National Academies Press: OpenBook
« Previous: Chapter 3: Modeling Approaches
Page 22
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 22
Page 23
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 23
Page 24
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 24
Page 25
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 25
Page 26
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 26
Page 27
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 27
Page 28
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 28
Page 29
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 29
Page 30
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 30
Page 31
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 31
Page 32
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 32
Page 33
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 33
Page 34
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 34
Page 35
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 35
Page 36
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 36
Page 37
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 37
Page 38
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 38
Page 39
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 39
Page 40
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 40
Page 41
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 41
Page 42
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 42
Page 43
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 43
Page 44
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 44
Page 45
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 45
Page 46
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 46
Page 47
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 47
Page 48
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 48
Page 49
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 49
Page 50
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 50
Page 51
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 51
Page 52
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 52
Page 53
Suggested Citation:"Chapter 4: Project Database." National Academies of Sciences, Engineering, and Medicine. 2021. Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report. Washington, DC: The National Academies Press. doi: 10.17226/26393.
×
Page 53

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.

24 C H A P T E R 4 - P R O J E C T D A T A B A S E Project Database This chapter summarizes the data collected and assembled for predictive safety model development. The discussion addresses the following topics:  A description of the data collection process.  A summary of the resulting database that was used for modeling.  Findings from a preliminary examination of trends in the crash data. Data Collection Activities The data collection process consisted of a series of activities that culminated in the assembly of a highway safety database suitable for the development of a safety prediction method for freeways and speed-change lanes. It included the following activities:  Develop a data collection procedures guide.  Develop a database dictionary.  Acquire road inventory and traffic volume data from state DOT data sources.  Acquire crash data from state DOT data sources.  Merge road inventory, crash, and traffic volume data into a highway safety database. This section includes one subsection to summarize the steps taken for each of the activities included in the previous list. Development of Data Collection Procedures Guide The Data Collection Procedures Guide for this project was developed to provide instructions to individuals collecting roadway geometric data. The document includes instructions for use of a macro- powered MS Excel tool that processes .kml Google Earth files containing placemarks and folders named and sorted in a specific manner (as described in the guide). Use of this Google Earth/MS Excel system preserves the locations from which measurements were made to enable verification by senior staff as well as adjustments in the course of the project if needed. The application of this system also enhances the usefulness of the database to subsequent research endeavors. This system was previously developed for NCHRP Project 17-45 and modified for this project to enable data collection in one direction and measurement of PTSU-specific geometry. The Data Collection Procedures Guide also includes a section on the quality assurance/quality control (QA/QC) process that was used by senior researchers to verify the work of other team members. The QA/QC section describes features incorporated into the Google Earth/MS Excel system to minimize the likelihood of incorrect data values. The Data Collection Procedures Guide also provides a means for future researchers to collect data in a manner consistent with the 17-89 research team, should they wish to expand this project’s dataset. The Data Collection Procedures Guide is provided in Appendix A.

25 Development of Database Dictionary The Database Dictionary lists variables in the Project 17-89 dataset and provides a descriptive variable name, format of the variable value, type of data (integers, characters, etc.), method of measurement, dimensional units (when applicable), potential values and meaning of values for more complex variables, and the types of sites for which the variable is collected. The Database Dictionary was updated throughout the project. It is provided in Appendix B. Assembly of Road Inventory Data A six-step process was undertaken to identify facilities, define evaluation periods, locate study sites, and collect data for each of the study sites. These six steps are identified in the following list. 1. Identify PTSU facilities and comparison facilities for study. 2. Define the evaluation period. 3. Geo-locate each site. 4. Acquire and summarize volume data. 5. Acquire and summarize hours of operation data. 6. Measure geometric elements for each site. Details regarding each of these steps are provided in the following sections. Early steps were undertaken in Phase I of the project, and later steps were conducted in Phase II. Step 1 – Identify PTSU Facilities and Comparison Facilities for Study The research team had a list of known PTSU and bus-on-shoulder (BOS) facilities from prior research efforts. The list was updated during this step to reflect recent changes. In general, many states with BOS facilities are part of the FHWA’s Highway Safety Information System (HSIS). As a result, crash data for the BOS dataset was obtained from HSIS. Additionally, most BOS facilities were opened in the 2000s or early 2010s, and there was a high likelihood that 5 years of crash data would be available. Ohio and Minnesota were chosen for BOS analysis. Minnesota has approximately as many miles of BOS as the rest of the United States combined and has a PTSU facility that could be included in the PTSU dataset, so it was a natural choice. Ohio was chosen because it has BOS facilities implemented in the 2000s (and thus their facilities are old enough to have 5 years of crash data), has both right-side and left-side BOS facilities, and has BOS facilities in multiple cities. In general, states with PTSU facilities are not part of the HSIS. As a result, crash data for these facilities was obtained from various state departments of transportation (DOTs). PTSU facilities opened in 2017 or later were excluded because of the likelihood that insufficient crash data would be available. State DOTs were contacted to assess the viability of obtaining crash data of sufficient quality to meet the project’s modeling needs. Colorado, Georgia, Hawaii, Minnesota, and Virginia were ultimately selected, although data collection activities in Colorado were later stopped (for reasons described later) and Colorado data were not used for model development. The research team chose not to collect data from other potential states with PTSU facilities opened prior to 2017. Massachusetts DOT indicated that the milepost locations of freeway crashes in their data are not reliable and tend to reflect the milepost of the nearest gore point rather than the crash location. It was unclear if New Jersey DOT would be able to provide the crash data necessary for research purposes. The state of Washington (prior to 2017) and Texas each had one relatively short facility, and the key characteristics of both facilities (side of shoulder used for travel, static versus dynamic hours of operation, and static versus dynamic signs) were already reflected in the database available from other states.

26 For each PTSU and BOS facility, a comparison facility was identified. The research team strove to select comparison facilities that were similar to study facilities in terms of the number of lanes, spacing and complexity of interchanges, role in freeway network (such as radial route versus beltway), and travel patterns served (such as inbound/AM peak period versus outbound/PM peak period and commuter- oriented route versus intercity route). Table 5 presents the PTSU and BOS facilities and comparison facilities selected for study; lengths presented in the table are approximate. Table 5. Study facilities. PTSU Type Location PTSU/BOS Facility Comparison Facility Corridors One-Way Length (miles) Corridors One-Way Length (miles) S-, D-PTSU Alpharetta, GA GA 400 9.3 GA 400 3.9 S-, D-PTSU Virginia Beach, VA I-264 8 I-264 6.8 S-, D-PTSU McLean, VA I-495 NB 2.1 I-495 SB 1.8 S-, D-PTSU Honolulu, HI I-H1 6.9 I-H1 3.4 S-, D-PTSU Minneapolis, MN I -35W 1.9 I-35W 1.3 S-, D-PTSU Fairfax County, VA I-66 12.3 I-66 13.9 S-, D-PTSU Gwinnett Co., GA I-85 NB 1.6 I-85 NB 1.7 PTSU Total: 42.1 32.8 BOS Minneapolis, MN I-35W, I-94, US 169 33.5 MN 610, US 52, I-35W 16.1 BOS Columbus, OH I-670, I-70 7.3 I-70, I-71 6.3 BOS Cincinnati, OH I-71 19.4 I-275 7.6 BOS Total: 60.2 30.0 NB = northbound; SB = southbound Step 2 – Define the Evaluation Period The evaluation period of each facility was dictated by the most recent year of crash data available in each state and the opening date of PTSU and BOS facilities. Ohio and Minnesota had data available through 2015, and BOS facilities in these states were open prior to 2011, so 2011–2015 was used as the analysis period for BOS in both of these states. The PTSU facility in Minnesota also opened prior to 2011, so that 2011 to 2015 data was also used. For PTSU states, the evaluation periods varied. Hawaii had data through 2016 available, and PTSU facilities opened prior to 2012, so 2012 to 2016 was used as the analysis period. PTSU operation was removed on I-66 in 2018 as part of a major widening project. Crash data from 2013–2017 were used for Virginia; this provided a 5-year study period for I-66 and I-264 (PTSU operation and comparison portions) and an approximately 2.5-year study period for I-495 (PTSU operation and comparison portions). PTSU operation on I-495 opened on July 7, 2015. Georgia had crash data through the early months of 2019 available but indicated that data for recent months are sometimes incomplete. Crash data from years 2014 to 2018 were therefore used for Georgia, but less than 5 years of data were used for some portions of some facilities, as necessary. I-85 (PTSU portion and comparison portion) was analyzed from July 1, 2015 (opening date of PTSU operation) through December 31, 2018. Different portions of GA 400 were analyzed for different years, corresponding to when PTSU operation was open.

27 Step 3 – Geo-locate Each Site The research team collected the majority of the project’s geometric data from Google Earth. This provided the research team with a high degree of confidence in the values of the geometric variables. The basic process for geo-locating sites was as follows:  Identify known milepost locations on the freeway. HSIS provided ArcGIS shapefiles for Minnesota and Ohio that could be viewed in Google Earth and indicated the starting mileposts of HSIS segments. This project used different segmentation criteria than HSIS, so HSIS segments were not used for analysis. However, the HSIS mileposts were used to establish known mileposts along the freeway. Virginia DOT had similar shapefiles online that were used to establish Virginia mileposts. For Hawaii and Georgia, mileposts were located in Google Streetview. Hawaii DOT also provided a listing of mileposts at features such as bridges and gores, and these mileposts were consistent with the mileposts located in Streetview. Virginia shapefiles indicated mileposts were different (typically by several tenths of a mile) on each direction of the freeway, so different milepost systems were established for each direction; a single milepost system for both directions of the freeway was used in other states.  Create the “alignment” of a freeway being studied. The alignment was created by placing a series of Google Earth placemarks (pins) along a consistent reference line, specifically the left edge line of the freeway. These placemarks identify features that will define the start and end of sites, including the start and end of curves, the start and end of speed-change lanes (i.e., gore and taper points), changes in lane width, changes in shoulder width, and the start and end of PTSU operation.  Identify site boundaries. The segmentation criteria used for this project include the segmentation criteria for the freeway procedure in the Highway Safety Manual Supplement (HSM Supplement) (AASHTO 2014) plus one additional criterion. In this project, site boundaries were also established at the start and end of horizontal curves. As a result, each site for this project is entirely on a tangent section or entirely on a curve section.  Assign a milepost to the start and end of each site. Milepost assignments were based on the distance along the alignment from a known milepost that was established in the manner described in the first step.  Collect geometric data for each site. Data were collected using a combination of manual observation using Google Earth measurement tools and further pin placement. This process is further described in Appendix B of the NCHRP Project 17-45 final report (Bonneson et al., 2012a). Some modifications to the process have been made for this project to account for single- direction modeling and PTSU-specific geometry. Specific Google Earth pin placement is further described in the Data Collection Procedures Guide. Step 4 – Acquire and Summarize Volume Data This section describes the volumes assembled for the study sites and the process used to assemble this data. For both the BOS and PTSU sites, one-way freeway annual average daily traffic (AADT) volume and “nearest ramp” AADT volume data were determined for each site. For the sites with PTSU operation, a representative 24-hour distribution of the AADT volume was also obtained for representative weekdays, Saturdays, and Sundays. These volume data were obtained for each of the 5 years for which crash data were provided by the state agency.

28 The representative 24-hour volume distributions were obtained using data from the permanent automatic traffic recording (ATR) station nearest to each site. The nearest ATR was typically within 10 miles of the study sites. AADT Volume Computing Process. The freeway AADT volume data were obtained from a combination of permanent ATR stations and short-term count stations. The latter stations tended to be located every 1 or 2 miles along the freeway. These short-term count stations were often located between successive interchanges. The process for computing the freeway AADT volume for each site in the database was to use the reported freeway AADT volume nearest to the site. Given that each site represented one direction of travel, the reported one-way AADT volume associated with the direction of travel was selected when available. If the agency reported only two-way AADT volumes for the freeway, the volume in each travel direction was assumed to equal one-half of the two-way AADT volume. In those instances where one or more ramps were located between a site of interest and the nearest station, the ramp AADT volume was used to adjust the freeway AADT to obtain a more accurate estimate of the freeway AADT beyond the ramp. This process is illustrated in Table 6, which illustrates the freeway AADT volume estimation process for sites that have one or more ramps located between the count station and sites of interest. The map shown in the right column of the table corresponds to a short section of GA 400. Table 6. Example one-way AADT volume calculation (GDOT 2020). Milepost Ramp AADT (veh/day) Increasing Direction AADT (veh/day) 10.734 0 87,000 (=174,000/2) 12.008 −15,700 (exit) 71,300 12.340 +11,300 (entrance) 82,600 Milepost Ramp AADT (veh/day) Decreasing Direction AADT (veh/day) 10.734 0 87,000 (=174,000/2) 11.532 −13,000 (entrance) 74,000 11.847 +11,900 (exit) 85,900 Source: https://gdottrafficdata.drakewell.com/publicmultinodemap.asp veh/day = vehicles per day The milepost locations increase from the bottom to the top of the map presented in Table 6. The count station, which is located at milepost 10.734, is shown as a blue dot at the bottom of the map with a two- way AADT volume label of “174,000.” This value is appropriate for all sites north of this station, up to a

29 point just before the interchange ramps shown near the top of the map. The northbound and southbound directions are both assumed to have a one-way AADT volume of 87,000 vehicles per day (veh/day) (= 174,000/2). Proceeding in the northbound (increasing milepost) direction, the loop exit ramp at milepost 12.008 is shown to have an AADT volume of 15,700 veh/day. This value is subtracted from the value of 87,000 to yield a one-way AADT volume 71,300 veh/day. Any site located between milepost 12.008 and the next ramp to the north is assumed to have a one-way AADT volume of 71,300 veh/day. Proceeding farther north, the entrance ramp at milepost 12.340 has a volume of 11,300 veh/day. It adds to the current AADT estimate to produce a one-way freeway AADT volume of 82,600 veh/day. All northbound sites beyond milepost 12,340 are assumed to have an AADT of 82,600 veh/day. AADT Volume Acquisition Challenges. Representative ramp AADT volumes are available for most states included in the database. However, ramp volumes were not available from the state of Hawaii. As a result, these volumes were estimated using the default values offered in Table 5 of the ISATe User Manual (Bonneson et al. 2012b). These default values are based on consideration of the number of ramp lanes, area type, and the adjacent freeway AADT volume. Interstate H1 (I-H1) in Hawaii contains right side PTSU operation (primarily right side eastbound), left side HOV lanes separated from general-purpose lanes by paint, and a “zipper lane” system in which a movable barrier is used to convert some westbound lanes to eastbound travel during the weekday AM peak. The zipper lane system is concurrent with a majority of the PTSU operation mileage. It differs from other types of HOV lanes because it is barrier-separated and only operational during some hours of the day. To remove the effect of zipper lane presence from the data, the crashes in the zipper lane were excluded from the crash data used for model development. Similarly, the zipper lane volume was subtracted from the freeway AADT volume. The Minnesota ramp volumes were only available in the form of daily raw detector data from a Minnesota DOT (MnDOT) web source (MnDOT, 2020). The data were obtained for every day of the 5- year evaluation period. They were then summarized using software code to obtain representative ramp AADT volumes for each year of interest. HSIS AADTs for Minnesota freeways were found to differ from values available from the MnDOT web source. Based on a comparison of these values with those from the MnDOT web source, the researchers concluded that those from the web source are more reliable. This determination was made primarily based on the changes in freeway AADT volume that occurred before and after ramp locations. The volumes from the MnDOT web sources are more consistent with the kinds of volume changes expected when the process described in the previous section is applied. As a result, the research team also downloaded and summarized the daily raw detector data from the MnDOT web source to compute the freeway AADT volumes for each study site. Step 5 – Acquire and Summarize Hours of Operation Data The status of the shoulder (open or closed to traffic) on PTSU facilities was obtained from each state. These data will be used for sub-annual models comparing periods of time when the shoulder is open and when the shoulder is closed. S-PTSU facilities, by definition, have fixed hours of operation, and the data were generally straightforward to collect. The shoulders on I-264 in Virginia, for example, are open to traffic from 6 AM to 8 AM westbound and 4 PM to 6 PM eastbound on weekdays, except for federal holidays, and are not open on weekends. The hours of operation of GA 400 were more complex than other S-PTSU facilities. GA 400 has six different segments in which the hours of operation (or start and end date of PTSU operation) varied. Additionally, the hours of operation on GA 400 have varied over the years, and Georgia DOT has modified signs displaying the hours. Different individuals and news media accounts provided consistent information on the dates of PTSU operation but gave varying information on the hours of operation. Ultimately, the research team located signs depicting hours of operation on GA 400

30 and used historical Google Streetview images to determine how the hours varied over the years. Generally, after initial implementation, the duration of shoulder use was increased during the peak period in the peak direction and/or the shoulder was also opened in the off-peak direction during a second peak period. D-PTSU facilities, by definition, have variable hours of operation. MnDOT provided a log of lane control signal displays for I-35W, and Virginia DOT provided a partial log of I-66 lane control signal displays for I-66 (after conversion from S-PTSU to D-PTSU operation). Data from some months were unavailable for I-66. Therefore, for each day of the week in these months without data, hours of operation were assumed to be equal to the average hours of operation for that weekday during the months in which data were available. Generally, the shoulder is open or closed for the entire length of a PTSU facility in one direction. However, sometimes portions of the shoulder are closed if they are blocked by incidents, and sometimes portions of the shoulder may be opened in low-volume periods to maintain capacity when general-purpose lanes are blocked by an incident or maintenance activities. Two facilities were converted from S-PTSU to D-PTSU operation during the study period. On September 15, 2015, I-66 was converted from S-PTSU operation to D-PTSU operation and shoulders were frequently opened in the off-peak direction and on weekends. In 2017, Georgia DOT began occasionally opening the shoulder on I-85 outside of the 3 PM to 7 PM weekday time period. Georgia DOT was unable to provide data on specific days and times in which this was done, so the hours of operation were assumed to be static on I-85 for the entire study period. The status of the shoulder (i.e., open or closed) for each weekday and weekend hour are presented in the latter section of this chapter titled Hours of Operation. Step 6 – Measure Geometric Elements for Each Site As described in the previous sections, Google Earth placemarks were created and assigned milepost values corresponding to the start and end of each site. A .kml Google Earth file containing these placemarks was then generated for every site and populated with additional placemarks named and stored in specific formats. For example, a series of placemarks were placed on a cross-sectional line at the start of the site to measure right shoulder width, lane width, left shoulder width, median width, and the distance to the face of the median barrier, if present. The distances between placemarks were computed with MS Excel macros. If a site was greater than 0.05 miles long, a second set of cross-sectional placemarks were placed at the end of the site and the average of the two values (shoulder widths, lane widths, etc.) was used in the project’s database. Other geometric variables were collected manually or computed in a spreadsheet. For example, the number of lanes on ramps was simply observed in aerial photographs, and distance from the end of a PTSU site to the end of the PTSU facility was computed in a spreadsheet using the end milepost of the site and the end milepost of PTSU operation. The number of variables measured and different measurement techniques (e.g., placemark placement, manual measurements, etc.) exceeds what can be committed to short-term memory and what can accurately and efficiently replicated from site to site. Because of this, variables were grouped into small subsets and analysts collected data for just the subset variables for all sites before moving onto another subset of variables. For example, outside (roadside) barrier data measured with Google Earth placemarks and MS Excel macros were collected in one subset. Rumble strip presence was collected in a different subset through Google Streetview. The full list of geometric variables collected, and collection techniques can be found in the Database Dictionary and, more concisely, in Table 1 of the Data Collection Procedures Guide.

31 Assembly of Crash Data This section describes how each state’s crash data were processed into a consistent format. The following variables were assigned to each crash using information in the state’s databases:  State  County (for Ohio only, where mileposts reset to zero at each county line)  Crash ID number  Route  Direction of freeway  Milepost  Year  Month  Day  Hour  Minute (except for Ohio, which did not provide this data)  Type (single vehicle or multi vehicle)  Severity (K, ABC, or O for Hawaii because Hawaii’s data only provided three severity levels; K, A, B, C, or O for other states)  Construction presence Many of these variables, such as route and year, matched fields in each state’s data and were trivial to assign. Techniques used by the research team to assign the direction of freeway and milepost varied from state to state and are described in the following section. Crash Location Assignment This section describes, on a state-by-state basis, how the direction of freeway and milepost were assigned to crashes. Colorado. Colorado DOT provided the research team with crash data in tabular (MS Excel) format. The location of crashes is described with a milepost. Latitude/longitude values were also provided, but Colorado DOT indicated these are assigned by the DOT based on milepost when processing the data, as opposed to being captured by police in the field; this was consistent with the research team’s finding from a review of a sample of the crashes. A review of the data by the research team found that some crashes are located to the nearest 0.01 mile or 0.1 mile, but approximately 31 percent of crashes are coded only to the nearest mile (Milepost XXX.00). To a lesser degree, the percent of crashes coded to the nearest 0.1 mile was greater than would be expected as well. The research team discussed this issue with Colorado DOT, which indicated it is a known issue with their freeway crash data that has existed for many years and there is no means for more precisely locating the majority of these crashes. Colorado DOT staff indicated they review individual police reports and assign more precise mileposts when possible, and they felt there would be limited value in a review of police reports by the research team. The research team also reached out to a Colorado- based firm that has previously done safety modeling of Colorado freeways and they confirmed the existence of this issue. The research team concluded that the resolution of mileposts assigned to crashes was not sufficient for modeling in Project 17-89. Data collection activities for Colorado were stopped, and data from Colorado were not included in the project dataset. Georgia. Georgia DOT provided the research team with crash data in GIS format. The location of crashes is described with latitude and longitude. Mileposts are provided for some crashes, but Georgia DOT indicated a lower degree of confidence in the milepost values, when present, than latitude/longitude. Georgia DOT indicated crashes are assigned a latitude/longitude corresponding to the location of the

32 police officer’s car at the time the report was filed. GDOT staff cleaned these data to remove crashes associated with ramps and crossroads. Prior to providing the data to the research team, Georgia DOT staff also identified and removed data for crashes where I-85 or GA 400 is listed as the route on which the crash occurred, but the latitude/longitude are located off of the freeway. In these cases, the latitude/longitude typically corresponds to a parking lot near an interchange and indicates the police officer filed the report from a parking lot after departing the scene of the crash, and Georgia DOT indicated they do not have a means of locating these crashes at specific locations on the freeway. The travel directions of the first two vehicles involved in the crash were assigned by Georgia DOT in the “DirVeh1” and “DirVeh2” fields. GA 400 and I-85 are both signed as north/south roads. Less than 1 percent of the crashes had no direction, an east or west direction, or different direction assigned to each vehicle. The research team manually assigned a direction to these crashes based on the side of the freeway onto which they were geocoded. About 9 percent of the crashes appeared to be geo-located on a portion of the roadway not corresponding to their travel direction (i.e., a northbound crash was geo-located on the southbound lanes). The majority of these crashes are in one of the clusters (discussed later) that were ultimately excluded from analysis for other reasons. For these 9 percent of crashes, the route field was reviewed. Police sometimes indicate the direction in this field (such as “I-85 NB”, “85N”, or “Interstate 85 North”), and if the direction differed from the DirVeh1 and DirVeh2 direction, the direction was manually changed to match where the crash was geo-located. This change was made to 3 percent (one-third of the 9 percent) of the crashes. So, the direction was not changed for the vast majority of these crashes. About 1 percent of the crashes appeared to be geo-located off the mainline freeway (typically on ramps or overpasses) and were removed from the dataset. The majority of these crashes had one or more attributes suggesting they were not located on a mainline freeway, such a “ManvrVeh1” value of “Turning Left” or “Turning Right,” or a “Route” value suggesting the crash was not on the mainline. The data were found to have several clusters of crashes, where many crashes had the exact same latitude and longitude assigned to them. The milepost, when reported, is the same for all of the crashes in a cluster, but the route description suggests the crashes are in different locations (such as route descriptions of “I-85 NB N of Jimmy Carter Blvd” and “I-85N South of Jimmy Carter” within the same cluster). The crashes had different “AccidentNo” values and different times and dates. Crashes in clusters are 14 percent of the crashes in the Georgia data and were disproportionally located on I-85. Based on the “route description,” it is believed these crashes were generally in or near the study area (as opposed to many miles away in a different portion of the state) but not at the exact location to which they are geocoded. Georgia DOT was contacted about these clusters, and they in turn discussed them with the vendor that manages their crash database. Neither Georgia DOT nor the vendor was previously aware of this issue, but upon investigating it, they indicated crashes in the clusters appeared to be from certain police agencies and the geocoded location is not correct. Georgia DOT was unable to provide a means of correctly locating these crashes, and as a result they were removed from analysis. The research team developed a script to assign mileposts to Georgia crashes based on (1) the latitude and longitude of mileposts used for segmentation and collection of geometric data and (2) the north/south and east/west distances from known milepost coordinates to crash coordinates. The results of the milepost assignment were visually verified in Google Earth by viewing the latitude and longitude of crashes, with mileposts assigned, and comparing them to the mileposts created for geometric data collection. This assignment enabled crashes to be assigned to sites as described in the next section. Hawaii. Hawaii DOT has provided the research team with crash data in tabular format. The location of crashes is described in three ways:

33  A milepost with a resolution of 1/10th of a mile  A reference distance/unit/direction/name (such as “700 ft w of wb off>kunia sb,” meaning 700 feet west of the westbound off-ramp to Kunia southbound)  A latitude/longitude Hawaii DOT staff indicated that the milepost values are approximate, and the reference distance/unit/direction/name and the latitude/longitude are more reliable. The research team reviewed a sample of data and found that locating a crash using the reference distance/unit/direction/name generally resulted in a very similar crash location as the latitude/longitude. In some locations, though, it was challenging to establish the location from which reference distances were being measured (such as “wb off>kunia sb” in the example above). The research team assigned mileposts to crashes based on the latitude and longitude and used the same script that was developed for Georgia crashes. The direction of the freeway on which a crash occurred was assigned to all crashes by Hawaii DOT in the “Street Highway” field. No other directional information was provided, and the geocoded locations placed crashes on the direction of the freeway indicated with the “Street Highway” field. Geocoded locations also did not place crashes on ramps or crossroads. A portion of Interstate H1 being studied in NCHRP 17-89 has a zipper lane on the westbound side of the roadway. The lane normally serves westbound traffic, but during the AM peak, a movable barrier and median crossovers enable eastbound traffic to use the lane. The research team excluded this lane from analysis. Geometric data were not collected from this lane, and volume it in was excluded from the directional volumes assembled for the Project 17-89 dataset. Hawaii DOT provided a list of crashes that occurred in the zipper lane, and these were excluded from the dataset. Minnesota. HSIS provided the research team with Minnesota crash data in tabular format. The location of crashes is described with the “milepost” field. The direction of the freeway on which the crash occurred is provided with the “trvl_dir” field. Attributes for this field are: N, S, E, W, and Z, where Z indicates an unknown direction. Approximately 5 percent of crashes had an unknown direction, and these crashes were excluded from the dataset. No other variables indicating directionality were available. The research team reviewed values in several fields to assess if ramp crashes were included in the dataset. All crashes had a “DIV_CODE” value of Freeway-Mainline or Other Divided Highway, and all crashes had a blank value for “RAMP.” The research team concluded that few or no ramp crashes were in the dataset. Ohio. HSIS provided the research team with Ohio crash data in tabular format. An “acc” file has one record per crash, and a “veh” file has one record per vehicle. Both files have a “caseno” variable that is unique to each crash and was used to link the files. The HSIS guidebook states that approximately 10 percent of the crashes in Ohio’s crash database could not be matched with segments, and HSIS deleted these crashes in the course of assembling their databases because there was not means of locating them. The location of a crash is described with the “milepost” field. The research team reviewed values in several fields to assess if ramp crashes were included in the dataset and determined that this likely occurred in some cases. Specifically, approximately 10 percent of the crashes with a value in the “loc_type” field of 1 (Intersection), 2 (Intersection-related), 3 (Driveway Access), or 8 (Private Property) were removed from the dataset. A review of other fields for crashes with these “loc_type” values indicated a much greater proportion of characteristics associated with intersection crashes—angle crashes, pedestrian crashes, and vehicles travelling in different directions—than in the dataset as whole. A review of the dates, times, and mileposts of crashes revealed that approximately 1.5 percent of crashes occurred at the same date and time (hour) as other crashes. Although the “caseno” variable associated with each crash was unique, the 1.5 percent rate of occurrence was approximately double that found in other states, raising the possibility they were duplicate records of the same crash rather than secondary crashes. The value for each of the 41 attributes in Ohio’s crash database was compared for the potentially duplicate crash pairs. Approximately half of these pairs had identical values for 38 or more of

34 the attributes. One of the crashes in each of these pairs was removed from the dataset, as it was assumed to be a duplicate pair. I-71 in Warren County, a portion of which is being studied for this project, had approximately 3 percent of its crashes coded at milepost 0.0, which is outside of the study area. Crash frequency on study sites on I-71 in Warren County appeared to be approximately 3 percent lower than on other sites in other counties. Other studied routes in other counties (I-71 in Hamilton County, I-275 in Hamilton County, and I-70 in Franklin County) had less than 0.05 percent of crashes assigned to milepost 0.0. Ohio’s data do not include a field indicating the direction of the freeway on which the crash occurred, but it does provide the direction each vehicle involved in the crash was coming from and going to. The research team established facility-specific rules to assign a direction to each crash based on the orientation of the facility within the study area. These assignment rules are described in Appendix C. A small percentage of crashes—about 2 percent of the total–were removed from the dataset because their directionality could not be determined. Virginia. VDOT crash data are available to the public on a website and can be downloaded in geographic information system (GIS) format. The statewide database was initially filtered to crashes where the “Route_Or_Street_Nm” was some variation of I-66, I-495, or I-264, the geographic area was larger but roughly equivalent to the study area, and the “Rte_Category_Cd” was IS and not ISRMP (i.e. the crash was on an Interstate, not an Interstate ramp. The location of crashes is described with a milepost field named “Coted_Mp”. Crashes also have latitude/longitude values. Crashes at the same centerline location (when mapped with the latitude and longitude) but in opposite directions of travel appeared to have different mileposts, with differences in the range of several tenths of a mile. These differences, and the crash mileposts themselves, were generally consistent with directional Virginia shapefiles downloaded from the Virginia DOT website. The use of different milepost systems for each direction of Virginia freeways enabled crashes to be accurately merged with analysis segments, as described later in this report. While the mileposts of most crashes were generally consistent with the directional shapefiles, the plotting of crashes based on latitude/longitude revealed a small percentage of crashes with mileposts that were inconsistent with the shapefiles and other crashes. Differences up to several tenths of a mile were observed, and a review of a sample of the crash narratives suggested the latitude/longitude was a more reliable indicator of a crash’s location than the assigned milepost. The research team reassigned mileposts to Virginia crashes based on the latitude and longitude and used the same script that was developed for Georgia crashes. The “Cotedrouteid” field provided the direction of the freeway on which the crash occurred, and all crashes have a known direction. The values in this field were generally consistent with directional values provided for individual vehicles involved in the crash. Further investigation was undertaken to determine if the dataset included crashes not on the mainline freeway. Approximately 2 percent of the crashes had a “Relation_To_Roadway” value, which suggested the crash was on a ramp, C-D road, crossroad, or ramp terminal intersection. The locations of these crashes were imported into Google Earth using the latitude and longitude and reviewed. About 30 percent (of the 2 percent) of these crashes appeared to be located off the mainline freeway and were removed from the analysis. Study Years A study period was established for each facility in each of the five states and was ideally 5 years in length. The study period was defined by a start date and an end date. The dates established for the start and end of the study period were subject to the following criteria:  For those sites with PTSU or BOS operation, the study period had to bracket the range of years during which PTSU was operating with regularity.

35  For those facilities that served as a comparison facility to a PTSU or BOS facility, the study period was desirably the same as that established for the nearby PTSU or BOS facility.  The study period should be at least 1 year in duration and desirably 5 years in duration.  The end date of the study period should coincide with the most recently available data from the associated state DOT. The following exceptions applied to these criteria: – Georgia and Virginia provided partial 2019 data but indicated that, even for the months provided, the data might be incomplete. As a result, the 2019 data were not used for this project. – 2018 data were also not used for Virginia because one of the three Virginia facilities (I-66) had PTSU operation removed during 2018 when a major widening project was initiated.  The study period should exclude time periods where long-term construction activities were present. Using the aforementioned criteria, the study period dates shown in Table 7 were established. The study period at the collective set of facilities ranged from 1 to 5 years in duration. For some facilities, the study period varied along the length of the facility because PTSU operation was implemented at different dates along the facility. Multiple study period dates are shown in the table for these facilities. Table 7. Study period dates by state and facility. State Facility Number of Sites Study Period Study Period Duration (years) Start Date End Date Georgia GA 400 GA 400 29 23 6/15/2015 1/1/2014 12/31/2018 12/31/2018 3.55 5.00 I-85 11 7/1/2015 12/31/2018 3.50 Hawaii I-H1 52 1/1/2012 12/31/2016 5.00 Minnesota S.R. 610 20 1/1/2011 12/31/2015 5.00 U.S. 52 41 1/1/2011 12/31/2015 5.00 U.S. 169 95 1/1/2011 12/31/2015 5.00 I–35W 81 1/1/2011 12/31/2015 5.00 I–94 26 1/1/2011 12/31/2015 5.00 Ohio I–70 69 1/1/2011 12/31/2015 5.00 I–71 I-71 I-71 66 5 4 1/1/2011 1/1/2011 1/1/2015 12/31/2015 12/31/2013 12/31/2015 5.00 3.00 1.00 I–275 26 1/1/2011 12/31/2015 5.00 Virginia I–66 92 1/1/2013 12/31/2017 5.00 I–264 63 1/1/2013 12/31/2017 5.00 I–495 25 7/7/2015 12/31/2017 2.49 Construction Presence Construction presence was incorporated into the database to enable the research team to screen for sites where long-term construction may have been present and not detected in the initial screening of historical aerial photos conducted during site selection and segmentation. Sites with a high frequency of construction-related crashes, particularly in the same year, were reviewed on an individual basis in Google Earth historical imagery and historical Google Streetview. If changes between images or construction activities visible in images led the research team to suspect the presence of long-term construction, the site was removed from the database or the period of analysis was modified to exclude the year(s) in which long-term construction was present. The research team did not remove sites or years

36 of study from the dataset if crashes where construction was present were believed to have occurred during short-term construction or maintenance activities. Police Reporting Thresholds In general, state laws do not require the police to file a report for all property-damage-only (PDO) crashes to which they respond. States generally require a report be filed if the estimated damage of the crash exceeds a certain dollar amount. These different “reporting thresholds” can explain differences in PDO crash frequencies from state to state. The reporting thresholds for each state are shown in Table 8. Table 8. Police-reporting thresholds for PDO crashes by state. State Police-Reporting Threshold for PDO crashes Georgia $500 Hawaii $3,000 Minnesota $1,000 Ohio $400 prior to 9/7/2011, $1,000 on and after 9/7/2011 Virginia $1,500 Crash Times Sub-annual models for PTSU operation and comparison facilities, described in Chapter 8, require the hour of the crash to be known. Table 9 shows the percent of crashes within the study area of each state that did not have an assigned hour of the day in the state’s crash data. These crashes were used for the development of annual crash models but excluded from the development of sub-annual models. The value shown for Minnesota was computed from the PTSU and PTSU comparison study segments only (BOS and BOS comparison segments are not included). Table 9. Percent of crashes without hour of day by state. State Percent of Crashes Georgia 0.3% Hawaii 1.1% Minnesota 0.1% Virginia 0.0% Merging of Road Inventory, Volume, and Crash Data After the databases described above were populated, they were initially merged into master PTSU and master BOS databases. Each record in the geometric, volume, and crash databases has a field for the state, the route number, the direction of travel (increasing milepost direction or decreasing milepost direction), and milepost. Ohio records also have a field for county because mileposts reset to 0.0 at county lines. The research team developed scripts to assign site identification numbers to each crash, and to assign crashes to the applicable site based on the state, route, direction of travel, milepost, and (when applicable) county. The master datasets include a file with one record per site and one record per crash. Modeling of BOS facilities, described in Chapter 8, determined that BOS presence did not have a statistically significant effect on crash frequency. Based on this finding, the BOS database was combined

37 with the PTSU database, and study sites on BOS and BOS comparison facilities were added to the database to increase the sample size and the range of values for key independent variables. Database Summary This section summarizes the data assembled for the purpose of estimating the predictive models. Initially, the site sample size is described. Then, the geometric and traffic characteristics are summarized. Finally, the crash data are summarized. The purpose of this summary is to provide information about the range of data included in the database. The discussion in this section is not intended to indicate conclusive results or recommendations. The proposed predictive models (and associated trends) are documented in subsequent chapters. Facilities with PTSU or BOS operation were considered treatment facilities. Facilities not having PTSU or BOS operation but located near the treatment facilities and having the same route designation were considered comparison facilities. Comparison facilities were used to specifically address the research question of whether PTSU or BOS operation has an effect on crash frequency or severity. Finally, facilities not having PTSU operation and located some distance away from treatment facilities were considered supplemental facilities. These facilities were added to the database to increase the sample size and the range of values for key independent variables. Data for supplemental facilities were initially collected to investigate the possible influence of BOS operation on safety (i.e., they were BOS or BOS comparison facilities). The findings from this investigation are described in Chapter 8. The database includes data for 14 facilities collectively located in five states. Treatment and comparison facilities are located in Georgia, Hawaii, Minnesota, and Virginia. Supplemental facilities are located in Ohio and Minnesota. Sample Size Table 10 lists the sample size represented in the highway safety database. The database includes data for 728 study sites. The distribution of these sites by PTSU operation and site type is provided in the top half of the table. The 728 sites have a total length of 164.8 miles, as shown in the last row of the table. About 25 percent of the mileage consists of facilities with PTSU operation during one or more hours of the day.

38 Table 10. Database sample size. Category PTSU Operation Site Type Sample Size by State Total Georgia Hawaii Minn. Ohio Virginia Number of sites Yes a Freeway segment 30 27 11 0 72 140 Ramp entrance speed-change lane 5 5 0 0 8 18 Ramp exit speed-change lane 3 3 1 0 12 19 No Freeway segment 18 12 164 123 63 380 Ramp entrance speed-change lane 4 2 44 25 14 89 Ramp exit speed-change lane 3 3 43 22 11 82 Total: 63 52 263 170 180 728 Site length (miles) Yes a Freeway segment 10.3 6.1 1.8 0.0 18.4 36.5 Ramp entrance speed-change lane 0.5 0.6 0.0 0.0 1.8 2.8 Ramp exit speed-change lane 0.1 0.1 0.1 0.0 1.8 2.1 No Freeway segment 4.8 2.8 43.5 35.6 19.6 106.3 Ramp entrance speed-change lane 0.6 0.3 5.4 2.8 2.0 11.2 Ramp exit speed-change lane 0.2 0.2 2.1 2.2 1.2 5.8 Total: 16.5 10.2 52.9 40.6 44.7 164.8 a Includes a transition zone located between, just upstream of, or just downstream of PTSU segments. Geometric and Traffic Characteristics Table 11 summarizes the geometric and traffic characteristics of the sites in the highway safety database. The sites are located in urban areas and have lane counts that range from 2 to 7 in the subject direction of travel. The alignments of 250 sites (34 percent) are curved. Shoulder rumble strips are used at most of the sites; however, they are not used on a shoulder when it is used for PTSU operation. A total of 141 sites (of 728) have a PTSU lane on the outside shoulder. Most of the remaining 587 sites have rumble strips on the outside shoulder. For these 587 sites, the average proportion of the site length with outside shoulder rumble strips is 0.35 (this average includes 208 sites that do not have outside shoulder rumble strips). Similarly, a total of 36 sites have a PTSU lane on the inside shoulder. About 40 percent of the remaining 692 sites have rumble strips on the inside shoulder. For these 692 sites, the average proportion of the site length with inside shoulder rumble strips is 0.37 (this average includes 419 sites that do not have inside shoulder rumble strips). The data in Table 11 indicate that barrier is quite prevalent along the study sites. Notably, median barrier is present on 675 segments (93 percent). The average proportion of the site length with median barrier is 0.88. Outside (roadside) barrier is present on 505 segments (69 percent) and has an average proportion of site length of 0.43. Clear zone width on the outside of the roadway was not explicitly measured for each site, partly because most sites were lined with outside (roadside) barrier. However, when the other cross section elements were measured using Google Earth, the clear zone along those portions of each site not protected by outside (roadside) barrier was visually examined and an assessment made as to whether the site met the “30-ft” clear zone width recommended by the AASHTO Roadside Design Guide (1996). All sites that were not fully lined with outside (roadside) barrier were judged to have a clear zone width of at least 30 feet wherever barrier was not present.

39 Table 11. Summary geometric and traffic characteristics. Variablea Average Minimum Maximum Freeway Segment Volume Variables Directional AADT volume of freeway, AADTfs (veh/day) 60,010 15,370 147,210 Horizontal Curve Variables Number of sites with horizontal curve 250 n.a. n.a. Horizontal curve radius, R (ft) 6,210 1,430 24,170 Cross Section Variables Number of lanes (in subject travel direction), n 3.3 2 7 Median width, Wm (feet) 35.0 5.1 90.9 Paved inside shoulder width, Wis (feet) 7.6 0.7 11.0 Inside PTSU lane width (left side), Wptsu,in (feet) 11.7 3.5b 14.0 Lane width (excluding PTSU lane), Wl (feet) 11.9 10.5 14.4 Outside PTSU lane width (right side), Wptsu,o (feet) 11.2 5.0b 16.8 Paved outside shoulder width, Ws (feet) 9.8 0.7 14.0 Rumble Strip Variables Proportion of site with rumble strips on the inside shoulder, Pirc 0.35 0.0 1.0 Proportion of site with rumble strips on the outside shoulder, Porc 0.37 0.0 1.0 Barrier Variables Proportion of site with barrier present in the median, Pib 0.88 0.0 1.0 Distance from inside shoulder to barrier face, Wicb (feet) 2.5 0.7 20.0 Distance from outside shoulder to barrier face, Wocb (feet) 1.7 1.0 20.0 Proportion of site with barrier present on the outside (roadside), Pob 0.43 0.0 1.0 Turnout Variables Number of sites with turnout beyond the shoulder 37 n.a. n.a. Length of turnout (from start of taper to end of taper), Lturnout (feet) 520 75 1,500 Speed-Change Lane Site Variables AADT volume of entrance ramp, AADTen (veh/day) 6,890 450 25,210 AADT volume of exit ramp, AADTex (veh/day) 7,750 450 30,680 Length of entrance ramp, Len (miles) 0.16 0.06 0.32 Length of exit ramp, Lex (miles) 0.09 0.02 0.27 n.a. = not applicable. a Variable names and definitions are consistent with those used in Chapter 18 of the HSM Supplement (AASHTO 2014). b PTSU tapers from full width to zero width over segment; table value is an average width over length of segment. c Proportion of site with rumble strips is reported for sites that do not have PTSU operation on the associated side. Roadside turnouts are used with PTSU facilities in three states and average 520 feet in length (measured from start of taper to end of taper). Turnouts provide emergency refuge spaces for disabled vehicles beyond the shoulder. Jenior et al. (2016) recommend that turnouts should be constructed at desirably one-half mile intervals. Table 12 identifies the PTSU facilities that do (and do not) have turnouts along the length each facility included in the highway safety database. The average distance between turnouts is shown in the last column of the table. This distance is shown to vary from 0.14 miles to more than 1.52 miles.

40 Table 12. Summary of turnout characteristics. State Facility Number of Turnouts Total Length of Turnouts (miles) Total Length of Segments (miles)a Turnout Spacing (miles/turnout)b Georgia GA 400 8 0.28 8.29 0.89 I–85 4 0.12 0.80 0.14 Hawaii I–H1 3 0.13 5.43 1.32 Minnesota I–35W 0 0.00 1.32 >1.32 Virginia I–66 12 1.58 11.50 0.76 I–264 10 1.54 6.77 0.48 I–495 0 0.00 1.52 >1.52 a Length includes only segments with PTSU operation (i.e., a shoulder being used by vehicles) during some part of the day. b Turnout spacing = (total length of segments – total length of turnouts) / (number of turnouts + 1). The following subsections present additional information on the characteristics of the study sites. This information is categorized by facility and site type. Segment Characteristics Table 13 presents a summary of the key attributes of the freeway segments on PTSU facilities. Table 14 presents a summary of the key attributes of the freeway segments on comparison and supplemental facilities. All weaving sections were one-sided ramp weaves according to the definition in the Highway Capacity Manual (HCM) (AASHTO 2010). Transition segments include freeway segments immediately prior to the start and beyond the end of PTSU operation, and this distinction is indicated in project’s dataset. The number of transition segments number varies from facility to facility because transitions can occur with or without a taper. In some cases, the start or end of a curve or speed-change lane necessitated that a transition segment be subdivided to remain consistent with segmentation criteria.

41 Table 13. Freeway segment data summary for PTSU facilities. Facility Number of Lanes Directional Volume (veh/day) Segment Length (miles) Total Segments Weaving Segmentb Transition Segmentb Total Length (miles) Min Max Avga Min Max Avga Min Max Avg GA 400 GA 4 4 4.0 68,000 92,650 84,128 0.02 0.74 0.36 24 0 11 8.656 I-264 VA 4 5 4.0 65,406 101,917 82,087 0.05 0.61 0.26 26 3 7 6.711 I-495 VA 4 4 4.0 90,299 99,215 95,766 0.03 0.36 0.17 11 4 6 1.874 I-H1 HI 3 5 4.3 44,019 114,263 70,600 0.03 0.93 0.23 26 2 9 5.958 I-35W MN 4 5 4.3 89,099 95,106 90,291 0.02 0.46 0.16 11 1 0 1.776 I-66 VA 3 6 3.5 46,980 92,451 81,879 0.04 0.86 0.27 37 0 10 10.145 I-85 GA 6 6 6.0 146,625 147,350 147,229 0.02 0.52 0.26 6 0 5 1.585 a Average per segment (not per mile) b Included in count of total segments Table 14. Freeway segment data summary for comparison and supplemental facilities. Facility Number of Lanes Directional Volume (veh/day) Segment Length (miles) Total Segments Weaving Segmentb Total Length (miles) Min Max Avg a Min Max Avga Min Max Avg GA 400 GA 3 4 3.2 68,825 75,700 72,427 0.04 0.38 0.21 15 0 3.151 I-264 VA 4 4 4.0 31,400 72,922 52,015 0.03 0.84 0.25 23 5 5.863 I-495 VA 4 4 4.0 98,330 112,329 106,278 0.08 0.37 0.20 7 0 1.372 I-H1 HI 4 5 4.2 50,121 91,876 75,776 0.09 0.48 0.24 12 0 2.834 I-35W MNc 5 6 5.3 81,889 90,237 85,684 0.12 0.49 0.29 4 0 1.173 I-66 VA 4 5 4.2 46,773 82,016 71,542 0.03 1.23 0.39 31 2 12.158 I-85 GA 7 7 7.0 145,500 146,125 145,917 0.07 1.02 0.52 3 0 1.562 US 52 MN 2 2 2.0 15,370 27,592 22,772 0.11 0.51 0.23 23 0 5.403 MN 610 MN 2 2 2.0 15,844 29,279 23,735 0.15 0.56 0.31 11 0 3.39 I-35W MNd 2 2 2.0 17,008 27,483 21,300 0.03 1.20 0.36 13 0 4.69 I-70 OH 2 4 3.1 26,717 68,486 42,898 0.06 0.89 0.26 20 0 5.221 I-275 OH 3 4 3.1 33,758 52,532 44,579 0.06 0.61 0.35 19 2 6.735 I-35W MNe 2 5 2.9 20,106 91,267 45,204 0.10 1.20 0.40 27 1 10.804 I-94 MN 3 4 3.8 65,940 80,464 74,123 0.02 0.51 0.25 17 3 4.274 US 169 MN 2 2 2.0 33,278 46,250 38,689 0.03 0.69 0.20 69 15 13.795 I-70 OH 3 4 3.5 42,104 67,735 55,023 0.07 0.79 0.26 23 1 5.883 I-71 OH 3 4 3.4 34,350 71,041 56,099 0.02 0.77 0.29 61 1 17.766 a Average per segment (not per mile) b Included in count of total segments c Comparison portion of I-35W d Supplemental portion of I-35W initially used for BOS comparison analysis e Supplemental portion of I-35W initially used for BOS analysis Table 15 presents a summary of the key attributes of the ramp entrance speed-change lane sites on PTSU facilities. Table 16 presents a summary of the key attributes of the ramp entrance speed-change lane sites on comparison and supplemental facilities. The values shown are similar to the corresponding freeway segment tables because the speed-change lane sites are adjacent to the freeway segments. The number of ramp entrance speed-change lane sites was lower than initially anticipated because many on- ramps on PTSU facilities were followed by a “speed-change lane” greater than 1,600 feet long. Consistent with the speed-change lane procedure in the HSM Supplement, such sites were classified as freeway

42 segments (not speed-change lane sites) having an additional basic lane and a lane drop. Three of the PTSU facilities do not have ramp entrance speed-change lanes. Table 15. Ramp entrance speed-change lane site data summary for PTSU facilities. Facility Number of Lanes Directional Volume (veh/day) Site Length (miles) Total Sites Weaving Segmentb Transition Segmentb Total Length (miles) Min Max Avga Min Max Avga Min Max Avg GA 400 GA 4 4 4.0 65,723 86,900 81,382 0.06 0.12 0.09 5 0 0 0.463 I-264 VA 4 4 4.0 72,885 91,600 77,721 0.04 0.27 0.19 4 0 2 0.743 I-495 VA - - - - - - - - - - - - - I-H1 HI 4 5 4.5 64,375 77,442 72,171 0.07 0.18 0.13 6 0 1 0.76 I-35W MN - - - - - - - - - - - - - I-66 VA 3 6 3.8 81,899 92,258 87,630 0.09 0.33 0.22 5 0 1 1.091 I-85 GA - - - - - - - - - - - - - a Average per site (not per mile) b Included in count of total sites Table 16. Ramp entrance speed-change lane site data summary for comparison and supplemental facilities. Facility Number of Lanes Directional Volume (veh/day) Site Length (miles) Total Sites Weaving Segmentb Total Length (miles) Min Max Avga Min Max Avga Min Max Avg GA 400 GA 3 3 3.0 68,825 75,700 73,408 0.11 0.23 0.19 3 0 0.558 I-264 VA 4 4 4.0 36,800 73,800 54,533 0.15 0.25 0.20 3 0 0.608 I-495 VA 4 4 4.0 106,628 112,329 109,479 0.02 0.14 0.10 4 0 0.403 I-H1 HI 4 5 4.5 74,650 83,877 79,263 0.12 0.23 0.17 2 0 0.347 I-35W MNc 5 5 5.0 86,677 86,677 86,677 0.14 0.14 0.14 1 0 0.141 I-66 VA 4 4 4.0 62,582 79,600 71,997 0.03 0.22 0.16 6 0 0.969 I-85 GA 7 7 7.0 134,418 134,418 134,418 0.08 0.08 0.08 1 0 0.08 US 52 MN 2 2 2.0 15,825 27,592 24,244 0.03 0.17 0.10 10 0 0.964 MN 610 MN 2 2 2.0 19,537 29,279 25,173 0.03 0.17 0.13 5 0 0.63 I-35W MNd 2 2 2.0 19,103 27,483 22,935 0.09 0.15 0.13 3 0 0.385 I-70 OH 2 4 3.3 32,738 71,645 53,969 0.04 0.13 0.07 7 0 0.517 I-275 OH 3 4 3.3 33,758 50,163 42,460 0.08 0.18 0.14 4 0 0.546 I-35W MNe 2 4 3.1 22,300 95,106 50,960 0.06 0.16 0.11 8 0 0.894 I-94 MN 4 4 4.0 76,900 88,623 80,572 0.07 0.17 0.12 4 0 0.474 US 169 MN 2 2 2.0 38,662 43,750 40,301 0.05 0.21 0.15 13 0 1.892 I-70 OH 3 4 3.3 47,107 67,167 57,949 0.02 0.17 0.11 6 0 0.676 I-71 OH 3 4 3.3 43,271 65,152 59,846 0.08 0.24 0.14 8 0 1.091 a Average per site (not per mile) b Included in count of total sites c Comparison portion of I-35W d Supplemental portion of I-35W initially used for BOS comparison analysis e Supplemental portion of I-35W initially used for BOS analysis

43 Table 17 presents a summary of the key attributes of the ramp exit speed-change lanes on PTSU facilities. Table 18 presents a summary of the key attributes of the ramp exit speed-change lanes on comparison and supplemental facilities. Table 17. Ramp exit speed-change lane site data summary for PTSU facilities. Facility Number of Lanes Directional Volume (veh/day) Site Length (miles) Total Sites Weaving Segmentb Transition Segmentb Total Length (miles) Min Max Avga Min Max Avga Min Max Avg GA 400 GA 4 4 4.0 45,475 92,650 75,370 0.04 0.05 0.04 3 0 0 0.133 I-264 VA 4 4 4.0 76,325 96,003 89,683 0.06 0.19 0.13 4 0 0 0.529 I-495 VA 4 4 4.0 99,215 99,215 99,215 0.08 0.10 0.09 2 0 0 0.184 I-H1 HI 4 5 4.3 59,665 77,442 65,804 0.04 0.06 0.05 3 0 0 0.148 I-35W MN 4 4 4.0 95,106 95,106 95,106 0.08 0.08 0.08 1 0 0 0.076 I-66 VA 3 4 3.5 49,200 92,451 78,918 0.13 0.28 0.18 6 0 1 1.068 I-85 GA - - - - - - - - - - - - - a Average per site (not per mile) b Included in count of total sites Table 18. Ramp exit speed-change lane site data summary for comparison and supplemental facilities. Facility Number of Lanes Directional Volume (veh/day) Site Length (miles) Total Sites Weaving Segmentb Total Length (miles) Min Max Avga Min Max Avga Min Max Avg GA 400 GA 4 4 4.0 71,440 75,625 73,533 0.04 0.10 0.07 2 0 0.144 I-264 VA 4 4 4.0 49,400 68,600 58,300 0.04 0.13 0.09 4 0 0.356 I-495 VA - - - - - - - - - - - - I-H1 HI 4 4 4.0 67,147 91,876 78,158 0.04 0.10 0.07 3 0 0.199 I-35W MNc - - - - - - - - - - - - I-66 VA 4 5 4.1 67,612 81,773 76,350 0.02 0.18 0.11 7 0 0.748 I-85 GA 6 6 6.0 145,500 145,500 145,500 0.05 0.05 0.05 1 0 0.051 US 52 MN 2 2 2.0 15,825 26,350 23,417 0.03 0.06 0.04 8 0 0.35 MN 610 MN 2 2 2.0 19,966 29,279 24,149 0.04 0.05 0.04 4 0 0.173 I-35W MNd 2 2 2.0 19,233 27,100 22,878 0.05 0.05 0.05 3 0 0.145 I-70 OH 2 4 3.2 31,055 72,359 48,176 0.09 0.12 0.11 5 0 0.572 I-275 OH 3 3 3.0 43,443 49,316 46,595 0.08 0.11 0.10 3 0 0.288 I-35W MNe 2 5 3.1 22,794 91,267 51,667 0.02 0.10 0.05 10 0 0.547 I-94 MN 4 4 4.0 72,687 77,845 75,785 0.04 0.06 0.05 5 0 0.245 US 169 MN 2 2 2.0 39,103 43,963 40,631 0.03 0.08 0.05 13 0 0.641 I-70 OH 3 4 3.5 46,445 66,721 60,068 0.02 0.13 0.09 8 0 0.732 I-71 OH 3 3 3.0 54,823 66,268 60,653 0.05 0.13 0.10 6 0 0.573 a Average per site (not per mile) b Included in count of total sites c Comparison portion of I-35W d Supplemental portion of I-35W initially used for BOS comparison analysis e Supplemental portion of I-35W initially used for BOS analysis The values shown in Table 17 and Table 18 are similar to the corresponding freeway segment tables because the ramp exit speed-change lane sites are adjacent to the freeway segments. Similar to ramp

44 entrance speed-change lane sites, the number of ramp exit speed-change lane sites was lower than initially anticipated because some off-ramps on PTSU facilities were preceded by a “speed-change lane” greater than 1,600 feet long. Consistent with the speed-change lane procedure in the HSM Supplement, such sites were classified as freeway segments (not speed-change lane sites) having an additional basic lane and a lane add. One of the PTSU facilities does not have a ramp exit speed-change lane. Barrier and PTSU Lane Data Table 19 and Table 20 present data on the barriers as well as several PTSU-related variables on freeway segments on PTSU facilities and comparison/supplemental facilities, respectively. The percent of median barrier and outside (roadside) barrier types reflect barrier lengths within the limits of the study, and average widths were computed on a per-site basis (not weighted to the length of the segments). A barrier type of “none” indicates no barrier was present or the barrier was more than 30 feet from the travelled portion of the roadway. Most segments have a median barrier, often concrete, which is not surprising given the urban nature of the dataset. Interstate H1 in Hawaii has a mix of right-side and left- side PTSU operation, but other facilities have exclusively one or the other. The PTSU lane, or portion of the shoulder used for travel, is generally between 11 and 12 feet wide. All the studied PTSU facilities have a second edge line and one or more feet of paved “shoulder beyond the shoulder.” Table 19. Freeway segment barrier and PTSU characteristics for PTSU facilities. Facility Percent Median Barrier Type Percent Outside (Roadside) Barrier Percent PTSU by Side Average Width of PTSU Lanea (ft) Average Width of Shoulder beyond PTSU Lane (ft) Concrete Guardrail Cable None Right Left GA 400, GA 100% 0% 0% 0% 43% 100% 0% 11.3 1.8 I-264, VA 100% 0% 0% 0% 86% 100% 0% 11.1 6.0 I-495, VA 100% 0% 0% 0% 70% 0% 100% 11.8 1.9 I-H1, HI 75% 22% 0% 2% 60% 73% 27% 10.3 1.5 I-35W, MN 93% 0% 0% 7% 23% 0% 100% 13.4 3.3 I-66, VA 75% 5% 0% 19% 75% 100% 0% 11.9 5.1 I-85, GA 100% 0% 0% 0% 66% 100% 0% 11.1 3.1 a Distance between first and second edge lines for PTSU operation, entire width of shoulder for BOS (no second edge lines present)

45 Table 20. Freeway segment barrier characteristics for comparison and supplemental facilities. Facility Percent Median Barrier Type Percent Outside (Roadside) Barrier Concrete Guardrail Cable None GA 400, GA 100% 0% 0% 0% 67% I-264, VA 100% 0% 0% 0% 62% I-495, VA 100% 0% 0% 0% 84% I-H1, HI 62% 38% 0% 0% 63% I-35W, MNa 100% 0% 0% 0% 44% I-66, VA 68% 19% 0% 12% 66% I-85, GA 100% 0% 0% 0% 99% US 52, MN 0% 2% 40% 58% 6% MN 610, MN 0% 0% 19% 81% 3% I-35W, MNb 0% 0% 51% 49% 4% I-70, OH 100% 0% 0% 0% 50% I-275, OH 0% 11% 16% 73% 35% I-35W, MNc 30% 0% 29% 41% 10% I-94, MN 100% 0% 0% 0% 17% US 169, MN 45% 13% 14% 28% 13% I-70, OH 71% 2% 5% 22% 38% I-71 OH 98% 0% 0% 2% 41% a Comparison portion of I-35W b Supplemental portion of I-35W initially used for BOS comparison analysis c Supplemental portion of I-35W initially used for BOS analysis Table 21 and Table 22 present data on the barriers and several PTSU-related variables on ramp entrance speed-change lane sites on PTSU facilities. The percent of median barrier and outside (roadside) barrier types reflect barrier lengths within the limits of study, and average widths were computed on a per-site basis (not weighted to the length of the sites). A barrier type of “none” indicates no barrier was present or the barrier was more than 30 feet from the travelled portion of the roadway.

46 Table 21. Ramp entrance speed-change lane site barrier and PTSU characteristics for PTSU facilities. Facility Percent Median Barrier Type Percent Road-side Barrier Percent PTSU by Side Average Width of PTSU Lanea (ft) Width of Shoulder beyond PTSU Lane (ft) Concrete Guardrail Cable None Right Left GA 400, GA 100% 0% 0% 0% 56% 100% 0% 12.4 3.8 I-264, VA 100% 0% 0% 0% 86% 100% 0% 10.8 4.7 I-495, VA - - - - - - - - - I-H1, HI 100% 0% 0% 0% 69% 50% 50% 10.9 1.9 I-35W, MN - - - - - - - - - I-66, VA 77% 13% 0% 10% 58% 100% 0% 12.4 4.3 I-85, GA - - - - - - - - - a Distance between first and second edge lines for PTSU operation, entire width of shoulder for BOS (no second edge lines present) Table 22. Ramp entrance speed-change lane site barrier characteristics for comparison and supplemental facilities. Facility Percent Median Barrier Type Percent Outside (Roadside) Barrier Concrete Guardrail Cable None GA 400, GA 100% 0% 0% 0% 86% I-264, VA 100% 0% 0% 0% 71% I-495, VA 100% 0% 0% 0% 42% I-H1, HI 65% 5% 0% 29% 55% I-35W, MNa 100% 0% 0% 0% 65% I-66, VA 78% 22% 0% 0% 92% I-85, GA 100% 0% 0% 0% 92% US 52, MN 0% 0% 49% 51% 3% MN 610, MN 0% 0% 50% 50% 0% I-35W, MNb 0% 0% 24% 76% 0% I-70, OH 100% 0% 0% 0% 52% I-275, OH 0% 9% 6% 84% 55% I-35W, MNc 53% 0% 30% 17% 13% I-94, MN 100% 0% 0% 0% 8% US 169, MN 42% 12% 20% 27% 9% I-70, OH 66% 0% 20% 14% 47% I-71, OH 92% 0% 0% 8% 39% a Comparison portion of I-35W b Supplemental portion of I-35W initially used for BOS comparison analysis c Supplemental portion of I-35W initially used for BOS analysis Table 23 and Table 24 present data on the barriers and several PTSU-related variables on ramp exit speed-change lane sites on PTSU facilities. The percent of median barrier and outside (roadside) barrier types reflect barrier lengths within the limits of study, and average widths were computed on a per-site

47 basis (not weighted to the length of the sites). A barrier type of “none” indicates no barrier was present or the barrier was more than 30 feet from the travelled portion of the roadway. Table 23. Ramp exit speed-change lane site barrier and PTSU characteristics for PTSU facilities. Facility Percent Median Barrier Type Percent Outside (Roadside) Barrier Percent PTSU by Side Average Width of PTSU Lanea (ft) Average Width of Shoulder beyond PTSU Lane (ft) Concrete Guardrail Cable None Right Left GA 400, GA 100% 0% 0% 0% 53% 100% 0% 11.1 1.8 I-264, VA 100% 0% 0% 0% 99% 100% 0% 11.2 2.5 I-495, VA 100% 0% 0% 0% 58% 0% 100% 11.9 1.4 I-H1, HI 100% 0% 0% 0% 47% 67% 33% 10.2 0.8 I-35W, MN 100% 0% 0% 0% 62% 0% 100% - - I-66, VA 74% 17% 0% 9% 57% 100% 0% 11.7 7.1 I-85, GA - - - - - - - - - a Distance between first and second edge lines for PTSU operation, entire width of shoulder for BOS (no second edge lines present) Table 24. Ramp exit speed-change lane site barrier characteristics for comparison and supplemental facilities. Facility Percent Median Barrier Type Percent Outside (Roadside) Barrier Concrete Guardrail Cable None GA 400, GA 100% 0% 0% 0% 90% I-264, VA 100% 0% 0% 0% 87% I-495, VA - - - - - I-H1, HI 60% 22% 0% 18% 14% I-35W, MNa - - - - - I-66, VA 75% 25% 0% 0% 81% I-85, GA 100% 0% 0% 0% 100% US 52, MN 0% 18% 46% 36% 71% MN 610, MN 0% 0% 50% 50% 77% I-35W, MNb 0% 0% 36% 64% 0% I-70, OH 100% 0% 0% 0% 78% I-275, OH 0% 17% 0% 83% 62% I-35W, MNc 65% 0% 32% 3% 50% I-94, MN 100% 0% 0% 0% 23% US 169, MN 56% 2% 20% 22% 19% I-70, OH 67% 9% 5% 20% 53% I-71, OH 100% 0% 0% 0% 54% a Comparison portion of I-35W b Supplemental portion of I-35W initially used for BOS comparison analysis c Supplemental portion of I-35W initially used for BOS analysis

48 Crash Characteristics The crash data for the study sites are summarized in Table 25. As shown in the bottom row of the table, the sites are collectively associated with 4,807 fatal-and-injury (FI) crashes and 11,937 PDO crashes. These counts are combined with the exposure value in the fourth column to obtain the corresponding crash rates shown in the right two columns of the table. The overall FI crash rate (shown in the bottom row of the table) of 0.28 FI crashes per million vehicles-miles (cr/mvm) is consistent with the urban freeway FI crash rates reported by Bonneson et al. (2012a) for the combined states of California, Maine, and Washington. Table 25. Summary of crash data by state and site type. State PTSU Operation Site Type Exposure (mvm) Reported Crashes Crash Rate (cr/mvm) FI PDO FI PDO Summary by State Georgia Yesa All 1,498 318 928 0.21 0.62 No All 758 172 519 0.23 0.68 Total: 2,256 490 1,447 0.22 0.64 Hawaii Yesa All 863 226 130 0.26 0.15 No All 472 47 44 0.10 0.09 Total: 1,335 273 174 0.20 0.13 Minnesota Yesa All 307 161 519 0.52 1.69 No All 3,658 718 2,103 0.20 0.57 Total: 3,965 879 2,622 0.22 0.66 Ohio Yesa All n.a. n.a. n.a. n.a. n.a. No All 3,698 1,055 3,146 0.29 0.85 Total: 3,698 1,055 3,146 0.29 0.85 Virginia Yesa All 3,133 1,383 3,058 0.44 0.98 No All 2,710 727 1,490 0.27 0.55 Total: 5,843 2,110 4,548 0.36 0.78 Summary by Site Type All Yesa Freeway segment 5,071 1,833 4,030 0.36 0.79 Ramp ent. speed-change lane 427 125 259 0.29 0.61 Ramp exit speed-change lane 303 130 346 0.43 1.14 Total: 5,801 2,088 4,635 0.36 0.80 No Freeway segment 9,674 2,350 6,276 0.24 0.65 Ramp ent. speed-change lane 1,035 203 590 0.20 0.57 Ramp exit speed-change lane 586 166 436 0.28 0.74 Total: 11,295 2,719 7,302 0.24 0.65 All Sites Combined Total: 17,096 4,807 11,937 0.28 0.70 a Includes transition zone located between, just upstream of, or just downstream of PTSU segments mvm = million vehicle-miles; n.a. = not applicable; cr/mvm = crashes per million vehicle-miles Further examination of the rates in Table 25 indicates that the PDO crash rates in Hawaii are quite low. This trend reflects the fact that Hawaii has a higher crash reporting threshold than the other states. It is a reminder that the development of CPMs based on multi-state crash data should separately estimate models for FI and for PDO crashes.

49 Speed-change-lane-related crashes were identified as those crashes that occurred: (a) between the painted gore point and the end of taper and (b) on the same side of the freeway as the speed-change lane. All other crashes were identified as segment-related crashes and were assigned to a freeway segment. The data in the lower half of Table 25 indicate that ramp exit speed-change lanes tend to have a higher crash rate than a freeway segment. This trend is likely due to the lane changing that occurs in the vicinity of the exit ramp as exiting vehicles shift lanes to reach the lane associated with the exit ramp. The data in the lower half of Table 25 also indicate that the sites without PTSU operation have a lower crash rate than those with PTSU operation. This trend is also illustrated in Table 26, which lists the FI crash rates for each facility with PTSU operation and for each comparison facility. The right column of Table 26 shows the crash rate ratio, which is computed as the FI crash rate for the PTSU facility divided by the FI crash rate for the comparison facility. The ratios range from 1.09 to 3.13. They also indicate that sites without PTSU operation have a lower crash rate than those with PTSU operation. This trend is discussed more in the next section. Table 26. Summary of FI crash rates for PTSU and comparison facilities. State Facility PTSU Facilities Comparison Facilities Crash Rate Ratio (PTSU/Comp.) Exposure (mvm) Reported FI Crashes FI Crash Rate (cr/mvm) Exposure (mvm) Reported FI Crashes FI Crash Rate (cr/mvm) Georgia GA 400 1,200 178 0.15 444 36 0.08 1.88 I–85 298 140 0.47 314 136 0.43 1.09 Hawaii I–H1 863 226 0.26 472 47 0.10 2.60 Minnesota I–35W 307 161 0.52 207 58 0.28 1.86 Virginia I–66 1,793 815 0.45 1,822 523 0.29 1.55 I–264 1,160 484 0.42 703 176 0.25 1.68 I–495 180 84 0.47 184 28 0.15 3.13 mvm = million vehicle-miles; cr/mvm = crashes per million vehicle-miles Table 27 summarizes the crash rates for sites in the highway safety database based on the PTSU feature that is present. The rates in the table can be compared to provide some insight into possible features or operation that have an influence on safety. For example, the FI crash rates for inside versus outside shoulder PTSU operation are fairly similar in value, which suggests that the side on which PTSU operates has no influence on safety. In contrast, the crash rates for PDO crashes suggests that PTSU operating on the inside shoulder may have more crashes than PTSU operating on the outside shoulder.

50 Table 27. Summary of crash rates for various PTSU features. Feature Exposure (mvm) Reported Crashes Crash Rate (cr/mvm) FI PDO FI PDO PTSU operation on inside shoulder (left side)a 773 268 770 0.35 1.00 PTSU operation on outside shoulder (right side)a 5,028 1,820 3,865 0.36 0.77 PTSU lane present (including tapered transitions) 4,962 1,833 4,012 0.37 0.81 Transition zone between, upstream of, or downstream of PTSU 839 255 623 0.30 0.74 PTSU without turnouta 3,795 1,509 3,269 0.40 0.86 PTSU with turnout beyond the shouldera 2,006 579 1,366 0.29 0.68 All Sites Combined Total: 17,096 4,807 11,937 0.28 0.70 a Includes transition zone located between, just upstream of, or just downstream of PTSU segments. mvm = million vehicle-miles; cr/mvm = crashes per million vehicle-miles Hours of Operation As discussed in a previous section, the research team collected the status of the shoulder (open or closed to traffic) for each hour through the duration of the study period. These results are presented in Table 28 and Table 29. The “WD” columns indicate average weekday PTSU status, and “WE” columns indicate the average weekend day PTSU status. The numbers in each cell indicate the proportion of time within an hour (for the duration of the study period) that the shoulder was open. Interstate H1, for example, is an S-PTSU facility with fixed hours of operation. Beyond milepost 15, H1 is open from 5:30 AM to 8:30 AM. The cells for the 5 AM and 8 AM hours on weekdays have a value of 0.50, which indicates the shoulder was open for half of these hours throughout the study period, and values of 1.00 for the 6 AM and 7 AM hours. Values of 0.96 for I-264 and I-495 reflect closure of the shoulder on federal holidays. Dynamic facilities have more variability. The shoulder on I-66 and I-35W was occasionally open during overnight hours, possible to due to short-term construction or incidents in other lanes. Additionally, MnDOT indicated the shoulder on I-35W is kept open during snowfall events because the presence of traffic helps to prevent the buildup of snow and ice. Zero values (hours in which the shoulder was never open) are omitted from the table to improve legibility.

51 Table 28. Proportion of each hour PTSU operation is open, except GA 400. Hour of Day I-85 I-H1 Prior to Milepost 15 I-H1 Beyond Milepost 15 I-66 WB I-66 EB I-264 WB I-264 EB I-495 I-35W WD WE WD WE WD WE WD WE WD WE WD WE WD WE WD WE WD WE 12A .01 .01 .06 .05 1A .01 .06 .05 2A .01 .06 .05 3A .01 .06 .05 4A .06 .05 5A 1 .50 .47 .52 .05 6A 1 1 .03 .95 .96 .12 .94 .05 7A 1 1 .11 .95 .96 .96 .95 .05 8A .50 .50 .25 .01 .96 .01 .96 .95 .05 9A .24 .04 .96 .08 .96 .95 .06 10A .16 .14 .95 .21 .96 .81 .21 11A .14 .25 .19 .30 .81 .49 12P .21 .32 .25 .37 .81 .53 1P .29 .37 .26 .39 .82 .54 2P .96 .39 .30 .40 .96 .95 .56 3P 1 .97 .40 .34 .40 .96 .96 .57 4P 1 .97 .40 .37 .40 .96 .96 .96 .58 5P 1 .97 .38 .40 .39 .96 .96 .96 .58 6P 1 .96 .38 .39 .37 .96 .95 .42 7P .96 .28 .27 29 .96 .24 .07 8P .66 .09 .05 .08 .08 .06 9P .04 .03 .02 .02 .07 .05 10P .02 .01 .06 .05 11P .01 .02 .06 .05 WD = average weekday PTSU status; WE = average weekend PTSU status

52 Table 29. Proportion of each hour PTSU operation is open, GA 400. Hour of Day Northbound MM 8.8-11.9 Northbound MM 19.9-20.7 Southbound MM 8.5-9.5 Southbound 9.5-11.4 Southbound 11.4-11.8 Southbound 12.1-15.0 WD WE WD WE WD WE WD WE WD WE WD WE 12A 1A 2A 3A 4A 5A 6A 1 .81 1 .88 .90 .90 7A 1 1 1 1 1 1 8A 1 1 1 1 1 1 9A 1 1 1 1 1 1 10A 11A 12P 1P 2P 3P 4P 1 1 1 .76 .80 .80 5P 1 1 1 .76 .80 .80 6P 1 .85 1 .76 .80 .80 7P 8P 9P 10P 11P MM = mile marker; WD = average weekday PTSU status; WE = average weekend PTSU status Table 30 aggregates the data in Table 28 and Table 29 to the daily level. The hours of operation were converted into the equivalent “proportion of time PTSU operating” value shown in the last column of the table. This value represents the proportion of time during the average day that PTSU is operating. If PTSU operated 24 hours a day, every day of the week, then the proportion would equal 1.0. If PTSU facility did not operate during the year, then the proportion would equal 0.0.

53 Table 30. Proportion of a typical day PTSU facility is open. State Facilitya Direction of Travel Milepost Hours of PTSU Operation Proportion Time PTSU Operatingb Start End Weekday Weekend Georgia GA 400 Northbound 8.8 11.9 7.0 0 0.208 19.9 20.7 6.7 0 0.198 Southbound 8.5 9.5 7.0 0 0.208 9.5 11.4 6.2 0 0.184 11.4 15.0 6.3 0 0.188 I–85 Northbound 98.9 100.6 4.0 0 0.119 Hawaii I–H1 Eastbound 5.4 6.8 3.5 0 0.104 8.8 12.9 3.5 0 0.104 15.1 17.9 3.0 0 0.089 Minnesota I–35W Northbound 14.2 16.3 13.2 5.3 0.455 Virginia I–66 Eastbound 57.3 63.9 8.2 3.7 0.287 Westbound 58.6 64.4 8.0 3.5 0.278 I–264 Eastbound 14.9 17.6 1.9 0 0.057 Westbound 14.9 18.8 1.9 0 0.057 I–495 Northbound 42.4 44.6 9.7 0 0.290 a I-35W has dynamic PTSU operation. I-85 and I-66 transitioned from static to dynamic PTSU operation during the study period. All other facilities have static PTSU operation. b Proportion time PTSU operating = (weekday hours × 5/7 + weekend hours × 2/7)/24 Exploratory Analysis As a precursor to model development, the database was examined using simple crash rates to identify the possible association between specific PTSU-related site characteristics and crash rate. The insights obtained from this examination were used to (1) determine which characteristics are likely candidates for representation in the model as an AF and (2) guide the functional form development for individual AFs. The discussion in this section is not intended to indicate conclusive results or recommendations. The proposed predictive models (and associated trends) are documented in a subsequent section. Table 31 documents a comparison of the FI crash rates provided in the previous section. The crash rates for various PTSU operating conditions are compared by ratio to provide some insight into the direction and magnitude of a possible safety influence. In the section titled Summary by Site Type, in Table 25, the FI crash rate for all sites with PTSU operation is 0.36 cr/mvm. The FI crash rate for all sites without PTSU operation is 0.24 cr/mvm. The ratio of these two crash rates is 1.50 (= 0.36/0.24). This ratio implies that FI crashes on a PTSU facility are 50 percent more frequent than those on a facility without PTSU operation, for the same volume level and length. This result is shown in the last column for the first row of Table 31. The PTSU facility can be disaggregated into those sites with a PTSU lane and those sites that are between, upstream of, or downstream of sites with a PTSU lane. This latter group of sites includes a PTSU transition zone where vehicles in the main lanes interact with those vehicles preparing to enter (or having just exited) the PTSU lane. The FI crash rate for sites with a PTSU lane is shown in Table 31 as 0.37 cr/mvm. This rate can be compared to that for sites without PTSU operation (of 0.24 cr/mvm) to produce a crash rate ratio of 1.54. This value is shown the last column of Table 31, near the middle of the table.

54 Table 31. Examination of FI crash rate ratios associated with two conditions. Comparison of Two Conditions (base condition is “without PTSU operation”) FI Crash Rate by Condition Inferred Change in FI Crash Frequencya Without (cr/mvm) With (cr/mvm) … vs. with PTSU operation (PTSU lane or transition zone) 0.24 0.36 1.50 … vs. with PTSU operation on inside shoulder 0.24 0.35 1.46 … vs. with PTSU operation on outside shoulder 0.24 0.36 1.50 … vs. with PTSU lane 0.24 0.37 1.54 … vs. with PTSU transition zone between, upstream of, or downstream of PTSU 0.24 0.30 1.25 Change in Condition From (cr/mvm) To (cr/mvm) Inferred Change in FI Crash Frequencyb Shift PTSU operation from outside to inside shoulder 0.36 0.35 0.97 Add turnout to segment 0.40 0.29 0.73 a Inferred change in crash frequency = crash ratewith / crash ratewithout b Inferred change in crash frequency = crash rateto / crash ratefrom The second-to-last row of Table 31 compares the FI crash rates for PTSU facilities with inside or outside shoulder operation. The rates suggest that a conversion from outside to inside PTSU operation would be associated with a crash rate ratio of 0.97. This ratio implies that FI crashes on a PTSU facility with the PTSU lane on the inside are 3 percent less frequent than those on a facility with the PTSU operation on the outside. The last row of Table 31 compares PTSU facilities that have one or more turnouts with those that do not have turnouts. The FI crash rates listed are from Table 27. Their ratio is 0.73, which implies that there are 27 percent fewer FI crashes on a PTSU facility with turnouts, as compared with a PTSU facility without turnouts. The association between turnout spacing and FI crash rate ratio was examined using the data in Table 12 and Table 26. This association is shown in Figure 2. Each data point in the figure corresponds to one of the facilities listed in each of the two tables. The solid line shown is a line of best fit. The trend suggests that FI crash rate ratio is higher for those PTSU facilities with longer turnout spacing. Figure 3. Relationship between turnout spacing and FI crash rate ratio.

55 The association between the “proportion of time during the average day that PTSU facility operates” and FI crash rate was examined using the data in Table 30 and Table 26. A length-weighted average “hours of operation” was computed for those facilities that operate at a different time period at various sections of their length. The association is shown in Figure 4. Each data point in the figure corresponds to one of the facilities listed in each of the two tables. The solid line shown is a line of best fit. The trend suggests that FI crash rate is about 0.30 cr/mvm when the PTSU facility does not operate. The crash rate increases with increasing hours of operation. Figure 4. Relationship between proportion of time PTSU operating and FI crash rate.

Next: Chapter 5: HSM Predictive Model »
Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Part-time shoulder use is a congestion relief strategy that allows use of the left or right shoulders as travel lanes during some, but not all, hours of the day.

The TRB National Cooperative Highway Research Program's NCHRP Web-Only Document 309: Safety Performance of Part-Time Shoulder Use on Freeways, Volume 2: Conduct of Research Report describes the development of crash prediction models for freeways with PTSU operation.

Supplemental to the document is a Freeway Analysis Tool, which includes BOS Data, S D PTSU Data, and a Prediction Tool, as well as NCHRP Web-Only Document 309: Safety Performance of Part-Time Shoulder Use on Freeways, Volume 1: Informational Guide and Safety Evaluation Guidelines.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!