A recent pilot effort in one central Virginia region—the Roanoke Valley-Alleghany Regional Commission—sought to strengthen the incorporation of safety into the transportation-planning process.
Based on that experience, this article identifies four steps that a transportation planner with limited computing and staff resources may undertake to support such integration. The steps are: (1) obtaining a regionally complete crash data set; (2) identifying crash clusters; (3) identifying locations that are over-represented in terms of crash frequency; and (4) applying safety-related performance measures that support a comprehensive transportation plan and do not rely exclusively on crash data.
Although they represent only modest progress toward linking safety and planning, these steps illustrate a tangible way to start such a linkage.
The rationale for linking safety and planning is that substantial crash reductions may be achieved by considering safety explicitly within the planning process. For example, methods for predicting crash frequency by transportation analysis zone may be integrated with travel-demand forecasts, leading to prioritization of improvements based on their crash-reduction potential, just as modelers explicitly consider improvements based on ability to reduce congestion. Ideally, safety considerations could fully inform the elements of the transportation-planning process, such as establishing a vision statement, selecting performance measures and prioritizing investments based on those measures.
Searching for links
However, establishing this safety-planning connection can be difficult organizationally because individuals with different skill sets (e.g., convening stakeholders for a public meeting or performing statistical analysis) may be housed in different functional units. One approach to initiate this integration is to perform a few simpler analyses that link planning and safety in a limited manner and then use any successes from such steps to enhance safety-planning integration.
The four steps described above are not a complete safety-planning integration, rather they provide a tangible way to initiate such a process.
Regional entities such as metropolitan planning organizations (MPOs) or state department of transportation construction districts may receive crash data from multiple jurisdictions. In some cases, the quality, completeness or format of location information for each crash vary depending on the jurisdiction or year of the data. To facilitate such sharing in the pilot effort, it was necessary to develop a 10-step procedure to incorporate crash data into a geographic information system (GIS).
For instance, there were some instances where crashes appeared to occur at integer mileposts (e.g., milepost 7.0) despite locations being reported to the nearest hundredth of a mile (e.g., milepost 7.18).The problem involved the order in which individual crash records were imported into the GIS: If the first imported crash truly occurred at an integer milepost, then the software assumed this field consisted solely of integers rather than decimals. (The problem was solved by formatting all records as decimals [e.g., a crash at milepost 7 became 7.00.])
As another example of a step that was documented, the state crash database includes a field called CollisionType for which a code of 13 indicates a bicycle. However, the use of this field to identify all bicycle crashes will not give a complete answer. Instead, one also must identify bicycle crashes based on an additional variable (the type of vehicle); thus, the structured query language (SQL) statement needed to identify such crashes was included in the documentation.
Visual inspection in GIS of select crashes showed up to a 50-ft difference between the location of the crash and the line segment representing the roadway or some differences between the location given on the crash report and the location indicated in the GIS. Such differences suggest that the GIS representation of crashes may be useful for determining areas of interest but may need to be supplemented with a review of crash reports if detailed engineering studies are undertaken.
Crashes within a certain distance of intersections (usually 150 to 250 ft) can be defined as a cluster. The question of how to define a cluster at a nonintersection location is not as clear.
For example, if several crashes are located within 100 ft of one another, is this a cluster? If several crashes are located within 1,000 ft of one another, is this also a cluster? A threshold distance that defines a crash cluster may be calculated, simplified based on a relatively strong (p = 0.001) confidence level. Crashes below the calculated distance threshold were presumed to be clustered; crashes above the threshold were not presumed to be clustered. For example, for an area of 429 million sq ft that had a total of 292 nonintersection crashes, the equation suggests that the clustering distance is 544 ft.
One may also stratify crashes by severity, alcohol use, collision type or other explanatory factors. For example, for enforcement purposes, clusters of intersection crashes involving younger drivers, defined as persons aged 18 or younger, may be identified as shown in Figure 1.
Hotspots and countermeasures
An expected-value analysis may be used to identify the number of crash clusters that should be studied in a given region. For the particular case study area, there were 519 intersections with most (435) having no injury crashes such that the 173 injury crashes were spread among the remaining 84 intersections. An expected-value analysis suggested that intersections with three or more injury crashes might warrant further attention. Such a criterion indicates there were 24 intersections (with three or more injury crashes).
At these locations, countermeasures may be identified based on the literature. For example, the two intersections shown in Figure 2 were both in the list of 24 high-injury crash locations. Both intersections had substantial angle and rear-end crashes. The crash diagram and narrative for the intersection on the right showed that four rear-end crashes had occurred when vehicles slowed or stopped specifically for the signal. Thus, steps to improve signal visibility might be appropriate at that intersection. Advance warning signals may be suitable at both intersections, given that rear-end crashes occurred when drivers may not have been aware of the intersection or the queue resulting from the intersection.
The Safe Routes to School (SRTS) initiative is emphasized in the MPO’s draft 2035 long-range transportation plan and suggests that an assessment of crash risk for pedestrians and bicyclists is appropriate. Although there was only one pedestrian- or bicycle-related crash within a mile of each of several schools examined in the case study area, this lack of crashes may result from a lack of nonmotorized travel rather than safe conditions for such travel.
Performance measures can be used to identify specific deficiencies whose correction would make travel more conducive to pedestrians. Such measures may be simple (e.g., the presence of sidewalks, the speed limit or an ADT) or more involved (e.g., the bicycle level of service). For example, a section of a facility that has a bicycle level of service of E (where A is highest and F is lowest), no sidewalks and a speed limit of 45 mph may be a deficient section in terms of nonmotorized travel.
With judgment, performance measures may also help prioritize improvements. For example, the ratio of the population to total roadway mileage for a given census block may be computed. Conceptually, in an area where no sidewalks are present and all roads are equally beneficial to pedestrians, an investment of a given dollar amount in a block having a higher ratio would serve more residents than the same level of investment in a block having a lower ratio. Yet in practice, such a ratio has limitations: Not all roadway facilities have the same need for sidewalks; not all populations will use sidewalks to the same degree; sidewalks benefit the pedestrians passing through an area in addition to those residing there; and there may be blocks where retrofitting some, but not all, roads with sidewalks is effective. Accordingly, population density can also be examined.
With those limitations in mind, Figure 3, left, suggests that areas near the periphery of the 1-mile radius from the school—that is, the northwest and southeast areas—would benefit from pedestrian facilities, whereas Figure 3, right, suggests areas to the southeast and to the south. Both parts of Figure 3, however, suggest that the region to the southeast of Burlington Elementary School (circled) appears promising for determining where pedestrian facilities may serve a large number of people.
The examples presented here provide one way of bringing crash considerations into the planning process. The countermeasures at specific locations may be used to address shorter-term problems, whereas Figure 3 shows how safety considerations may impact longer-term planned development or infrastructure retrofits.
Several individuals provided insights for this effort: Cristina Finch and Jake Gilmer (RVARC), Ivan Rucker (Federal Highway Administration), and Linda Evans, Michael Gray, Robin Grier, Jim Hopkins, Ning Li, In-Kyu Lim, Amy O’Leary, Bob Rasmussen, and Tien Simmons (Virginia Department of Transportation). TM&E