Quantitative Safety Analyses for Highway Applications (2025)
Chapter: 5 Modifications and Additions to HSM1 Materials for Use in HSM2
CHAPTER 5
Modifications and Additions to HSM1 Materials for Use in HSM2
This chapter summarizes the modifications and additions that the research team made to the HSM draft material under NCHRP Project 17-71A for use in HSM2. It does not include the modifications and additions to the chapters made to the HSM draft material under NCHRP Project 17-71. At a high level, this chapter explains the changes that were made and, in some cases, provides the sources of the results of research that developed or supported those changes. The chapter also identifies some changes that were not made because of lack of research results or unresolved issues concerning such results.
It is important to note that the second edition is not a complete rewrite of the manual, but rather the manual was updated to incorporate relevant ongoing and completed research and to expand the scope and quality of the manual to increase its application and improve its usability. The outline and structure of HSM2, as provided in Table 4, are similar to those of the first edition, but several organizational changes have been made. The HSM2 still has four parts (Parts A, B, C, and D). There is an introductory chapter to the HSM2 as a whole, which comes before Part A. In the HSM1, this introductory chapter was included in Part A.
Changes to and Proposed Structure of HSM2
Six new chapters are incorporated into the second edition. There is one new chapter in HSM2 Part A—Fundamentals that addresses pedestrians and bicyclists (Chapter 4). Two new chapters are included in HSM2 Part B—Roadway Safety Management Process. Chapter 5 addresses an areawide approach to roadway safety management, and Chapter 12 addresses the systemic approach to roadway safety management. HSM2 Part C—Predictive Methods includes a new introductory chapter that presents general concepts for applying the Part C predictive methods (Chapter 13). HSM2 Part D (Crash Modification Factors) is completely new. Rather than presenting information on HSM-approved CMFs for roadway segments, intersections, interchanges, special facilities and geometric situations, and road networks, HSM2 Part D consists of two new chapters on selecting (Chapter 19) and applying (Chapter 20) CMFs that are a key part of the processes and methods described in HSM2 Parts B and C.
The HSM2 text was reviewed and revised for consistent usage of language and terminology throughout. A further editorial review will be undertaken by AASHTO prior to publication.
Structurally, all appendices from the first edition have been removed. The information and material that were presented in appendices in the first edition have either been incorporated into the main body of the chapters or removed from the manual. Additionally, the blank worksheets that were provided at the end of the HSM1 Part C chapters to assist with hand calculations are no longer included in the manual.
The glossary for HSM2 will be updated under a separate contract with AASHTO.
Table 4. Final outline/structure of HSM2 as developed in NCHRP Project 17-71A.
| HSM2 Chapter | HSM1 (2010) Chapter | Chapter Title |
|---|---|---|
| Front Matter | ||
| Preface | ||
| 1 | 1 | Introduction and Overview to the Highway Safety Manual |
| Part A—Fundamentals | ||
| Introduction to Part A | ||
| 2 | 3 | Road Safety Principles |
| 3 | 2 | Human Factors |
| 4 | N/A | Pedestrians and Bicyclists |
| Part B—Roadway Safety Management Process | ||
| Introduction to Part B | ||
| 5 | N/A | Areawide Approach to Roadway Safety Management |
| 6 | 4 | Network Screening |
| 7 | 5 | Diagnosis |
| 8 | 6 | Countermeasure Selection |
| 9 | 7 | Economic Appraisal |
| 10 | 8 | Project Prioritization |
| 11 | 9 | Countermeasure Effectiveness Evaluation |
| 12 | N/A | Systemic Approach to Roadway Safety Management |
| Part C—Predictive Methods | ||
| Introduction to Part C | ||
| 13 | N/A | General Concepts for Applying the Part C Predictive Methods |
| 14 | 10 | Predictive Method for Rural Two-Lane, Two-Way Roads |
| 15 | 11 | Predictive Method for Rural Multilane Highways |
| 16 | 12 | Predictive Method for Urban and Suburban Arterials |
| 17 | 18 | Predictive Method for Directional Freeway Segments |
| 18 | 19 | Predictive Method for Ramps |
| Part D—Crash Modification Factors | ||
| N/A | Introduction to Part D | |
| 19 | N/A | Selecting CMFs |
| 20 | N/A | Applying CMFs |
| Glossary | ||
Note: N/A = “not applicable” because chapter is not present in HSM1.
The following subsections discuss the major additions and changes to the individual HSM2 chapters. It is not specifically stated in each chapter but, as discussed previously, the language in all materials and chapters for HSM2 was updated based on the style and usage guide to avoid language that might raise tort liability issues for transportation agencies.
Front Matter
Preface
The preface was updated to include a new section on what is new in the second edition.
HSM2 Chapter 1—Introduction and Overview of the Highway Safety Manual
The content of this chapter was updated according to the revised outline and structure of HSM2.
HSM2 Part A—Fundamentals
Part A provides foundational information related to the evaluation and interpretation of crash frequency and severity. This content sets the stage for HSM use and explains the relationship of
applying the HSM methods to transportation planning, design, operations, and maintenance activities and presents an overview of the safe-system approach. The title of Part A in HSM1 was Introduction, Human Factors, and Fundamentals. For HSM2, the title of Part A has been simplified to just Fundamentals.
Introduction to Part A
The Introduction to Part A is new. HSM1 did not have an introduction to this part. The outline and structure of this introductory section is consistent with the introductions to Parts B, C, and D.
HSM2 Chapter 2—Road Safety Principles
The title of this chapter has changed from Fundamentals to Road Safety Principles, and this chapter is now Chapter 2 rather than Chapter 3. A section on the safe-system approach has been added to the chapter. The fundamental safety performance analysis terms have been expanded to be more inclusive of pedestrians and bicyclists. Content on the application of CMFs that was originally intended to be in HSM2 Part D as specified under NCHRP Project 17-71 has been incorporated into this chapter. Additionally, content from the chapter appendices in the first edition has been blended into Chapter 2 or removed.
HSM2 Chapter 3—Human Factors
This chapter has largely been restructured and rewritten to emphasize practical aspects, and it is now Chapter 3 rather than Chapter 2. The primary revisions to this chapter are described in the following:
- The chapter is now focused on helping the practitioner successfully integrate human factors into planning, design, and operations.
- The style and language have been shifted to a more conversational and less academic tone. That is, the chapter now emphasizes the practical implications of human factors on safety, and less emphasis is given to theories and high-level concepts. The goal of this approach is to improve the accessibility and approachability of the chapter for the typical end user.
- The chapter describes and discusses the differences between “human-factors issues” and “aberrant driver behaviors,” especially as they relate to the nature of the problems and available countermeasures.
- The chapter includes additional references to content in NCHRP Report 600: Human Factors Guidelines for Road Systems (Campbell et al., 2012) and the HSM/HFG Primer (Campbell et al., 2018).
- The chapter provides specific, actionable content and emphasizes the practical ways in which human factors play a role in highway safety, as well as the practical solutions available through the application of human-factors data.
- Examples have been developed to illustrate the relationships between human-factor issues and both driver behavior and safety. These examples focus on the effects of crash contributing factors, such as visibility, workload, expectations, and time, on driver performance and safety.
- An extended discussion of countermeasures has been developed for typical crash types and where/how typical countermeasures affect the demands placed on the road user.
- The chapter illustrates the application of human factors to common design projects, and the everyday utility of the chapter has been increased.
In summary, this chapter infused the following three themes into the narrative:
- Key contributors to roadway crashes are the human-factors issues that lead to driver errors. Therefore, understanding and applying the differences between human-factors issues and aberrant driver behaviors is crucial to correctly diagnosing the key contributing factors to errors and crashes and in identifying and selecting targeted countermeasures.
- Roadways are communication devices and, at all times, are sending messages to the road user; the job of the roadway designer is to design the road in such a way that the right messages are being sent to the road user at the right time.
- Many crashes happen when the demands of the roadway environment exceed the capabilities of the roadway user; practitioners will benefit from understanding the various ways that roadway design and operations can increase demands on drivers and then identifying and selecting countermeasures targeted toward such demands.
HSM2 Chapter 4—Pedestrians and Bicyclists
This is a new chapter. The topics covered in this chapter include (a) factors contributing to pedestrian and bicycle collisions, (b) safety data for pedestrians and bicyclists, (c) indirect safety measures for pedestrians and bicyclists, (d) integrating pedestrian and bicycle considerations into roadway safety management and predictive methods, and (e) special considerations for pedestrians and bicyclists.
HSM2 Part B—Roadway Safety Management Process
Introduction to Part B
The outline and structure of this introductory section was modified to be consistent with the introductions to Parts A, C, and D. In addition, the description of the roadway safety management process was modified to include information on areawide and systemic approaches to roadway safety management.
HSM2 Chapter 5—Areawide Approach to Roadway Safety Management
This is a new chapter developed under NCHRP Project 17-81, “Proposed Macro-Level Safety Planning Analysis Chapter for the Highway Safety Manual.” This chapter presents a method to predict areawide crash totals within geographical areas of various sizes, using a predictive method based on macro-level safety performance analysis. An areawide evaluation can serve as a first step in the roadway safety management process. The draft chapter provided by the NCHRP Project 17-81 team was modified to fit the format and structure of HSM2.
HSM2 Chapter 6—Network Screening
Two network screening performance measures were removed: excess predicted average crash frequency using method of moments, and excess predicted average crash frequency using SPFs. Critical crash rate was moved to a new section titled “Measures Used for Reporting or Other Purposes.” An error in the peak searching method example was corrected. Pedestrian and bicycle SPFs for network screening were added as recommended from NCHRP Project 17-84, “Pedestrian and Bicycle Safety Performance Functions for the Highway Safety Manual.” The example used to demonstrate the performance measure calculations was modified.
HSM2 Chapter 7—Diagnosis
Material from Chapter 3 on human factors was integrated into Chapter 7 where practical, and a new sample problem was added.
HSM2 Chapter 8—Countermeasure Selection
Material from Chapter 3 on human factors was integrated into Chapter 8 where practical, and a new sample problem was added.
HSM2 Chapter 9—Economic Appraisal
Material that was deleted from HSM1 under the NCHRP Project 17-71 contract was replaced. Updated crash costs by severity were provided along with how to adjust the crash costs. Methods
for estimating a change in crash frequency for a proposed project were clarified. A new sample problem was added.
HSM2 Chapter 10—Project Prioritization
A new sample problem was added to this chapter.
HSM2 Chapter 11—Countermeasure Effectiveness Evaluation
The title for this chapter was changed from Safety Effectiveness Evaluation to Countermeasure Effectiveness Evaluation. This chapter has largely been restructured and rewritten to incorporate material on developing CMFs, prepared as part of NCHRP Project 17-63, “Guidance for the Development and Application of Crash Modification Factors.” The content on developing CMFs that was integrated into the chapter included:
- Fundamental concepts related to CMFs,
- Relevant statistical concepts and terminology,
- Study design and analysis approaches to develop CMFs,
- Other approaches for developing CMFs,
- Crash modification functions, and
- CMF reporting.
Items 1, 2, 4, 5 and 6 became stand-alone sections in Chapter 11. Item 3 was merged into a section on study design and methods for countermeasure effectiveness evaluation. To integrate this content, several additional structural changes were made to Chapter 11:
- The overview, introduction, and definition and purpose of safety effectiveness evaluation sections were consolidated into an introduction section.
- Content from the following HSM1 sections was moved and integrated into new Section 11.4: Study Design and Methods for Countermeasure Effectiveness Evaluation.
- HSM1 Section 11.5: Evaluating a Single Project at a Specific Site to Determine Its Safety Effectiveness,
- HSM1 Section 11.6: Evaluating a Group of Similar Projects to Determine Their Safety Effectiveness, and
- HSM1 Section 11.7: Quantifying CMFs as a result of a Safety Effectiveness Evaluation.
Other significant changes included:
- Discussion of experimental design was removed from the chapter.
- Text was added in various places to connect safety effectiveness evaluation to estimating crash modification factors.
- Terminology was made more consistent (e.g., using only “sites” not “sites” and “entities,” and using “nontreatment” versus “comparison” versus “control sites”).
HSM2 Chapter 12—Systemic Approach to Roadway Safety Management
This is a new chapter, and much of the information comes from FHWA’s Reliability of Safety Management Methods: Systemic Safety Programs (Gross et al., 2016). The topics covered in this chapter include (a) the five general steps of the systemic approach, (b) allocating funds for systemic projects, and (c) data used for the systemic approach. While working on and editing this chapter, the overview and introduction sections were expanded to discuss how the systemic approach fits into the roadway safety management process considering the areawide approach and crash-based approach. The five-step systemic approach that was previously discussed remains, but pedestrian- and bicycle-specific information was added when appropriate. A paragraph was added to discuss flexibility within the systemic approach. The examples in Chapter 12 were updated to be consistent with terminology and crash cost values in Chapter 9. The section
Systemic Safety Management Approach in Practice was removed, and material was moved to appropriate places in Section 12.3. Material in the section Integrating Systemic Approaches in a Comprehensive Safety Plan was integrated into Section 12.3. A new section, Section 12.4, was added, as was a new sample problem.
HSM2 Part C—Predictive Methods
HSM2 Part C presents the predictive methods presented for quantifying crash frequency and severity. The equivalent part of the HSM1, also designated as Part C, was titled “Predictive Method.” The HSM1 developers used the singular term of the word “method” because they wanted to emphasize that there was a single predictive method or philosophy used throughout the HSM. However, there were substantial variations in predictive methods between chapters in the HSM1, and there are even more pronounced variations in HSM2. These variations have occurred because of differences in the crash frequencies and severities between facility types and differences in the research approaches that were found to be appropriate to address those differences. Given these variations in predictive methods, it appeared most appropriate to title HSM2 Part C as “Predictive Methods,” using the plural form of the word.
The results of recent research were reviewed to determine which results would fit appropriately within the HSM and would be compatible with existing HSM procedures and other new research results. The selected new research results were incorporated into HSM2. Research topics for which the available results did not appear appropriate for use in HSM2 were added to the list of potential topics for future research (see Chapter 7).
Changes Common to All HSM2 Part C Chapters
In HSM1, the term “crash modification factors” was used to refer to two somewhat different concepts:
- Factors presented in HSM1 Part D that represent the effects of specific design or traffic control elements or crash countermeasures in reducing (or increasing) crashes.
- Factors used in HSM1 Part C predictive methods to adjust predicted crash frequencies for the effects of specific design or traffic control features.
For HSM2, it was decided that it could be confusing to continue using the term CMFs for both of these applications, so a decision was reached to change the name given to the latter application (i.e., the application in HSM2 Part C) from CMFs to adjustment factors (AFs). This change has been made throughout the HSM2 Part C chapters. The term CMF has been retained for the former application (representing the effect on crashes of specific design or traffic control changes or crash countermeasures outside the context of an HSM Part C method).
HSM2 Chapter 13—General Concepts for Applying the Part C Predictive Methods
This is a new chapter, and much of the information comes from (a) the Introduction to Part C from HSM1, (b) chapter appendices from HSM1, (c) FHWA’s Safety Performance Function Decision Guide: SPF Calibration vs. SPF Development (Srinivasan et al., 2013), (d) FHWA’s Safety Performance Function Development Guide: Developing Jurisdiction-Specific SPFs (Srinivasan and Bauer, 2013), and (e) the User’s Guide to Develop Highway Safety Manual Safety Performance Function Calibration Factors (Bahar and Hauer, 2014). The purpose of this chapter is to introduce the HSM2 Part C predictive methods for estimating predicted and expected average crash frequencies and general concepts related to the predictive methods. The primary topic areas this chapter addresses are (a) the general steps of HSM2 Part C predictive methods, (b) SPFs, (c) AFs, (d) assignment of crashes, (e) the empirical Bayes method, (f) calibration, (g) development of
jurisdiction-specific SPFs, and (h) methods for estimating the change in crashes for a proposed project.
HSM2 Chapter 14—Predictive Method for Rural Two-Lane, Two-Way Roads
HSM2 Chapter 14, which presents a crash prediction method for rural two-lane, two-way roads, was adapted from HSM1 Chapter 10, incorporating the results of new research. While the structure of the predictive method remained similar to that used in HSM1, the following new features were added to HSM2 Chapter 14:
- New SPFs for total crashes, KABC crashes, and KAB crashes (from KABCO: K = fatal injury, A = suspected serious injury, B = suspected minor injury, C = possible injury, and O = no apparent injury) were added based on the results of NCHRP Project 17-62 (Ivan et al., 2018) for rural two-lane roadway segments and three intersection types (3ST, 4ST, and 4SG intersections; see Table 5 for explanation of abbreviations). Existing CMFs from the crash prediction method in HSM1 Chapter 10 were retained for use as AFs in the updated chapter since NCHRP Project 17-62 did not recommend any revised AFs.
- New SPFs developed in NCHRP Project 17-68 (Torbic et al., 2020) for three intersection types (3STT, 3SG, and 4AST intersections; see Table 5) were incorporated into the chapter.
- A predictive method with SPFs and AFs for roundabouts on rural two-lane roads, developed in NCHRP Project 17-70 (Ferguson et al., 2018), was incorporated into the chapter (see later section pertaining to predicting crashes for roundabouts).
- Many of the SPFs from recent research mentioned previously were adjusted, where appropriate, through calibration to data for a common state using results from NCHRP Project 17-72 (Srinivasan et al., 2022), as described in the later section pertaining to calibration of related SPFs to a common state. In addition, the sets of original and calibrated SPFs were reviewed for consistency and credibility, as described in the later section pertaining to calibration of related SPFs to a common state. In individual cases, this review led to one of the following decisions: to use the calibrated SPFs; to adjust the calibrated SPFs, as appropriate, for consistency with one another; or to retain the original SPFs.
- New crash prediction methods for pedestrian and bicycle crashes on rural two-lane roads developed in NCHRP Project 17-84 (Torbic et al., 2022) have been incorporated into the chapter (see later sections pertaining to predicting pedestrian collisions and predicting bicycle collisions).
- To illustrate the changed predictive methods in the chapter, existing sample problems have been updated, and new sample problems have been added (see later section pertaining to sample problems).
Some of the research performed for HSM2 was not compatible with the other procedures in the chapter and, for that reason, decisions were reached to not incorporate some new models into the chapter. For example, the crash-type–specific SPFs developed in NCHRP Project 17-62 used crash-type definitions that were not compatible with those used in other HSM procedures or not compatible with Model Minimum Uniform Crash Criteria (MMUCC) definitions (NHTSA, 2017) and, for those reasons, were not used in the chapter.
Predicted crash frequencies for individual crash severity levels are determined using tables of crash severity proportions, except for roundabouts, for which separate SPFs for KABC and property damage only (PDO) crashes are provided along with severity distribution functions (SDFs) that separate the predicted KABC crash estimate into estimates for individual injury severity levels.
Predicted crash frequencies for specific collision types are determined using tables of collision type proportions. The SPFs and AFs used in HSM2 Chapter 14 to estimate predicted average crash frequencies (excluding pedestrian and bicycle collisions) are shown in Table 5.
Table 5. SPFs and AFs used in HSM2 Chapter 14: Rural two-lane, two-way roads.
| Site Type | SPFs | AFs |
|---|---|---|
| Roadway segments |
|
|
| Conventional intersections |
|
|
| Roundabouts |
|
* SDFs are provided for these facility types.
The results of NCHRP Project 17-54, “Consideration of Roadside Features in the Highway Safety Manual,” were considered for inclusion in HSM2 Chapter 14. However, it was found that the NCHRP Project 17-54 results provided predictions for run-off-road crashes. Neither the HSM1 nor HSM2 procedures distinguish run-off-road crashes from other crash types. Other inconsistencies in the NCHRP Project 17-54 models include (a) overly high predicted levels of run-off-road crashes compared to historical crash data, (b) inconsistency of base conditions for CMFs, and (c) inconsistencies in the CMF for isolated fixed-object density. These findings made it impractical to use the NCHRP Project 17-54 results in HSM2 Chapter 14. For the same reasons, it was found to be impractical to use the NCHRP Project 17-54 results in HSM2 Chapters 15 and 16 as well.
HSM2 Chapter 15—Predictive Method for Rural Multilane Highways
HSM2 Chapter 15, which presents a crash prediction method for rural multilane divided and undivided highways, was adapted from HSM1 Chapter 11, incorporating the results of new research. While the structure of the predictive method remained similar to that used in HSM1, the following new features were added to HSM2 Chapter 15:
- New SPFs for total crashes, KABC crashes, and KAB crashes were added based on the results of NCHRP Project 17-62 (Ivan et al., 2018) for undivided roadway segments, divided roadway segments, and three intersection types (3ST, 4ST, and 4SG intersections). Existing CMFs from the crash prediction method in HSM1 Chapter 11 were retained for use as AFs in the updated chapter since NCHRP Project 17-62 did not recommend any revised AFs.
- New SPFs developed in NCHRP Project 17-68 (Torbic et al., 2020) for one intersection type (3SG intersections) were incorporated into the chapter.
- A predictive method with SPFs and AFs for roundabouts on rural multilane roads, developed in NCHRP Project 17-70 (Ferguson et al., 2018), was incorporated into the chapter (see later section pertaining to predicting crashes for roundabouts).
- Many of the SPFs from recent research mentioned previously were adjusted, where appropriate, through calibration to data for a common state using results from NCHRP Project 17-72 (Srinivasan et al., 2022), as described in the later section pertaining to calibration of related SPFs to a common state. In addition, the sets of original and calibrated SPFs were reviewed for consistency and credibility, as described in the later section pertaining to calibration of related SPFs to a common state. In individual cases, this review led to one of the following decisions: to use the calibrated SPFs; to adjust the calibrated SPFs, as appropriate, for consistency with one another; or to retain the original SPFs.
- New crash prediction methods for pedestrian and bicycle crashes on rural multilane roads developed in NCHRP Project 17-84 (Torbic et al., 2022) have been incorporated into the chapter (see later sections pertaining to predicting pedestrian collisions and predicting bicycle collisions).
- To illustrate the changed predictive methods in the chapter, existing sample problems have been updated, and new sample problems have been added (see later section pertaining to sample problems).
Some of the research performed for HSM2 was not compatible with the other procedures in the chapter and, for that reason, decisions were reached to not incorporate some new models in the chapter. For example, the crash-type–specific SPFs developed in NCHRP Project 17-62 used crash-type definitions that were not compatible with those used in other HSM procedures or not compatible with MMUCC definitions (NHTSA, 2017) and, for those reasons, were not used in the chapter.
Predicted crash frequencies for KABC, KAB, and PDO crashes are determined using separate SPFs for each crash severity level, except for roundabouts, for which separate SPFs for KABC and PDO crashes are provided along with SDFs that separate the predicted KABC crash estimate into estimates for individual injury severity levels.
Predicted crash frequencies for specific collision types are determined using tables of collision type proportions. The SPFs and AFs used in HSM2 Chapter 15 to estimate predicted average crash frequencies (excluding pedestrian and bicycle collisions) are shown in Table 6.
HSM2 Chapter 16—Predictive Method for Urban and Suburban Arterials
HSM2 Chapter 16, which presents a crash prediction method for urban and suburban arterials, was adapted from HSM1 Chapter 12, incorporating the results of new research. While the
Table 6. SPFs and AFs used in HSM2 Chapter 15: Rural multilane highways.
| Site Type | SPFs | AFs |
|---|---|---|
| Roadway segments |
|
|
| Conventional intersections |
|
|
| Roundabouts |
|
* SDFs are provided for these facility types.
structure of the predictive method remained similar to that used in HSM1, the following new features were added to HSM2 Chapter 16:
- New SPFs for total crashes, KABC crashes, and KAB crashes were added based on the results of NCHRP Project 17-62 (Ivan et al., 2018) for five types of roadway segments (2U, 3T, 4U, 4D, and 5T; see Table 7 for expansions of abbreviations). Existing SPFs from HSM1 were retained for four intersection types (3ST, 3SG, 4ST, and 4SG intersections) on roadways with five or fewer lanes. Existing CMFs from the crash prediction method in HSM1 Chapter 12 were retained for use as AFs in the updated chapter since NCHRP Project 17-62 did not recommend any revised AFs.
- New SPFs developed in NCHRP Project 17-68 (Torbic et al., 2018) for seven intersection types (3ST-HS, 3AST, 3STT, 3SG-HS, 4AST, 4SG-HS, and 5SG intersections; see Table 7) were incorporated into the chapter.
- SPFs for four roadway segment types on urban and suburban arterials with six or more lanes (6U, 6D, 7T, and 8D), three one-way roadway segment types on urban and suburban arterials (2O, 3O, and 4O; see Table 7), and four related intersection types on six-lane and one-way arterials (3ST, 3SG, 4ST, and 4SG intersections) were incorporated into the chapter based on the results for NCHRP Project 17-58 (Lord et al., 2019).
- A predictive method with SPFs and AFs for roundabouts on urban and suburban arterials, developed in NCHRP Project 17-70 (Ferguson et al., 2018), was incorporated into the chapter (see later section pertaining to predicting crashes for roundabouts).
- Many of the SPFs from recent research mentioned previously were adjusted, where appropriate, through calibration to data for a common state using results from NCHRP Project 17-72 (Srinivasan et al., 2022), as described in the later section pertaining to calibration of related SPFs to a common state. In addition, the sets of original and calibrated SPFs were reviewed for consistency and credibility, as described in the later section pertaining to calibration of related SPFs to a common state. In individual cases, this review led to one of the following decisions: to use the calibrated SPFs; to adjust the calibrated SPFs, as appropriate, for consistency with one another; or to retain the original SPFs.
- New crash prediction methods for pedestrian and bicycle crashes on urban and suburban arterials developed in NCHRP Project 17-84 (Torbic et al., 2022) have been incorporated into the chapter (see later sections pertaining to predicting pedestrian collisions and predicting bicycle collisions).
- To illustrate the changed predictive methods in the chapter, existing sample problems have been updated and new sample problems added (see later section pertaining to sample problems).
Some of the research performed for HSM2 was not compatible with the other procedures in the chapter and, for that reason, decisions were reached to not incorporate some new models into the chapter. For example, the crash-type–specific SPFs developed in NCHRP Project 17-62 used crash-type definitions that were not compatible with those used in other HSM procedures or not compatible with MMUCC definitions (NHTSA, 2017) and, for those reasons, were not used in the chapter.
Predicted crash frequencies for either KABC, KAB, and PDO crashes or KABC and PDO crashes are determined using separate SPFs for each crash severity level, except for roundabouts for which separate SPFs for KABC and PDO crashes are provided along with SDFs that separate the predicted KABC crash estimate into estimates for individual injury severity levels.
Predicted crash frequencies for specific collision types are determined using tables of collision type proportions. The SPFs and AFs used in HSM2 Chapter 16 to estimate predicted average crash frequencies (excluding pedestrian and bicycle collisions) are shown in Table 7.
HSM2 Chapter 17—Predictive Method for Directional Freeway Segments
The crash prediction method for freeway segments in HSM2 Chapter 17 is essentially the same as the method in HSM1 Chapter 18, except that the method has been converted from a
Table 7. SPFs and AFs used in HSM2 Chapter 16: Urban and suburban arterials.
| Site Type | SPFs | AFs |
|---|---|---|
| Roadway segments |
Two-way roadway segments with five lanes or fewer (2U, 3T, 4U, 4D, and 5T):
Two-way roadway segments with six or more lanes (6U, 6D, 7T, and 8D):
One-way roadway segments (2O, 3O, and 4O):
|
|
| Conventional intersections |
Intersections between two two-way arterials each with five lanes or fewer (3ST, 3ST-HS, 3AST, 3STT, 3SG, 3SG-HS, 4ST, 4ST-HS, 4AST, 4SG, 4SG-HS, and 5SG):
Intersections between two-way arterials with six or more lanes (3ST, 3SG, 4ST, and 4SG):
Intersections between two one-way arterial streets (3ST, 3SG, 4ST, and 4SG):
Intersections between one-way arterial streets and two-way arterials (3ST, 3SG, 4ST, and 4SG):
|
|
| Roundabouts |
|
* SDFs are also available; a applicable to two-way roadway segments with five lanes or fewer; b applicable to two-way roadway segments with six or more lanes; c applicable to one-way roadway segments; d for two-way roadway segments with five lanes or fewer, driveways are addressed as part of the SPF, not with an AF; e applicable to intersections on two-way roadway segments with five lanes or fewer; f applicable to intersections on two-way roadway segments with six or more lanes; g applicable to intersections of two one-way roadways; h applicable to intersections of a one-way roadway and a two-way roadway; i used only in conjunction with the SPF for 3STT intersections.
procedure that predicts crashes for two-way segments to an equivalent procedure for directional segments. A directional freeway segment consists of the roadway for one direction of travel (i.e., on one side of the median). The reasons for converting the method to provide a directional segment analysis are as follows:
- Many analysts appear to find a directional segment procedure to be more useful than a two-way segment procedure; it is understood that analysts seeking crash predictions for a directional segment have been applying the two-way segment procedure in HSM1 Chapter 18 assuming identical roadways in both directions of travel and then dividing the crash prediction in half.
- Use of a directional analysis procedure allows elimination or simplification of many awkward assumptions in the two-way procedure about how to analyze roadways in the opposing
- directions of travel that are not identical. With the change from a bidirectional to a directional analysis procedure, the following issues no longer complicate the roadway segmentation process:
- Independent alignments in the two directions of travel that differ in length
- Differing horizontal curve radii and lengths in the two directions of travel
- Differing speed-change lane types, lengths, and locations in the two directions of travel
- Variations in median width
- Variations in barrier placement or barrier distance from the traveled way in the two directions of travel
- A directional segment procedure could possibly be more appropriate for integration of part-time shoulder use (PTSU) and managed lanes [i.e., high occupancy vehicle (HOV) and high occupancy toll (HOT) lanes] procedures if they are added to HSM2 Chapter 17 in the future.
The changes to the existing two-way segment crash prediction method used to implement the directional segment procedure were as follows:
- Revise text descriptions and figures to address directional segments
- Use directional annual average daily traffic (AADT) rather than two-way AADT
- Multiply directional AADT by two
- Use existing two-way segment models with directional segment characteristics in the revised crash prediction procedure that has been modified as described previously
- Divide resulting crash frequency predictions by two
This approach of simple modifications to the HSM1 two-way segment procedure to create the directional segment procedure in HSM2 Chapter 17 is fully compatible with the HSM2 Chapter 18 procedure for ramps.
All of the changes discussed here have been integrated into the equations used in the directional segment procedure, so no special understanding of the procedural details of the change from two-way to directional segments is needed by the HSM user.
Some reorganizing of the chapter and some revisions to figures made in NCHRP Project 17-71 were retained in the chapter developed in NCHRP Project 17-71A.
The research team assessed the crash prediction method for freeway segments with PTSU developed in NCHRP Project 17-89 (Jenior et al., 2021). For reasons discussed later in the section pertaining to predicting crashes for part-time shoulder use, it was concluded that the NCHRP Project 17-89 results were not compatible with the available HSM2 crash prediction method for freeway segments, so the NCHRP Project 17-89 method for freeway segments with PTSU has not been incorporated into HSM2 Chapter 17.
The research team assessed the crash prediction method for freeway segments with managed lanes (i.e., HOV/HOT lanes) developed in NCHRP Project 17-89A (Himes et al., 2021). For reasons discussed later in the section pertaining to predicting crashes for managed lanes, it was concluded that the NCHRP Project 17-89A results were not compatible with the available HSM2 crash prediction method for freeway segments, so the NCHRP Project 17-89A method for freeway segments with managed lanes has not been incorporated into HSM2 Chapter 17.
An assessment by FHWA of the freeway segment severity distribution function in HSM1 Chapter 18 found that the method indicates an increase in fatal (K) and serious-injury (A) crashes when shoulder rumble strips are provided. This appears counterintuitive given that the current understanding of shoulder rumble-strip effects on crashes indicates that shoulder rumble strips reduce crashes. FHWA has been investigating but has not yet resolved this issue. Therefore, to avoid counterintuitive results, the AF for shoulder rumble strips was removed from the predictive method.
The SPFs and AFs used in HSM2 Chapter 17 are identified in Table 8.
Table 8. SPFs and AFs used in HSM2 Chapter 17: Directional freeway segments.
| Site Type | SPFs | AFs |
|---|---|---|
| Directional freeway segments |
Multiple-vehicle crashes (KABC and PDO)*:
Single-vehicle crashes (FI and PDO)*:
|
|
| Directional freeway segments containing speed-change lanes |
Directional segments with ramp entrance speed-change lanes (KABC and PDO)*:
Directional segments with ramp exit speed-change lanes (KABC and PDO)*:
|
|
* SDFs are also available; a used with SPFs for KABC crashes only; b used with SPFs for multiple-vehicle crashes only; c used with SPFs for single-vehicle crashes only; d used with SPFs for KABC single-vehicle crashes only.
HSM2 Chapter 18—Predictive Method for Ramps
The crash prediction method for freeway ramps and collector–distributor (C-D) roads in HSM2 Chapter 18 is essentially the same as the method in HSM1 Chapter 19, with the addition of SPFs for single-point and tight-diamond crossroad ramp terminals developed in NCHRP Project 17-68. Some reorganizing of the chapter and revisions to figures made under NCHRP Project 17-71 were also retained in the chapter.
The SPFs and AFs used in HSM2 Chapter 18 are identified in Table 9.
Predicting Crashes for Roundabouts
New crash prediction methods for roundabout intersections were developed in NCHRP Project 17-70 (Ferguson et al., 2018) and have been incorporated into HSM2 Chapters 14, 15, and 16. The research developed methods to predict crash frequencies for a roundabout intersection as a whole or for specific roundabout legs. For consistency, only the methods for roundabout intersections as a whole were incorporated into the HSM, since no other intersection type addressed in the HSM has leg-specific models.
Table 9. SPFs and AFs used in HSM2 Chapter 18: Ramps.
| Site Type | SPFs | AFs |
|---|---|---|
| Roadway segments on ramps |
Rural ramp segments (MV-KABC, MV-PDO, SV-KABC, SV-PDO)*:
Urban ramp segments (MV-KABC, MV-PDO, SV-KABC, SV-PDO)*:
|
|
| Roadway segments on C-D roads |
Rural C-D road segments (MV-KABC, MV-PDO, SV-KABC, SV-PDO)*:
Urban C-D road segments (MV-KABC, MV-PDO, SV-KABC, SV-PDO)*:
|
|
| One-way stop-controlled crossroad ramp terminals |
Three-leg terminals with diagonal exit ramp (D3ex) (KABC and PDO)*:
Three-leg terminals with diagonal entrance ramp (D3en) (KABC and PDO)*:
Four-leg terminals at conventional diamond (D4cd) (KABC and PDO)*:
Four-leg terminals at four-quadrant partial cloverleaf A (A4) (KABC and PDO)*:
Four-leg terminals at four-quadrant partial cloverleaf B (B4) (KABC and PDO)*:
Three-leg terminals at two-quadrant partial cloverleaf A (A2) (KABC and PDO)*:
Three-leg terminals at two-quadrant partial cloverleaf B (B2) (KABC and PDO)*:
|
|
| All-way stop-controlled crossroad ramp terminals |
|
|
| Signal-controlled crossroad ramp terminals |
Three-leg terminals with diagonal exit ramp (D3ex) (KABC and PDO)*:
Three-leg terminals with diagonal entrance ramp (D3en) (KABC and PDO)*:
|
|
| Site Type | SPFs | AFs |
|---|---|---|
Four-leg terminals at conventional diamond (D4cd) (KABC and PDO)*:
Four-leg terminals at four-quadrant partial cloverleaf A (A4) (KABC and PDO)*:
Four-leg terminals at four-quadrant partial cloverleaf B (B4) (KABC and PDO)*:
Three-leg terminals at two-quadrant partial cloverleaf A (A2) (KABC and PDO)*:
Three-leg terminals at two-quadrant partial cloverleaf B (B2) (KABC and PDO)*:
Six-leg terminal with diagonal ramps at a single-point diamond (SP) (KABC and PDO)c Combined four-leg terminals with diagonal ramps at tight diamond (D4td)(KABC and PDO)c |
|
|
| Roundabouts | See applicable SPFs in HSM2 Chapters 14, 15, and 16 |
|
* SDFs are also available; a used with SPFs for KABC multiple-vehicle crashes only; b applies to urban ramp terminals only; c the listed AFs do not apply to this terminal type.
Crash prediction models were developed in NCHRP Project 17-70 for intersections as a whole for all combinations of the following features:
- For both three-leg and four-leg roundabouts.
- For roundabouts with both one and two circulating lanes.
- For both KABC and PDO crash severity levels.
The crash prediction models for roundabouts include both SPFs and AFs. Some AFs apply to the roundabout as a whole, while others apply to individual roundabout legs and are aggregated to the intersection level by combining the AFs for all legs present at the roundabout. The roundabout models include an AF for the presence of one outbound-only leg. This configuration generally only occurs for roundabouts at crossroad ramp terminals or on one-way streets. It might seem logical to present the procedures for crossroad ramp terminals in HSM2 Chapter 18. However, the research team decided that it made little sense to include the entire roundabout procedures from HSM2 Chapters 14, 15, and 16 with only one AF value in HSM2 Chapter 18 potentially differing from those other chapters. Therefore, HSM2 Chapter 18 includes a reference to HSM2 Chapters 14, 15, and 16 for the applicable predictive methods for roundabouts at crossroad ramp terminals. The SPFs and AFs included for roundabouts in each chapter are summarized in Tables 5 through 7.
Each of the chapters in which roundabouts were considered incorporated the roundabout types most typically used on the roadways addressed in that chapter. Specifically:
- Only single-lane roundabouts are considered for rural two-lane roads in HSM2 Chapter 14.
- Only two-lane roundabouts are considered for rural multilane highways in HSM2 Chapter 15.
- Both single- and two-lane roundabouts are considered for urban and suburban arterials in HSM2 Chapter 16.
In the unusual case that a two-lane roundabout should be found on a rural two-lane road, the HSM2 Chapter 15 procedure for two-lane roundabouts may be applied. In the unusual case that a single-lane roundabout should be found on a rural multilane highway, the HSM2 Chapter 14 procedure for two-lane roundabouts may be applied.
An issue identified in the review of the NCHRP Project 17-70 research results is that the crash prediction models for roundabouts predict many more crashes than the comparable HSM1 and HSM2 models for conventional intersections. Available research results suggest that the opposite should be the case. For example, HSM1 Part D shows that typical CMFs for total crashes for converting conventional intersections to roundabouts for all settings and crash severity levels are 0.52 for signalized intersection to roundabout conversions in HSM1 Table 14-3 and 0.56 for stop-controlled intersection to roundabout conversions in HSM1 Table 14-4. The base condition for these CMFs was the absence of a roundabout. Such CMFs were used in adjusting the roundabout SPFs from NCHRP Project 17-70 to reflect the known effectiveness of roundabouts to be more appropriate relative to other intersection types (see discussion later in the section pertaining to calibration of related SPFs to a common state).
Predicting Crashes for Part-Time Shoulder Use
The research team assessed the crash prediction method for freeway segments with PTSU developed in NCHRP Project 17-89 (Jenior et al., 2021). This research developed AFs representing the effect on crashes for:
- PTSU lanes and transition zones and
- Turnouts provided for stopped vehicles outside shoulders with PTSU.
The researchers in NCHRP Project 17-89 concluded that these AFs were most appropriately used with a directional segment prediction method, rather than the two-way segment method presented in HSM1 Chapter 18. Therefore, NCHRP Project 17-89 developed a new predictive method for directional freeway segments, independent of the HSM1 Chapter 18 method, to which the PTSU AFs were intended to be applied. Because this new directional freeway segment method was developed with data from different states than were used to develop the crash prediction models in HSM1 Chapter 18, this new method had the potential to provide crash frequency predictions that were incompatible with crash frequency predictions from HSM1 Chapter 18 and the new HSM2 Chapter 17. Therefore, it appeared that the PTSU AFs developed in NCHRP Project 17-89 would need to be modified in some way yet to be determined to be implemented in HSM2 Chapter 17. Furthermore, there was an issue that the PTSU AFs developed in NCHRP Project 17-89 sometimes predicted fewer crashes with PTSU than without PTSU, and sometimes predicted more crashes with PTSU than without PTSU. No independent data source was available to confirm whether these varying effects were accurate. Therefore, a decision was reached not to incorporate the NCHRP Project 17-89 effects for PTSU into HSM2 Chapter 17.
If the issues described previously concerning PTSU effects are resolved in the future, the conversion of the crash prediction method for freeway segments to a directional procedure should make HSM2 Chapter 17 a more suitable candidate for incorporating PTSU effects than HSM1 Chapter 18.
Predicting Crashes for Managed Lanes
The research team assessed the crash prediction method for managed lanes on freeway segments, including HOV and toll HOT lanes, developed in NCHRP Project 17-89A (Himes et al., 2021). This review found that NCHRP Project 17-89A had developed a crash prediction method for HOV/HOT lanes, but this crash prediction method had been developed independently of
the existing crash prediction method for freeway segments without HOV/HOT lanes in HSM1 Chapter 18, which served as the basis for the directional freeway segment method in HSM2 Chapter 17. This independent development of models makes it unlikely that the NCHRP Project 17-89A results can be used reliably to forecast the difference in crash frequency between managed and general-purpose lanes or the effect on crashes of adding managed lanes to an existing freeway segment. For this reason, a decision was reached not to incorporate the NCHRP Project 17-89A models for HOV/HOT into HSM2 Chapter 17.
If the issues described here concerning the effect on crashes of providing HOV/HOT lanes on an existing freeway are resolved in the future, the conversion of the crash prediction method for freeway segments to a directional procedure should make HSM2 Chapter 17 a more suitable candidate for incorporating HOV/HOT lane effects than HSM1 Chapter 18.
Predicting the Effect on Crashes of Freeway Rumble Strips
An unpublished investigation by FHWA has found that the current SDF for freeway segments predicts an increase in fatal (K) and serious-injury (A) crashes when shoulder rumble strips are installed on an existing freeway segment. This appears counterintuitive given that the current understanding of shoulder rumble-strip effects on crashes indicates that shoulder rumble strips reduce crashes. FHWA has been investigating but has not yet resolved this issue. Therefore, to avoid counterintuitive results, the AF for shoulder rumble strips was removed from the predictive method in HSM2 Chapter 17.
Predicting Pedestrian Collisions
The HSM1 included only limited capabilities for predicting pedestrian collisions. HSM1 Chapter 12 included a multiplicative factor to increase the predicted frequencies of crashes not involving pedestrians or bicyclists to account for motor-vehicle collisions with pedestrians. These factors provided typical or average estimates of pedestrian crash frequencies but were not sensitive to the presence/absence or characteristics of pedestrian facilities. HSM1 Chapter 12 also included a crash prediction method for motor-vehicle collisions with pedestrians at four-leg signalized intersections on urban and suburban arterials.
HSM2 Chapters 14, 15, and 16 include prediction methods for pedestrian collisions developed in NCHRP Project 17-84 for a broad range of facility types. Pedestrian collisions are defined in the HSM2 as collisions involving a motor vehicle and a pedestrian. Several candidate approaches for predicting pedestrian crash frequencies were considered in NCHRP Project 17-84, but the most appropriate method selected in that research was based on the crash prediction method developed by the International Road Assessment Programme (iRAP) with support from its U.S. partner, the U.S. Road Assessment Program (usRAP). The iRAP procedures were developed using worldwide research on pedestrian collisions and have been used in more than 70 countries around the world. The iRAP crash prediction method was adapted to U.S. conditions in NCHRP Project 17-84 and was adapted in ways to best fit into the HSM2 in NCHRP Projects 17-84 and 17-71A.
The crash prediction methods from NCHRP Project 17-84 (Torbic et al., 2022) were incorporated into HSM2 Chapters 14, 15, and 16. The crash prediction methods address the frequency of crashes between motor vehicles and pedestrians in which a pedestrian was injured and the number of pedestrians injured in such crashes; other types of pedestrian collisions or incidents are not addressed because, under the laws of most states, these are not considered reportable traffic crashes and, for that reason, data on such occurrences are typically incomplete or unavailable. The models are not intended to estimate miscellaneous pedestrian collisions related to pedestrian actions without forecast volumes, such as:
- Police officers standing in the traveled way for traffic control or enforcement reasons,
- Construction workers in the traveled way in work zones, and
- Vehicle occupants whose vehicle break down and who leave their vehicles to work on repairs.
No pedestrian crash prediction methods were incorporated into HSM2 Chapters 17 and 18 because most transportation agencies prohibit pedestrian travel on freeways and ramps.
Four types of pedestrian crash prediction models for specific HSM2 chapters were obtained from NCHRP Project 17-84, including models for collisions involving:
- Pedestrian movements along the left side of the road,
- Pedestrian movements along the right side of the road,
- Pedestrian crossing movements at midblock locations (including marked midblock crossings and unmarked locations), and
- Pedestrian crossing movements at intersections (including marked and unmarked locations).
The general form of the prediction model to estimate the frequency of pedestrian collisions is as follows:
N = Crash Likelihood Factors × Crash Severity Factors × Motor-Vehicle Speed Factor × Motor-Vehicle Volume (AADT) Factor × Peak-Hour Pedestrian Volume Factor × Calibration Factor
Where:
N = annual number of predicted pedestrian collisions.
The crash likelihood factors include:
- Factors related to the likelihood that motor vehicles will depart from their lane or leave the roadway (and, therefore, might potentially strike a pedestrian) and
- Factors related to the direct effects of pedestrian facilities on pedestrian collisions.
The crash severity factors all relate to the direct effects of pedestrian facilities on pedestrian collisions.
The speed- and volume-related factors have effects in the expected direction. Crash frequencies for specific crash severity levels increase with increasing motor-vehicle speed, increasing motor-vehicle volume, and increasing pedestrian volume. If pedestrian volume data are not available for particular locations, such volumes should be estimated.
The pedestrian models have already been calibrated to typical U.S. conditions in NCHRP Project 17-84. The calibration procedures in HSM2 Chapter 13 can be used by individual agencies to calibrate the models to their local conditions. Table 10 identifies the specific types of pedestrian facilities whose effects on pedestrian collisions are addressed in the crash prediction methods.
Table 10. Pedestrian facility types and related features whose effects on pedestrian collisions are addressed in the HSM2 crash prediction methods.
Presence/absence of sidewalk
Presence and width of paved shoulder
|
Advance visibility of crossing Number of traffic lanes to be crossed Presence of pedestrian fencing Type of median present Type of intersection present School zone crossing:
|
The quantitative results for pedestrian collisions provided by the crash prediction method include:
- Number of fatal (K) collisions,
- Number of A-injury collisions,
- Number of B-injury collisions,
- Number of C-injury collisions,
- Number of pedestrians fatally injured,
- Number of pedestrians with A injuries,
- Number of pedestrians with B injuries, and
- Number of pedestrians with C injuries.
These quantitative results are computed with crash severity proportions and average numbers of pedestrians killed or injured per crash from Torbic et al. (2022).
The crash prediction method for pedestrian collisions at four-leg signalized intersections on urban and suburban arterials in HSM1 Chapter 12 was not retained because it was incompatible with the NCHRP Project 17-84 method, which addressed the same intersection type as well as several others.
Predicting Bicycle Collisions
The HSM1 included only limited capabilities for predicting bicycle collisions. HSM1 Chapter 12 included a multiplicative factor to increase the predicted frequencies of crashes not involving pedestrians or bicyclists to account for motor-vehicle collisions with bicycles. These factors provided typical or average estimates of bicycle crash frequencies but were not sensitive to the presence/absence or characteristics of bicycle facilities.
HSM2 Chapters 14, 15, and 16 include prediction methods for bicycle collisions developed in NCHRP Project 17-84 (Torbic et al., 2022) for a broad range of facility types. Bicycle collisions are defined in the HSM2 as collisions involving a motor vehicle and a bicycle but not involving a pedestrian. Several candidate approaches for predicting bicycle crash frequencies were considered in NCHRP Project 17-84, but the most appropriate method selected in that research was based on the crash prediction method developed by iRAP with support from usRAP. The iRAP procedures were developed using worldwide research on bicycle collisions and have been used in more than 70 countries around the world. The iRAP crash prediction method was adapted to U.S. conditions in NCHRP Project 17-84 and was adapted in ways to best fit into the HSM2 in NCHRP Projects 17-84 and 17-71A.
The crash prediction methods from NCHRP Project 17-84 were incorporated into HSM2 Chapters 14, 15, and 16. The crash prediction methods address the frequency of collisions between motor vehicles and bicycles in which a bicyclist was injured and the number of bicyclists injured in such collisions; other types of bicycle collisions or incidents are not addressed because, under the laws of most states, these are not considered reportable traffic crashes and, for that reason, data on such occurrences are typically incomplete or unavailable. No prediction methods for estimating the frequency of bicycle collisions were incorporated into HSM2 Chapters 17 and 18 because most transportation agencies prohibit bicycle travel on freeways and ramps.
Two types of bicycle crash prediction models for specific HSM2 chapters were obtained from NCHRP Project 17-84, including models for collisions involving:
- Bicycle movements along the road and
- Bicycle movements through intersections.
The general form of the prediction model to estimate the frequency of bicycle collisions is as follows:
N = Crash Likelihood Factors × Crash Severity Factors × Motor-Vehicle Speed Factor × Motor-Vehicle Volume (AADT) Factor × Peak-Hour Bicycle Volume Factor × Calibration Factor
Where:
N = annual number of predicted bicycle collisions.
The crash likelihood factors include:
- Factors related to the likelihood that motor vehicles will depart from their lane or leave the roadway (and, therefore, might potentially strike a bicycle) and
- Factors related to the direct effects of bicycle facilities on bicycle collisions.
The crash severity factors all relate to the direct effects of bicycle facilities on bicycle collisions.
The speed- and volume-related factors have effects in the expected direction. Crash frequencies for specific crash severity levels increase with increasing motor-vehicle speed, increasing motor-vehicle volume, and increasing bicycle volume. If bicycle volume data are not available for particular locations, such volumes should be estimated.
The bicycle models have already been calibrated to typical U.S. conditions in NCHRP Project 17-84. The calibration procedures in HSM2 Chapter 13 can be used by individual agencies to calibrate the models to their local conditions. Table 11 identifies the specific types of bicycle facilities whose effects on bicycle collisions are addressed in the crash prediction methods.
The quantitative results for bicycle collisions provided by the crash prediction method include:
- Number of fatal (K) collisions,
- Number of A-injury collisions,
- Number of B-injury collisions,
- Number of C-injury collisions,
- Number of bicyclists fatally injured,
- Number of bicyclists with A injuries,
- Number of bicyclists with B injuries, and
- Number of bicyclists with C injuries.
The quantitative results listed here are computed with crash severity proportions and average numbers of bicyclists killed or injured per crash from Torbic et al. (2022).
Calibration of Related SPFs to a Common State
As was done in developing HSM1, the crash prediction methods provided in HSM2 that may be used to directly compare design alternatives have been calibrated, where appropriate, to a common state. The reason for calibration to a common state was to reduce the possibility that crash
Table 11. Bicycle facility types and related features whose effects on bicycle collisions are addressed in the HSM2 crash prediction methods.
|
|
prediction methods for design alternatives that might be compared to one another (e.g., two-lane undivided and four-lane divided roadway segments on rural highways) might provide misleading results when compared not because of inherent differences in crash experience between the design alternatives, but because the models were developed with data from different states. Calibration to a common state was implemented by changing the value of the constant term (i.e., the value of the a coefficient) in the SPF using a calibration factor whose value was determined with the calibration procedure presented in HSM2 Chapter 13. Such calibration increases the likelihood that the models for various design alternatives provide crash predictions whose differences are realistic in at least a relative sense, such that the design alternatives likely to experience fewer crashes than other alternatives can be identified.
Pairs of crash prediction models with and without calibration to a common state were compared in NCHRP Project 17-71A, and the calibrated models were used whenever the revised SPFs appeared to improve the relative magnitudes of predicted crash values. Specifically, calibration factors derived in NCHRP Project 17-72 (Srinivasan et al., 2022) were used to adjust the value of the a coefficients in the SPFs used in HSM2 Chapters 14 and 15. These calibration factors determined in NCHRP Project 17-72 were based on roadway characteristics and crash data for locations on state highways in Ohio. The models without calibration were retained in HSM2 Chapter 16 because the relative magnitudes of the models without calibration appeared to be more realistic than the models with calibration; in some cases, the SPFs from HSM1 were retained. No calibration to a common state was needed for the freeway and ramp models in HSM2 Chapters 17 and 18 because, with minor exceptions, all of the models in these two chapters were originally developed in NCHRP Project 17-45 (Bonneson et al., 2012) with data from a common set of states.
Another issue that was addressed in conjunction with calibration of models to a common state concerned the crash prediction models for HSM2 Chapters 14, 15, and 16 developed in NCHRP Project 17-62 (Ivan et al., 2018). Part of the research in that project was a decision to exclude animal-related crashes from the data used to develop revised SPFs. This may (or may not) have been an appropriate decision, but it created a potential inconsistency between the SPFs from NCHRP Project 17-62 and the SPFs in those same chapters from other sources that did include animal-related crashes in the SPF development. The SPFs from NCHRP Project 17-62 predicted crash totals that were slightly smaller in comparison to the SPFs developed in other projects. An adjustment was made to the value of the a coefficient in the NCHRP Project 17-62 SPFs based on typical proportions of animal-related crashes for specific facility types, based on crash-type proportion data for those facility types from the HSM1. This adjustment was generally very limited in magnitude except for roadway segment crashes on rural two-lane highways in HSM2 Chapter 14.
After implementing the results of calibration to the common state that was done in NCHRP Project 17-72, some differences between crash prediction models for configurations that would likely be considered as design alternatives remained. For example, the roundabout models developed in NCHRP Project 17-70 (Ferguson et al., 2018) generally predicted higher crash frequencies than the models for comparable signalized and unsignalized intersections. CMFs for conversion of conventional intersections to roundabouts show that just the opposite would be expected; roundabouts generally experience about half as many crashes as comparable signalized or unsignalized intersections. Since no calibration had been performed to address such differences, the a coefficients in the roundabout SPFs used in HSM2 Chapters 14, 15, and 16 were adjusted based on known CMFs for roundabout conversion projects from the FHWA CMF Clearinghouse. The resulting roundabout models provide realistic results relative to the comparable signalized and unsignalized intersection models.
Appendix B presents representative plots of calibrated and noncalibrated models to illustrate the results of the reviews that were performed. Even where calibration to a common state has been implemented, calibration of the crash prediction models by individual agencies using the
procedures presented in HSM2 Chapter 13 is appropriate because of differences in crash experience between states and between time periods.
Sample Problems
Sample problems are included in HSM Part C to illustrate how the crash prediction calculations are intended to be performed. As part of NCHRP Project 17-71A, the existing sample problems were updated in response to changes in the crash prediction procedures, and new sample problems were added for selected new facility types and for pedestrian and bicycle collisions.
HSM2 Part D—Crash Modification Factors
Introduction to HSM2 Part D
The Introduction to Part D combined material from the HSM1 Introduction to Part D and a draft introductory chapter developed under NCHRP Project 17-71. It includes an updated description of the roadway safety management process to include material on areawide and systemic evaluations and how CMFs are used in the project development process.
HSM2 Chapter 19—Selecting CMFs
This is a new chapter developed under NCHRP Project 17-63. The topics covered in this chapter include (a) identifying the most appropriate CMF, (b) converting CMFs for appropriate application, (c) searching for CMFs, (d) developing a custom CMF list, and (e) estimating countermeasure effects without applicable, high-quality CMFs. As part of NCHRP Project 17-71A, text and tables on the new CMF rating system used in conjunction with FHWA CMF Clearinghouse have been integrated into the chapter, and text and a figure on adjusting CMFs to local conditions have also been integrated into the chapter.
HSM2 Chapter 20—Applying CMFs
This is a new chapter developed under NCHRP Project 17-63. The topics covered in this chapter include (a) applying CMFs to baseline crashes, (b) estimating a confidence interval, (c) using results, and (d) common misapplications of CMFs. As part of NCHRP Project 17-71A, text on combining CMFs for the same countermeasure has been integrated into the chapter. In addition, the research team edited, modified, and combined material to reduce redundancy between Chapter 2: Road Safety Principles, Chapter 11: Countermeasure Effectiveness Evaluation, and Chapter 19: Selecting CMFs. Two sample problems were also updated.
