attributes that require advanced techniques to address beyond agency resources and would require support from other resources such as universities and consultants. Another important consideration is the implementation of the performance measures to affect decision-making. For example, many modern pavement management systems include capabilities for cost benefit analysis. Including models that describe the effect of preserva- tion on the pavement condition into pavement management systems would help support decision-making. Finally, the calculations for assessing preservation effectiveness can be automated and implemented into many existing software tools. Although there are no specific hardware or software requirements associated with implementing the measures, these analytical tech- niques lend themselves to the use of specialized analytical software (e.g., Microsoft Excel™, Matlab™, etc.) to facilitate the implementation. References Amekudzi, A., and M. Meyer. 2011. Best Practices in Selecting Performance Measures and Standards for Effective Asset Management. Georgia Department of Transportation Office of Materials and Research, Forest Park, GA, 2011. ASTM International Standard D6433. 2009., Standard Practice for Roads and Parking Lots Pavement Condition Index Surveys. ASTM International, West Conshohocken, PA.: ASTM International, 2009. C H A P T E R A - 5 Summary

A-32 Quantifying the Effects of Preservation Treatments on Pavement Performance Bergdahl, M., Ehling, M., Elvers, E., Földesi, E., Körner, T., Kron, A., Lohauß, P., Mag, K., Morais, V., Nimmergut, A., Viggo Sæbø, H., Timm, U., João Zilhão, M. 2007. Handbook on Data Quality Assessment Methods and Tools. European Commission, Eurostat, Brussels. Cambridge Systematics. 2006. Performance Measures and Targets for Transportation Asset Management. “NCHRP Report 551: Performance Measures and Targets for Transportation Asset Management,” Transportation Research Board of the National Academies, Washington, 2006. Carvalho, R., Ayres, M., Shirazi, H., Selezneva, O., & Darter, M., 2011. Impact of Design Features on Pavement Response and Performance in Rehabilitated Flexible and Rigid Pavements. Washington: Federal Highway Administration, Final Report FHWA-HRT-10-066. Washington, D.C. Costello, S.B., L. Bargh, T. Henning and M. Hendry. 2013. “Proposed New Performance Indicator – Vehicle Operating Cost Index (VOCi) Due to Road Roughness.” Presented at the 93rd Annual Conference of the Transportation Research Board. Washington, D.C. January 2013. de Leon Izeppi, E., S. Katicha, G. Flintsch, R. McCarthy, K. McGhee. 2016. Continuous Friction Measurement Equipment as a Tool for Improving Crash Rate Prediction: A Pilot Study. Virginia Transportation Research Council. Charlottesville, VA. Elkins, G. E., T. M. Thompson, J. L. Groeger, B. Visintine and G. R. Rada. 2013. “Reformulated Pavement Remain- ing Service Life Framework.” Federal Highway Administration, Report No. FHWA-HRT-13-038.McLean, VA, 2013. Federal Register. 2017. National Performance Management Measures: Assessing Pavement Condition for National Highway Performance Program and Bridge Condition for National Highway Performance Pro- gram. 23 CFR Part 490, Federal Register Number 2017-00550. January 18, 2017. Washington, D.C. FHWA. 2016. Information: Guidance on Highway Preservation and Maintenance. Federal Highway Administra- tion Associate Administrator for Infrastructure. Washington, D.C. Available at https://www.fhwa.dot.gov/ preservation/memos/160225.cfm as of April 9, 2017. FHWA. 2014. Highway Performance Monitoring System Field Manual. Federal Highway Administration Office of Highway Policy Information. Washington, D.C. Federal Highway Administration. 2014. Highway Perfor- mance Monitoring System (HPMS) Field Manual, Office of Highway Policy Information. Florida DOT. 2015. Asset Management Plan: Preserving the States Infrastructure. Florida Department of Trans- portation Office of Policy Planning, Tallahassee, FL. Gelman, A. and Hill, J. 2017. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press, Cambridge, UK. Jackson, N.C. 2008. Development of Revised Pavement Condition Indices for Portland Cement Concrete Pavements for the WSDOT Pavement Management System. Washington State Transportation Center, Seattle, WA. Luhr, D., C. Kinne, J. Uhlmeyer, and J.P. Mahoney. 2010. “What We Don’t Know About Pavement Preservation.” Conference Proceedings from the First International Conference on Pavement Preservation. Newport Beach, CA, pp. 611–627. Luhr, D.R., and T. C. Rydholm. 2015. “Economic Evaluation of Pavement Management Decisions.,” in Com- pendium of Papers from the 9th International Conference on Managing Pavement Assets, Alexandria, VA, 2015. Maryland State Highway Administration, 2014. Pavement and Geotechnical Design Guide. Maryland DOT, Baltimore, MD. McGhee, K. H. 2002. Development and Implementation of Pavement Condition Indices for the Virginia Department of Transportation. Virginia Department of Transportation, Richmond, VA. Pipino, L., Lee, Y., Wang, R. 2002. Data Quality Assessment. Communications of the ACM, Vol. 45, No. 4. pp. 211–218. Proctor, G.D., S. Varma and S. Varnedoe. 2012. Asset Sustainability Index: A Proposed Measure for Long-Term Performance. Federal Highway Administration, Report No. FHWA-HEP-12-046. Washington, D.C. Rada, G., Visintine, B., Bryce, J., Thyagarajan, S., Elkins, G., 2016. Application and Validation of Remaining Ser- vice Interval Framework for Pavements. Federal Highway Administration Report No. FHWA-HRT-16-053, Washington, DC. Rada, G. R., Bryce, J., Visintine, B.A., Hicks, R.G., and Cheng, D. 2017. Pavement Performance Measures that Consider the Contributions of Preservation Treatments. Final Project Report. NCHRP Project 14-33, Washington, 2016. Transportation Research Board of the National Academies, Washington, D.C. (under preparation) Rydholm, T.C. and D.R. Luhr. 2014. “Modeling and Analyzing Budget Constrained Pavement Preservation Strat- egies.” Presented at the 93rd Annual Conference of the Transportation Research Board. Washington, D.C. Simpson, A., G. Rada, B. Visintine, J. Groeger, and J. Guerre. 2013. Improving FHWA’s Ability to Assess Highway Infrastructure Health: Development of Next Generation Pavement Performance Measures. Federal Highway Administration, Report No. FHWA-HIF-13-042, Washington, D.C.

Attachment A-33 Speir, R., T. Puzin, R. Barcena and P. Desaraju. 2009. Development of Friction Improvement Policies and Guidelines for the Maryland State Highway Administration. Maryland State Highway Administration, Baltimore, MD. Texas DOT. 2009. “Texas Department of Transportation PMIS Technical Manual.” Texas Department of Trans- portation, Austin, TX. Washington DOT Materials Division. 2010. WSDOT Strategies Regarding the Preservation of the State Road Net- work, A report to the State Legislature in Response to SB 6381. Seattle, Washington. World Road Association. 2010. “Towards Development of a Risk Management Approach.,” WRA, La Défense cedex, France. Zietsman, J., T. Ramani, J. Potter, V. Reeder, and J. DeFlorio. 2011. A Guidebook for Sustainability Performance Measurement for Transportation Agencies. Final Report. NCHRP Project 08-74. Transportation Research Board of the National Academies, Washington, D.C.

A-35 This appendix presents 10 case studies that illustrate important concepts for consideration in implementing performance measures to assess the effect of preservation treatments on pave- ment performance. Four case studies illustrate the effect of data quality on the initial condition jump and long-term performance model development, and the benefits of collecting additional explanatory variables. Three case studies illustrate how to model the initial condition jump and long-term performance changes resulting from the application of preservation treatments. One case study describes how to estimate changes in performance, service life, and life cycle costs. Two case studies illustrate the implementation of the initial condition jump and long- term performance models within a pavement management system using data from two state highway agencies. AA-1 Initial Condition Jump and Long-Term Performance Changes As noted throughout the guide, assessing the effect of pavement preservation on pavement performance requires information on the initial condition jump and the long-term changes in performance. The process of calculating the condition jump and changes in long-term perfor- mance resulting from the application of preservation treatments is illustrated in this section for three cases: • Using data with several independent variables (factors that could potentially influence con- dition jump and long-term performance) and no control segments; • Listing data with few independent available variables and no control segments; and • Using data that includes several independent variables and control segments. AA-1.1 Using Data with Multiple Independent Variables The ideal scenario for assessing the effects of preservation on performance measures is the case in which a large data set with many explanatory variables are available to use in the regres- sion. For example, the Virginia DOT maintains a database of information regarding traffic, hard pavement thickness, pavement structural number, and functional class along with the pavement performance data. In addition to assessing pavement performance as a function of preservation treatment type, the effect of such additional variables can be estimated. To illustrate the development of model for the condition jump and changes in long-term performance, two cases are evaluated: one for microsurfacing using data from the Virginia DOT; and another for diamond grinding using data from Washington State DOT. A P P E N D I X A Case Studies

A-36 Quantifying the Effects of Preservation Treatments on Pavement Performance Microsurfacing data from three Virginia districts was gathered to evaluate transverse cracking and rutting. The data provided condition for data on 0.1-mile segments from 2007 to 2015, as well as data pertaining to pavement functional class, average annual daily truck traffic, pavement thickness, and pavement structural number as obtained from deflection testing (the structural number and thickness information were provided only for interstate pavements on a 0.2-mile interval). The first step in analyzing the data was to identify the locations and dates of the microsurfac- ing applications. This information was contained within a construction history database pro- vided by the Virginia DOT that tracked treatment and layer information for each segment on the pavement network. Using this database, and available condition data, several projects that were constructed between 2008 and 2014 were identified. All these projects were used to evalu- ate the immediate jump in condition, and the projects constructed between 2008 and 2010 were used to evaluate the performance following a treatment. The next step in analyzing the data was to match the condition data over time with the struc- tural and traffic data. However, different designations were given to the start and end locations of the 0.1-mile segments annually, making it impossible to compare the performance of the same pavement segment over time. To resolve this issue, a set of reference locations against which each of the variables could be mapped was selected; these locations were those where the structural number was estimated from deflection measurements. The condition data for each 0.1-mile segment that had a centroid located within 0.1 miles of the location was averaged for each of the years. In addition, the thickness and traffic data were also similarly mapped onto these reference locations. A sample of the data compiled in conjunction with transverse crack- ing is shown in Table AA-1. Once the data were compiled, the effect of microsurfacing on the immediate change in con- dition was estimated by subtracting the condition measurement after treatment application from that before treatment application. For example, the data shown in Table AA-1 indicates a reduction in transverse cracking due to microsurfacing application is 13 ft (4 m), effectively resetting the transverse cracking back to zero. Transverse cracking data microsurfacing effectively resets the length transverse cracking back to zero feet in 93 percent of the segments. Figure AA-1 shows the rutting after micro- surfacing application versus the rutting before application. The rutting following a micro- surfacing application was slightly higher than that before application by an increase average of 0.028 inches (0.7 mm) due to rut growth in the year that elapsed between measurements. The analysis showed that no variables other than the treatment type and condition prior to treatment were related to the condition following microsurfacing. The performance of the pavements following microsurfacing application was evaluated in terms of transverse cracking. The same approach can be used for assessing the performance in terms of rutting or other measures. Given that data were available for a relatively short time, it Table AA-1. Sample data from a microsurfacing segment on Interstate 81. Year of Condition Measurement Begin Milepost End Milepost Year of Micro- surfacing Transverse Cracking (ft) 2009 118.480 118.580 2009 13 2010 118.500 118.600 2009 0 2011 118.500 118.600 2009 1 2012 118.500 118.575 2009 1

Attachment A-37 was assumed that cracking deteriorated linearly when assessing whether the growth rate of trans- verse cracking was only a function of the treatment type or a function of additional explanatory variables. For this evaluation, a robust linear regression model was fit to each segment for which at least five years of data following treatment application was available to estimate a growth rate trans- verse cracking following the process described in Section AA.1.2. Once slopes were fitted to each segment, it was combined with the remainder of the data (e.g., thickness, traffic, etc.) to perform a regression; a sample of data is shown in Table AA-2. Regression was then performed to assess whether the growth rate of transverse cracking was a function of any of the considered independent variables collected (the last five columns in Table AA-2). The model developed to predict the growth rate of transverse cracking following microsurfacing application is presented by the equations     TR crack growth 8.09 0.47 PreTreatmentCrack 0.70 PreTreatmentCrack PostTreatmentCrack 0.88 StructuralNumber + 0.05 PreTreatmentCrack PostTreatmentCrack StructuralNumber ( ) ( ) = + − − − − A comparison of the values predicted using the model (y-axis) to the measured rate is shown in Figure AA-2. The most relevant variable for describing the growth rate following treatment is the value of the distress prior to treatment application. Rutting Prior to Microsurfacing application (in.) R ut tin g Fo llo w in g M ic ro su rf ac in g ap pl ic at io n (i n. ) Figure AA-1. Change in rutting following a microsurfacing application. Transverse Cracking Growth Rate After Treatment (ft/year) Transverse Cracking Before Treatment (ft) Change in Cracking Immediately After Treatment (ft) Structural Number Pavement Thickness (in.) AADTT 3.7 14.5 14.5 9.1 13.5 5427 0.9 0.0 0.0 7.3 11.5 5393 12.7 52.5 28.5 4.6 10 2431 8.3 206.5 206.0 5.3 10.5 2436 14.8 556.5 555.0 5.1 10.5 2436 Table AA-2. Sample data from microsurfacing segment on Interstate 81 in Virginia.

A-38 Quantifying the Effects of Preservation Treatments on Pavement Performance No control segments were provided by the Virginia DOT to enable assessment of the differ- ences in the distress growth rates between treated and control segments. However, analysis on seg- ments that are selected as controls would determine whether the distress growth rates following preservation are significantly different from those for unpreserved pavements. The control seg- ments should be selected in a manner as to not bias the results. For example, selecting pavements that have never received a treatment will minimize the effect of preservation because pavements that require no treatment perform much better than pavements for which some treatment is required. Pavements that were identified as needing preservation, but not scheduled for preserva- tion due to budget constraints or other considerations, are good candidates for control segments. Washington State DOT is a state agency that maintains a database with many potential inde- pendent variables. A similar approach was used to assess the influence of diamond grinding using data from this state. The data provided included construction history, condition, thickness, and traffic information, and other information that was not considered in this analysis (e.g., treat- ment cost). For this case study, the effect of diamond grinding (e.g., dowel rod replacement) on the condition jump and changes in performance for faulting was investigated (independent of other preservation). The locations where diamond grinding was performed were identified and data for all relevant variables was gathered. Once diamond grinding was performed on a pavement section, its begin- ning and end milepost remained the same (until the next work was performed). In addition, thickness and traffic information of the segments were matched to those in the construction record and average values were used for each segment (if different measurement locations were identified within the segment). The effect of diamond grinding on the immediate change in faulting was evaluated by sub- tracting the posttreatment measurement from the pretreatment measurement. Faulting reduc- tion as a function of initial faulting, shown in Figure AA-3, indicates that most of the faulting was eliminated for many segments. Linear regression was performed to assess whether the improvement in faulting is related to any other variables. It was found that initial faulting was the primary factor in explaining the reduction in faulting, and thus the reduction in faulting following diamond grinding can be estimated using only the initial faulting. Figure AA-2. Transverse crack growth rate as a function of many variables.

Attachment A-39 To assess the performance in terms of faulting for each segment following diamond grinding, a robust linear regression model was fit to the data for each segment. A linear fault growth was assumed to assess whether the growth rate of faulting was only a function of the treatment type or additional independent variables. Although fault growth is not expected to be linear, linear model will give a good estimation of the relative rate of faulting growth for each segment over a relatively short time frame. To further evaluate whether the growth rate was a function of additional variables, a linear regression model was fit to the data for each segment for which data for at least five years fol- lowing treatment was available. The slope of the regression of each segment was used with the remainder of the data (e.g., thickness and traffic) to perform a regression to determine the extent of importance of the other variables. In this case, it was found that none of the potential inde- pendent variables was important in explaining the variance in faulting growth rates following diamond grinding. The faulting growth rate versus the initial faulting indicates that the average rate of faulting growth is almost zero. AA-1.2 Use of Data Where Limited Variables are Available Models can be developed even when data for all desirable variables (e.g., pavement thickness and deflection data, etc.) are not available, using available data. To illustrate this, data provided by the Texas DOT for chip seals and pavement condition were used to develop the condition jump and performance models. Data for four districts (Yoakum, Atlanta, Bryan, and Dallas) spanning from 2005 to 2015 for chip seals (separated out as surface treatment, 1-layer, 2-layer, and rubberized chip seal within the dataset) were used. The distresses included percent patching, percent block cracking, percent fatigue cracking, percent longitudinal cracking, quantity of transverse cracks, and percent rut- ting (shallow and severe levels). The data for surface treatment in the years 2006 to 2014 were identified for 6,757 pavement segments. The first step in assessing the effect of chip seals on the pavement was to identify the initial condition jump resulting from its application. In this example, transverse cracking and rough- ness were evaluated (assessing the effects using other measures can follow the same steps). The condition jump was calculated as the change in cracking and roughness following treatment application from that prior to treatment application on the same segment. The data showed that the application of the chip seal eliminates transverse cracking for 98 percent of the segments. Several data points (equaling two percent of the data) fall below the line of equality, but given Figure AA-3. Effect of diamond grinding on faulting.

A-40 Quantifying the Effects of Preservation Treatments on Pavement Performance measurement errors, it should not be expected that all segments would be shown to have elimi- nated transverse cracking completely. The change in roughness following a chip seal is shown in Figure AA-4. The figure shows a significant variability in the IRI measurements. For example, the changes in IRI value following the chip seal application can be up to ± 100 in./mile (with a mean change of zero) for segments having IRI value before the chip seal application of 150 in./mile Figure AA-4 shows that, with the exception of segments with an initial roughness greater than 200 in./mile (3.16 m/km), the chip seal is not expected to affect roughness. For segments with an initial roughness greater than 200 in./mile (3.16 m/km), a roughness is reduced following the application of the chip seal. The effect of chip seals on pavement performance in terms of transverse cracking and rough- ness was then evaluated. The growth rate of each measure following a treatment was related to the pretreatment condition values. The growth rate was calculated as the slope fit to the data points as a function of time following treatment, and the slope was calculated using regression. No control segments were provided to compare the results from preserved and not preserved pavements. The growth rate of transverse cracking following preservation is shown in Figure AA-5. The figure shows that the growth rate is related to the initial number of transverse cracks, although Figure AA-4. Change in roughness following chip seal. Figure AA-5. Transverse crack growth rate following chip seal.

Attachment A-41 there is significant variability in the data. The figure shows that transverse cracking is expected to grow more quickly when a chip seal is placed on a pavement with a higher number of transverse cracks. Because 94 percent of the segments that received a chip seal had no initial transverse and 4 percent had one or two cracks per 100 feet, transverse cracking performance could not be related to the initial transverse cracking values. In order to more accurately model the perfor- mance of transverse cracking, several segments with high initial cracking should be identified and included with the model data. Otherwise, additional variables (e.g., traffic and thickness) should be considered to model the transverse cracking performance following a chip seal application. An investigation of the roughness growth rate following a chip seal application revealed that the roughness change was not a function of the initial roughness; a large number of segments with at least 5 years of IRI measurements showed negative IRI growth rates changes. The mean value of roughness change was zero, but there was significant variance in these measured value ranges. Based on the analysis of the data provided by the Texas DOT, it can be concluded that chip seals only have an effect on cracking and not roughness. AA-1.3 Use of LTPP Data If adequate data is not available to develop the condition jump and change in performance following a treatment, data from other sources (e.g., LTPP program) can be used. LTPP data is available at the FHWA webpage LTPP InfoPave™ that provides a data query interface to allow users to specify data to be downloaded. To demonstrate the use of these queries, LTPP data were extracted for two cases: 1. Data from the LTPP SPS-3 experiment dealing with the preservation treatments for AC pave- ments; each treatment was accompanied by a control. In this case study, data from the wet non-freeze climate zone were extracted to evaluate the effect of preservation treatments on pavement performance. Data on pavement construction history, climatic information, traffic information, condition information and structural information were extracted with thin over- lays. Measures evaluated were rutting and transverse cracking. 2. Data from the General Pavement Studies (GPS)-3 and -4 experiments, dealing with jointed concrete pavements. Pavement test sections in these experiments were not accompanied by control segments. Sections treated with diamond grinding were evaluated, and the measures considered were faulting and roughness. Case 1 – Thin Overlay on AC Pavement Assessing the effect of preservation on pavements using the selected performance measures required several steps. The first step was to assess the immediate change in condition following the thin overlay application. Not every segment within the wet no-freeze region had distress measurements for each distress prior to the application of the thin overlay. The change in trans- verse cracking following the application of the thin overlay is shown in Figure AA-6. All of the data points fall on the diagonal that defines where the improvement is equal to the initial crack- ing value, indicating that the thin overlay sets the transverse cracking back to zero. The change in rutting following the application of the thin overlay is shown in Figure AA-7. The thin overlay reduced the rutting to zero for most (but not all) test sections (many points coincide with the diagonal). It is conservatively estimated that the thin overlay reduces rutting to 0.05 in. (the mean rut depth following the thin overlay). The effect of a thin overlay application on long-term performance was then calculated. The growth rate for each measure was calculated using the process shown in Figure AA-4. Slope of

A-42 Quantifying the Effects of Preservation Treatments on Pavement Performance the line fitting the data points as a function of time following treatment application was calcu- lated using a robust regression technique that iteratively applies weights to the individual resid- ual values. The median values for traffic, number of freeze-thaw cycles, yearly rainfall, structural number, and yearly temperature were also gathered for each test section. A sample of the data compiled for transverse cracking performance is shown in Table AA-3. A regression procedure was then followed to determine if the growth rate of each measure is a function of the other independent variables that were considered. The growth rate of trans- verse cracking was found to be statistically related to the average annual temperature and the structural number. However, a model to relate transverse cracking growth rate to the available independent variables could not be developed. The data from the LTPP SPS-3 experiment allows a pairwise analysis to be conducted because each control test section is paired with a treatment test section and the difference in performance can be compared on a test section-by-test section basis; Figure AA.8 shows such a comparison. Further evaluations determined that the performance of the control test section and treatment test section was significantly different. Although growth rates for the majority of treatment test sections are lower than for the control test section, one treatment test section had a growth rate much higher than the control test section (greater than 60 ft/year), as shown in Figure AA-8. Transverse cracking on this test section was initially detected within a year of the thin overlay application, and grew rapidly thereafter. T ra ns ve rs e C ra ck in g R ed uc tio n (f t) Figure AA-6. Transverse cracking reduction following thin overlay. R ut tin g R ed uc tio n (i n. ) Figure AA-7. Rutting reduction following thin overlay.

Attachment A-43 Rutting following the overlay application was evaluated for both control test sections and treatment test sections; the comparison is shown in Figure AA-9. The figure shows that the growth rate of the control test sections tend to be higher than for treatment test sections (i.e., more data are below the line of equality). Further analysis determined that the performance of the control segment and treatment segment was different. The average rut growth rate for the control segments is 0.04 in./yr, compared to an average of 0.03 in./yr. for the treatment segments. A model was developed for both control and treatment test sections to relate rut growth rate to the median values of traffic, number of freeze-thaw cycles, yearly rainfall, structural number, and yearly temperature. Details are shown in Table AA-4. The rut growth rate (in inches per year) can be estimated as the sum of the products of the coefficients and variables. A comparison of modeled growth rates to measured growth rates is shown in Figure AA-10 for both control and treatment test sections. Figure AA-10 indicates that the measured rut performance for both treatment and control test sections can be predicted using the model in Table AA-4. Case 2 – Diamond Grinding on PCC Pavements The first step in evaluating the effect of diamond grinding was to find the immediate change in condition following the treatment. The change in IRI following diamond grinding is shown in Figure AA-11. The regression model is shown in Figure AA-11. The relationship between the reduction in IRI and the initial IRI value has a slope close to one, indicating a nearly constant reduction in IRI with a negative reduction in IRI (i.e., roughness worse after grinding) for seg- ments with the IRI less than 68 in./mile before grinding. Transverse Crack Growth Rate (ft/yr) Annual Precipitation (in.) Annual Number of Freeze-Thaw Cycles Average Annual Temperature (degrees F) Structural Number Annual ESALS (Thousands) 2.71 45.63 24 64 5.5 309 3.01 39.38 73 58 6.1 110 12.36 25.96 22 61 5.4 8 9.38 41.38 79 56 7.2 560 0 47.07 86 55 3.8 5 Table AA-3. Sample data for LTPP transverse cracking model. Figure AA-8. Transverse cracking growth rate for control and treatment segments.

A-44 Quantifying the Effects of Preservation Treatments on Pavement Performance Figure AA-9. Rut growth rate for treatment and control segment. Variable Treatment Model Coefficient Control Model Coefficient Constant -0.72 -0.22 Yearly Precipitation (inches) 0.003 0.003 Number of Freeze Thaw Cycles 0.003 0.003 Average Annual Temperature (degrees C) 0.031 0.006 Structural Number -0.001 -0.071 Average Annual ESALs (in Thousands) 1.35e-05 4.14e-05 Average Annual Temperature (degrees C) * Structural Number 0 0.0032 Number of Freeze Thaw Cycles * Average Annual ESALs (in Thousands) -3.53e-06 0 Structural Number * Average Annual ESALs (in Thousands) 5.55e-05 0 Table AA-4. Rut growth rate models for control and treatment segments. Figure AA-10. Model rut growth rate versus measured rut growth rate.

Attachment A-45 The change in faulting in the wheelpath resulting from diamond grinding is shown in Fig- ure AA-12. The figure shows that the majority of test data are not on the line of equality that defines the situations in which faulting is eliminated following diamond grinding indicating that some faulting remains after grinding. The regression model shown in Figure AA-12 indicates that faulting is reduced by approximately half due to diamond grinding. It should be noted that only a relatively few segments had faulting measurements prior to diamond grinding; many of these sections had a significant gap in time of more than two years between the pre-treatment and post-treatment measurements. Therefore, the data shown in Figure AA-12 show the effect of diamond grinding in addition to the immediate change that occurred in the elapsed time between the measurements. However, to limit the effect of faulting growth between the two measurements, segments with time between the two measurements of no more than three years were considered. To assess the effect of diamond grinding on the pavement performance in terms of faulting and roughness, data pertaining to pavement thickness, traffic loading, and climate were obtained in addition to the roughness and faulting data over time. These data revealed that the calculated ESAL values were not available for many of the test sections. The rate of growth of faulting and roughness over time for each pavement segment was then determined by fitting a robust regression line to the distress versus time values (simi- lar to the method as shown in Figure AA-4) for those pavement test sections that had more than three measurements at different points in time following diamond grinding application. Figure AA-11. Effect of diamond grinding on IRI. Figure AA-12. Change in wheelpath faulting following diamond grinding.

A-46 Quantifying the Effects of Preservation Treatments on Pavement Performance However, the majority of pavements did not have an adequate number of faulting measure- ments following diamond grinding. Because a small number of pavements (less than 10) had an adequate number of faulting measurements a performance model could not be developed; faulting performance could not be assessed. Table AA-5 presents an example of data available for evaluating the performance in terms of roughness. Calculated ESAL data were available for 17 sections; 44 sections lacking ESAL data had all other information. Therefore, attempts to develop models describing performance in terms of roughness with each set of data (including and excluding ESALs) were made. Stepwise linear regression was performed using the rate of IRI growth as the dependent variable and the other items shown in Table AA-5 as independent variables. In both cases (including and exclud- ing ESALs), no relevant model could be developed to describe IRI performance. AA-2 Calculating Changes in Performance, Service Life, and LCC This section illustrates the calculation of changes in performance, service life, and life cycle costs using data from the LTPP SPS-3 experiment. AA-2.1 Estimating Changes in Performance To demonstrate the change in performance calculations, IRI data for thin overlays were extracted from the LTPP SPS-3 experiment. The first step was to identify the time period over which the performance would be assessed. Because the control test sections and the treatment test sections are paired (i.e., many of the independent variables are the same over time), the beginning of the time period for both control and treatment test sections was taken when the thin overlay was applied. The analysis was limited to control and treatment test sections that had values of IRI following treatment within 15 in./mile of each other in order to minimize the effect of the initial condition jump; 28 test sections meeting this criteria were identified. Also only pavements with very similar initial conditions were included in the analysis. The performance of the pavement test sections is then estimated for the desired time period using two methods. In the first method, a regression model was fitted to the data following the application of preservation, and the slope (roughness growth rate versus time) was taken as indication of per- formance. Regression models were fit to the data for the test section (see Table AA-6), and the results are shown in Figure AA-13. In the second method, the weighted distress value was calculated for each of 28 test sections using Equation 4-1 (see Chapter A-4.2). The weighted distresses were calculated for the test Rate of IRI Growth (in./mile/yr) IRI Before Grinding (in./mile) PCC Thickness (in.) Average ESALs Per Year (in Thousands) Average Annual Precipitation (in.) Mean Annual Temperature (Fahrenheit) Freeze Index (degree F-Days) Number of Freeze Thaw Cycles 0.23 159.86 9.30 228 25.90 44.36 1817 86 8.92 143.06 10.80 130 33.10 47.16 995 83 0.07 66.33 10.10 479 20.38 48.03 898 119 8.30 195.94 10.10 469 21.37 48.01 876 120 Table AA-5. Example data for assessing roughness performance.

Attachment A-47 section. The weighted distress for the treatment test section is 70.0 in./mile, and the weighted distress for the control test section is 69.4 in./mile. Thus, this approach also identified the control test section as performing slightly better than the treatment test section. The performance of the control test section is then compared to that of the treatment test sections to determine the effect of preservation, based on the relative change in performance of thin overlays and not based on the absolute change in performance. Therefore, the comparison was made by calculating the average ratio of performance of the treatment test sections to that of the control test section. A comparison of the IRI growth rate for the treatment to that of the control group shown in Figure AA-14 shows that the data points lie above the line of equality indicating that the treat- ment test sections perform better. The ratio of the IRI growth rate of the treatment test section performance to that of the control test section was calculated for each test section, and the average value for all sections was 0.93, indicating that the IRI growth rate for the treatment test section is 7 percent less than that for the control test section. Alternatively, the weighted distress of the treatment was compared to that of the control group as shown in Figure AA-15. The figure shows similar results, with the majority of the data points above the line of equality, indicating that the treatment test sections perform better on average. The average ratio of the weighted distress of treatment test section performance to that of the control test section for all test sections was 0.89, indicating that IRI performance for the treatment test section is 11 percent less than that for the control test section. The results shown in Figure AA-14 and Figure AA-15 are very close with a minimal difference due to the different approaches used to estimate performance; either approach can be used. However, if data variability is relatively small, the weighted differences approach is preferred. State Treatment Segment Control Segment Treatment Date Date of IRI Measurement IRI of Treatment Segment (in./mile) IRI of Control Segment (in./mile) AL C310 C340 July 1990 August 1992 58 61 AL C310 C340 July 1990 August 1994 71 70 AL C310 C340 July 1990 April 1996 67 69 AL C310 C340 July 1990 April 1999 65 67 AL C310 C340 July 1990 April 2003 75 73 Table AA-6. Example of data extracted from the LTPP. Figure AA-13. Example of using regression to evaluate the growth rate of IRI.

A-48 Quantifying the Effects of Preservation Treatments on Pavement Performance AA-2.2 Calculating Performance, Service Life, and LCC The following example illustrates the procedure for estimating changes in performance and service life for the thin overlay examples presented in the previous section. The IRI performance model is given by Equation A-1; IRI is expressed in terms of in./mile. 40 exp (A-1)0.05IRI t t( ) =  IRI threshold value was assumed to be 120 in./mile. The thin overlay was applied to a pavement with a measured IRI of 100 in./mile. Immediate improvement in IRI following a thin overlay application is given by Equation A-2 (Section A.4.1). 0.9 64.0 (A-2)IRI IRIimprovement Initial= − Rate of deterioration is 7 percent less than for the pavement without treatment after the application of the thin overlay. Figure AA-14. Comparison of IRI performance with regression slope method. Figure AA-15. Comparison of IRI performance with weighted distress method.

Attachment A-49 The following steps were taken to estimate the effect of the thin overlay on pavement performance: Step 1: Define t1, t2 and t3 The value for t1 is defined as the effective age of the pavement when preservation is applied. This can be calculated using Equation A-1: 100 40 exp 1 0.05 ln 100 40 18.30.05 11 tt= → =    =  The value for t2 is defined as the effective age of the pavement when the deterioration of the control segment reaches the threshold value; also calculated from Equation A-1: 120 40 exp 1 0.05 ln 120 40 22.00.05 21 tt= → =    =  The value for t3 is defined as the time required for the treated segment to reach the deterioration threshold value. In order to calculate the value for t3, the deterioration model for the preserved pavement segment will be assumed to have the same function of Equation A-1, but with modi- fied coefficients and the understanding that application of a thin overlay will reduce the rate of deterioration by 7 percent. To estimate the coefficients, the following assumptions are made. IRI immediately after application of preservation (year 18.3) is calculated by subtracting the IRI improvements estimated using Equation A-2 as follows: 100 0.9 100 64 74IRI IRI IRIinitial Improvement ( )= − = − − = A 7 percent reduction in the rate of deterioration following overlay application for a linear deterioration function equates to a 7 percent reduction in the slope of the line. For a nonlinear model, it will be assumed that 7 percent more time will be required for the IRI of the preserved pavement to reach the defined threshold value, as follows: a. The time required for IRI to change from 74 to 120 in./mile was calculated above as 22.0 years: The effective pavement age, to when IRI was 74 in./mile, is determined from Equation A-1: 74 40 exp 1 0.05 ln 74 40 12.30.05 00 tt= → =    =  Thus, the time required for IRI of untreated pavement to change from 74 to 120 in./mile is: 22.0 12.3 9.7 years− = and the time required for IRI of the treated pavement to change from 74 to 120 in./mile is: 9.7 1.07 10.4 years= b. The value of t3 is then calculated as the time required for the treated segment to reach the IRI threshold value (totals the effective pavement age when the overlay is applied plus the time required for IRI to change from 74 to 120 in./mile). 18.3 10.4 28.73t years= + =

A-50 Quantifying the Effects of Preservation Treatments on Pavement Performance Step 2: Calculate the Change in Service Life The change in service life can be estimated using the values for t2 and t3 as follows: Change in service life 28.7 22.0 6.73 2t t years= − = − = Plus, the application of the preservation treatment extended the time during which pavement IRI would be less than 120 in./mile by nearly 7 years, as shown in Figure AA.16. Step 3: Calculate the Change in Performance To calculate the changes in performance (shaded area in Figure AA.16), the deterioration function for the preserved pavement segment is required. It is assumed that this deterioration model will have the same functional form as the original deterioration model (Equation A-1) but with modified coefficients. These modified coefficients can be found by using the two known data points: the IRI at year 18.3 which is 74 and the IRI at year 28.7 which is 120. Thus: 74 74 18.3 18.3 ae a e b b = → =  120 74 12028.7 18.3 28.7ae substitute a e eb b b= → → =   Solving these two equations for a and b results in 0.047, 31.27b and a= = Thus, the deterioration function for the preserved pavement will be 31.27exp (A-3)0.047IRI t t( ) =  As noted in Equation A-3, the value of the coefficient within the exponent (0.047) is 7 per- cent less than the value for the original deterioration model, indicating a reduction in the rate of deterioration that can be reflected by multiplying the coefficient of the exponent by 1 − x (with x in decimal form). 0 20 40 60 80 100 120 140 160 180 10 12 14 16 18 20 22 24 26 28 30 IR I (i n. /m il e) Effective Pavement Age (Years) Control (Untreated Pavement) Preserved Pavement Immediate Condition Jump Service Life Extension IRI Threshold Improvement in Performance Figure AA-16. Illustration of example results.

Attachment A-51 Using the above information, the change in performance resulting from the application of the thin overlay shown in Figure 4.3 can be calculated using Equations A-1 and A-3 as follows: 40exp 31.27exp 120 28.7 22.0 218 . 0.05 18.3 22.0 0.047 18.3 28.7 dt dt in mile yearst t∫ ∫ ( )− + − = The life extension and increase in performance from preservation can both be used to inform pavement management decisions. An example of using life extension as part of a treatment selection objective function as used by the Maryland State Highway Administration is presented in Section AA.3.1. AA-2.3 Estimating Effects of Preservation on LCC using EUAC The data for the thin overlay example described earlier was used to illustrate the estimation of LCC using the marginal EUAC. The IRI deterioration model given by Equation A-1, the change in IRI following the thin overlay application given by Equation A-2, and the estimated reduction in the rate of deterioration after the overlay application were used. Three pavements, each with a different initial IRI value, were compared. These pavements, designated Pavement 1, 2, and 3, had an initial IRI value of 85, 100 and 115 in./mile, respectively. The cost of applying the thin overlay for all three cases was assumed to be $75,000. The first step was to calculate the life extension associated with each treatment. The approach to calculating the life extension was described in the previous example, and the results are pre- sented in Figure AA-17. These results show that the maximum life extension is achieved just prior to reaching the 120 in./mile IRI threshold value. This can be explained by the relatively similar IRI value application of the thin overlay, as calculated using the model developed using state agency data (presented in Section AA.4.1). The pavement with the higher initial IRI will have a larger decrease in IRI after overlay application, and will therefore have the largest value for the life extension. The effective pavement age shown in Table AA-7 is the time at which IRI reaches the values listed in the table as predicted by the model. The condition jump and life extension values for the pavements are illustrated in Figure AA-17, which shows that the major- ity of the life extension is a result of the condition jump. The next step is to calculate the marginal EUAC for each of the alternatives. In this example, the EUAC was calculated only for the life extension (i.e., initial construction costs were not 40 60 80 100 120 140 160 180 200 220 10 15 20 25 30 IR I (i n. /m il e) Effective Pavement Age (years) Control Pavement Pavement 1 Pavement 2 Pavement 3 Life Extension Pavement 3 Life Extension Pavement 2Life Extension Pavement 1 Threshold Figure AA-17. Condition jump and life extension for the three pavements.

A-52 Quantifying the Effects of Preservation Treatments on Pavement Performance considered), and a discount rate of 3 percent was assumed. The marginal EUAC for each of the three pavements is:   75,000 0.03 1 0.03 1 0.03 1 $20,71585 3.9 3.9EUAC ( ) ( ) = + + − =   75,000 0.03 1 0.03 1 0.03 1 $12,506100 6.7 6.7EUAC ( ) ( ) = + + − =   75,000 0.03 1 0.03 1 0.03 1 $9,577115 9.1 9.1EUAC ( ) ( ) = + + − = These results agree with the life calculated extensions in that the pavement with the highest initial IRI has the lowest uniform costs. One important note is that, in order to compare the application of preservation to the do-nothing scenario, some initial cost must be associated with the pavement. This example demonstrates a case where the IRI immediately after overlay application is almost the same regardless of the pre-treatment condition, and the change in the rate of dete- rioration does not depend on the pre-treatment condition. Therefore, the thin overlay was more cost effective when placed on a pavement in the worse condition. However, other performance measures for some treatments may show a difference trend such that preservation will be more effective when application pavement is in relatively good condition. AA-2.4 Estimating LCC using Cost Benefit Analysis Use of cost benefit analysis to evaluate preservation treatment effectiveness is illustrated for the thin overlay example described in the previous section. The deterioration model for IRI given by Equation A-1, the change in IRI following the thin overlay application given by Equation A-2, and the assumed 7 percent reduction in rate of deterioration after the overlay application was used. The same three pavement segments with initial IRI values 85, 100, and 115 in./mile were considered and the cost of applying the thin overlay to each segment was taken as $75,000. The approach described in Section AA.3.2 was followed, and the results are given in Table AA-8. The trend of cost to increase in performance is the same as obtained from the EUAC analysis Initial IRI (in./mile) Effective Pavement Age When OL is Applied (years) IRI After Overlay (in./mile) Effective Pavement Age After Overlay (years) Effective Age When IRI of 120 in./mile is Reached Life Extension (years) 85 15.1 72.5 11.9 25.9 3.9 100 18.3 74.0 12.3 28.9 6.7 115 21.1 75.5 12.7 31.0 9.1 Table AA-7. Life extension for the three pavements. Initial IRI (in./mile) IRI After Overlay (in./mile) Life Extension (years) Increase in Performance (in./mile - years) Cost Divided by Benefit (dollars per (in./mile - years) 85 72.5 3.9 103 728 100 74.0 6.7 218 344 115 75.5 9.1 336 223 Table AA-8. Increase in performance resulting from thin overlay on each pavement.

Attachment A-53 because a similar form of the deterioration model was used resulting in similar correlation between life extension and increase in performance. AA-3 Implementation of Performance Measures Implementing performance measures into pavement management practices will help con- sider the contributions of preservation on performance in the agency decision-making pro- cesses. This section describes the implementation process considering like data available at two agencies: Maryland SHA and Virginia DOT. AA-3.1 Maryland SHA The Maryland SHA provided information regarding their decision-making processes along with the data. Using this information and data, implementation of the performance measures and models would require the following steps: 1. Selection of the performance measures. Performance measures are selected and models developed to support the treatment selection process which is one step in the network-level optimization procedure that is designed to group and recommend projects to implement in a given year. In order to select the treatment for each pavement segment, an analysis is per- formed to identify the treatment that maximizes life extension, while minimizing treatment costs. The measures selected for this case study are those used by the Maryland SHA in the decision-making process: – Functional cracking index (FCI) which combines non-load-related cracks – Structural cracking index (SCI) which describes load-related-cracks – IRI – Rutting – Skid number 2. Development of Models. The development of the models to describe initial condition jump and changes in performance can be completed as described earlier in this appendix. This step begins with gathering data for each measure and assessing whether the appropriate data exists to support developing models based on data within the agency database. A distinct pos- sibility is that there is adequate data to support the development of condition jump models, but inadequate information to support development of the models describing changes in performance. Thus, data may also need to be gathered from other sources (e.g., LTPP data) to support development of a full set of models. An example of assessing a thin overlay is detailed in this section to demonstrate the imple- mentation, and information from the previous sections of this appendix (assessing thin over- lays using LTPP SPS-3 experiment data) is used to support this demonstration. In addition to the rutting and cracking models that were assessed in Section AA-2 of this appendix, data were gathered from the LTPP SPS-3 experiment to assess the effect of thin overlays on pave- ments in terms of the IRI. The data gathering and model development followed the approach discussed previously. The immediate change in IRI resulting from applying a thin overlay application versus the IRI before the overlay application is shown in Figure AA-18. A model showing that the improvement in IRI is a function of the IRI before the overlay application was developed. The change in performance over time was also investigated. Figure AA-19 shows the IRI growth rate of the treated sections versus that of the control sections. The majority of the data fall below the line of equality, indicating a lower IRI growth rate following a thin overlay applica- tion than for untreated sections. The development of a model that explains the difference

A-54 Quantifying the Effects of Preservation Treatments on Pavement Performance in performance was also investigated, but no reliable model could be developed. Therefore, the relative difference in performance was used in the decision-making process. 3. Incorporate Measures into Business Processes. Information obtained from the models is used to support the network-level treatment selection process. The objective function used by the Maryland SHA in the treatment selection process is the following:  0.5 (A-4)Max z LifeExtension Cost VMT VMT Lane Miles Lane Miles Segment Total Segment Total = + − −     where z = the objective function value; LifeExtension = the life extension provided by the treatments in years; Cost = the treatment costs in dollars; VMTSegment = the vehicle miles traveled on the segment; VMTTotal = the vehicle miles traveled on the network; Lane-MilesSegment = the length of the segment in lane miles; and Lane-MilesTotal = the length of the network in lane miles. Figure AA-18. Immediate IRI change after a thin overlay application versus pre-overlay IRI. Figure AA-19. Difference in performance – control segments compared to segments that received a thin overlay.

Attachment A-55 The calculation of the life extension is described later in this section. The remaining ser- vice life is calculated as the lowest of the times it takes for each distress to reach a predefined threshold value. The following information was used to support this example; – The cost of applying a thin overlay was provided by the Maryland SHA and is a function of the district, functional class, road class, and the initial IRI. The pavement class informa- tion and initial condition are given in Table AA-9, and the cost was taken as $98,600 per lane-mile. – The baseline FCI and SCI deterioration models are given by Equations A-5 and A-6, respec- tively. It was assumed that the thin overlay resets the FCI and SCI back to 100, and that the SCI deterioration follows the baseline SCI curve. Recognizing that transverse cracking growth rate is reduced following application of a thin overlay (see Figure AA-18), the FCI growth rate was reduced by 20 percent following application of the thin overlay.   ( ) ( ) = − ≤ = − > 100 0.65 7 104.4 1.2725 7 (A-5) FCI age age for age FCI age age for age where age = the age of the pavement (time following treatment) in years, and FCI = the functional cracking index defined by Maryland SHA. 100 0.241 12 114.5 1.446 12 (A-6) SCI age age for age SCI age age for age ( ) ( ) = − ≤ = − >   – The IRI deterioration curve is given by Equation A-7, and it was assumed that the IRI growth rate is reduced following the thin overlay application (see Figure AA-19). The reduction in IRI following the thin overlay application is given by Equation A-8 40 (A-7)0.035IRI age e age( ) = µ   where μ is 1 for the do nothing case and 0.8 following the thin overlay application. ( )= − −0.9 64 (A-8)IRI IRI IRIAfterThinOL Initial Initial where IRIAfterThinOL is the IRI following application of the thin overlay, and IRIInitial is the IRI immediately preceding the thin overlay application. – The baseline rutting model is given by Equation A-9. It was assumed that rutting was set to zero immediately following the thin overlay application and the growth rate was reduced by 10 percent thereafter. ( ) = 0.008 (A-9)Rut age age District Functional Class Road Class Length (Lane-Miles) AADT IRI (in./mile) FCI SCI Rut Depth (in.) Skid Number 1 1 Urban 2 5000 110 85 90 0.1 40 Table AA-9. Initial pavement condition.

A-56 Quantifying the Effects of Preservation Treatments on Pavement Performance – The skid resistance model is given by Equation A-10. Maryland SHA data showed that the skid number following a thin overlay application could be taken as 47. ( ) ( )= − +3.161 log (A-10)Skid Number age age SkidInitial where SkidInitial is the initial value of the skid number. – To calculate the remaining service life, the following threshold values were suggested: 170 in./mile for IRI, 65 for SCI, 50 for FCI, 0.3 in. for rutting, and 35 for the skid number. Once the models were defined, the effect of applying a thin overlay on the pavement seg- ment described in Table AA-9 was investigated. First, the life extension was calculated using the models described previously. As illustrated in Figure AA-20 for IRI, the application of the thin overlay immediately reduces the IRI, and slows the deterioration of IRI over time. The life extension based on IRI is calculated as the time difference between that required of the treated pavement to reach the threshold IRI and that remaining for the untreated pave- ment to reach the same threshold. To calculate the overall life extension, the life extension for all performance measures is calculated, and the lowest of these values is taken as the overall life extension resulting from the application of the thin overlay. Based on the models described previously, the life extension gained by applying the thin overlay is 10 years. This information can be put in the objective function, as is demonstrated in Equation A-11. The ratio of relative VMT and segment length to total values (see the denominator in Equation A-11) was obtained from the Maryland SHA. The resulting value of the objective function is 3.14. This value can be used to compare multiple treatments (e.g., chip seal, etc.), and the treatment that results in the largest value of the objective function can be selected as the best treatment to apply. In addition, the value of the objective function can be used to rank all treatments for a pavement network, where the pavement segments with treatments that have the largest objective function value are selected as the highest priority segments to implement into a work plan.    ( ) = − = − 10 98,600 2 0.5 3.225 10 3.14 (A-11) 5 z lane miles AA-3.2 Virginia DOT Another method for implementing the performance measures and models into agency pave- ment management is detailed in a report prepared for the Virginia DOT (de Leon Izeppi et al. 2016). In conducting the network-level needs assessment, the DOT develops recommendations 50 70 90 110 130 150 170 190 0 5 10 15 20 25 30 35 IR I (i n. /m ile ) Time (Years) Relative to Begining of Analysis Initial IRI Deterioration IRI Deterioration After Overlay IRI Threshold = 170 Life Extension Relative to IRI Initia l Red uctio n in IRI from Thin Ove rlay Figure AA-20. Calculating the life extension in terms of IRI.

Attachment A-57 for each pavement segment regarding the type of maintenance category that should be per- formed (e.g., preventive maintenance, rehabilitation, etc.), not a specific treatment at the net- work level. The maintenance categories for a given year are selected for each pavement segment based on condition thresholds. The network-level analysis is conducted centrally to provide budget recommendations and the number of lane-miles for each maintenance category to district engineers for use in selecting projects and their treatment. The treatment types for the pavement segments are selected based on a number of factors described in the report (de Leon Izeppi et al. 2016) and an associated tool. The tool includes a decision analysis framework for implementing performance measures and models. A simplified version of this framework is summarized as follows (de Leon Izeppi et al. 2016): 1. Select the performance measures. The Critical Condition Index (CCI) is used as the perfor- mance measure. The CCI is a composite pavement condition measure expressed on a 100 point scale with 100 being the best possible score and 1 being the worst possible score. 2. Select the feasible treatments. This is done for each pavement segment based on the initial condition of the pavement segments. 3. Calculate the benefit of each preservation treatment. This is based on the area between the do-nothing deterioration model and the preservation deterioration model (as illustrated in Figure AA-21). To calculate the benefit, the condition jump and change in deterioration resulting from preservation are required. The effect of preservation on the CCI can be calculated using the approaches described previously and using the CCI instead of IRI or other measures. Alter- natively, the condition jump and change in performance can be calculated independently for each metric that defines the CCI, and then the metrics are aggregated up to determine the effect of preservation application on the CCI. The latter approach enables isolating sources of variance in the models before aggregating the metrics into a single value. 4. Select treatment for segments. The final step is to select the treatment types to apply to each pavement that maximize the cost effectiveness of the preservation program for the pavement network and schedule application. To illustrate the implementation of the measures, a set of example projects was selected, and the cost benefit ratio of applying preservation at different conditions was calculated. The data provided was used to calculate the initial condition jump, in terms of the CCI, resulting from the application of a thin overlay; the results are shown in Figure AA-22 (i.e., the CCI following application of the thin overlay minus the CCI before its application). The line of maximum improvement shown on Figure AA-22 defines the case where the highest possible improve- ment results from the thin overlay (i.e., it brings the CCI back to 100). As shown, the majority Time C on di tio n Threshold Condition for Benefit Cutoff Treatment Benefit Do Nothing Deterioration Preservation Deterioration Preservation Condition Jump Figure AA-21. Calculating the benefit of a preservation treatment (de Leon Izeppi et al. 2016).

A-58 Quantifying the Effects of Preservation Treatments on Pavement Performance of segments fall on the line of maximum improvement, indicating that thin overlays reset the CCI value back to 100. The remainder of the required information was extracted from the report (de Leon Izeppi et al. 2016): • Thin overlay cost is $33,077 per lane-mile. • CCI threshold value for calculating the benefit area is 60. • Do nothing deterioration was assumed to follow Equation A-12, and deterioration following the thin overlay application was assumed to follow Equation A-13. 100 5.2 (A-12)CCI age age( ) = −  where age is the time after the last pavement treatment application in years. 100 1.09 (A-13)CCI age age( ) = −  The results of applying a thin overlay at several initial conditions is shown in Table AA-10. The table shows the initial CCI values, and the estimated benefit calculated as the area between the treatment performance curve and the do-nothing performance curve (see Fig- ure AA-21). The benefit can be calculated analytically, using integral calculus or numeri- cally. In this case, the benefit was calculated analytically because the performance models are linear and the bounds of the integral are defined. Table AA-10 also shows that the most benefit is gained by applying the thin overlay at the lowest initial CCI value. This is because the jump in condition is the same regardless of the initial CCI value and the deterioration rate is not dependent on the initial CCI value. Evaluat- ing multiple treatments will help assess the cost benefit of the treatments and select the most Figure AA-22. Immediate effect of thin overlay on CCI. Initial CCI Benefit of Thin Overlay (CCI-Years) 90 647 85 674 80 695 75 712 Table AA-10. Benefits of applying a thin overlay at different initial CCI values.

Attachment A-59 effective. In addition, the reported models (de Leon Izeppi et al. 2016) indicate that seg- ments that receive a thin overlay deteriorate at 20 percent of the rate of an equivalent control segment. AA-4 References de Leon Izeppi, E., Morrison, A., Flintsch, G., McGhee, K. 2016. Best Practices and Performance Assessment for Preventive Maintenance Treatments for Virginia Pavements. Virginia Department of Transportation Final Report VCTIR 16-R3. Richmond, VA Rada, G., Visintine, B., Bryce, J., Thyagarajan, S., Elkins, G., 2016. Application and Validation of Remaining Ser- vice Interval Framework for Pavements. Federal Highway Administration Report No. FHWA-HRT-16-053, Washington, DC.