Feasibility Study for Highway Hazardous Materials Bulk Package Accident Performance Data Collection (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1 Feasibility Study for Highway Hazardous Materials Bulk Package Accident Performance Data Collection Bulk packages are a common means of transporting hazardous materials. Accurately estimating the safety performance of these packages in an accident is critical in the evalua- tion of transportation risks. Such an assessment requires reliable data on accidents, design characteristics of bulk packages, and failure of a package or its components. Accurate data would help carriers, shippers, and regulators make better decisions about package selection and design and would help guide overall risk analysis related to transportation of hazardous materials. HMCRP Project 07 was developed to address this research need. The objectives of HMCRP Project 07 were the following: • Review data currently being collected about accidents involving U.S. DOT-specified haz- ardous material bulk packages (i.e., portable tanks, cargo tank motor vehicles). • Determine what data are needed to develop a satisfactory bulk package accident perfor- mance database. • Develop methodologies for collection and analysis of accident performance data. • Identify and evaluate institutional barriers to development of such a database and make recommendations for overcoming them. Information Needed to Assess Bulk Package Accident Performance Information about accident performance and the conditional probability of release of haz- ardous materials when a bulk package is involved was identified by reviewing cargo tank clas- sifications and specifications, industry practices, existing data collection methodologies, risk analysis strategies, container performance studies, and cargo tank accident investigations. In addition, potential database users were asked to identify components and information that should be collected. The information was ranked by importance in predicting bulk package accident perfor- mance. The HMCRP Project 07 panel then specified a level of detail for a pilot study of a data collection program. The resulting data fields are categorized as follows: • Bulk Package Design Information. This information enables prediction of how well a bulk package will withstand damage, includes design information from bulk package specification plates, and indicates whether the bulk package was jacketed. • Basic Commodity Information. This information enables a risk-based decision-making approach to determine whether additional damage-protection measures are appropriate for a particular commodity; it includes commodity identification information typically found on shipping papers, and information on the packaged amount by compartment. S U M M A R Y

2• Bulk Container Damage Information. This information describes the severity of the accident and enables identification of components or locations that are prone to damage and/or release of hazardous materials; bulk container damage information also can be used as a proxy for impact energy and forces involved in the accident. This informa- tion includes identification of damaged components; damage location, type, and size; indication of whether the damage sustained by the component resulted in a release; and, if applicable, the amount released and breach size. • Basic Accident Information. This information accounts for variability in predicting the conditional probability of release and amount released from impact forces in different kinds of accidents and includes identification of a rollover and/or collision with other motor vehicles, the speeds of the vehicles involved, and identification of the physical object(s) that damaged the bulk package. Data Collection Methodologies Two options for recording cargo tank performance in accidents were considered: either modifying or emulating existing databases and data collection processes. To increase the quality of reported data and the ease of reporting, technical implementation and security methods were explored in further detail. The following databases were considered for modification or emulation: • FMCSA’s Motor Carrier Management Information System (MCMIS) Crash File. • PHMSA’s Hazardous Materials Incident Reporting System (HMIRS). • NHTSA’s Trucks Involved in Fatal Accidents (TIFA). • NHTSA’s National Automotive Sampling System (NASS) General Estimates System (GES). • Railway Supply Institute (RSI)-Association of American Railroads (AAR) Tank Car Accident Database (TCAD). The programs were examined and the following dichotomies were identified: • Participation in the data collection program could be either voluntary or mandatory. • The program could be sponsored by either a consortium of industry organizations or by a government organization. • The database could either be a standalone new data collection program or an extension of an existing one. These options were presented in a survey of stakeholders. Although opinions were fairly evenly split regarding mandatory or voluntary participation, respondents were in favor of an industry-sponsored extension of an existing program. PHMSA’s HMIRS was determined to be the most appropriate existing program because it collects approximately 70% of the desired information. HMIRS could be extended by modifying the existing Form DOT F 5800.1 to collect additional data or by importing the necessary information from Form DOT F 5800.1 and requesting additional information separately. The following four options were considered for further analysis: • A program of improved compliance with the existing Form DOT F 5800.1 modified to collect information about component performance, with mandatory participation. • A government-sponsored extension of Form DOT F 5800.1 that collects all information required to evaluate component performance, with mandatory participation.

3 • An industry-sponsored extension of Form DOT F 5800.1 that collects all information required to evaluate component performance, with voluntary participation. • A government-sponsored new database (independent of Form DOT F 5800.1) that collects all information required to evaluate component performance, with mandatory participation. From these four options, the project panel recommended that the research team investigate the feasibility of implementing a government-sponsored extension of Form DOT F 5800.1, with mandatory participation. Accordingly, a pilot study was undertaken to explore collec- tion of accident damage data. Results of the pilot study were used to evaluate the expected quality of data, identify improvements to the data collection system, demonstrate the types of analyses that the database would facilitate, and estimate how long it would take to collect incident data sufficient to support reasonable statistical analyses. To facilitate data collection for the pilot study, an online tool was developed. To reduce errors and reporting time and to improve data quality, questions were dynamically adjusted based on logic and previous responses—for example, certain questions were displayed only if a response was logically expected. Drop-down menus, check-off boxes, and radio buttons provided consistency and uniformity in responses, and text-based response options were provided only when necessary. In addition, logical quality checks upon report submission helped ensure that responses were realistic and congruent. (These quality checks must be conducted whether or not responses are adjusted dynamically.) Industry participants were solicited to provide bulk package performance data for the pilot study. Despite extensive efforts to involve the industry, the participation level was unsatisfactory. Therefore, data from NTSB accident reports, PHMSA HMIRS reports, FMCSA MCMIS reports, and news articles were used to populate the pilot database. Through this process, several improvements to the data collection tool were made, including improving the logic for presenting possible responses and increasing the robustness of the system so that it could handle unexpected actions (such as a respondent using the browser’s refresh button). In any broader implementation, data integrity measures will help ensure that collected information is valid. Accordingly, three levels of access are recommended: administrator, reporter, and public. Administrator access allows database owners to grant access to other users, create backups of the datasets, implement data quality checks, correct errors, track the number of views the dataset generates, and link to datasets generated by other organiza- tions. Reporter access is granted to individuals or companies that are required to submit a report. The rights granted under reporter access must be sophisticated enough to allow the reporting individuals or companies to view and update their reports at a later date. Pub- lic access, granted by sharing either raw or processed data, requires careful consideration because it could influence the success of the program. Data Analysis Analytical statistical models for estimating bulk package performance were explored by first examining how these models relate to the risk of transporting hazardous materials. Once metrics of interest had been determined, data analysis methods were examined. The methods chosen emulated others that analyze similar data collected as part of the RSI-AAR TCAD. Because of anticipated variability in conditional probabilities of release, separate regression equations can be developed for each component-accident scenario pair using the methodologies presented in this report.

4Although the pilot study did not have sample sizes sufficient to evaluate performance based on bulk package design, the information offered a preliminary understanding of com- ponent performance by location on the tank, under various accident scenarios. For example, a single incident could damage components in different areas of a tank; therefore, a com- ponent in one location might be less likely than one located in another area of the tank to result in a release of hazardous materials. Preliminary analysis of pilot study data showed that damage is most likely to occur to the top front passenger-side tank shell and to the piping and/or undercarriage below the tank. In general, damage to the top front passenger- side results from rollover accidents, while damage to the piping and/or undercarriage below the tank results from accidents involving other vehicles. Crushing damage to the tank shell, including dents, is the most prevalent type of damage; however, it is the least likely to result in a release of hazardous materials. More information is necessary, however, to determine the type of damage that will most likely result in a release. Pilot test data also showed that larger releases occur in accidents that involve crossing the median or centerline, running off the road, overturning, and catching fire, and that the greatest number of releases occurred when the bulk package struck both the roadway and the adjacent ground. Finally, the time required to collect incident data sufficient to support reasonable sta- tistical analyses was determined. Based on data from PHMSA HMIRS reports, FMCSA MCMIS reports, and news articles over a 7-month period, it was determined that approx- imately 132 accidents can be expected per month, with about 34 resulting in release of hazardous materials. Based on this expected accident rate, the minimum sample size to acquire statistically significant results was estimated using pilot study data. Because pilot study data were insufficient to estimate variances for all types of bulk pack- ages, sample size estimates were generated only for records corresponding to hazardous materials transported in MC 306 or DOT 406 containers. The analysis indicated that there are three tiers of variables. Tier I variables, requiring sample sizes of less than 800 accident records, consist of the following: • Design variables, including total capacity of the bulk package and head or shell thickness. • The component that was damaged. • Damage locations of the front head above the centerline and the passenger-side bottom middle. • Damage type, including crushing, bending, and gouging or cutting. • Whether the units separated in the crash (if the tractor-trailer was also towing a pup trailer). • Accident characteristics, including whether a personal vehicle or heavy vehicle was involved, whether the bulk package crossed the centerline or median, ran off the road, rolled over, or struck the ground or a concrete barrier. • Speeds of the vehicle(s) involved in the accident. Between 858 and 1,286 accident records are required for the Tier II variables: • Packaged amount. • Damage locations of the driver-side bottom front and top rear and passenger-side top middle. • Whether the bulk package struck the roadway. Variables requiring sample sizes above 5,000 are classified as Tier III: • Bulk package design pressure. • Burst or ruptured damage.

5 • Damage locations of the rear head below the centerline; the bottom middle, top middle, and/or top front driver-side; and/or the bottom front, top front, and top rear passenger-side. • Whether the bulk package struck the guardrail. Compared to the rate of data acquisition for the corresponding variables and assuming an accident-reporting rate of 20%, these minimum sample sizes indicate that meaningful statistical analyses of Tier I variables can start within 2 years of program implementation. By the fourth year of data collection, the significance of Tier II variables could be tested. The research team assumes that other types of bulk packages will require similar sample sizes, and that they can be expected to take longer to acquire in proportion to their usage rates. Institutional Barriers Several industry concerns and institutional barriers must be addressed in order to have a successful data collection program. These barriers were identified through stakeholder surveys and discussion with industry representatives and are classified into three categories: implementation, participation, and information gathering. Barriers associated with implementation include regulatory hurdles and cost. A rea- sonable basis for evaluating the costs, risks, and benefits of the proposed database can be extrapolated from the RSI-AAR TCAD. Rail industry associations that sponsor that database report savings of at least 11 times the cost of the implementation, bringing positive returns on investment in terms of improved safety and business operations. The credibility of the TCAD program has grown such that the rail industry is consistently successful in providing fact-based responses to regulatory proposals. Thus, institutional barriers can be diminished by developing the reporting form in cooperation with carriers already committed to risk- based approaches to safety. Industry associations and member companies are reluctant to participate in a package damage database for a number of reasons, including the belief that accident data would be inappropriately used against a carrier in legal proceedings. Three options to address that concern are proposed: • Implementation of a no-fault provision. • Adoption of processes for storing and sharing data that prevent disclosure of trade or security-sensitive data (includes strict authorization procedures for access to database information). • Adoption of a program that stores anonymous accident data separately from carrier/ reporter information but requires carriers to provide proof of reporting during com- pliance checks. An additional disincentive to participation is lack of compensation for time spent filling out a report. Therefore, the burden of reporting must be placed on the company rather than the person in possession of the hazardous material at the time of the accident. An alternative is to provide a monetary incentive for accident reporting. Another concern is that the data would be used by shippers as a reason to select other modes of transportation and/or other types of bulk packages. To overcome these and other participation barriers, enforcement could be increased and/or safety ratings could be linked to accident-reporting performance. Information-gathering challenges include the inability to access the accident scene to col- lect information because of safety concerns and the desire to clear the roadway as quickly as possible. To assist with data collection, crash-clearing procedures could be modified to

6include taking photos before removing the bulk package. In addition, if damage informa- tion must initially be estimated due to a lack of access to equipment following the accident, carriers should be allowed, and even encouraged, to revise that information when they are able to access the accident scene. Conclusion This project offers a methodical approach to developing and implementing a database system to collect information about damage to U.S. DOT–specified hazardous material bulk packages involved in accidents, regardless of whether the damage resulted in a leak of con- tents, as well as the characteristics of an accident. Institutional barriers to data collection were identified, and strategies for overcoming those obstacles were proposed. A successful data collection system requires cooperation from multiple stakeholders in the industry. If implemented, the system could provide a comprehensive source of information to help the industry to proactively assess potential improvements in package design. The system would also inform and improve package design and policy decisions by providing quantitative data on safety and risks.