CHAPTER 4

Case Examples

Follow-up interviews with select state DOTs were conducted to gather further details on individual state DOT practices on the development of calibration factors and jurisdiction-specific SPFs. The DOT survey respondents were asked to participate in the case example and were encouraged to invite individuals with direct experience with practices on calibration factors and jurisdiction-specific SPFs to participate in or contribute to developing the case examples. The semi-structured interviews followed the questions outlined in Appendix A but often expanded to a discussion of the unique experiences of each state DOT. The interviewees were advised that the intent of the case example questions was to prompt thoughts for discussion, and they were encouraged to provide a narrative and details as they deemed appropriate.

Caltrans and Florida, Nebraska, North Carolina, and Washington State DOTs were selected for case examples based on their survey responses. This chapter provides a summary of these case examples.

California

Caltrans was contacted for its willingness to participate as a case example due to its responses to the survey indicating that it applied SPFs without calibration and did not develop its own state-specific SPFs. (Contact: https://dot.ca.gov/programs/safety-programs.) The reason for that choice is that California crash data were used in the development of the original HSM SPFs; thus, DOT staff felt that the SPFs already reflected conditions in California. However, Caltrans recently completed a scoping project to determine the level of effort and benefits that would be obtained from calibrating the HSM SPFs or developing state-specific SPFs. Based on the results of this project, Caltrans will likely seek to calibrate existing SPFs in lieu of developing state-specific SPFs. This choice is being made due to the use of California data in the development of the HSM SPFs and the feeling that developing state-specific SPFs would require significant resources but not offer sufficient additional precision in safety estimates to warrant the cost. Regionalization will be considered in the calibration process to account for the diverse conditions experienced in California; this will likely be done at the county level, where individual counties will be aggregated together to form a region based on their area type (metropolitan, mountainous, coastal, inland) and availability of sufficient crash data/sites. This future calibration effort will not be done in-house; Caltrans will seek either a university partner or consultant to assist. The timing of this future calibration effort is not certain; Caltrans staff indicated that the upcoming publication of the second edition of the Highway Safety Manual (HSM2) may lead them to delay calibration until newer SPFs are available. Caltrans staff are still discussing whether they will calibrate the existing HSM1 SPFs, calibrate available HSM2 SPFs before HSM2 publication, or wait until the publication of HSM2 before embarking upon calibration.

Caltrans is primarily using design-level SPFs for the following purposes: alternatives analysis, design exceptions, interstate access request justifications, and intersection control evaluations (ICE). However, Caltrans’ Division of Safety Programs is currently working with a local university partner to develop SPFs to support network screening efforts. Caltrans staff are not currently aware of the development or use of SPFs by local agencies within California. The prevailing thought is that this is due to a lack of awareness of HSM methods and/or acceptance of HSM methods by these local agencies. The future calibration effort is also being used as an opportunity to promote these methods across the state and help local agencies better understand how these quantitative tools can be used to improve decision-making. This may spur the development of SPFs by some local agencies.

Florida

Florida DOT (FDOT) was contacted for its willingness to participate as a case example due to its responses to the survey that indicated that FDOT utilized calibrated versions of the HSM SPFs and developed its own state-specific SPFs. (Contact: https://www.fdot.gov/Safety/default.shtm.) Calibration factors were originally developed for design-level SPFs in the HSM when originally published. The results were integrated into the 2024 FDOT Design Manual (FDOT, 2024), which is used throughout the state to better understand the impacts of various design decisions. These calibration factors have not been updated since the original effort, but the calibrated SPFs are still included in the design manual and used by FDOT engineers and consultants today. The calibration factors included regionalization in the form of unique calibration factors for individual FDOT districts for which sufficient data were available.

After the publication of the HSM, a committee within FDOT considered two options for the integration of HSM network screening methods in Florida: calibration of HSM SPFs and development of state-specific SPFs. The committee decided to develop state-specific network-screening-level SPFs since this provided predictions on safety performance that better reflected Florida conditions. Segment SPFs were developed a few years ago using crash data from 2014 to 2018 and have not been updated since. Intersection SPFs are developed/updated annually as a part of the network screening process. The intersection SPFs are updated more frequently due to convenience and ease. Intersections are fixed units, whereas FDOT uses a sliding window for roadway segments, making updates to segment SPFs more cumbersome and labor/data intensive. However, FDOT plans to develop and update calibration factors annually for these state-specific segment design-level SPFs to reflect more recent safety performance within Florida. Regionalization is not being considered for these as network screening happens at the state level and this information is then passed on to individual FDOT districts for diagnosis and countermeasure selection.

FDOT has also developed SPFs for a specific facility type (restricted U-turn crossing intersections, RCUTs) using data from other states. This was done to better understand the safety performance of this facility type before application in Florida. Additionally, FDOT developed additional design-level SPFs for suburban residential and suburban commercial intersections. These context classes are more in line with what is typically used by FDOT than the traditional rural/urban–suburban subdivision in the HSM. All SPFs are integrated into an Excel-based FDOT SPF tool. This tool also provides CURE plots and other goodness-of-fit measures to ensure that the SPFs provide a good fit to the observed data. Poor-fitting SPFs are typically due to small sample sizes and are not used. Local municipalities typically coordinate with individual FDOT districts for assessment of safety performance and the districts rely on state-level SPFs.

Nebraska

The Nebraska DOT was contacted for its willingness to participate as a case example due to its responses to the survey that indicated that it did not currently apply SPFs in the HSM or develop their own SPFs. (Contact: https://dot.nebraska.gov/safety/hso/.) However, the case example discussion disclosed that Nebraska DOT does apply part of the HSM in its safety management process. For example, the DOT follows the network screening process proposed in the HSM using the Rate Quality Control Method performance measure. In this process, crash rates are estimated for each site and compared against the critical crash rate, which is equal to the average crash rate for all sites of a given type. Sites that have larger rates than this critical value are then identified and prioritized based on other metrics: ratio of crash rate to the critical value and total societal cost. The former is used to identify larger outliers, while the latter accounts for higher crash frequency and severity.

Nebraska DOT is currently working to calibrate HSM SPFs to integrate into the network screening activities. When completed, the calibrated SPFs will be used to obtain excess expected crash frequencies with EB adjustments, which will be used as one of the performance measures for network screening. This calibration process is being done by Nebraska DOT personnel in-house after the DOT reached out to local university partners and found that these partners did not have the requisite experience to assist in SPF calibration. The calibration process will make use of the AASHTOWare Safety software, which will pull the necessary inputs (e.g., traffic volumes and crash frequencies) of all sites on the roadway network of a given type from data feeds and will provide the estimated calibration factors. The AASHTOWare Safety software also facilitates the implementation of these calibrated SPFs for network screening that will be performed by DOT personnel, making it an attractive option for calibration. While this process has not yet been completed, the DOT expects that the agency will update calibration factors once every year in line with updates to the roadway management system database. This update factor is subject to change after the initial calibration process is complete, based on the level of effort required and DOT personnel availability. However, the DOT indicated that the annual updates are temporary: this process will be revisited after examining how much calibration factors change from year to year.

Further, Nebraska DOT indicated that design-level SPFs for roadway segments and intersections of two-lane, two-way rural roads were currently being used by members of the roadway design team. These SPFs are used to better understand the safety impacts of different design decisions, identify potential alternatives, and justify decision-making when design exceptions are being sought. Currently, these SPFs are uncalibrated versions of the HSM SPFs; however, once calibration factors are developed for the network screening exercise, the DOT plans to apply these to the design-level SPFs as well. The temporary use of uncalibrated SPFs was done with the knowledge that calibration efforts are planned for the near future and that the national standard of the HSM helps justify their use within the state.

North Carolina

The North Carolina DOT (NCDOT) was contacted for its willingness to participate as a case example due to its responses to the survey that indicated that the DOT currently applies calibrated versions of the HSM SPFs and did not develop jurisdiction-specific SPFs, even though the review of the literature identified a project sponsored by the NCDOT for the development of state-specific SPFs for network screening purposes. (Contact: https://apps.ncdot.gov/dot /directory/authenticated/UnitPage.aspx?id=1338.) NCDOT personnel confirmed that jurisdiction-specific SPFs were developed in Srinivasan and Carter (2011) as a first attempt at implementing HSM methods in North Carolina. However, this project only developed network-screening-level

SPFs due to a lack of detailed data availability and confidence in data accuracy to support design-level SPFs. Currently, NCDOT applies SPFs exclusively for design-decision purposes; network screening is performed using a warrant-type method that was fine-tuned over several years. NCDOT currently has a project that will compare the outcomes of their existing network screening process with other methods, including SPF-based methods in the HSM. As a part of this project, NCDOT hopes to determine if state-specific network screening SPFs are needed and if they provide a quantifiable improvement over the existing screening methods. Further, a project is underway to standardize data for individual roadway elements; this project would be able to facilitate the development of state-specific SPFs in the future, if desired. However, NCDOT members revealed that they would need to see proof that these state-specific SPFs are clearly better for their needs.

Due to (a) their comfort with the existing network screening approach, (b) the lack of accurate data to support the development of state-specific SPFs, (c) the lack of clear evidence that state-specific SPFs would provide a significant benefit, and (d) the focus of the NCDOT program on fatal + serious injury crashes only, the DOT staff have primarily considered calibration of the HSM (and other national-level) design-level SPFs. For these reasons, the original network-screening-level SPFs that were developed in 2010 have not been updated or calibrated further and they are not currently applied. To date, five projects have been performed to develop (and update) calibration factors for the application of HSM design-level SPFs in North Carolina. The most recent effort was completed in 2021, and NCDOT is planning for another update to begin in 2024. These efforts have developed calibration factors for:

• Two-lane rural roadway segments and intersections
• Multi-lane rural roadway segments and intersections
• Urban–suburban arterial roadway segments
• Urban and rural freeway sections

The newest project will include additional design-level SPFs for freeway ramps, ramp terminals, and roundabouts, among others. The calibration factors are estimated according to the definition provided in the HSM and are typically done as a part of a research project by a university partner.

The primary use of the calibrated design-level SPFs is to compare the safety outcomes of different design decisions on a project-by-project basis to make more informed and holistic decisions. However, NCDOT personnel indicated that they understand SPF outputs are not highly accurate but instead provide a “ballpark” estimate of safety performance. Thus, they apply these calibrated SPFs to obtain a better understanding of the relative safety performance of different design decisions. This is one reason that the accuracy provided by the calibrated HSM SPFs is sufficient for their purposes and that state-specific SPFs have not received more attention in North Carolina. The calibration factors are updated approximately every four years. NCDOT personnel indicated that this frequency was selected because it followed general guidance to update approximately every three years and it fit well with research funding and contracting. However, the DOT is actively interested in guidance on how often these SPFs should be calibrated. Of particular interest is how much the calibration factors change from year to year; NCDOT personnel felt that a new update with little change in annual calibration factors would likely lead to a reduced frequency of calibration efforts.

NCDOT staff indicated that southeastern states have noted calibration factors that suggest significant differences between the HSM SPF predictions and observed safety performance. The use of jurisdiction-specific SPFs from nearby states has not been considered, but the agency would be open to this if evidence suggested these provide more accurate predictions. Lastly, NCDOT has in-house staff to estimate CMFs for a variety of roadway countermeasures or features. They also maintain a list of select “approved” CMFs from other sources (e.g., the CMF

Clearinghouse) that were vetted by the staff. These are all integrated into the design decision process and HSM-calibrated SPFs.

Washington State

Washington State DOT (WSDOT) was contacted for its willingness to participate as a case example due to its responses to the survey that indicated that WSDOT applied SPFs without calibration and also developed its own state-specific SPFs. (Contact: https://wsdot.wa.gov/engineering-standards/planning-guidance/planning-study-guidance/safety-guidance-planning-studies.) The use of uncalibrated SPFs was a conscious choice of WSDOT leadership since Washington State data were used in the development of many HSM SPFs. The prevailing thought was that Washington State conditions are already well-reflected in the HSM SPFs. Both network-screening-level and design-level SPFs are used in Washington State. Policies and requirements regarding the use of design-level SPFs in the HSM at WSDOT are provided in the WSDOT Safety Analysis Guide (WSDOT, 2020). These design-level SPFs are being used by agency engineers and consultants to support the analysis of alternatives and design decisions for individual facilities. Another pertinent example is the use of two-lane, two-way rural roadway segment SPFs from the HSM to quantify the societal crash benefit of rumble strip installation at individual sites across all two-lane rural state highways.

State-specific SPFs were estimated for various roadway segment types, including two-lane, two-way rural roads and urban–suburban arterials. These state-specific SPFs allowed WSDOT to (a) explore different modeling approaches from those used to develop the HSM SPFs (specifically, the integration of random parameters), (b) integrate data elements available in Washington that were not available for use in the HSM SPFs (e.g., geometric and roadside data), and (c) answer specific safety-related questions within Washington State. One example of the latter was the development of SPFs to assess the safety impacts of continuous lighting on limited-access freeways. This study showed that continuous lighting in this specific context did not significantly affect fatal and serious injury crashes, and thus it supported the removal of lighting in some areas to reduce roadside objects that can be struck, reduce energy consumption, and reduce exposure of maintenance personnel. During the development of state-specific SPFs, adjustment factors are considered to reflect safety differences in the eastern versus western parts of the state. The use of state-specific tools from other neighboring states has not been considered due to differences in reporting thresholds, crash reporting practices, and safety management across agencies. State-specific SPFs for intersections have not been developed because WSDOT does not currently have an intersection database: work has been initiated to create such a database given the upcoming Model Inventory of Roadway Elements (MIRE) requirements for 2026.

The fit of uncalibrated HSM SPFs to Washington data was assessed several years ago via CURE plots obtained from the Calibrator Tool, and the results suggested a good fit to local conditions. Updated goodness of fit using more recent data has not been performed. One reason noted was that WSDOT has been awaiting the publication of HSM2; any calibration or assessment of the goodness of fit will be obsolete with the new and updated HSM2 methods. WSDOT personnel also noted that a balance needs to be achieved between research and policy: while frequent updates/assessments/calibration might be more scientifically appropriate and provide more accurate predictions, there is a downside to changing policy too often using newer models. WSDOT personnel also noted that a lack of funding to purchase software to implement the upcoming HSM2 and related tools may present a barrier or challenge to implementation.