Assessing Interactions Between Access Management Treatments and Multimodal Users (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.

15 C H A P T E R 3 : R E S E A R C H A P P R O A C H Research Approach Introduction This chapter describes the approach used to develop performance relationships for commonly used access management (AM) techniques for which there is little existing information. The approach focuses on (1) identifying the techniques for which performance information is needed most by practitioners, (2) assessing the resources needed for relationship development, and (3) determining the most cost-effective combination of techniques for study in Phase 2. An overview of the research approach is described in this chapter. Additional details of the approach are provided in Appendix C. The approach used to identify and develop the most cost-effective combination of performance relationships consisted of four activities. These activities are described in the following list: 1. Assessment of the reliability of existing relationship definitions. The objective of this activity was to (1) identify performance relationships documented in the literature and (2) assess the degree to which they were “practice ready.” The information used in this assessment was gathered during the literature review. The findings from this review are summarized in Chapter 2. The more detailed findings are provided in Appendix A. 2. Prioritize AM techniques. The objective of this activity was to identify the relationships that were not available in the literature and for which information was most needed by practitioners. These relationships were prioritized using (1) the information gathered during the agency survey and (2) the relationship’s judged amenability to study. The findings from the survey are summarized in Chapter 2. The detailed findings are provided in Appendix B. The product of this activity was a list of 20 high-priority techniques. 3. Develop an initial study design for high-priority AM techniques. The objective of this activity was to develop an initial study design for each high-priority technique. Each initial study design outlined an approach for collecting the data needed to develop the performance relationship identified in Activity 2. The study design included sufficient detail as to provide a reasonable basis for estimating the cost of developing the associated performance relationships. These designs are provided in Appendix D. The product of this activity was a list of four recommended techniques for study in Phase 2. 4. Develop a final study design for the recommended AM techniques. The objective of this activity was to develop a final study design for the recommended techniques. Each final study design expanded the initial study design by incorporating additional information about site characteristics, sample size, and data collection methods. These designs are provided in Appendix E. This chapter consists of three sections. The first section describes results of activities 1, 2, and 3 in the previous list. The second section provides an overview of the initial study design for each technique of interest. The initial study design outlines the study objectives, method, scope, and output results. The third section provides an overview of the final study designs for each of the techniques selected for study

16 in Phase 2. The final study design expands on the initial study design by identifying site selection factors, potential test sites, specific database variables, sample sizes, and data collection techniques suitable for quantifying the specified performance relationships. This (third) section describes the results of Activity 4 in the previous list. Assessment of Data and Information Needs A total of 74 AM techniques were identified during the literature review. These techniques were previously shown in Table 3. Most of these techniques are expected to have some effect on the safety or operation of the four travel modes (i.e., pedestrians, bicycles, buses, or trucks) of interest. However, as project resources were finite, the objective of the assessment process was to identify the subset of techniques for which developed performance relationships would collectively provide the information most needed by practitioners for a total cost that was within the project budget. The assessment process included an evaluation of the utility and development cost of new performance relationships for each candidate technique. In this regard, utility describes the value a performance relationship indirectly provides to the traveling public by more reliably informing decision-makers during the project development process. The assessment process was conducted on a technique-by-technique basis. The overall utility associated with a technique is based on consideration of the collective set of associated performance relationships (note: only those relationships that are classified as “Trend” or “Possible” were considered). The assessment on a technique-by-technique basis recognized that an economy-of-scale is likely when research is undertaken to develop multiple needed relationships for given technique (e.g., some of the data collected to develop one relationship may be used to develop another relationship). The assessment process occurred in three stages. The level of evaluation detail increased with each stage, and the number of candidate techniques was incrementally reduced at the end of stages 1 and 2. The staged approach minimized the time required to evaluate a large number of performance relationships while maximizing the likelihood that the most-needed relationships were identified, while staying within the constraints of project resources. The three stages are outlined in the following list.  Stage 1 – Initial selection of techniques. The objective of this stage was to develop and apply a selection procedure that justified reducing the number of candidate techniques from 74 to 20. The 20 selected techniques are identified in Table 4.  Stage 2 – Final selection of techniques. The objective of this stage was to develop and apply a refined selection procedure that justified reducing the number of candidate techniques from 20 to six.  Stage 3 – Detailed assessment of selected techniques. The objective of this stage was to develop a prioritized list of techniques for which performance relationships could be developed in Phase 2. Table 4. Techniques identified as candidates for further study during stage 2. Category Technique (listed by ID code) Primary techniques identified in NCHRP Report 420 1a. Establish traffic signal spacing criteria (A-1-3) 1b. Establish spacing for unsignalized access (A-1-4) 1c. Establish corner clearance criteria (A-1-5) 2a.& 2b. Install non-traversable median on undivided highway and replace TWLTL with non-traversable median (B-3-2, B-3-3, & B-3-4) 2c. Close existing median openings (B-3-5)

17 Category Technique (listed by ID code) 2d. Replace full median opening with median designed for left turns from the major roadway (B-3-6) 3c. Install continuous two-way left-turn lane on undivided highway (B-3-11) 3d. Install U-turns as an alternative to direct left turns (B-3-18) 4a. Install right-turn deceleration lane (B-4-3) 4b. Install continuous right-turn lane 5a. Consolidate driveways (B-5-1-1) 5b. Channelize driveways to discourage or prohibit left turns on undivided highways (B-7-3) 6b. Locate/relocate the intersection of a parallel frontage road and a crossroad farther from the arterial–crossroad intersection (B-2-4) Selected from other techniques identified in NCHRP Report 420 B-3-1 Install median barrier with no direct left-turn ingress or egress B-4-6 Move sidewalk-driveway crossing laterally away from highway B-5-2-2 Require access on collector street (when available) in lieu of additional driveway on highway B-5-2-3 Relocate or reorient access B-6-8 Replace curb parking with off-street parking B-6-10 Install Roundabout B-7-11 Improve driveway sight distance and B-7-12 Regulate minimum sight distance Note: Technique ID codes are referenced to NCHRP Report 420 (Gluck et al., 1999). Additional details of the activities associated with each stage (and the findings that were produced) are described in Appendix C. As described in the subsequent section titled Supplemental Techniques, the prioritized list was revised based on panel guidance, and a final list of techniques was produced. Preliminary List of Techniques As a result of the assessment process, six techniques were identified as having the highest priority based on consideration of their utility and development cost. The expected utility score and the estimated development cost were used to determine the priority rank for each technique. The utility-to-cost ratio was used to define the cost effectiveness of developing relationships for a given technique. The technique with the largest ratio was determined to be the most cost-effective technique, so it was given a rank of 1. The technique with the next highest ratio was given a rank of 2, and so on for the remaining techniques. The ranked list of techniques is provided in Table 5.

18 Table 5. Researcher’s prioritized list of techniques to study in Phase 2. Technique (listed by ID code1) Selected Techniques Rank Expected Utility Cost of Study Ratio3 Total: 76.3 $336,125 B-6-10 Install roundabout2 1 13.7 $45,625 0.30 B-3-1 Install median barrier with no direct left-turn ingress or egress. 2 14.0 $48,125 0.29 1c. Establish corner clearance criteria 3 13.7 $48,750 0.28 6b. Locate/relocate the intersection of a parallel frontage road and a crossroad farther from the arterial–crossroad intersection 4 13.7 $55,625 0.25 4a. Install right-turn deceleration lane or right-turn lane 5 9.3 $45,500 0.21 1a. Establish traffic signal spacing criteria 6 12.0 $92,500 0.13 Notes: 1 – Technique ID codes are referenced to NCHRP Report 420 (Gluck et al., 1999). 2 – Technique added or modified by the NCHRP 03-120 research team. 3 – Ratio = expected utility / cost of study × 1000. Supplemental Techniques The recommended techniques listed in Table 5 were discussed with the project panel at an interim point of the project. The panel identified four additional techniques to be considered for study in Phase 2. The four techniques that the panel identified include:  Two-way left-turn lane (TWLTL) versus restrictive (i.e., non-traversable) median, to include: o Install TWLTL on undivided highway o Install non-traversable median on highway with TWLTL o Install non-traversable median on undivided highway  Replace curb parking with off-street parking  Install appropriate driveway width and radius  Limit access in the functional intersection area (signalized) Each of the four techniques identified by the panel was judged to be very similar to one or two of the original 74 techniques. As a result, the findings from the original assessment process were used as the basis for assessing the four supplemental techniques. The computed utility and development cost for each of the four techniques is listed in Table 6. The techniques listed in this table are not ranked; their order of presentation does not imply any priority of one technique relative to another. The estimated total development cost for the four techniques is $474,625.

19 Table 6. Estimated utility and development cost for four supplemental techniques. Supplemental Technique Original Technique (listed by ID code1) Expected Utility Cost of Study Ratio2 Total: 57.8 $474,625 TWLTL vs. non- traversable median 2a.& 2b. Install non-traversable median on undivided highway and replace TWLTL with non-traversable median 19.3 $214,000 0.09 3c. Install continuous two-way left-turn lane on undivided highway Replace curb parking with off-street parking B-6-8 Rep lace parallel on-street parking with off-street parking. 7.7 $101,250 0.08 Install appropriate driveway width and radius Install driveways with the appropriate return radii, throat width, and throat length for the type of traffic to be served2 3.5 $55,000 0.06 Limit access in the functional intersection area 1c. Establish corner clearance criteria 27.3 $104,375 0.26 6b. Locate/relocate the intersection of a parallel frontage road and a crossroad farther from the arterial–crossroad intersection Notes: 1 – Technique ID codes are referenced to NCHRP Report 420 (Gluck et al., 1999). 2 – Ratio = expected utility / cost of study × 1000. Final Assessment and Prioritization of Techniques A follow-up meeting was convened with the panel to discuss the recommended six techniques listed in Table 5 and the four new techniques in Table 6. This discussion led to the panel’s ranking of these techniques in order of priority for Phase 2 research. The decisions reached by the panel during this meeting are summarized in this section. Based on guidance from the panel, revisions were made to the study design for three techniques. The nature of the revisions is described in Appendix C (in the discussion associated with Table 82). The revised study designs are provided near the end of Appendix D, in the section titled Study Designs for Supplemental Techniques. These revisions also altered the study scope and cost for the three techniques. The computed utility and development cost for each of the nine techniques is listed in Table 7. The estimated total development cost for the nine techniques is $654,275. One technique (i.e., 1c-Establish corner clearance criteria) was common to the two lists. As a result, there were only nine uniquely titled techniques in the combined list. The first column of the table lists the “short title” for each technique. This title was used during the panel meeting. The third column lists the corresponding original title.

20 Table 7. Final prioritized list of techniques considered for study in Phase 2. Short Title Supplemental Technique Original Technique (listed by ID code1) E xp ec te d U til ity C os t o f S tu dy A cc um ul at ed C os t Total: 101.5 $654,275 1. Driveway design Install appropriate driveway width and radius Install driveways with the appropriate return radii, throat width, and throat length for the type of traffic to be served2 3.5 $75,000 $75,000 2. Right-turn deceleration -- 4a. Install right-turn deceleration lane or right-turn lane 9.3 45,500 $120,500 3. TWLTL vs. non- traversable median TWLTL vs. non-traversable median 2a.& 2b. Install non-traversable median on undivided highway and replace TWLTL with non-traversable median 14.0 $93,150 $213,650 3c. Install continuous two-way left-turn lane on undivided highway 4. Corner clearance Limit access in the functional intersection area (Part 1) 1c. Establish corner clearance criteria 13.7 $97,500 $311,150 5. Signal spacing -- 1a. Establish traffic signal spacing criteria 12.0 $92,500 $403,650 6. Median barrier w/no lefts -- B-3-1 Install median barrier with no direct left-turn ingress or egress. 14.0 $48,125 $451,775 7. Replace parking Replace curb parking with off- street parking B-6-8 Rep lace parallel on-street parking with off-street parking. 7.7 $101,250 $553,025 8. Relocate access Limit access in the functional intersection area (Part 2) 6b. Locate/relocate the intersection of a parallel frontage road and a crossroad farther from the arterial–crossroad intersection 13.7 $55,625 $608,650 9. Roundabout -- B-6-10 Install roundabout2 13.7 $45,625 $654,275 Notes: 1 – Technique ID codes are referenced to NCHRP Report 420 (Gluck et al., 1999). 2 – Technique added or modified by the NCHRP 03-120 research team. Further discussion occurred at the panel meeting to identify those techniques for which (1) performance information is most needed by practitioners and (2) the total development cost was nearly equal to the Phase 2 budget of $325,000. After this discussion, a series of straw polls were taken of the panel members to identify the relative priority of the nine techniques. The priorities that were established in this manner are indicated in Table 7 by the order in which the techniques are listed. The technique in the first row of the table was given highest priority. The accumulated cost of the ranked techniques is provided in the last column of Table 7. This column was used to identify the number of technique studies that could be undertaken in Phase 2, given that $325,000 was available for this purpose. Examination of this column indicates that the estimated development cost for the first four techniques nearly equals the available $325,000. It was also noted by the panel and research team that final study designs and associated resource needs for these four techniques may further reduce the number of treatment ultimately examined in Phase 2. Resources were not believed to be sufficient to develop performance relationships for the last five techniques listed in Table 7. These techniques are identified by grey shaded cells in the table. They are described in Chapter 5 as areas that are deserving of future research.

21 Overview of Initial Study Designs This section provides an overview of the initial study design for each technique of interest. The objective of a study design is to describe the process for collecting and analysing the data needed to produce a desired performance relationship. In fact, the study design for a given technique will typically address performance relationships for several combinations of travel mode (e.g., pedestrian) and performance measure category (e.g., safety). The study design was used to estimate the cost of developing relationships for a specified technique. This estimate was then used to compute the cost effectiveness of each technique being considered for development in Phase 2 of the project. The initial study design provided a study description and identified the major elements of the study design. Specifically, the study description summarizes the study objectives, method, scope, and output results. The major elements include the analysis scale, data source, key independent variables, and performance measures. The initial study designs for 20 high-priority techniques are provided in Appendix D. A final study design for each technique studied in Phase 2 was developed at the start of Phase 2. The final study design expanded upon the initial study designs by identifying site selection factors, potential test sites, specific database variables, sample sizes, and data collection techniques suitable for quantifying the specified performance relationships. An overview of the final study designs is provided in the next section. This section consists of two subsections. The first subsection describes the typical spatial scale of application for each AM technique. This descriptor of technique application is used to determine the analysis scale for a study. The second section describes the major elements of a technique’s initial study design. Technique Application Scale Column 11 of Table 3 (shown in Chapter 2) describes the application scale for each of the techniques listed. Three scales are identified: system, corridor, and site. Techniques are considered to be system applications if they specify strategic changes in agency policy (or changes in legal statutes) that are likely to be applied on a system-wide basis (and perhaps are implemented over a long time). Techniques are considered to be corridor applications if they specify a change in the design or operation of a major arterial roadway and this change extends for a significant length of the facility (e.g., access spacing, median type). Finally, techniques are considered to be site applications if they improve access at a specific location on the roadway (e.g., an intersection or driveway), or along a short length of roadway (e.g., a street segment). These techniques are generally tactical in nature and are considered to be site- specific applications. Techniques that are implemented along some portion of the corridor (or at some points) but not others represent a mixture of corridor and site-specific applications. For the purpose of evaluating technique effect, this type of corridor application is more challenging to study for two reasons. First, the variation of technique presence (or level) within the study area will create inconsistencies in operation (as drivers move between segments “with” and “without” the technique). Second, it will likely dilute technique effect making it more difficult to define a statistically valid relationship. In summary, for the purpose of evaluating corridor applications, the technique of interest should be implemented consistently along the length of the corridor being studied, and this study length should extend for a distance that is sufficient for drivers to become accustomed to the presence of the technique (i.e., a mile or more). The initial study design elements described in the next section include the importance of technique application scale when developing the study design. In general, the effect of a technique on safety or operation can be most accurately quantified by matching the study area with the application scale. In this

22 regard, the evaluation of a technique with system-wide scale would likely require the collection of system-level data (i.e., data describing the entire system). In contrast, the evaluation of a technique implemented at a specific site would likely require only the collection of site-specific data. In some instances, a technique can be applied at a site level, but have a secondary impact on the safety or operation of adjacent road segments. In this instance, the study design should consider collecting data at the corridor level to ensure that secondary impacts are also quantified. Initial Study Design Elements This section describes the initial study design elements for a given technique. Initial study designs were developed for each of the 20 techniques selected in Stage 1 of the assessment process (described in Appendix C and listed in Table 74). Each study design describes the data needed to quantify the influence of a technique on safety and operations. The reduction and analysis of these data are intended to lead to the development of a performance relationship for a given travel mode and performance measure category (i.e., safety or operations). The major elements of the study design are described in the remaining subsections. Table 8 lists these elements in a worksheet format that was completed for each technique considered. Table 8. Study design worksheet. Technique: Analysis Scale: (check one) ___Corridor ___Site-specific Performance Measure Category Data Source (check one) Key Independent Variables Travel Mode Performance Measures (list in space provided) Operations __ Field __ Simulation Pedestrian Bicycle Transit Truck Safety __ Field __ Simulation __ Crash reports Pedestrian Bicycle Transit Truck Analysis Scale The analysis scale describes the spatial limits of the study area. The study area is dictated largely by the extent to which a technique has an effect on traffic operations or safety. In this regard, the analysis scale is often dictated by the technique’s application scale, as discussed previously. For this reason, the analysis scale is specified as corridor or site. A technique that is implemented at a corridor level typically requires a corridor-level analysis scale. Data are collected to describe the overall corridor (e.g., annual average daily traffic (AADT), driveway density, signal density, adjacent land use). A technique that is implemented at a specific site typically requires a site-level analysis scale. The data for site analysis describe the site’s traffic, roadway, and crash history (e.g., lane width, bay length, turn movement volume).

23 Study Method for Corridor Analysis Analysis scale can also influence the choice of study method. Two study methods for corridor analysis were considered, they include: before–after, and cross-sectional. Each method is described in the following paragraphs. Before–After Study Method. The before–after method is based on the assumption that there are no changes at the treated corridor or site, except for the implementation of the technique. This assumption is very difficult to satisfy for corridor analyses (unless the data are obtained from a corridor simulation model) because agencies often make many changes to the roadway during a single resurfacing, rehabilitation, or reconstruction project. Even if it were satisfied, the resulting value that describes the change in performance will typically have limited transferability because of the unique nature of each corridor. For example, consider the technique “close existing median openings along the corridor.” A specific corridor of interest is studied using simulation. The only change made is to close the median openings. A 10 percent increase in signalized intersection delay is found. The existing condition included a unique combination of conditions (e.g., number of median openings, signal spacing, land use). The 10-percent increase can only be reliably applied to other corridors with very similar design and traffic conditions; however, there are likely to be very few similar corridors. Cross-Sectional Study Method. The cross-sectional study method is based on a large database that collectively describes the traffic, road, and crash history of several corridors or sites. This method is based on the assumption that the differences in performance among the corridors or sites in the database can be explained by the differences in the measured variables associated with each corridor or site. Thus, if several streets are found that are very similar except for differences in travel speed, volume, and driveway spacing, then they could be evaluated using regression analysis where the two independent variables are volume and driveway spacing. The resulting regression coefficients could be used to quantify the individual effects of volume and spacing on travel speed. The challenge with this study method is to ensure that all influential variables are included in the database and that the corridors or sites in the database are carefully selected to minimize correlation among the independent variables. For example, if the streets used for regression analysis also varied in lane width but this variable was not measured (and if it turns out that streets with wider lanes also tend to have longer driveway spacing), then the regression coefficient for driveway spacing will be biased to include the effect of lane width. As a result, users of the regression model would not obtain reliable estimates of the effect of driveway spacing on speed. If the issues underlying the assumption are minimized through careful corridor selection, the resulting regression model developed from this study should be transferable to other corridors (as can be demonstrated through model validation). Study Method Summary. For the aforementioned reasons, a cross-sectional study method was used for the corridor analysis scale. This study method has the advantage of allowing multiple techniques to be evaluated using a common database. The safety or operational effect of one or more techniques can be quantified using multiple-variable regression analysis. This method was used by Persaud et al. (2011) in their development of multiple-variable regression models for estimating the influence of various techniques on corridor safety. Study Method for Site Analysis The assumption of no changes at the treated site can often be satisfied for before–after studies involving a change to a specific site. Moreover, the resulting value that describes the change in performance will typically have acceptable transferability because it is easier to find similarities among sites. This positive quality stems from the fact that a site (e.g., intersection or segment) is smaller than a corridor and thus has fewer attributes to be considered when searching for similar locations. In other

24 words, it is very possible that a city has many signalized intersections with two adjacent through lanes on each approach and no turn bays. However, any given one-mile length of arterial in a city is likely to be unique in terms of the cross section, driveway spacing, and signal density. To illustrate the before–after study method, consider the technique “install right-turn deceleration lane.” A specific site of interest is studied using crash data. The only change made is to install the turn lane. A 10 percent reduction in crash frequency is found. The existing condition included a unique combination of conditions (e.g., signalized intersection, two adjacent through lanes). The 10 percent reduction can be reliably applied to other intersections that are signalized with two adjacent through lanes, of which there are likely to be many. The cross-sectional study method is also applicable to the site analysis scale for the same reasons that it is applicable to the corridor analysis scale. However, the issues associated with missing variables and correlated variables are also present and must be minimized through careful site selection. Of the two study methods, the before–after study is likely to provide more reliable results because it eliminates extraneous differences among sites, while also eliminating the issue of missing or correlated variables. For the aforementioned reasons, the before–after study method was used as the first-choice study method for the site analysis scale. However, a cross-sectional study method was also considered in those cases where a before–after study could not be conducted cost-effectively within the project schedule. Data Sources There are several alternative sources of data that can be used to quantify most performance relationships. The sources considered for this project include field measurements, simulation, and crash records. The choice of data source is influenced by the technique being studied and the performance measure category (i.e., operations or safety). The data sources proposed for each technique are listed in the last two columns of Table 3 (shown in Chapter 2). The rationale for the proposed sources is described in the following two subsections. Operations Data For operations-based relationships, viable data sources include field measurements and simulation. Field data includes the measurement of traffic, roadway, and control data at one or more locations. Field data are expensive and time-consuming to collect, relative to the budget and schedule of most projects. The costs increase with an increase in analysis scale. Nevertheless, field data are often necessary to collect when the technique being studied is intended to encourage changes in traveler behavior (e.g., use of painted island channelization to prohibit some driveway turn movements). Simulation data is obtained from an appropriate microscopic traffic simulation software product (e.g., VISSIM). Simulation is attractive because (1) the performance data are relatively cost-effective to acquire and (2) it allows the analyst to control changes in the traffic and roadway conditions similar to a before– after study. Several simulation software products can be used to estimate a variety of operational performance measures for the pedestrian, bicycle, transit, and truck travel modes. Simulation models are a viable data source when the evaluation is focused on the influence of traffic control or roadway elements for which travelers are largely responsive and compliant (e.g., changes in intersection traffic control, number of lanes, bay presence, intersection location). Simulation is less viable when traveler response and compliance is less predictable or when it may evolve over time. For the aforementioned reasons, simulation was considered the first-choice data source for (1) techniques requiring a corridor analysis scale and (2) techniques requiring a site analysis scale and for which travelers are largely responsive and compliant. Field measurements were the data source for techniques whose performance is largely influenced by the extent to which travelers are responsive to (and compliant with) the technique’s intended function.

25 Safety Data For safety-based performance relationships, viable sources of data include crash reports, field measurements, and simulation. Crash reports may be either a summary crash record (possibly in electronic format) or a copy of the actual crash report, depending on the information needed to evaluate a specific technique. Field measurements and simulation can be used to quantify the number of vehicle-related conflicts for each travel mode of interest (e.g., vehicle-pedestrian conflicts). The challenge of using conflicts is that their relationship to crash frequency or severity is not typically known. This deficiency is true for the vehicle-related conflicts that are of interest to this project. If conflicts were used, some project resources would need to be expended to investigate (and quantify if possible) the relationship between the measured conflicts and crash frequency or severity. If this connection was established, then the effect of a technique on conflict frequency could be extrapolated to an expected change in crash frequency or severity. Without this connection, the safety effect of a technique is limited to statements about its ability to change conflict frequency. Relationships of this nature are intentionally excluded from the Highway Safety Manual (AASHTO, 2010). Simulation is attractive for the reasons mentioned in the previous subsection. Simulation models predict conflict frequency (e.g., pedestrian-vehicle conflicts) as a surrogate measure of safety. Several software products (e.g., Paramics, VISSIM) output the time–space trajectories of pedestrians, bicycles, buses, and trucks as they move along a street or through an intersection. The FHWA SSAM software can be used to examine trajectory data for conflicts between vehicles and pedestrians, vehicles and bicycles, bicycles, vehicles and buses, or vehicles and trucks (FHWA, 2011). Agarwal (2011) used the FHWA SSAM software to obtain estimates of vehicle-pedestrian conflicts. Crash reports are recognized as the best method for quantifying the effect of a technique on safety (AASHTO 2010). However, the frequency of crashes involving pedestrians, bicycles, buses, and trucks is sufficiently small that it is often difficult to quantify the effect of a technique on safety using only available crash reports. This challenge is particularly true when the technique is applied to an individual site because crashes in general are few in number at most sites. In recognition of this challenge, Persaud et al. (2011) focused their evaluation of safety effect on techniques having a corridor application scale. They found that the frequency of crashes along an entire corridor was sufficient to quantify plausible relationships between corridor-based techniques (e.g., establish traffic signal spacing) and safety (Gross et al., 2013). Based on the preceding discussion, crash data was considered the first-choice data source for all techniques. However, it was recognized that some techniques that are implemented at a point along the roadway (e.g., a driveway, median opening) would not likely be amenable to evaluation using crash data. In these instances, the use of conflict data was considered. The conflict data were to be obtained from simulation if simulation was also the data source for the operations data. Similarly, the conflict data were to be obtained from field measurement if field measurement was the data source for the operations data. Candidate Corridors and Sites The locations of the corridors and sites selected for study were based on the research team’s knowledge of candidate locations, suggestions from the project panel, and the information obtained from the agency survey. The criteria for selecting these locations were established when the final study design was developed at the start of Phase 2. These criteria were envisioned to include consideration of the following factors:  No major construction activity during the study period.  Availability of crash, traffic volume, and road inventory data.  Arterial functional classifications: principal arterial and minor arterial.

26  Area types: urban and suburban.  Primary land uses: residential, industrial, commercial business, and office.  Median treatments: none (undivided), flush paved, and raised curb. Speed was not included in the preceding list because it is indirectly represented by the functional classification and area type factors. High-speed arterials will likely have a principal arterial classification and suburban area type. Flush paved medians include paint-striped medians and two-way left-turn lanes. Key Independent Variables The independent variables were listed for each study design (using column 3 of Table 8). These variables were thought likely to have some association with the safety or operational effect of a technique. The list includes “key” variables for this point in the development of the study design (i.e., the initial study design stage); a key variable was one that was judged to have a strong association with technique effect and to be under the control of the operating agency. These variables were used to guide site selection or simulation model testbed development. The key variables identified for each technique are provided in Appendix D. Performance Measures A range of performance measures were also listed for each study design. These measures were used to evaluate each combination of technique and travel mode. Key considerations for selecting the candidate measures were: (1) the measure is meaningful to the operating agency, (2) the measure can be used to diagnose issues and evaluate alternative techniques, (3) the measure can be cost-effectively applied and monitored, and (4) the measure can be described to administrators and the public. Measures that satisfied these considerations are listed in Table 9.

27 Table 9. Candidate performance measures. Performance Measure Category Travel Mode Candidate Performance Measures Operations Pedestrian Travel speed (corridor or segment) Delay (intersection or driveway) Average pedestrian space (for sites in central business district) Bicycle Travel speed (corridor or segment) Delay (intersection or driveway) Stop rate Transit Travel speed (corridor or segment) Person delay (intersection or driveway) Wait–ride score (corridor or segment) based on HCM, Chapter 17 Truck Travel speed (corridor or segment) Delay (intersection or driveway) Stop rate Safety Pedestrian Vehicle-pedestrian crash frequency, by severity Vehicle-pedestrian conflict frequency Bicycle Vehicle-bicycle crash frequency, by severity Vehicle-bicycle conflict frequency Pedestrian-bicycle conflict frequency Transit Bus-related crash frequency, by severity and type Bus-related crash involvements, by severity and type Bus-vehicle conflict frequency Truck Truck-related crash frequency, by severity and type Truck-related crash involvements, by severity and type Truck-vehicle conflict frequency Overview of Final Study Designs This section provides an overview of the final study designs for each of the techniques selected for study in Phase 2. The final study design expands on the initial study design by identifying site selection factors, potential test sites, specific database variables, sample sizes, and data collection techniques suitable for quantifying the specified performance relationships. The final study designs are provided in Appendix E. Final study designs were implemented for only two of the four high-priority techniques identified in Table 7. The two techniques for which a final study design was developed were right-turn deceleration and TWLTL vs. non-traversable median. A final study design was developed for the driveway design technique. However, the study associated with this technique was abandoned after it was determined that finding a sufficient number of study sites with the needed pedestrian and bicycle volumes would exceed the available time and budget. The Phase 2 work plan recommended delaying the start of the study associated with the corner clearance technique. This strategy was used to maximize the probability of success for the three higher priority studies (i.e., those studies identified in the previous paragraphs). As time passed, challenges associated with these three studies were encountered, which required additional project resources to overcome. As a result, the Corner Clearance study was not undertaken in Phase 2.

28 This section provides an overview of the final study designs associated with the techniques identified in Table 10. The research approach used for a given technique is based on the performance measure category and corresponding data source (i.e., operations or safety). The data source for each technique is listed in the last two columns of the table. Table 10. Final list of techniques for further study in Phase 2. Short Title Supplemental Technique Original Technique (listed by ID code1) Appli- cation Scale (Study Method2) Data Source O pe ra tio ns Sa fe ty Right-turn deceleration -- 4a. Install right-turn deceleration lane or right-turn lane Site (B-A) Simula- tion Simula- tion TWLTL vs. non- traversable median TWLTL vs. non-traversable median 2a.& 2b. Install non-traversable median on undivided highway and replace TWLTL with non-traversable median Corridor (C-S) Simula- tion Crash Reports 3c. Install continuous two-way left-turn lane on undivided highway Notes: 1 – Technique ID codes are referenced to NCHRP Report 420 (Gluck et al., 1999). 2 – Study method: C-S = cross-sectional study; B-A = before–after study. As indicated in column 5 of Table 10, simulation was the proposed data source for the operations relationships. In contrast, column 6 indicates that a combination of crash reports and simulation was proposed for the collective set of safety relationships. Given the fundamental differences between operations and safety performance, and between simulation and crash data sources, three different research approaches are outlined in this subsection. They are identified in the following list and discussed in the next three subsections:  Operations relationships based on simulation  Safety relationships based on simulation  Safety relationships based on crash data Operations Relationships Based on Simulation This subsection provides an overview of the research approach for the development of operations relationships based on simulation data. The concepts discussed in this subsection are common to both techniques listed in Table 10. The VISSIM simulation model was used to obtain the desired performance data. This model has the capability of simulating cars, pedestrians, bicycles, transit, and trucks. The research approach is described as a sequence of activities that lead to the development of a database suitable for calibrating a relationship between technique design (or presence) and operational performance. Performance relationships for pedestrian operations were found in the literature, so the simulation model was used to produce the data needed to estimate relationships for the bicycle, transit, and truck travel modes.

29 Site Selection and Data Collection for Simulation Calibration The focus of the site selection and data collection activities was obtaining the data needed to calibrate the model to replicate bicycle, transit, and truck operation. Initially, a small number of sites (i.e., intersection or segment) were identified to provide the data needed to calibrate VISSIM. Intersection sites were selected for the Right-Turn Deceleration technique. Segment sites were selected for the TWLTL vs. Non-Traversable Median technique. The set of sites included a representative set of street segments bounded by signalized intersections. All sites were known to have moderate to high levels of bicycle, transit, and truck traffic. The segment sites were selected to have either a non-traversable median or a two-way left-turn lane median. Also, the set of sites were selected to include at least one of the following features:  A crossroad between a signalized crossroad-ramp terminal and the adjacent signalized intersection, or  A transit stop at a mid-segment location. The intersection sites were selected to include some locations with a right-turn lane on a major street approach and some locations without right-turn lanes. Similarly, the signalized intersections were selected to collectively include some locations with a driveway adjacent to the major street and within the functional area of the intersection; other signalized intersections had no driveways in the functional area. At least one intersection was selected to include a transit stop on the departure leg of the intersection. Field data were collected at the selected sites for the purpose of calibrating key parameters in the simulation model. One global set of calibration parameters was obtained from the collective set of sites. Experimental Design A key component of the study design is the experimental design. The experimental design describes the independent and dependent variables included in the database that is used to develop the desired performance relationships. The experimental design also describes the number of unique simulation runs formed by the factorial combination of inputs, and the number of replications associated with each unique run. One experimental design was developed and one database was assembled for each technique evaluated. Influential independent variables (IIVs) represent those independent variables likely to have a “practically significant” effect on the performance measure of interest. A variable was considered to be “practically significant” if its effect was statistically significant, of logical direction, and made a tangible change in performance for a typical design application. One or more IIVs were used to describe the geometric design of the subject technique. Other IIVs were identified that were believed to directly or indirectly affect the effect of the technique on performance. Those independent variables that were unlikely to affect performance were set at a representative value for all simulation runs. The variables likely to be considered as IIVs were identified for each technique studied in Phase 2. A range of values was identified and represented in the database for each IIV. Collectively, these input values describe the range of reasonable operating conditions likely to be encountered at sites for which the technique is considered for implementation. The dependent variable values were obtained from the simulation model’s output performance measures. Simulation Testbed The phrase “simulation testbed” is used to describe the simulation model input data file set up for the evaluation of a specific technique and containing the values of the calibration parameters. The testbed

30 describes the traffic characteristics, geometric design elements, and traffic control device features of the site of interest. Selected input variables are considered to be independent variables, as described in the previous paragraphs. The values of the independent variables in the testbed were then changed, in conformance with the experimental design, to form supplemental input data sets. Testbed development was influenced by the study method (i.e., cross-sectional or before–after). Specifically, if a before–after study method was used, then two testbeds were developed. One testbed described the geometric design elements and traffic control features associated with the “before” condition. The second testbed described the elements and features associated with the “after” condition. In contrast, if a cross-sectional study method was used, then one or more independent variables in the testbed were varied in a discrete manner to describe a range in (1) the dimension of a geometric design elements (e.g., increase signal spacing by X feet), or (2) the signal settings (e.g., increase cycle length by X seconds) that describe the technique’s design and operation. Thus, a series of testbeds were established to represent the range of possible dimensions or settings associated with the technique’s typical design and operation. Performance Relationships The performance measures obtained from the simulation model were used to calibrate the identified performance relationships. The exact analytic form of the relationship was dependent on whether interactions exist such that the technique’s effect is influenced by other factors. If no interactions existed, then the technique’s effect on a given performance measure was expressed as a constant (e.g., a delay reduction factor). If a simple interaction existed, then the technique’s effect was expressed as an equation. For example, the effect of right-turn lane installation on pedestrian delay might be influenced by the right- turning vehicle volume. Thus, the relationship would be expressed as an equation (e.g., delay reduction proportion = function of [right-turn volume]). The experimental design used to guide the development of the simulation testbed facilitated the exploration of these interactions and, thereby, the development of transferrable performance relationships. Safety Relationships Based on Simulation This subsection provides an overview of the research approach for the development of safety relationships based on simulated conflict data. The concepts discussed in this subsection apply to the right-turn deceleration technique listed in Table 10. The Surrogate Safety Assessment Model (SSAM) was used to evaluate the vehicle trajectory data output from VISSIM. Specifically, it was used to identify conflicts associated with the relevant travel modes. The research approach is described as a sequence of activities that lead to the development of a database suitable for calibrating a relationship between technique design (or presence) and safety performance. Performance relationships for pedestrian and bicycle safety were found in the literature, so the simulation model was used to produce the data needed to estimate relationships for the transit and truck travel modes. Site Selection and Data Collection for Simulation Calibration The sites used to provide VISSIM calibration data for operations relationships were also used to obtain the SSAM calibration data. The criteria used to select these sites and the data collected are described in the previous section titled Operations Relationships Based on Simulation.

31 The field data needed to calibrate the SSAM software tool included four different types of conflict for each of the transit and truck travel modes. The conflict data were extracted from the video recordings made during the field studies conducted to obtain the VISSIM calibration data. Crash history data were also obtained for each of the calibration sites. These data obtained described all crashes in the most recent five consecutive years at each site. Only those crashes that involved a bus or truck were requested. A copy of the crash diagram and narrative was also requested for each crash of interest. These data were reviewed to confirm their association with the subject site. Assessing Relationship of Surrogates to Safety A surrogate is considered valid for safety-based decisions if it is shown to be correlated to crash frequency or severity. The conflicts extracted from the video recordings were compared with the reported crash frequency on a site-by-site basis. A statistical analysis was used to quantify the correlation between the selected surrogates and crash frequency or severity. The finding of a plausible correlation provided support for the use of conflict frequency to assess the relative safety of various techniques. SSAM Calibration SSAM was not developed to specifically identify pedestrian, bicycle, transit, or truck conflict frequency. However, the VISSIM output files can be screened to include only those trajectories that (1) potentially conflict and (2) involve a travel mode (i.e., pedestrian, bicycle, bus, or truck) of interest. The screened output file is submitted to SSAM and the time-to-collision (TTC) and post-encroachment time (PET) parameters are used to identify the desired conflicts. The TTC is defined as the time distance to a collision of two road users, assuming that their travel direction and velocity is unchanged. PET is defined as the period of time from first road user’s departure from the conflict area to the second road user’s arrival at the conflict area. The TTC and PET parameters were adjusted to calibrate SSAM. To calibrate the TTC and PET parameters, the trajectory files were submitted to SSAM using specified values of TTC and PET. The predicted conflict frequency was then compared with the frequency obtained from field data on a site-by- site basis. The process was repeated for a range of values for TCC and PET. The pair of values providing the best agreement between the predicted and observed conflicts were used to calibrate SSAM. Experimental Design and Simulation Testbed The experimental design and simulation testbed used for conflict simulation was the same as that described in the subsection titled Operations Relationships Based on Simulation, as applied to the right- turn deceleration technique. Performance Relationships The performance measures obtained from the simulation model were used to calibrate the identified performance relationships. The exact form of the relationship was dependent on whether interactions existed such that the technique’s effect was influenced by other factors. If no interactions existed, then the technique’s effect on a given performance measure was expressed as a constant (e.g., a conflict reduction factor). If a simple interaction existed, then the technique’s effect was expressed as an equation.

32 Safety Relationships Based on Crash Data This subsection provides an overview of the research approach for the development of safety relationships based on crash data. The concepts discussed in this subsection apply to the TWLTL vs. non- traversable median technique. A cross-sectional database was used for the development of safety performance measures. This database included data describing many sites with a TWLTL and many sites with a non-traversable median. The database also included the corresponding transit- and truck-related crash frequency (and severity) for each of these sites. The research approach is described as a sequence of activities that lead to the development of a database suitable for calibrating a relationship between technique design (or presence) and safety performance. Performance relationships for pedestrian and bicycle safety were found in the literature, so the database included only the data needed to estimate relationships for the transit and truck travel modes. Site Selection and Experimental Design The types of sites represented in the database were considered representative of typical streets with a TWLTL or non-traversable median. The selected sites collectively represented a range in design dimensions (e.g., median width, number of through lanes). Most importantly, the selected sites collectively represented a range in the independent variables considered to have some influence on crash frequency (e.g., average daily traffic volume, driveway density, etc.). For this study, a site represented one street segment bounded by signalized intersections (with no signalized intersection along the segment length). A large number of sites were needed in the crash database because of the relatively infrequent occurrence of transit- and truck-related crashes. A minimum number of 260 sites were estimated as needed to produce statistically valid results (i.e., 130 sites with a TWLTL and 130 sites with a non- traversable median). Field Data The data obtained for each of the sites included traffic characteristics, geometric design, signal control settings, and transit route operation present at each site. The traffic characteristics and transit operation data were obtained from the local jurisdiction. The geometric design data were obtained from aerial images available on the internet. The speed limit was obtained from the local jurisdiction or from video log information available on the internet. Transit route operation data included average transit headway and average occupancy. Crash history data were obtained for each of the study sites. These data described all crashes in the most recent five consecutive years at each site. Only those crashes that were bus-related or truck-related were requested. A copy of the crash diagram and narrative was requested for each crash of interest. These data were reviewed to confirm their association with the subject site. References Agarwal, N. (2011). Estimation of Pedestrian Safety at Intersections using Simulation and Surrogate Safety Measures. College of Engineering, University of Kentucky, Lexington, Kentucky. FHWA (2011). Surrogate Safety Assessment Model Software. Version 2.1.6 Release Notes. Accessed 12/20/2015 at: http://www.fhwa.dot.gov/downloads/research/safety/ssam/ssam2_1_6_release_notes.cfm

33 Gluck, J., H. Levinson, and V. Stover. (1999). NCHRP Report 420: Impacts of Access Management Techniques. TRB, National Research Council, Washington, D.C. Gross, F., C. Lyon, B. Persaud, and J. Gluck. (2013). “Safety Evaluation of Access Management Policies and Techniques.” Paper No. 13-3921. Presented at the 92nd Annual Meeting of the Transportation Research Board. Washington, D.C. Highway Safety Manual. (2010) American Association of Highway Transportation Officials, Washington D.C. Persaud, B., F. Gross, A. Hamidi, and C. Lyon. (2011). “Safety Effects of Access Management Techniques: State of Knowledge and Recent Research.” Proceedings of the 1st International Conference on Access Management. Athens, Greece. Accessed December 20, 2015 at: http://www.accessmanagement.info.