The losses in these accidents provide a part of the basis for the risk estimates for this transportation. The events set analyses supported this risk assessment process, as described below, and provided a basis for determining the new estimated risks if changes are introduced to the present regulator scheme.

3.1.1. Risks in transportation of explosives

The accident cases, when analyzed using the events set approach, show how many events sets involved in Class A explosives transportation accidents are common elements of other accident processes. They also indicate which events sets are not common to other accidents, and where hazmat risk control measures could be focused. Lastly, they also suggest where existing regulations did not achieve the desired protection. None of the cases appeared to involve noncompliance with packaging regulations designed to control risks.

An established risk assessment coding system was selected to assess the risks posed by these types of accidents.

3.1.2. The Risk Assessment Code (RAC) system

The analysis method, focusing on mishap processes and outcomes, lends itself to the assignment of risk assessment codes using the method pioneered by and used in the Department of Defense (DOD) to establish priorities for risk assessment, risk control actions and follow-up efforts.

The Risk Assessment Code (RAC) values are shown in Figure 3, below. For a detailed description of the criteria for selecting the mishap probability and severity codes, see the terms Mishap probability rating and Mishap severity rating in the Glossary.

Figure 3-1.

3.1.3. Application of RAC system to Class A Explosives Transportation Risks

Applying this risk estimating system for the three accident types defined above, the resultant RAC values are shown in the following table. The column title "Present RAC " shows the probability and severity ratings which lead to the assigned RAC from the above matrix. Where historical data are available, these probability and frequency ratings can be assigned from experience, if no changes are introduced. Where no historical data are available, as with a new system, the probability and frequency are assigned on the basis of preliminary or extensive risk analysis results.

TABLE 3-1 RAC VALUES FOR ACCIDENT TYPES ANALYZED

ACCIDENT TYPE	Present RAC
1. Vehicle collision followed by fire and en masse detonation of Class A explosive cargo	(IB) 1
2. Single vehicle fire followed by detonation of a cargo including a Class A explosive component	(IB) 1
3. Single vehicle fire in which Class A explosive only burned in the fire.	(IIIB) 3

The RAC 1 values require effective controls to reduce them to -- at worst -- a RAC 3 if accepted at the proper level of an Agency, or, if reasonably feasible, a RAC 4 (which can be accepted at a lower level of an agency) before they can be considered acceptable from a regulatory perspective. Do present regulatory controls accomplish this? The loss history indicates that in the aggregate they do not for accident types 1 and 2.

To achieve a reduction in the RAC level, changes will have to introduced into the Class A explosives transportation system that will reduce the exposures to the mishaps, the frequency of mishaps, or the mishap consequences. To achieve a RAC 3, would require a massive reduction in the foreseeable severity of these mishaps if the frequency of mishaps is to remain the same, or a massive reduction in the frequency of mishaps, if the severity is not changed, or some combination of changes, as suggested by Figure 3-1.

This can be demonstrated by the RAC 3 for Type 3 accidents. Historically, the loss record suggests a marginal loss history for this type of accident. However, the events set analyses disclosed large gaps and uncertainties in understanding of these processes in past highway accidents, so it can not be stated the regulations are responsible for achieving this risk level or whether other events produced the result. Before any consideration is given to changing safety regulations for Type 1 and 2 mishaps, to reflect that historical RAC value, the gaps in the events sets show that the mechanisms which produced the relatively safe fire outcomes in these accidents need to be better understood. Then, historical experience indicating marginal losses and risks needs to be tested against this improved understanding, including some new hypothetical loss estimates, to assure that new requirements attack the proper events sets.

3.1.4. Effectiveness of regulations or guides.

Each individual regulation can be reviewed against the events sets generated in this study, and its specific effectiveness judged by whether it controlled or is likely to control the outcome when the event set occurs in the future. For example, the events sets relating to the use of fire extinguisher capabilities required on vehicles at the time of these accidents demonstrate that the regulation did not produce desired fire extinguishing outcomes. In the absence of data in the cases analyzed, this determination could be made by analyzing either other highway fire accidents or hypothesizing (with modern deductive logic charting analysis methods) possible events sets that produced the outcomes. Additionally, the events set analyses show each area where the accident loss control process or regulations may merit review, and suggest specific areas where new initiatives might be taken or additional understanding developed to achieve more effective control of these risks. These options are discussed in the sections displaying the individual events sets, and are summarized below.

3.2. Conclusions

3.2.1. Adequacy of the regulations

The discussions accompanying each events set were reviewed, and a conclusion drawn whether or not the safety regulation(s) and guidance documents issued by the OHMT, in place to control the risks, should be judged adequate or not, based on the outcomes of the set. The results for highway transportation are reported in Table 3-2, and for rail transportation in Table 3-3, below.

The regulations considered were the numerous regulations applicable to Class A explosives transportation published in the Code of Federal Regulations by the Federal Highway Administration (governing highway carriers, vehicles and drivers), the Federal Railroad Administration (governing railroad carriers, cars and operations) and by the Office of Hazardous Materials Regulations, covering most other requirements. Guidance considered was primarily the Emergency Response Guides prepared by the Department of Transportation.

Each judgment was made using as the main criterion the outcome of the events sets, based on the information reported. If the outcome was as apparently intended, the judgment was that it was adequate, and a "Yes" was entered into the appropriate row and column of the Table 3-2. If the outcome did not abort or disrupt the accident process at that point, it was judged inadequate, or dubious. If information from the accident contradicted the guidance it was judged inadequate. If insufficient information was reported to support any judgment, the judgment was listed as uncertain. Where the regulation or guidance was not applicable to a Type of accident or event set, a judgment was not appropriate and "n/a" was noted for that events set.

Considered in the aggregate, the Table 3-1 suggests the conclusion that overall risk control performance should be judged inadequate because of the loss history for the Types 1 and 2 accidents. The frequency and severity of Type 1 and 2 accidents, as presently controlled, currently merit a RAC 1. Where can changes be introduced to reduce the rating to a RAC 3?

Consideration of the individual events sets disclosed that dis-aggregated judgments produced differences in the adequacy of the specific regulations or guidance, and thus dis-aggregated judgments are more desirable, since the differences can help point to priorities for the development of more effective controls. By looking at the events in the general Type 1, 2 and 3 models, specific areas where regulations might be made more effective are suggested by the entries in the Tables.

TABLE 3-2 Judgment of Needs -Highway Events Sets

Events in model	Any Safety regulations or guidance Applicable?	Type 1,2 Risk Reg. Controls Adequate?	Type 3 Risk Reg. Controls Adequate?	Guidance Adequate?
Collision events	Yes	Uncertain	Uncertain	n/a
Initial fire events	Yes	N o	Uncertain	n/a
Driver actions after accident began.	Yes	N o	Uncertain	n/a
Driver fire fighting actions.	Yes	Uncertain	Uncertain	n/a
Driver rescue efforts	N o	N o	n/a	n/a
Police response	Yes	dubious	n/a	N o
Fire behavior after initiation.	N o	N o	Uncertain	N o
Driver traffic control actions	Yes	N o	probably	probably
Firefighter actions	Yes	Uncertain	Uncertain	N o
Driver-firefighter interactions	N o	N o	Uncertain	N o
Cargo behavior in fire	N o	N o	Uncertain	N o
Post-explosion firefighter actions.	N o	N o	n/a	N o
Post-explosion cargo behavior.	N o	N o	n/a	n/a
Explosion consequences	N o	N o	n/a	n/a
Shipper response actions	N o	Uncertain	Uncertain	probably
Evacuation	Yes	N o	Probably	N o
Accident reporting	Yes	N o	N o	N o

The model for the railroad cases differed from the highway cases in that no collision-type events were involved. Both cases resulted from fires initiated by the running gear on the rail cars. Also, trains differ from single-vehicle highway equipment in terms of load size, load distribution in vehicles, susceptibility to sympathetic detonation of adjacent cargoes and exposures to the mishaps. Additionally, different types of regulations are applicable to rail transportation of Class A explosives. Thus the events in the model listed in Table 3-3 differ slightly from those listed in Table 3-2. Despite some differences, several similarities remain.

It should be noted that the judgments in Table 3-3 are made on the basis of only two cases. Additional cases of fires involving transportation of Class A explosives are believed to have occurred, but records of such occurrences did not appear in the modest search effort conducted for this demonstration study. Further, consideration of guidance material did not address Department of Defense guidance or DOD requirements. Thus, the entries in Table 3-3 should be considered a starting point for further research into these issues, rather than definitive conclusions.

TABLE 3-3 Judgment of Needs - Railroad Events Sets

Events in model	Any safety regulations or guidance applicable?	Type 1,2 Risk controls adequate?	Type 3 risk controls adequate?	OHMT Guidance adequate
Initial fire events	Yes	N o	Uncertain	n/a
Crew actions after accident began.	N o	N o	Uncertain	n/a
Crew fire fighting actions.	N o	Uncertain	Uncertain	uncertain
Police response	Yes	dubious	n/a	uncertain
Fire behavior after initiation.	N o	N o	Uncertain	N o
Crew traffic control actions	N o	N o	n/a	probably
Firefighter actions	Yes	Uncertain	Uncertain	probably
Crew-firefighter interactions	N o	N o	Uncertain	uncertain
Cargo behavior in fire	N o	N o	Uncertain	N o
Post-explosion firefighter actions.	N o	N o	n/a	probably
Post-explosion cargo behavior.	N o	N o	n/a	probably
Explosion consequences	N o	N o	n/a	n/a
Shipper response actions	N o	Uncertain	Uncertain	probably
Evacuation	Yes	N o	Probably	N o
Accident reporting	Yes	N o	N o	N o

* AAR emergency response guidance also considered, because of special role in rail carrier accidents.

3.2.2. What the Events Sets analyses couldn't achieve

The analyses did not produce a complete model of any of the three accident types. The most comprehensive model was prepared for the Type 1 and 2 accidents, primarily because the NTSB had investigated these accidents with teams, and prepared extensive reports of its investigations. However, even the reports of accidents investigated by the NTSB, when subjected to the method used in this study, had gaps which prevented completion of the models.

The work performed in this study can be used, for future accidents, to guide the development of information needed to make the assessments of regulations in Table 3-2 and Table 3-3 more factual and less judgmental. However, until that information becomes available, the judgmental conclusions - based on the events sets and their sequential logic foundations- are more desirable than ignoring the data, or waiting for enough accidents to occur to develop trend analyses or traditional statistical models.

3.2.3. Data constraints

Throughout Chapter 2, "?" are used to indicate where data did not permit the events sets to be completed. Elsewhere, in the text, numerous missing data items are mentioned. This study did not attempt to bridge gaps in the data. Such gaps can be bridged with hypothesized scenarios, developed by use of logic trees or similar deductive or sequential logic display techniques. This would be very beneficial in several areas for the Type 1 and 2 accidents, and could be illuminating for the Type 3 accidents. Additional useful understanding of the Type 1 and 2 accident risks could be developed using such techniques, and they need to be applied to pre-investigate future accidents of these types as promptly as possible.

3.2.4. Reporting demands

One of the key observations during this work was the recognition that the data demanded by the 5800 reporting system does not provide sufficient information to assess the effectiveness of the risk controls established by the regulations. Form 5800 is presently designed to provide a census of incidents, and provide information about package performance. This system is presently not designed to provide information needed for assessment of the full range of regulations, particularly where infrequent events are involved, as is the case with Class A fires and explosions in transportation.

The effects of this system design can be observed readily when the accident processes displayed on the flow charts are compared with the reporting requirements of the 5800 system. The 5800 system does not address or provide guidance for reporting the full range of event types in accidents, and thus provides an incomplete basis for monitoring ongoing operations, isolating priority events set problems, assessing the effectiveness of specific regulatory requirements, predicting the effectiveness of changes in regulations, and verifying that risks have been estimated adequately. Supplementary investigations by NTSB and others satisfy these needs only in part. Unfortunately, the additional information available from the modal agencies' investigations does not cover these aspects of the accident process adequately either. Unless this fundamental reporting system deficiency is remedied, the DEPARTMENT DOES NOT HAVE AN ADEQUATE BASIS FOR EVALUATING THE PERFORMANCE OF ITS HAZARDOUS MATERIALS SAFETY REGULATIONS FOR THE TRANSPORTATION OF CLASS A EXPLOSIVES. Since the 5800 system and modal reporting system is identical for all materials, it appears likely that the system is inadequate for the assessment of most of these safety regulations.

To remedy this situation, collection of "more data" is not a feasible or necessary response. Collection of data by OHMT is limited by law and the law's implementation by OMB. Data collection can be burden unless it is collected and used to meet a particular objective. The analysis presented in this study suggests an alternative approach. By adapting a STEP-type risk modeling and Risk Assessment Coding approach to its program, the OHMT could modify its data collection practices by establishing priorities for acquisition of data, by limiting its reporting requirements to certain types of mishaps, and by providing for reporting of differing levels of detail commensurate with the RAC levels. The models would provide guidance for the information to be collected, and the RACs would provide a basis for selecting the level of detail requested for different kinds of hazardous materials transport activities. Because of the common interests of all parties in acquiring useful descriptive information about potentially serious accidents, and about the effectiveness of risk control measures for more serious risks, it is likely that OHMT could enlist the cooperation of other parties in this effort.

The models developed thus far could reduce the cost of gathering additional data. In future accidents, data gathering efforts could be aimed at verifying events sets documented in past accidents, and at acquiring new data to fill in gaps in RAC 1 and 2 accident risk models. By providing guidance in the form of relevant hazardous materials risk models and RACs to other DOT Administrations and to the NTSB, the efforts of those agencies could be guided to serve these important needs. The approach would impose a data collection burden commensurate with the risks introduced by the hazardous materials movements, which would be difficult to oppose. That would be a logical extension of the principle behind the relaxation in the reporting of incidents involving paint spills introduced several years ago by OHMT. Such action would seem within DOT's authority under the Hazardous Materials Transportation Act and amendments, and could be construed as an essential requirement to implement the Secretary's duty to regulate materials posing unreasonable risks. Since each accident would be unique, and no data collection forms per se would be used, OMB would not be involved, based on our understanding of their role.

Adoption of the multilinear events sequencing methodology as an experimental initiative to help OHMT and the modal administrations better manage hazardous materials transportation risks would have the advantage of showing OHMT's aggressive pursuit of better ways to manage risks, while allowing OHMT flexibility in its initial and ultimate application. Properly announced and implemented, this action has the potential to generate voluntary support from the private sector, and possibly from other Governmental agencies. Since the worksheets are descriptive, they could be shared with cooperating organizations, which would be beneficial to everyone.

The scope of these kinds of accidents, in terms of the number of events and outcomes that might have to be recorded, can become so large that the level of detail could be massive if the entire accident process from the beginning event to the last harmful event were to be fully documented This results from the complexity of this type of accident, rather than the complexity of the investigative methodology. However, simplifying the descriptions of those accidents by raising the level of abstraction poses the risk of overlooking possible risk control options that are more effective that those flowing from generalizations and abstractions such as accident cause or accident factors. By focusing on the process elements in terms of events sets, redundant sets as would be expected in the case of some repetitive injuries or damages, can be dealt with by a single events set. Other efficiencies are available, and can be demonstrated by experienced personnel.

3.3. Recommended actions.

The documentation of past accidents in a multilinear events sequence-based format, and the analyses of the events sets from that documentation, combined with the judgments of need appearing in Tables 3-2 and 3-3, suggest several new and potentially fruitful areas where actions might be initiated to reduce significantly the risks of fire in Class A explosives transportation.

1. To help establish priorities for allocation of its limited resources to reduction of the most significant hazardous materials transportation risks,

a. adopt and begin applying experimentally a risk assessment coding methodology, such as the Risk Assessment Coding system used in this report, as a screening tool for establishing risk control data gathering, analysis and action priorities.

2. To develop the best possible understanding of the highest level risks in Class A explosives transportation from past accidents,

a. adopt as an experimental analysis method the event-based methods presented in this study, and the models developed with this method.

b. perform additional inquiries to acquire or hypothesize data which would bridge gaps in the events descriptions displayed by the models of the accidents in this study, and then analyze the more complete events sets, particularly for Types 1 and 2 highway accidents, to determine the candidate events which reduce the loss outcomes resulting in the RAC 1 risks.

3. To develop specific regulatory or guidance action in the two most promising areas identified by the analysis,

a. Utilize new deterministic analysis methods and the worksheets already developed, to analyze truck driver decision process, task options and workload in Type 1 and 2 accidents, and to define the decision process and tasks which drivers should be programmed to perform during such accidents.

b. Utilize new deterministic analysis methods and knowledgeable subject matter experts, along with the events sets in the models, to develop the most likely description(s) of events that result in a no-explosion outcome when fire is present in motor vehicle accidents.

4. To capture needed new data about Class A explosives transportation risks from incidents or accidents that occur in the future,

a. publicize the adoption of this experimental method for Class A explosives in technical circles as a new initiative by OHMT to achieve reduced hazmat transportation risks.

b. request other DOT Administrations, NTSB, DOD and other governmental agencies and private parties to support this new OHMT initiative, by furnishing new mishap investigation data that OHMT will use with this method to determine new options for reducing Class A transportation fire and explosion risks.

c. take a lead role in the coordination efforts by making available a procedure and sample data needs for others to follow so they send OHMT data in a format compatible with the RAC 1 accident models and events sets, and follow up with others to ensure that all available data needed to describe Class A mishap and incident processes are captured.

---

Return to Section A