Step 8 in the scalable risk assessment process is to calculate risk values based on the outputs from previous steps. That is, Step 6 provides exposure estimates and Step 7 provides observed crashes, expected crashes, or additional risk indicators that are then used to calculate final risk values at the geographic scale chosen in previous steps.

Case studies are provided in this chapter to tie together the eight steps described in this guide. The case studies are based on actual examples of risk assessment for pedestrians and bicyclists.

Case Study: Pedestrian Risk Assessment in Michigan

The Michigan DOT partnered with the University of Michigan Transportation Research Institute to develop a risk assessment tool (https://www.cmisst.org/tools-resources/) for pedestrian crashes for all 83 counties throughout the state. Michigan DOT’s goal was to create a risk score, based on mapping crashes and the risk characteristics, for a defined area or network for the entire state of Michigan. We now present a fictional case example based on this project.

This case example focuses on pedestrian risk assessment to identify corridors in Detroit Michigan in need of countermeasures. Often the characteristics that make walking safe (or unsafe) persist over space. For example, along busy roads, land use features like business districts or the lack of lighting are often consistent over space. Due to this spatial continuity, transportation engineers often would like to improve the facilities in an entire corridor, not just one location.

Step 1: Determine use(s) of risk values. Network screening, Area Based.

The MDOT engineers were interested in estimating pedestrian risks to identify corridors in need of countermeasures.

Step 2: Select Geographic Scale. Areawide->Network–>Corridor

MDOT project team was interested analyzing risk at the corridor level.

Step 3: Select risk definition. Expected Crashes

The definition of risk combined various risk indicators to estimate the expected number of pedestrian-vehicle crashes.

Step 4: Select exposure measure. Trips made

The project team measured exposure in trips per day in a Pedestrian Analysis Zone (PAZ), which is a 400m x 400m unit of analysis. These units are aggregated up to the level of the corridor.

Step 5: Select analytic method to estimate exposure.

Demand Estimation –>Pedestrian Trip generation and flow models

The analytic method used a statewide travel survey, land-use data, and household characteristics to generate pedestrian trips at the PAZ level.

STEP 6. Use analytic method to estimate selected exposure measure

The project team used the Michigan household travel survey (MTC III) to fit our trip production and destination choice models. We divided trips into five categories, namely home-based other (HBOther), home-based shopping (HBShopping), home-based school (HBSchool) and non-home-based other (NHBO) and non-home-based work (NHBW) and run the regression separately. We highlight the results for home based other (HBO) trips. For home-based trips, we estimate the number of trips per day at the household level using a negative binomial regression of the of the form,

Number of HB walking trips = f(number of households+household characteristics + built environment).

Below are representative regression results for the “home-based other” trip purpose.

After completing all of the steps in the methodology, we obtained pedestrian exposure estimates. For more detailed information about the methodology, see (Cai et. al., 2018). Figure 33 shows the results for Wayne county Michigan where Detroit is located.

STEP 7. Compile other required data.

The approach required many other data sources to calculate the risk values. The schematic below shows the various data sources used in the risk model.

STEP 8. Calculate risk values

We used the Empirical Bayes framework from the HSM to create customized safety performance functions (SPFs) for both bicyclist and pedestrians (Cai et al., 2018).

The MDOT engineers also wanted to analyze corridors. Thus we applied the Level-of-Service-of-Safety (LOSS) metric from the highway safety manual. LOSS divides the risk-scores of candidate areas into 4 categories based on its standard deviation from the average risk score (Kononov et al., 2003; Kononov et al., 2015). The areas in the highest quantile are the most dangerous for pedestrians. Figure 36 shows that LOSS map for pedestrian risk. In order to calculate the risk values of the corridors, we add together the risk score of each PAZ in the Gratiot corridor to arrive at a cumulative risk score of 50. The corresponding cumulative risk fir the Woodward corridor was 91.

Lessons Learned

• Pedestrian Trip generation and flow models require significant technical capabilities.
• Methods from the Highway Safety Manual can be modified and used for non-motorized risk assessment.
• While these results are promising, extensive work is required to the validate exposure and risk models in order to integrate them into MDOT’s business processes.

Case Study Example for Arizona DOT Bicycle Safety Action Plan

The scalable risk assessment methods were applied in Arizona as part of the Arizona DOT Bicycle Safety Action Plan as part of a statewide review of segments and intersection with a high priority for bicycle safety improvement projects.

The study team applied a network planning analysis approach to identify priority corridor locations and countermeasures to provide safety improvements for bicyclists. Emphasis was placed on providing safe conditions for bicycle travel all along a corridor (segment) and within the bicycle travel network. To apply this network analysis approach to the 2018 Bicycle Safety Action Plan Update for the Arizona State Highway System, high-crash intersections and segments and high-crash potential segments were grouped into Priority Locations. A Priority Location may consist of one or more high-crash segments, intersection, or high-crash potential segments. The high crash potential segments were identified through a risk assessment methodology. These Priority Locations comprise 94% of the high-crash segments, 100% of the high-crash intersections, and 74% of the high-crash potential segments.

The approach included an initial review of high bicycle crash locations on the Arizona State Highway System. These locations were identified using GIS and subsequently verified by visual inspection. The locations are separated into highway segments and intersections/interchanges. A high-crash intersection/interchange and segment location includes at least three bicycle crashes within the five-year period. In addition, bicycle count data were included where available for the intersection/interchange and segment location. The count data were from the recent Arizona DOT efforts to develop a bicyclist and pedestrian count strategy plan for the State Highway System. The purpose of the counts is to provide insight into the bicycle exposure on these selected high-crash locations.

A key element of improving bicycle safety in Arizona is to proactively identify locations where bicycle improvements are needed, leading to projects to address the need. This section introduces a risk assessment methodology that can assist ADOT in identifying state highway segments and intersections where investment can help to lower the risk of bicycle crashes. The proposed methodology is similar to the process used in the 2016 ADOT Pedestrian Safety Action Plan. The assessment methodology represents an approach through which high-probability segments can be identified and addressed before bicyclist/motor-vehicle crashes occur.

Methodology

The methodology considers factors that are frequently identified as contributing factors or environmental/facility conditions that are common to bicycle crashes on the SHS. These factors are associated with the roadway facilities’ existing conditions that relate to the absence of sufficient bicycle accommodation and bicycle demand as data is available. Bicycle demand can be estimated based on the facilities’ proximity to specific land uses such as institutional areas that include schools, colleges, or universities, or being part of a known popular cycling route or corridor. Strava is a tool that can be used as a tool to help identify the popularity of cycling routes and corridors, although the Strava app data may be used more focused by recreational bicyclists.

Application of the methodology occurred through a GIS-based screening that utilized available statewide GIS data to identify and screen potential SHS locations where bicycle facilities should be considered, consistent with an established set of risk criteria. Note that interstates were excluded from the screening as the intent of this is application was to identify and direct resources to where they will be the most effective.

*The USBR is not so much as risk factor, but is used to gain higher priority for improvements with the designation.

Screening Results

A scale was developed based on the distribution of the overall scores assigned to the SHS. The scale is defined in Table 29. A total of 31 higher-risk locations were identified and are shown in Figure 37.

Lessons Learned

• Although there was a desire to incorporate an exposure measure into the risk assessment, it was not appropriate at the statewide level for this project. Bicycle count data would have been needed at enough locations to estimate for the entire network and that is not feasible at this time.
• The use of population data and/or national census results were evaluated but determined not to be an appropriate application for exposure on a statewide basis.
• The use of population density as a surrogate factor related to exposure within the risk assessment is being evaluated.