Highway Safety Manual Implementation A National Perspective Kim Kolody Silverman, PE September 28, 2017

Are you using the Highway Safety Manual? It is too difficult It takes too much time We like the way we are doing things. Why change? The results are not what we expected. We don’t like the answer. The predictive models did not address the scenario I was working with

The HSM is a tool to change how we consider safety Nominal Safety Substantive Safety The expected or actual crash frequency and severity for a highway or roadway Examined in reference to compliance with standards, warrants, guidelines and sanctioned design procedures Key Message: Highway engineers are used to thinking about safety in terms of adherence to design criteria such as are published in the AASHTO Green Book. This is referred to as ‘nominal safety’. The performance of a highway (either existing or expected) as determined by crash frequency and severity, is referred to as ‘substantive’ or quantitative safety. The HSM provides this added dimension to the knowledge base of highway designers. Background: The term “nominal safety” was coined by Dr. Ezra Hauer to describe characterization of a situation in terms of its adherence to design standards and practices. We can think of a road as “nominally safe” if it meets the minimum standard of care and is current with respect to published standards and guidelines. The term substantive safety (or perhaps quantitative safety) is its actual or expected performance in terms of crash frequency and severity. Note that substantive safety is a function not only of the basic characteristics of the road, it is also a function of maintenance, law enforcement and other resources we choose to devote to its operation. Interactivity: The instructor could ask participants whether the two dimensions of safety are the same? Or do they express related yet fundamentally different information about a highway? Notes: N/A *Ezra Hauer, ITE Traffic Safety Toolbox Introduction, 1999 3

Operations, Maintenance & Construction Ranking - Based on organizational policy Compare Safety Impact vs Other Impacts (e.g. environmental) Prioritization Incl. assessment of potential countermeasure Countermeasure Selection, B/C Site diagnosis, countermeasure selection, economic analysis Countermeasure Selection & B/C - Site diagnosis, countermeasure selection, economic analysis Operations, Maintenance & Construction Planning & Programming Network Screening Based on policy focus (e.g. SHSP, systematic approaches, risk-based (proactive) approaches, and reactive approaches; some as a result of STIP, TIP, route development process and corridor planning Pre-design & Scoping 3R vs 4R - (i.e. less restrictive design requirements vs Green Book new construction criteria) HSM Part B, C, and D Design exceptions/ deviations Evaluating Individual Projects Before-after studies Evaluate design alternatives HSM Part C, and D Evaluation & Performance Measurement Compare safety impact vs other impacts (e.g. environmental) Evaluating System Performance Performance Measures for Safety Design & Construction Evaluate design-build proposals - Using value-based evaluation that includes safety Evaluate Alternatives - Evaluate alternatives in operations, maintenance, and construction The Project Development Process at a state DOT, activities, and the relationship with the HSM

ELEMENTS OF Institutionalization Data, Model Development Tools, Analysis, Evaluation Training, Marketing Use it & Policy Champion Implementing the HSM requires each of these components 5

HSM Implementation Plans help manage the process to achieve full institutionalization HSM Implementation Framework Obtain Support Materials Increase Understanding of HSM within MoDOT Establish a HSM Implementation Team Schedule Training for …(FHWA, NHI, Lead States) Part D and CMF Clearinghouse Training Develop Policy/Guidance & Integrate into Processes Provide Technical Support/Monitor HSM Use Expand awareness to outside partners

Data collection can be manageable Roadway data Google Street View, Google Earth, ArcGIS, Field GPS data collection LiDAR Data maintenance Crash data Set up databases and processes so that 3-years of after-project implementation data can be complied easily Traffic Other data sources Need to have enough good quality data for model calibration, safety performance function development, analysis Many techniques for collecting highway inventory data have been used by state and local agencies in the U.S. These techniques include field inventory, photo/video log, integrated GPS/GIS mapping systems, aerial photography, satellite imagery, virtual photo tourism, terrestrial laser scanners, mobile mapping systems (i.e., vehicle-based LiDAR, and airborne LiDAR). These highway inventory data collection methods vary in terms of equipment used, time requirements, and costs. Each of these techniques has its specific advantages, disadvantages, and limitations. This research project sought to determine cost-effective methods to collect highway inventory data not currently stored in IDOT databases for implementing the recently published Highway Safety Manual (HSM). The highway inventory data collected using the identified methods can also be used for other functions within the Bureau of Safety Engineering, other IDOT offices, or local agencies. A thorough literature review was conducted to summarize the available techniques, costs, benefits, logistics, and other issues associated with all relevant methods of collecting, analyzing, storing, retrieving, and viewing the relevant data. In addition, a web-based survey of 49 U.S. states and 7 Canadian provinces has been conducted to evaluate the strengths and weaknesses of various highway inventory data collection methods from different state departments of transportation. To better understand the importance of the data to be collected, sensitivity analyses of input variables for the HSM models of different roadway types were performed. The field experiments and data collection were conducted at four types of roadway segments (rural two-lane highway, rural multi-lane highway, urban and suburban arterial, and freeway). A comprehensive evaluation matrix was developed to compare various data collection techniques based on different criteria. Recommendations were developed for selecting data collection techniques for data requirements and roadway conditions.

Safety Performance Functions need to be calibrated or developed for local conditions and specific analysis Below are the tables with goodness-of-fit evaluation results for calibrated state systems SPFs (see the attached technical memo for more details). As can be seen, for majority of calibrated SPFs (24 out of 36 SPFs for segments, and 21 out of 24 SPFs for intersection), the cumulative residual curve cannot be within 95% confident limit for all AADT/segment length ranges.   The reason is that the calibration process can only change the calibration factor in the crash predictive model but none of the SPF coefficients. The calibration process can make the total observed crashes the same as total predicted crashes, but it cannot guarantee the calibrated SPF curve can fit the crash data well. The function form of the calibrated SPF is still the same as that of uncalibrated SPF, which were determined based on old crash data. The SPF redevelopment process, however, calculate the SPF coefficients with the new crash data; therefore, the redeveloped SPF can fit the new crash data better. The advantage for SPF calibration lies in that it is more straightforward and can be understood by most engineers – no knowledge on statistics (negative binomial distribution) is required. A majority of calibrated SPFs (24 out of 36 SPFs for segments, and 21 out of 24 SPFs for intersection), the cumulative residual curve cannot be within 95% confident limit for all AADT/segment length ranges. .

Embed safety performance evaluation into processes Set up databases and processes so that 3-years of after-project implementation data can be complied easily Every research project should, where possible, have a crash modification factor development component Integrate project and program evaluation into transportation system management Below are the tables with goodness-of-fit evaluation results for calibrated state systems SPFs (see the attached technical memo for more details). As can be seen, for majority of calibrated SPFs (24 out of 36 SPFs for segments, and 21 out of 24 SPFs for intersection), the cumulative residual curve cannot be within 95% confident limit for all AADT/segment length ranges.   The reason is that the calibration process can only change the calibration factor in the crash predictive model but none of the SPF coefficients. The calibration process can make the total observed crashes the same as total predicted crashes, but it cannot guarantee the calibrated SPF curve can fit the crash data well. The function form of the calibrated SPF is still the same as that of uncalibrated SPF, which were determined based on old crash data. The SPF redevelopment process, however, calculate the SPF coefficients with the new crash data; therefore, the redeveloped SPF can fit the new crash data better. The advantage for SPF calibration lies in that it is more straightforward and can be understood by most engineers – no knowledge on statistics (negative binomial distribution) is required.

Many agencies are developing customized HSM tools SafetyAnalyst, IHSDM Excel based user friendly tools for crash prediction GIS online, web interface Customize State calibration factors, SPFs Crash Modification Factors, service life and project costs Many techniques for collecting highway inventory data have been used by state and local agencies in the U.S. These techniques include field inventory, photo/video log, integrated GPS/GIS mapping systems, aerial photography, satellite imagery, virtual photo tourism, terrestrial laser scanners, mobile mapping systems (i.e., vehicle-based LiDAR, and airborne LiDAR). These highway inventory data collection methods vary in terms of equipment used, time requirements, and costs. Each of these techniques has its specific advantages, disadvantages, and limitations. This research project sought to determine cost-effective methods to collect highway inventory data not currently stored in IDOT databases for implementing the recently published Highway Safety Manual (HSM). The highway inventory data collected using the identified methods can also be used for other functions within the Bureau of Safety Engineering, other IDOT offices, or local agencies. A thorough literature review was conducted to summarize the available techniques, costs, benefits, logistics, and other issues associated with all relevant methods of collecting, analyzing, storing, retrieving, and viewing the relevant data. In addition, a web-based survey of 49 U.S. states and 7 Canadian provinces has been conducted to evaluate the strengths and weaknesses of various highway inventory data collection methods from different state departments of transportation. To better understand the importance of the data to be collected, sensitivity analyses of input variables for the HSM models of different roadway types were performed. The field experiments and data collection were conducted at four types of roadway segments (rural two-lane highway, rural multi-lane highway, urban and suburban arterial, and freeway). A comprehensive evaluation matrix was developed to compare various data collection techniques based on different criteria. Recommendations were developed for selecting data collection techniques for data requirements and roadway conditions.

HSM Training should be tailored to the Target Audience(s) Elected Officials Local agencies, DOT engineers, Local Agency engineers Staff Local Agency Technical Staff

Adding Principles and Approaches Encourages the Use of HSM Tools State Policies and Procedures on Use of the Highway Safety Manual Design exceptions Access Justification Reports Environmental Impact Studies HSM expertise as a prequalification for consultant selection

Evaluate pavement, bridge condition with safety performance for programming projects Condition Rating System (CRS) – Structural: Loss of load carrying capacity or structural breakdown International Roughness Index (IRI) – Functional/Surface: Excessive roughness impacting functional usability and causing drive discomfort Safer Roads Index (SRI) – Safety Performance (PSI): Establishes safety risk based on historical severe crashes and exposure State of Repair CRS Range 9.0 to 7.6 Excellent 7.5 to 6.1 Good 6.0 to 4.6 Fair 4.5 to 1.0 Poor IRI Range (in/mi) 1 to 94 95 to 177 > 177 SRI Range Minimal Low Minor Medium Moderate High Severe Critical/5% 5%

Transportation System Performance Measures Integrate performance measures into decision making tools for state and local roadways Transportation System Performance Measures Legend

Project Examples

Omaha Major Investment Study used HSM to evaluate alternatives for the 2050 Long Range Plan Comprehensive study that evaluates freeways, arterials, local roads, and transit system needs Purpose: Develop comprehensive, multimodal plan for the interstate and major roads in the region Prioritize projects for short-term, mid-term, and long-term Consider shortfalls in existing sources of local, state, and federal funding Facilities analyzed: 80 miles of freeway 35 miles of interchanges 250 miles of arterial roads 260 ramp terminals and signalized intersections

MTIS study approach provides the opportunity for a performance-based evaluation Conducted in three phases PHASE 1: Existing safety assessment and network screening PHASE 2: Selection of strategies, future no build and alternative evaluation PHASE 3: Alternative design and development of implementation plan Technical analyses will be used for the MAPA 2050 Long Range Transportation Plan

Recommended locations of safety needs to be analyzed in the next phase 16 miles of freeway, (20%) 4 miles of interchanges, (9%) 35 miles of arterial roads, (14%) 33 ramp terminals and intersections (13)%

Regional strategy packages covered a wide variety of improvements

LA 3235 used HSM to evaluate access management strategies Many access management treatments are not found within the HSM’s Part C predictive methods. Proposed a corridor-based approach in which Part D CMFs were applied at the corridor level after using Part C to predict future no-build crashes. Corridor Features 16 mile length 4 lane divided highway Posted speed limit 50 to 65 mph Rural in character but urban designation 75 median openings Heavy truck traffic

The analysis provided a quantitative understanding of what types of crashes would be eliminated or reduced through access management.

Local agencies are using CMFs and some crash prediction models to assess safety Oregon DOT (ODOT) developed a new All Roads Transportation Safety (ARTS) Cost Effective Index Proposed countermeasures: - Pedestrian countdown signals - Intersection illumination - Bike boxes at conflict points - Buffered bike lane

What is next for the HSM?

AASHTO Highway Safety Manual 2nd Edition is under development Plan for 2019 New Content Planned Part B Roadway Safety Management Systemic Safety Analysis Human Factors enhanced Section on bicycle and pedestrian safety

AASHTO Highway Safety Manual 2nd Edition Part C Predictive Methods Safety Prediction Models for Six‐Lane and One‐Way Urban/Suburban Arterials Improved Prediction Models for Crash Types and Severities Intersection Crash Prediction Methods for the HSM Roundabout Crash Prediction Models and Methods for the HSM

AASHTO Highway Safety Manual 2nd Edition Part D Crash Modification Factors Development of Crash Reduction Factors for Uncontrolled Pedestrian Crossing Treatments Safety Impacts of Intersection Sight Distance Guidance for Development and Application of Crash Modification Factors Update of Crash Modification Factors for the HSM Part D will provide guidance but will the CMFs will be on the CMF Clearinghouse

Thank You