Safety Evaluation of Access Management Policies and Techniques

CHAPTER 4. DATA COLLECTION

Detailed data were collected for more than 600 mi of corridors across four regions of the United States. The regions included North Carolina (Raleigh, Cary, and Wake Forest), Minnesota (St. Paul and Minneapolis), Northern California (Oakland, Sacramento, San Francisco, and San Jose), and Southern California (Los Angeles and San Diego). This section identifies the procedures for selecting corridors, collecting and verifying data, and merging the various sources of data for analysis.

IDENTIFYING CANDIDATE CORRIDORS

State and local agencies were contacted to solicit candidate corridors for inclusion in the study. Guidance was provided on what constituted suitable corridors to assist the State and local agencies with this process. The critical factors for corridor selection included the following:

• No major construction activity during the study period.
• Availability of crash, traffic volume, and roadway inventory data.
• Arterial functional classification (e.g., principal arterial, minor arterial).
• Display of at least one of the target AM strategies.
• Area type of urban, suburban, or urbanizing.
• Land use of residential, commercial, or mixed-use.

The desired sample size was 150 mi for each region with a relatively equal number of miles for each area type (i.e., urban, suburban, and urbanizing) and land use category (i.e., commercial, mixed-use, and residential). Some agencies were able to provide a relatively large number of miles of candidate corridors, while others were only able to provide a partial list owing to limited staffing availability. Consequently, the study team identified additional corridors, extending existing corridors where possible, and vetted the list with the respective agencies.

The greatest challenge in this task was avoiding areas with construction activity because records of construction activity were not readily available. To overcome this challenge, the study team employed multiple search and confirmation methods. First, the team used historical aerial photography to check for major changes between past and present conditions. (Further details are provided in the Corridor Screening and Supplemental Data Collection and Verification sections below.) Past and present aerial photography was obtained from the United States Geological Survey (USGS), and Google® Earth™ was also employed using the historical imagery feature.

COLLECTING HIGHWAY SAFETY INFORMATION SYSTEM DATA

The Highway Safety Information System (HSIS) contains readily available crash, roadway, and traffic volume data for selected States. By design, the three States included in this study—California, Minnesota, and North Carolina—are all members of HSIS. The HSIS guidebooks were examined, and any potentially useful HSIS variables were requested. At the time of the HSIS data request, the most recent year of available data was 2008. Therefore, the study period for this project includes 2006 to 2008. Appendix A identifies the variables received from the roadway, crash, and vehicle files for each of the three States.

The HSIS data provided a starting point for the corridor-level databases and were also used to screen potential corridors. The team checked each candidate corridor for the availability of HSIS data; if a candidate corridor could not be found within the HSIS data, it was immediately rejected. To perform queries within HSIS data, the route number and mileposts were needed for each corridor. This task was accomplished with the help of the respective agencies.

CORRIDOR SCREENING

Three rounds of screening were employed to ensure no major construction activities or changes occurred along the corridors during the study period. As described previously, initial screening was conducted by the agencies through a review of construction records.

The study team performed a second phase of screening using HSIS data. The HSIS roadway files from 2006 to 2008 were compared to each other to detect changes that would indicate construction activity (e.g., number of lanes, lane width, shoulder width, median type, median width, and mileposts). In a small number of cases, changes were detected. Of these, the most common were small changes in median width. However, the changes from year to year appear to be the result of slight differences in data coding and not actual physical changes. As a result, no corridors were eliminated by this process.

The team performed a third round of screening using historical aerial imagery. The team identified high-resolution aerial imagery for the identified corridors from the USGS National Seamless Server. By comparing historical aerial images with current conditions, the team was able to identify where changes had taken place during the study period.

Where major changes were identified, the team did not select the candidate corridor for inclusion unless the start and end dates of construction activity could be determined. If the construction activity did not include the entire study period, then the corridor was included in the study using a subset of data that did not include the construction period. Examples of major changes include the addition of through lanes, the construction of a new interchange or new ramps to an existing interchange, and the development of large commercial properties or subdivisions along a corridor.

Corridors with minor changes remained in the dataset for analysis. An example of a minor change is the addition of a single commercial or residential driveway. The team did not want to eliminate an entire corridor from the study merely because of these types of isolated changes. Two approaches were considered to deal with instances of isolated changes. The preferred approach was to contact the agency responsible for the roadway and obtain maintenance records for that specific location. While this was the preferred approach, it was dependent on the availability of records and assistance from the State. The alternative approach was to exclude the section of the corridor around the change (e.g., new signal). In other words, the team segmented the corridor to avoid the section with the change.

SUPPLEMENTAL DATA COLLECTION AND VERIFICATION

The HSIS data as well as the area type and land use information obtained during the corridor identification provided a starting point for the database of features needed for this project. However, all information had to be verified and augmented with additional data from aerial photography and field visits. As described above, aerial imagery was obtained for corridor screening purposes (i.e., checking for changes indicating construction activity). The aerial imagery was also used to verify the land use, number of through lanes, and median type for each corridor.

In North Carolina, the team superimposed the HSIS data on aerial photos, using ArcMap™ to expedite the verification process. The objective was to visually display the median type and number of lanes from HSIS geospatially using color-coded symbols. This way, the HSIS data could be compared rapidly with the conditions shown by the high-resolution imagery. Figure 12 presents an example of this approach for one of the corridors (Hammond Road). The images on the left show the HSIS data, which were verified using aerial imagery of the corresponding segments on the right.

Source: FHWA	© U.S. Geological Survey.
A. Illustration of number of lanes.	B. Photo of number of lanes.
Source: FHWA	© U.S. Geological Survey.
C. Illustration of median type.	D. Photo of median type.

Figure 12. Illustrations and photos. Verifying HSIS data with aerial photos.⁽¹⁵⁾

The process of superimposing HSIS data on the aerial imagery was time-consuming and labor-intensive. In addition, the study team identified inconsistencies in the HSIS data, including primary variables such as number of lanes, median type, and posted speed. Subsequently, it was determined that this information should be obtained from field visits and aerial imagery rather than the HSIS database for the remaining regions.

For California and Minnesota, the basic roadway characteristics (i.e., number of lanes and median type) were obtained from aerial imagery that was taken during the study period. The data were then compared with video logs obtained during the field visits to identify any differences. Figure 13 displays an example of a high-resolution aerial image from one of the study corridors, and figure 14 shows the corresponding street view from the field review video log. If inconsistencies were identified between the imagery and video, the study team explored the segment further to determine when the change occurred.

©2013 Google® ©2013 Europa Technologies. Modified by authors.
Note: The double-headed arrow was inserted by authors to indicate the direction of which the image in figure 14 was captured.

Figure 13. Screen shot. Verifying data with aerial imagery.⁽¹⁶⁾

Source: FHWA.

Figure 14. Photo. Verifying data with video.

Aside from median type and number of lanes, aerial imagery was also used to collect information that was unavailable from HSIS. These data included the frontage type (fully, partially, or undeveloped), presence of a frontage or backage road, extent of internal cross connectivity, and condition of pavement markings (poor or not). Collecting this information was straightforward because these parameters did not change frequently along a given corridor. Consequently, the beginning and ending mileposts were rapidly noted for the above parameters.

Finally, the aerial imagery was used to collect information regarding access points, including the location, type, and density. Specifically, unsignalized access spacing, driveway spacing, interchange spacing, median spacing, and corner clearance were obtained from aerial imagery. However, collecting data on these features was more complicated and required the setup of an ArcGIS™ database as discussed in the following section.

SETUP OF GEOGRAPHIC INFORMATION SYSTEM (GIS) DATABASE TO FACILITATE DATA COLLECTION

ArcGIS™ feature classes were created for signalized intersections, unsignalized intersections, driveways, and medians. This allowed data collectors to insert symbols representing these objects on the aerial images of the corridors. Data fields were created for each object so its characteristics could be noted. The characteristics collected for each object are summarized in table 5.

Table 5. Objects and characteristics coded in ArcGIS™.

Figure 15 is an example of these objects for a 1-mi section of a study corridor (California Route1). In total, the study corridors contained more than 1,500 signalized intersections, 3,500 unsignalized intersections, and 15,000 driveways.

©U.S. Geological Survey.
Note: Overlay symbols and legend were inserted by project team to indicate driveway types and intersection traffic control.

Figure 15. Image. Example of point objects for a 1-mi corridor.⁽¹⁷⁾

FIELD DATA COLLECTION AND VERIFICATION

Field visits were used to collect additional information and verify the data obtained from HSIS and the aerial imagery. The HSIS data and aerial imagery provided many but not all of the required features data. The following information was still needed:

• Presence of street lighting.
• Presence of visual clutter (e.g., excessive signage, roadside advertisements, banners).
• Detailed signalized intersection data, including the following:
  
  
  • Turning restrictions on mainline.
  • Turning restrictions on cross street.
  • Left-turn phasing.

In the first two regions (North Carolina and Northern California), the ArcMap™ files (aerial imagery, layers, and feature classes) were transferred to a touch-screen tablet personal computer (PC) for fieldwork. The ArcMap™ files were used to create ArcPad™ versions, which enabled the team to insert corridor features onto the aerial imagery. The tablet PC was linked with a Global Positioning System (GPS) device to identify the precise location in the field relative to the aerial imagery. This method allowed the team to “drop” specific features (e.g., driveways and intersections) on the aerial images with the tablet PC and enter detailed characteristics for each feature.

Figure 16 shows the tablet PC and GPS equipment used for the field data collection and verification. The engineer is holding the tablet PC, noting specific characteristics of the intersection, while the GPS device is mounted atop a backpack system. Note that this photo was taken while verifying the functionality of the equipment; the actual field data collection was performed from a vehicle with the equipment mounted inside the vehicle.

Source: FHWA.

Figure 16. Photo. Data collection equipment used for field visit.

This method proved to be extremely cumbersome and time consuming. It was anticipated that the GPS unit would help synchronize the location of the vehicle with the location on the GIS map. This was generally not the case because the GPS unit needed several satellites in range to provide the level of accuracy needed for this project. In an attempt to overcome the satellite coverage issue, satellite “schedules” were used to identify the magnitude and time of coverage for the study corridors. However, the optimal schedule of the satellites did not always fit with the data collection times and routes. Therefore, the data collection procedure was revised.

The revised procedure included narrated video logs to document the attributes of each corridor, and the data were then entered in a spreadsheet in an office setting. Using this method, it was possible to conduct the field reviews with one driver and a digital video camera, which was mounted to the front windshield facing the direction of travel. The video captured the specific corridor details while the driver noted changes in the characteristics (e.g., number of lanes, median type, speed limit, and lighting). The narration was particularly useful for the data reduction process because the data analyst could listen to the video and readily identify changes. The analyst could then stop the video and enter the data in a spreadsheet as changes occurred along the route. To help ensure consistency in the field data collection process, a procedure (provided in appendix B) was formulated and provided to every team member participating in the process.

POST PROCESSING

From the previous data collection tasks, data were obtained in various formats. Some information was stored in Microsoft® Word documents (e.g., area type, land use, and frontage type), specifying the beginning and ending points. HSIS data were provided in Microsoft® Excel format. General corridor characteristics and specific attributes for signalized intersections were identified in the video logs. Other information was identified and stored in the form of ArcGIS™ feature datasets (e.g., intersections and driveways). The corner clearances and ramp spacing information required specific measurements to be made using ArcGIS™.

Transforming all these data sources into a well-integrated database to serve as the basis for statistical computation required a post-processing step. In this step, multiple tasks were performed, and the following identifies the general sequence of tasks:

1. General corridor information. A spreadsheet was developed and populated to identify the general corridor data, including a unique corridor identity (ID), route number, beginning milepost, ending milepost, beginning cross street, ending cross street, area type, and land use for each corridor. The video logs were then used to populate additional data fields for each corridor, including the type of operation, number of lanes, median type, posted speed limit, lighting presence, condition of pavement markings, and visual clutter. Data from the aerial images were then added, including the frontage type, internal cross connectivity, and presence of a frontage/backage road. HSIS data were then appended to identify the specific fields needed to match crash data.
   
   
2. Segmentation. Within each corridor, new segment links were created anytime one or more of the variables changed. The beginning and ending mileposts were identified for each individual link. The links were then aggregated based on area type and land use to create the study corridors. Specifically, each study corridor was consistent with respect to area type and land use, but other variables were allowed to change. Each study corridor was assigned a unique ID, and spreadsheets were then created for each corridor and linked using the unique ID.
   
   
3. Driveway spacing. The ArcGIS™ files were used to query the driveway information for each study corridor. The total number of driveways was identified for each study corridor as a whole and by direction. The driveway information was also provided by direction for commercial and residential driveways and also for full- and limited-movement driveways.
   
   
4. Unsignalized intersection spacing. The ArcGIS™ files were used to query the unsignalized intersection information for each study corridor. Several variables were created, including the total number of unsignalized intersections, minimum spacing, maximum spacing, number with right-turn lanes, number with left-turn lanes, and number of unsignalized intersections by legs (i.e., three-legged, four-legged, and five-legged) and turning restrictions (full movement, right-in/right-out, and left-from-major-only).
   
   
5. Signalized intersection spacing and corner clearance. The ArcGIS™ files were used to query the signalized intersection information for each study corridor. Several variables were created, including the total number of signalized intersections, minimum spacing, maximum spacing, number with right-turn lanes, number with left-turn lanes, number of signalized intersections by legs (i.e., three-legged, four-legged, and five-legged), number of interchange-related intersections, and number of signalized intersections by turning restrictions (full or limited movement). The corner clearance was also identified for each signalized intersection, including measurements of the distance to the nearest driveway for the approach and receiving lanes in both directions.
   
   
6. Median openings and crossovers. The ArcGIS™ files were used to query the median opening information for each study corridor with a median. The total number of median openings was identified for each study corridor along with the number of median openings with and without left-turn lanes.
   
   
7. Interchange-related spacing. The ArcGIS™ files were used to query the interchange-related spacing information for each study corridor. Specifically, the ramp location and type were identified for each interchange. For off-ramps, the distances to the first downstream driveway (on right), the first median opening allowing left turns, and the first major intersection were measured. For on-ramps, the distances from the last driveway (on right) and major intersection were measured.

As noted previously, the segmentation process required that homogeneous segment links be combined into study corridors to achieve a reasonable length for each site. In this way, some variables were summed over all links making up a study corridor (e.g., number of driveways). In other cases, new variables reflecting the percentage of the total length were created (e.g., number of lanes). This process also required that the county route and milepost information for each link be retained for matching with the crash and traffic volume data because, in some cases, the county routes change within a corridor. The crash data from 2006 to 2008 obtained from HSIS were queried to match crashes to the study sites. The annual average daily traffic (AADT) and percentage truck variables were calculated as weighted averages, weighting by the lengths of the links within a corridor. AADT represents the bidirectional traffic volume for a segment.