Safe System for Intersections Method Scoring: Comparing Intersection Alternatives Based on Exposure,Kinetic Energy Management,and Movement Complexity

Abstract

Transportation agencies across the U.S. are increasingly looking to the Safe System approach as a path toward achieving Vision Zero and other similar road safety performance goals. Vision Zero and the Safe System approach seek to eliminate traffic fatalities and serious injuries. The Federal Highway Administration (FHWA) funded an effort to create a Safe System for Intersections (SSI) method to provide a technical basis by which road planners and designers could apply Safe System principles on intersection projects, detailed in the report A Safe System-Based Framework and Analytical Methodology for Assessing Intersections. Applying the SSI method results in a set of SSI scores for each intersection design under consideration. The SSI scores characterize the extent to which an intersection alternative aligns with Safe System principles. This paper builds on the content presented in the FHWA report and describes how the key SSI method components of user exposure, conflict point severity, and intersection movement complexity are combined to produce the SSI scores. This additional background and detail on the development of the SSI scores will support future applications of the method and the identification of future research needs and method improvements.

Keywords

infrastructure roadway design roundabouts and other intersection design and control strategies safety safety performance and analysis transportation safety management systems quantitative analysis safe systems

The Safe System approach represents a paradigm shift in road safety management. The Federal Highway Administration (FHWA) has identified the Safe System approach as a means for realizing the goals of programs and partnerships such as Vision Zero and the Road to Zero: eliminating fatalities and serious injuries from the U.S. road system ( 1 ). The Safe System approach is based on the reality that humans make mistakes and the human body is vulnerable. Because of this, it recognizes that crashes will occur and focuses on managing the mechanical forces in those crashes to a level that will not result in fatalities and serious injuries to the people involved ( 2 ). This concept of managing the kinetic energy in a crash is foundational to the Safe System approach. It also underscores that responsibility to eliminate traffic fatalities and serious injuries is shared among all participants in the transportation system—from road users to managing agencies, designers to vehicle manufacturers—and that the safety of that system must be pursued proactively. Figure 1 depicts the six Safe System principles and five key elements, as identified by FHWA.

Figure 1.

The Safe System approach incorporates six principles and five elements ( 1 ).

Some Safe System principles have become part of the state of practice in U.S. road safety management, to some extent, over the last several decades. However, much of the early experience in developing and implementing the Safe System approach has occurred elsewhere. Sweden was among the first countries to implement a Safe System approach to national road safety management, passing the Road Traffic Safety Bill in 1997. The bill was founded on Vision Zero and based on the idea that a zero-fatality target was the only ethically justifiable target for road safety ( 3 ). Early Vision Zero documentation identified the foundational concept that mobility and safety cannot be “traded off” against each other, but instead the level of mobility follows from achieving the desired level of safety. This is a key paradigm shift in implementing Vision Zero.

The Safe System approach also contends that the desired level of safety cannot be attained using traditional road safety management techniques. Road design has historically accounted for mistakes on the part of road users, for example by widening roadways to allow drivers more space to recover in the event of a deviation in trajectory. However, research shows that wider roads often lead to higher vehicle speeds, increasing the kinetic energy in the occasion of a crash and thus the likelihood of a fatality or serious injury ( 4 ). Safe-System-based roadway design also seeks to accommodate road user mistakes—albeit in a different way—through managing the kinetic energy that is generated in and transferred to the human body in a crash. Belin et al. noted that this kinetic energy management approach is rooted in established scientific research, citing foundational kinetic energy research by DeHaven, Haddon, and Robertson ( 5 – 8 ). Sweden made this scientific framework the basis of its road design and management programs through the Road Traffic Safety Bill and Vision Zero.

Similarly, the Netherlands developed a vision for a Safe System approach, called Sustainable Safety, in the early 1990s. It was enacted through the Sustainable Safety Start-Up Program in 1997 ( 9 ). The goal of Sustainable Safety is to prevent traffic fatalities and serious injuries and it has five fundamental principles, as summarized in Table 1.

Table 1.

Five Sustainable Safety Principles ( 9 )

Sustainable safety principle	Description
Functionality of roads	“Monofunctional” roads (i.e., roads have a single function) as either through roads, distributor roads, or access roads in a road hierarchy
Homogeneity of mass and/or speed and direction	Equality of speed, direction, and mass at moderate and high speeds
Predictability of road user behavior by recognizable road design	Road environment and road user behavior that support road user expectations through consistency and continuity of road design
Forgiveness of road design and of road users	Injury limitation through a forgiving road environment and anticipation of road user behavior
State awareness by the road user	Ability to assess one’s capability to handle the driving task

The Australian Safety Council adopted the principle of the Safe System approach in 2004, based on Vision Zero and Sustainable Safety ( 10 ). To advance the understanding of the importance of kinetic energy management, Corben et al. developed the conceptual Kinetic Energy Management Model (KEMM) ( 11 ). KEMM featured five “layers” as shown in Figure 2. New Zealand subsequently drew on experience in Australia to develop Safer Journeys 2020: New Zealand’s Road Safety Strategy 2010–2020, based on a Safe System approach ( 12 ).

Figure 2.

The five layers of protection in the Kinetic Energy Management Model (KEMM), as adapted from Corben et al. ( 11 ).

Speed and speed management—along with other factors such as collision angle, road user behavior, vehicle characteristics, and roadway characteristics—are critical determinants of the extent to which a Safe System can be realized. Collision speed is the key determinant of kinetic energy in a crash; kinetic energy is related to mass and the square of velocity. Consolidating concepts of Layers 1, 2, and 3 of KEMM, Tingvall and Haworth provided the speed thresholds shown in Table 2 as representing an “inherently safe system” ( 2 ).

Table 2.

Possible Long-Term Maximum Travel Speeds Related to Infrastructure, Given Best Practices in Vehicle Design, and 100% Restraint Use ( 2 )

Type of infrastructure and traffic	Possible travel speed, km/h (mph)
Locations with possible conflicts between pedestrians and cars	30 km/h (20 mph)
Intersections with possible side impacts between cars	50 km/h (30 mph)
Roads with possible frontal impacts between cars	70 km/h (45 mph)
Roads with no possibility of a side impact or frontal impact (only impact with the infrastructure)	100+ km/h (60+ mph)

As locations with concentrations of planned points of conflict between different road users, at-grade intersections are particularly significant in the overall landscape of safety performance across the U.S. transportation system. In 2019, 28.2% of traffic fatalities in the U.S. happened at intersections, equating to 10,180 fatalities ( 13 ). To provide a technical basis by which intersection planners and designers can apply Safe System principles as part of intersection project planning and alternatives screening, FHWA funded the development of the Safe System for Intersections (SSI) method in 2021. The SSI method is documented in the research report A Safe System-Based Framework and Analytical Methodology for Assessing Intersections ( 14 ). As this report notes, the SSI method was designed to be readily implementable within typical U.S. project development processes and to use commonly available data inputs. It incorporates the concepts of conflict point identification and classification, exposure, conflict point severity based on the estimated level of kinetic energy transfer and management in a crash, and movement complexity. Applying the method produces a set of SSI scores that practitioners and researchers can use to assess how a given intersection alternative aligns with Safe System principles. The SSI method is not based on observed or predicted crash occurrence. Rather, it is designed to complement existing crash-based safety methods or to provide metrics that consider safety in cases where it is not possible to conduct crash-based analyses, such as for atypical or emerging intersection concepts that are not yet addressed by crash-based methods. The SSI method represents the first objective analysis approach that captures key Safe System concepts and is readily implementable in common intersection planning and design processes.

Objective and Scope

The objective of this paper is to present the approach by which the SSI method combines the concepts and resulting metrics related to user exposure at conflict points, conflict point severity, and movement complexity to create an interpretable SSI score. The paper begins with an overview of the method itself, drawing on the content originally presented in the FHWA report A Safe System-Based Framework and Analytical Methodology for Assessing Intersections ( 14 ), and then describes the steps behind the production of the SSI scores and presents an example application of the scoring methodology. The FHWA report that originally presented the SSI method provides detailed content on the concepts surrounding the method, its practical application (including thorough example calculations), and how to interpret its results ( 14 ). However, it did not describe the process used to develop the resulting SSI scores. This paper provides practitioners and researchers with additional background and detail on the mechanics of the SSI scores to support future applications of the method and the identification of future research needs and method improvements.

SSI Method Overview

This section of the paper provides a high-level overview of the SSI method to set the stage for the main purpose of the paper. Readers can reference the full FHWA report for additional detail and explanation ( 14 ).

The SSI method provides a way for intersection planners and designers to apply the concepts of the Safe System approach and kinetic energy management to intersection projects, with the goal of informing project decisions that contribute to reducing fatalities and serious injuries for all road users. As the first such method developed in the U.S. (i.e., a “version 1.0”), it will likely spur additional method expansions and enhancements. This first method focused on implementation within a Stage I Intersection Control Evaluation (ICE), when design alternatives are assessed to determined how well they align with a project’s purpose and need. The SSI method uses commonly available project-level data inputs to be easily implementable by practitioners at this early stage of project development. These include annual average daily traffic (AADT) volume in vehicles per day (vpd), posted speed limit, and number of through lanes on each of the intersecting roads. There are several other optional inputs that are central to Safe System principles but have not historically been as utilized or explored by the research and practitioner communities. If available, the optional inputs will improve the precision of the analysis. They include daily turning movement counts, nonmotorized AADT estimates, vehicle speeds for individual movements, left turn traffic signal phasing, and others. The SSI method offers default assumptions for these optional inputs if they are not available to the method user.

The SSI method has four main components that include different analysis steps: (1) conflict point identification and classification, (2) conflict point exposure, (3) conflict point severity, and (4) movement complexity.

The SSI method identifies conflict points as any points where road user paths coincide ( 15 ). Since this first version of an SSI method is intended for use in alternatives screening, when exact lane arrangements may not be known, the SSI method identifies conflict points on a movement basis as opposed to a lane-by-lane basis. The method assigns conflict points to one of four categories: crossing, merging, diverging, and nonmotorized. Figure 3 depicts the identification and classification of movement-based conflict points for a median U-turn (MUT) intersection.

Figure 3.

Diagram of movement-based conflict points for a median U-turn (MUT) intersection ( 14 ).

Once the conflict points are identified and classified, the next step of the SSI method estimates the exposure through each conflict point. The SSI method computes an exposure index for each conflict point based on work by Hakkert and Mahalel ( 16 ). The exposure index, $I_{c}$ , for a particular conflict point $c$ , is the product of the daily user volumes of the movements passing through that conflict point ( $Q_{1, c}$ and $Q_{2, c}$ ). The exposure index or indices can be examined as a sum across all conflict points at the intersection, or by individual conflict point type (e.g., exposure through all crossing conflict points, exposure through all nonmotorized conflict points).

Conflict point severity in the SSI method is the estimated probability that a crash at a given conflict point involving the conflicting users making typical movements though that conflict point will result in a fatal or serious injury. The SSI method defines this as $P (FSI)$ . The method defines a fatal or serious injury as any injury with a score of three or higher on the maximum abbreviated injury scale (MAIS). The SSI method computes conflict point severity for vehicle-vehicle conflict points based on the estimated speeds of the two conflicting movements and the angle between them. The method computes conflict point severity for nonmotorized conflict points using the estimated speed of vehicles passing through those points. The $P (FSI)$ estimates are based on models from previous research ( 17 – 19 ).

The final component of the SSI method includes a set of steps to characterize the complexity of different movements for all users of the intersection. These steps capture user behavior and workload for different movements through the intersection, and how these concepts are affected by the intersection design and operational characteristics. User behavior includes all the actions (both intentional and unintentional) that a driver performs while using the transportation system. User workload reflects the mental resources required to execute the tasks associated with navigating the transportation system. The SSI method includes two movement complexity factors, $L_{1}$ and $L_{2}$ . The first complexity factor, $L_{1}$ , applies to both vehicle-vehicle and nonmotorized conflict points, and it accounts for complexity added by the characteristics of the conflicting traffic as well as how the complexity is mitigated by traffic control devices that are intended to separate movements in time and space. The second complexity factor, $L_{2}$ , applies only to nonmotorized conflict points and represents characteristics that present additional complexity to pedestrians and cyclists, including the need to take indirect paths through the intersection and the presence of non-intuitive vehicle movements.

The components of the SSI method lead to a set of SSI measures of effectiveness (MOEs). These include the relative exposure at each intersection alternative through each conflict point type, the average severity—i.e., average $P (FSI)$ —for each conflict point type, and the average complexity of movements passing through different conflict point types. The scales of the three sets of MOEs are very different. Exposure indices are large numbers given that they present a product of daily volumes, $P (FSI)$ is a probability, and the complexity factors generally range from one to eight for most scenarios and are interpreted in relative terms much like a crash modification factor (CMF), representing interpretations such as “10% more complex,”“twice as complex,” and so forth. The SSI MOEs provide insights into how different intersection alternatives perform with respect to user exposure, kinetic energy management in the event of a crash, and movement complexity for all users. However, the creators of the method also sought a way to combine all three concepts together into one additional set of metrics, the SSI scores. The remainder of this paper focuses on the development of SSI scores.

SSI Method Scoring

The SSI method combines the concepts of exposure, conflict point severity, and movement complexity to create the SSI scores. The SSI scores can range from zero to 100, where scores closer to 100 represent project contexts and intersection designs that are closer to a Safe System (i.e., low probability of fatality or serious injury for users considered by the method). The method produces an SSI score for each conflict point type at an intersection alternative, as well as an SSI score for the overall intersection alternative.

Exposure-Severity-Complexity Product

Following the application of the method outlined in previous sections of the paper, each conflict point has an exposure index, a conflict point severity, and either one (for vehicle-vehicle conflict points) or two (for nonmotorized conflict points) movement complexity factors. The first step in creating the SSI scores is to calculate the sum of the exposure-severity-complexity products for all conflict points of a given type $t$ (i.e., crossing, merging, diverging, or nonmotorized). This sum of the exposure-severity-complexity products is $E_{t}$ , computed as shown in Equation 1.

\begin{matrix} E_{t} = \sum_{i = 1}^{n_{t}} [I_{i, t} \times P {(FSI)}_{i, t} \times L_{1, i, t} \times L_{2, i, t}] \end{matrix}

(1)

where

$I_{i, t}$ = the exposure index for conflict point $i$ of type $t$ ,

$P (FSI)_{i, t}$ = the probably of a fatality or serious injury occurring if a crash occurs at conflict point $i$ of type $t$ ,

$L_{1, i, t}$ = the conflicting traffic complexity factor at conflict point $i$ of type $t$ , and

$L_{2, i, t}$ = the nonmotorized movement complexity factor at conflict point $i$ of type $t$ .

Detailed information on computing each of these variables, as well as example calculations, can be found in the FHWA report ( 14 ).

The sum of exposure-severity-complexity products for each conflict point type, $E_{t}$ , is computed for all four conflict point types to produce $E_{crossing}$ , $E_{merging}$ , $E_{diverging}$ , and $E_{nonmotorized}$ .

Conflict Point Type SSI Scores and Intersection SSI Score

While the sum of the exposure-severity-complexity products, $E_{t}$ , accomplishes the goal of bringing together the various SSI method components, the multiplicative nature of the formula can result in very high numbers, limiting the potential usefulness of the measure in relation to interpretation of the comparisons across alternatives. Therefore, the SSI method creators determined an additional step was necessary to produce a more interpretable measure, the SSI scores. The following sections present the approach to transform the sum of exposure-severity-complexity products for each conflict point type, $E_{t}$ , to SSI scores for the individual conflict point types and for the overall intersection. The method used to make the transformations limits the SSI scores to be between zero and 100, where a score of 100 represents an intersection alternative and project context that is closest to a Safe System.

Once the sums of the exposure-severity-complexity products for each conflict point type have been obtained, SSI scores are computed for each conflict point type $t$ as well as for the whole intersection. The conflict point type SSI scores ( $SS I_{t}$ ) account for all conflict points at the intersection of type $t$ and are computed as shown in Equation 2.

\begin{matrix} SS I_{t} = 100 \times \exp (- \frac{1}{z} \times E_{t}) \end{matrix}

(2)

The primary input in Equation 2 is $E_{t}$ , and the exponential function transforms the sum of exposure-severity-complexity products for a given conflict point into the conflict point type SSI score. The variable $z$ in Equation 2 represents a scaling parameter. This parameter is a constant that non-linearly scales $E_{t}$ such that the resulting conflict point type SSI scores are bounded between zero and 100, allowing for a relative comparison of how an intersection in a given context performs from a Safe System perspective relative to all other potential alternatives and contexts. The following sections provide more detail on how the exponential transformation function works and the development of the scaling parameter, $z$ .

The use of the exponential transformation function means that the slope (or rate of change) of the respective SSI scores at a given input value of the sum of the exposure-severity-complexity products is nonlinear. In other words, a difference of one percentage point in the $SS I_{t}$ or $SS I_{int}$ score (e.g., the difference between 99% and 98% versus the difference between 51% and 50%) will correspond to different magnitudes of changes to input exposure-severity-complexity products depending on where that one percentage point difference occurs along the exponential function curve. This is illustrated in Figure 4. The shape of the curve results in several desirable properties of the SSI scores that are based on the exposure-severity-complexity products. Intersection alternatives will not have a large effect on the SSI scores at very low volumes, speeds, or combinations of both because such alternatives will already be approaching a Safe System due the low exposure and speeds in the given context. Conversely, intersection alternatives will also not have a large effect on the SSI scores at very high volumes, speeds, or combinations of both because geometric and operational strategies can only make a relatively small contribution to Safe System performance when large numbers of users, many of whom are traveling at high speed, are crossing at-grade. In these cases, the complete separation of users (e.g., separated or protected pedestrian or bicycle facilities) may provide the best Safe-System-aligned design option. Based on Figure 4, SSI scores will have the biggest sensitivity to the selection of intersection alternatives in the ranges of volumes and speeds in between these extremes. These represent contexts where managing speeds and impact angles and simplifying decision-making using geometrics and traffic control can make strong contributions to reducing the chances of fatalities and serious injuries.

Figure 4.

Exponential model for transforming the sum of the exposure-severity-complexity products into Safe System for Intersections (SSI) scores.

In addition to the conflict point type SSI scores from Equation 2, the SSI method also produces an intersection SSI score, $SS I_{int}$ , using a similar exponential transformation. This score accounts for all the identified conflict points at the intersection. It also uses the scaling parameter $z$ to bound the scores between zero and 100. The $SS I_{int}$ score is computed as shown in Equation 3.

\begin{matrix} SS I_{int} = 100 \times \exp [- \frac{1}{z} \times (E_{crossing} + E_{merging} + E_{diverging} + E_{nonmotorized}) / 4] \end{matrix}

(3)

The addition of the four conflict-type exposure-severity-complexity sums (i.e., $E_{t}$ values) divided by four may be interpreted as an average exposure-severity-complexity product sum across conflict points of all types.

The process for conducting the SSI computations and producing the conflict point type and intersection SSI scores is summarized in Figure 5.

Figure 5.

Flowchart illustrating the process for computing the Safe System for Intersections (SSI) scores.

Development of the SSI Scoring Equations and Scale Parameter

The development of the exponential transformation scale parameter introduced in the previous section was performed as follows. First, a batch test analysis was conducted on a range of intersection input values to obtain a representative distribution of potential $E_{t}$ values considering all different conflict point types. Second, the form of the transformation function between the $E_{t}$ and the SSI scores was determined based on a review of the distribution of the $E_{t}$ values. Finally, the exponential transformation function was optimized to find the best-fit scaling parameter $z$ for the range of potential $E_{t}$ results.

Batch Test Analysis

The purpose of the batch test analysis was to obtain a representative distribution of $E_{t}$ values for the four conflict point types to guide the development of the SSI scoring functions. This representative distribution constitutes an expansive universe of potential intersection contexts and data inputs designed to characterize most real-world intersection scenarios. The batching conditions were defined for each of the data inputs as shown in Table 3.

Table 3.

Batch Input Conditions

Data input	Min.	Max.	Possible values
Number of through lanes in one direction	1	4	[1 2 3 4]
Annual average daily traffic (AADT) (vpd)	200	70,000	[200 500 1,000 5,000 10,000 15,000 20,000 … 70,000]
Posted speed limit (mph)	15	60	[15 20 25 30 35 40 45 50 55 60]
Nonmotorized user volume (users/day)	0	4,000	[50 100 200 500 1,000 2,000 4,000]

Note: Min. = minimum; Max. = maximum; vpd = vehicles per day.

The process for selecting the batch inputs was as follows:

1. Randomly select major road volume.

2. Randomly select major road number of through lanes in one direction, satisfying the following conditions:

a) If major road volume is less than 20,000 vpd, number of through lanes in one direction is either one or two.

b) If major road volume is greater than or equal to 20,000 vpd and less than or equal to 40,000 vpd, number of through lanes in one direction is either two, three, or four.

c) If major road volume is greater than 40,000 vpd, number of through lanes in one direction is either three or four.

3. Randomly select major road posted speed limit.

4. Repeat steps 1 to 3 for the minor road, satisfying the following conditions:

a) Minor road volume must be less than or equal to 50,000 vpd.

b) Minor road volume must be less than or equal to major road volume.

c) Minor road number of lanes must be less than or equal to major road number of lanes.

d) Minor road posted speed limit must be less than or equal to major road posted speed limit.

5. Randomly select total volume of nonmotorized users at the intersection, satisfying the following conditions:

a) If major road volume is less than 20,000 vpd, nonmotorized user volume is less than or equal to 200 users per day.

b) If major road volume is greater than or equal to 20,000 vpd and less than or equal to 40,000 vpd, nonmotorized user volume is less than or equal to 2,000 users per day.

c) If major road volume is greater than 40,000 vpd, nonmotorized user volume is less than or equal to 4,000 users per day

This process was used to select SSI method inputs for 3,000 random intersection scenarios. The inputs for these 3,000 different scenarios were run through the SSI method calculations to produce a representative universe of possible sums of the exposure-severity-complexity products (i.e., $E_{t}$ values), covering every situation from low-speed, low-volume intersections to high-speed, high-volume intersections, and everything in between. The SSI method calculations were performed for all the intersection designs in the SSI library of intersection types. Currently, the SSI library of intersection types includes the following intersection alternatives: traditional (signalized, all way stop control [AWSC], minor road stop control [MRSC]), roundabout (1 × 1, 2 × 1, 2 × 2), restricted crossing U-turn (RCUT) (signalized, MRSC), MUT, bowtie, jughandle, quadrant roadway, partial displaced left turn (PDLT), and full displaced left turn (FDLT). These “raw” scores varied over an extremely large range, as was expected.

Exponential Function Determination

The results of the batch test analysis showed that the $E_{t}$ values can range over several orders of magnitude, between zero and $10^{9},$ and exhibit an exponential distribution. Given the large range of $E_{t}$ values and the observed distribution, the exponential transformation function was chosen for transforming the $E_{t}$ values into SSI scores. Two notable advantages of the exponential transformation are that it bounds the resulting SSI scores between zero and 100 and the maximum $E_{t}$ value need not be known. As previously noted, the shapes of the exponential curves also lead to several desirable properties of the sensitivity between intersection alternatives and SSI scores.

Optimizing the Scaling Parameter $z$

The SSI scaling parameter, $z$ , determines how fast the exponential function will change as a result of intersection inputs. For this application, the scaling parameter was optimized for the range of $E_{t}$ values determined through the batch test analysis. The scaling parameter affects the shape of the exponential curve and can be optimized using the mean of all potential $E_{t}$ values, as shown in Equation 4.

\begin{matrix} z = mean (all E_{t} values) \end{matrix}

(4)

By design, this $z$ value will give the best-fit coefficient for an exponential distribution ( 20 ). However, to minimize the effect of high outliers on computing $z$ , $E_{t}$ values greater than 125,000,000 (or roughly 8.9% of the $E_{t}$ values) were ignored in the $z$ determination. This was deemed reasonable given that these extremely high outlier cases represented intersections that are rare in practice; altogether, the combination of this filter on the $E_{t}$ values and the mean computation results in a $z$ value of 1.37 × 10⁷. Note that this $z$ value sets the shape of the exponential transformation function to be approximately flat in slope at the 1% and 99% ends as previously discussed in this paper.

Since the average $E_{t}$ input used in Equation 3 is not computed directly, the individual $E_{t}$ inputs are used to estimate $z$ instead. The possible range of average and individual conflict $E_{t}$ values are assumed to be similar. Overall, this assumption enables the use of the same $z$ for both Equations 2 and 3.

The value of $z$ presented here is not calibrated to local conditions. Rather, as explained in the prior sections, it is based on an expansive universe of potential intersection contexts and data inputs designed to represent most real-world intersection scenarios.

Example Scenario and Calculations

This section presents an example scenario designed to illustrate the computation of the SSI scores. For the purposes of this paper, the example will not go into detail on the early steps in the SSI method to compute the results of each method component (i.e., the exposure index $I$ , $P (FSI)$ , or the complexity factors $L_{1}$ and $L_{2}$ ). For an explanation of the process for performing these pieces of the SSI method, including detailed descriptions and thorough example calculations, refer to the FHWA report ( 14 ).

This example scenario involves a suburban signalized intersection of a four-lane arterial and a two-lane collector and is depicted in Figure 6. The intersection is being studied as part of a broader corridor planning effort. Design year traffic volumes are estimated at 25,000 and 20,000 vpd, respectively, for the major and minor roads. The posted speed limits are 45 mph on the major road and 35 mph on the minor road. The intersection serves a moderate number of nonmotorized users, estimated to be 2,400 users per day.

Figure 6.

Illustration of scenario intersection conditions.

A Stage I ICE analysis was conducted for this intersection as part of the corridor planning study. This involved a capacity-based operational review and a crash-based safety performance review in addition to the SSI analysis. Based on a volume-to-capacity analysis using the Capacity Analysis for Planning of Junctions (CAP-X) tool, the minor road stop control (MRSC) traditional, all-way stop control (AWSC) traditional, and 1 × 1 roundabout alternatives are dropped from further consideration. The crash-based safety performance review used the Safety Performance for Intersection Control Evaluation (SPICE) tool and determined that all the intersection types for which there are predicted methods available have fewer predicted crashes (both total and fatal/injury) than the existing signalized traditional intersection design. This example will examine the SSI method results for the signalized traditional, signalized RCUT, MUT, 2 × 1 roundabout, and 2 × 2 roundabout intersection designs.

Once the exposure, severity, and complexity calculations have been performed for all the conflict points at each intersection alternative, the SSI scores can be computed according to the approach outlined previously in this paper. In this case, these prior calculations result in sums of the exposure-severity-complexity products as shown in Table 4. The bottom row of the table contains the total of each column.

Table 4.

Scenario Totals of Exposure-Severity-Complexity Products

$E_{t}$	Signalized traditional	2 × 1 roundabout	2 × 2 roundabout	Signalized restricted crossing U-turn (RCUT)	Median U-turn (MUT)
$E_{crossing}$	22,557,321	1,057,110	1,445,314	4,155,769	8,933,214
$E_{merging}$	977,146	241,553	290,494	3,587,068	2,604,296
$E_{diverging}$	82	4,217	4,217	1,996,516	1,813,623
$E_{nonmotorized}$	54,551,197	34,679,521	45,788,807	39,798,121	31,899,331
Total	78,085,746	35,982,401	47,528,833	49,537,474	45,250,464

The overall intersection SSI score, $SS I_{int}$ , can be computed for each alternative according to Equation 3 introduced previously. Equations 5 through 9, below, show the calculation of $SS I_{int}$ for the five intersection designs in this scenario, using the $E_{t}$ values in Table 4.

\begin{matrix} SS I_{Signalized Trad} = 100 \times \exp [- \frac{1}{1.37 \times 10^{7}} \times (78, 085, 746) / 4] = 24 \end{matrix}

(5)

\begin{matrix} SS I_{2 x 1 Roundabout} = 100 \times \exp [- \frac{1}{1.37 \times 10^{7}} \times (35, 982, 401) / 4] = 52 \end{matrix}

(6)

\begin{matrix} SS I_{2 x 2 Roundabout} = 100 \times \exp [- \frac{1}{1.37 \times 10^{7}} \times (47, 528, 833) / 4] = 42 \end{matrix}

(7)

\begin{matrix} SS I_{Signalized RCUT} = 100 \times \exp [- \frac{1}{1.37 \times 10^{7}} \times (49, 537, 474) / 4] = 40 \end{matrix}

(8)

\begin{matrix} SS I_{MUT} = 100 \times \exp [- \frac{1}{1.37 \times 10^{7}} \times (45, 250, 464) / 4] = 44 \end{matrix}

(9)

The overall intersection SSI Scores shown in Equations 5 through 9 show that the 2 × 1 roundabout has the highest (i.e., best) SSI score at 52. The four intersection alternatives shown here (2 × 1 roundabout, 2 × 2 roundabout, signalized RCUT, and MUT) indicate an improved SSI score compared with the existing signalized traditional intersection.

These $E_{t}$ values can be plugged into Equation 2 to compute the conflict point type SSI scores. As an example, Equation 10 shows the computation of $SS I_{crossing}$ for the existing signalized traditional intersection design.

\begin{matrix} SS I_{crossing, Signalized Trad} = 100 \times \exp (- \frac{1}{1.37 \times 10^{7}} \times 22, 557, 321) = 19 \end{matrix}

(10)

Table 5 summarizes the conflict point type SSI scores for the four intersection alternatives.

Table 5.

Scenario Conflict Point Type Safe System for Intersections (SSI) Scores

$SS I_{t}$	Signalized traditional	2 × 1 roundabout	2 × 2 roundabout	Signalized restricted crossing U-turn (RCUT)	Median U-turn (MUT)
$SS I_{crossing}$	19	93	90	74	52
$SS I_{merging}$	93	98	98	77	83
$SS I_{diverging}$	100	100	100	86	88
$SS I_{nonmotorized}$	2	8	4	5	10

Some additional context explaining these overall intersection SSI scores can be derived by examining Table 5. The four alternatives to the existing condition shown here have improved conflict point type SSI scores for the nonmotorized conflict points and crossing conflict points compared with the existing signalized traditional intersection. These designs reroute one or more movements at the intersection, removing crossing conflict points and reducing vehicle speeds and angles at crossing conflict points, or both. Several of these designs, such as signalized RCUT and MUT, have lower merging and diverging scores than the signalized traditional intersection. This is because the rerouting of movements involved in these designs removes crossing conflict points but introduces additional merging and diverging conflict points. As expected, the range of nonmotorized and crossing conflict point SSI scores is lower than the merging and diverging scores, indicating that nonmotorized and crossing conflicts are more likely to lead to fatalities and serious injuries than merging and diverging conflicts.

Summary and Conclusion

The SSI method provides practitioners with a tool to integrate Safe System principles, such as exposure, kinetic energy management, and movement complexity, into the intersection project development process. Application of the method results in a set of SSI MOEs and SSI scores. The SSI scores are designed to bring together the exposure-severity-complexity factors addressed in the SSI method and package them into a single metric that is easy to understand, compare, and communicate. While the FHWA report associated with this research details the working of the method and its application to intersection projects, this paper provides supplementary background on the reasoning and process behind the development of the SSI scoring system. This additional background and detail on the development of the SSI scores will support future applications of the method and the identification of future research needs and method improvements.

The SSI score uses an exponential function to transform the sum of the exposure-severity-complexity products to a scale ranging from zero to 100, where 100 represents a situation approaching a Safe System. The shape of the exponential function, as influenced by the scaling parameter $z$ , was determined using a batching exercise that sought to represent the complete universe of potential intersection scenarios. The shapes of the exponential curves also lead to several desirable properties of the sensitivity between intersection alternatives and SSI scores.

Increasing application of the SSI method to real-world intersection projects will continue to reveal the effectiveness of the SSI scores in informing agency decision-making. This may lead to new opportunities for adjusting the SSI scores or calibrating them to local conditions. For example, the SSI method currently applies typical pedestrian paths and assumes that bicyclists follow the same paths as pedestrians (i.e., sidewalks or shared-use paths). Additionally, the SSI method does not currently consider rear-end conflicts that result from speed differentials caused by traffic congestion or traffic control devices. Similarly, it does not consider merging and diverging conflicts caused by lane changing, including weaving movements. Future research efforts should focus on incorporating more refined identification and analysis of pedestrian and bicyclist conflict points as well as rear-end and sideswipe conflicts. These additions may benefit from further refinement of the SSI score computation and calibration.

The exposure-severity-complexity products that lead to the SSI scores contain the same general components as would be expected in an intersection crash predictive method for fatal and serious injury crashes:

• Characterization of exposure

• Movement complexity factors that, in addition to exposure, are expected to be associated with higher or lower crash likelihoods

• The probability (or proportion) of crashes that occur resulting in a fatality or serious injury

Future efforts to validate and calibrate the SSI method using crash data could take multiple approaches:

• One approach could seek to establish statistical associations between the SSI MOEs and SSI scores and the average number of intersection crashes.

• A second approach could try to maximize the statistical associations between the SSI MOEs and SSI scores and the average number of intersection crashes. This could include calibrating some of the parameters that are part of the exposure-severity-complexity product using observed crash data.

Such efforts could further enhance the complementary nature of Safe System metrics and crash-based metrics and support steps toward future versions of resources such as the HSM incorporating the SSI method and other types of Safe System assessments.

Ideally, efforts to validate and calibrate the SSI method with crash data would be based on fatal and serious injury crashes, with serious injuries determined on an abbreviated injury scale (AIS) by trained medical professionals following an assessment of a patient’s injuries at the hospital. Such an approach, however, would require a traffic injury surveillance system in the state where police-reported crashes are linked to hospital records, which is rare. In the absence of such a system, it is possible that validation and calibration could occur using K and A crashes on the KABCO injury classification scale, where K corresponds to “fatal injury,” A to “suspected serious injury,” B to “suspected minor injury,” C to “possible injury,” and O to “no apparent injury.”

Footnotes

Acknowledgements

The research described in this paper was conducted as part of FHWA Office of Safety Task Order DTFH61-16-D-00005, A Safe System Approach to Intersection Planning & Design in the United States. The authors would like to express their gratitude to Mr. Jeffrey Shaw and Mr. Mark Doctor from the FHWA Office of Safety for their guidance.

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: R.J. Porter, I. Huang, M.R. Dunn; data collection: I. Huang, R.J. Porter, M.R. Dunn; analysis and interpretation of results: I. Huang, R.J. Porter, M.R. Dunn; draft manuscript preparation: M.R. Dunn, I. Huang, R.J. Porter. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Michael R. Dunn

Richard J. (R.J.) Porter

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the Federal Highway Administration

References

FHWA. The Safe System Approach. FHWA-SA-20-015. Federal Highway Administration, 2020. https://safety.fhwa.dot.gov/zerodeaths/docs/FHWA_SafeSystem_Brochure_V9_508_200717.pdf.

Tingvall

Haworth

Vision Zero – An Ethical Approach to Safety and Mobility. Proc., 6th ITE International Conference Road Safety & Traffic Enforcement: Beyond 2000, Melbourne, Australia, 1999. https://www.monash.edu/muarc/our-publications/papers/visionzero.

Johansson

Vision Zero – Implementing a Policy for Traffic Safety. Safety Science, Vol. 47, No. 6, 2009, pp. 826–831.

Fitzpatrick

Carlson

Brewer

Wooldridge

Design Factors That Affect Driver Speed on Suburban Streets. Transportation Research Record: Journal of the Transportation Research Board, 2001. 1751: 18–25.

Belin

Tillgren

Vedung

Vision Zero – A Road Safety Policy Innovation. International Journal of Injury Control and Safety Promotion, Vol. 19, No. 2, 2012, pp. 171–179.

De Haven

. Mechanical Analysis of Survival in Falls From Heights of Fifty to One Hundred and Fifty Feet. Injury Prevention, Vol. 6, No. 1, 2000, pp. 62–68.

Haddon

Advances in the Epidemiology of Injuries as a Basis for Public Policy. Public Health Reports, Vol. 95, No. 5, 1980, pp. 411–421.

Robertson

L. S.

Injuries – Causes, Control Strategies, and Public Policy. Journal of Public Policy, Vol. 4, No. 4, 1983, pp. 359–360.

SWOV. Sustainable Safety: Principles, Misconceptions and Relations with Other Visions. SWOV Institute for Road Safety Research, Leidschendam, The Netherlands, 2013.

10.

Australian Transport Council. National Road Safety Strategy 2011–2020. Australian Transport Council, Canberra, 2011.

11.

Corben

Van Nes

Logan

Archer

Kinetic Energy Management Model and Safe Intersection Design Principles. Report No. 316c. VicRoads, Victoria, Australia, 2010.

12.

New Zealand Ministry of Transport. Safer Journeys 2020: New Zealand’s Road Safety Strategy 2010–2020. New Zealand Ministry of Transport, Wellington, 2010.

13.

NHTSA. Traffic Safety Facts: Preview of Motor Vehicle Traffic Fatalities in 2019. DOT HS 813 021. National Highway Traffic Safety Administration, Washington, D.C., 2020. https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/813021.

14.

Porter

R. J.

Dunn

Soika

Huang

Coley

Gross

Kumfer

Heiny

A Safe System-Based Framework and Analytical Methodology for Assessing Intersections. FHWA-SA-21-008. Federal Highway Administration, Washington, D.C., 2021. https://safety.fhwa.dot.gov/intersection/ssi/fhwasa21008.pdf.

15.

FHWA. Crossover-Based Intersections. Federal Highway Administration, Washington, D.C., 2019. https://safety.fhwa.dot.gov/intersection/innovative/crossover/

16.

Hakkert

A. S.

Mahalel

Estimating the Number of Accidents at Intersection From a Knowledge of Traffic Flows on the Approaches. Accident Analysis & Prevention, Vol. 10, 1978, pp. 69–79.

17.

Evans

Driver Injury and Fatality Risk in Two-Car Crashes Versus Mass Ratio Inferred Using Newtonian Mechanics. Accident Analysis & Prevention, Vol. 26, No. 5, 1994, pp. 609–616.

18.

Tefft

B. C.

Impact Speed and a Pedestrian’s Risk of Severe Injury or Death. Accident Analysis & Prevention, Vol. 50, 2013, pp. 871–878.

19.

Chidester

A. B.

Isenburg

R. A.

The Pedestrian Crash Data Study – Final Report. National Highway Traffic Safety Administration, Washington, D.C., 2001.

20.

Bertsekas

D. P.

Tsitsiklis

J. N

. Introduction to Probability. Athena Scientific, Nashua, NH, 2008.