View/print Graphics that accompany this document.
A RISK‚ÄëBASED ANALYSIS METHOD
FOR HAZARDOUS MATERIALS
RAIL TRANSPORTATION SAFETY
Internal Working Draft 11/2/72
Prepared by Ludwig Benner Hazardous Materials Specialist
Bureau of Surface Transportation Safety
Henry H. Wakeland, Director
This is a copy of an internal NTSB staff working paper that was being developed to demonstrate the feasibility of implementing the Board's recommendations to adopt a risk-based approach to regulation of hazardous materials transportation safety in its Concepts of Risk special study. It is offered to provide some insight into the technical thinking during early hazmat risk-based safety analysis efforts.
I found this document among old papers while rummaging through material to discard. It was in the form of a yellowed Xerox copy of a typewritten manuscript I had sent the Dir. Wakeland's assistant, for discussion of the next steps in the study. The study was overtaken by events, as I recall, especially by a heavy accident investigation workload about that time. By the time we were ready to resume work on the project, some progress in the private sector was evolving, in the form of risk analysis studies, relieving the need for the Board to demonstrate their feasibility.
While the study was not completed for the Board, much of the thinking that it generated found its way into subsequent safety initiatives especially for the firefighting community. It generated the seeds for the GEBMO and other models I subsequently published for that community, and provided a framework for hazmat investigations and many recommendations that followed.
The original manuscript was digitized as is, and was only edited to include handwritten charts, drawings, tables and notations on the original that confused the digitization process. Thus this is about 99.9% verbatim content of the original. Some of the original figures, and faded tables had to be completely redrawn, but the original content was preserved faithfully. All illustrations are found at the end of the text.
In reading the document, please bear in mind that it was a work in progress, with additrional development planned for how to predict the exposure to the hazmat threat after an accident, as indicted by the the rough pencil sketches of dispersion patterns at the end of the graphics section (best reproduction possible.) They models were later refined and published for the Montgomery College Fire Science curriculum. Eventually the technical content would be subject to a substantial review and editing processes before presentation to the Board.
Oakton VA May 14 2007
TABLE OF CONTENTS
II. Scope of this Study
III. General Approach
A. Activity Relationships
B. Risk Analysis Framework
C. Pathway Modules
D. Verification of Predicted Risks
E. Risk‚ÄëPeak Shaving Rationale
F. Commonality of Hazards
IV. Risk Analysis
A. Historical Risk Levels
1. Overall Risk
2. Risk by Classes of Person
3. Risk by Commodity
B. Predicting Changes in Risks
1. Probability of Occurrence,
2. Probable harm in accident
4. Predictive Analysis for Catastrophe (not shown in original)
C. Demonstration of method
1. Individuals at risk
In recent years accidents involving railroad tank car transportation of hazardous materials have resulted in numerous casualties. Some of these accidents have been of catastrophic proportions. In 1959, 23 persons lost their lives near Meldrin, Georgia. Three persons were injured fatally and 278 others were injured in varying degrees at Laurel, Mississippi, in 1969. In 1970, the business district of Crescent City, Illinois, was destroyed and 66 persons were injured. In 1971, one fatality and 50 injuries occurred at Houston, Texas, in such a railroad accident. In 1972, 247 persons were injured during an incident at St. Louis, Illinois.
The Safety Board is aware of 46 fatalities and over 1, 200 injuries attributable to hazardous materials being transported by railroad during the last 25 years (See Figures 1-A and 1-B.),
The "peaking" characteristics of both these curves illustrate the need for concern about the occurrence of catastrophic accidents in such transportation. Of the 46 fatalities, 37 occurred in three tank car accidents. Thus over 80% of the fatalities occurred in less than 1/3 of the fatal accidents. Of the over 1, 200 injuries, in the 25 year span, 60% occurred in only five accidents involving tank cars, and 40% occurred in the last three years of the period.
The Safety Board is not aware of any changes in railroad tank car equipment, operations, or materials carried which would substantially reduce the likelihood of the occurrence of future catastrophes. The railroad industry has undertaken extensive research and testing programs to determine what safety modifications might be made to tank car equipment, but no substantial retrofitting of existing cars or redesign of new cars has been accomplished. The Federal Railroad Administration has recently adopted track maintenance standards and is developing maintenance standards for railroad rolling stock. The Safety Board has recommended special precautions for firemen at the scene of railroad wrecks. However, a comprehensive predictive safety analysis which attempts to determine the interrelationships and combined safety effect of these changes has not yet been performed.
These circumstances suggest that
1. while rail transportation of hazardous materials has caused relatively few fatalities in recent years, the potential for a catastrophic hazardous material rail transportation accident continues to exist
2. this potential is not diminishing.
No systematic analysis for such catastrophic accident has yet been prepared, because a method for analysis is lacking. Thus the problem remains: How should the transportation of hazardous materials be analyzed for catastrophes on a predictive basis? The purpose of this special study is to suggest a method for such an analysis and illustrate its use. With a suitable analytical method, it should be possible to discover the effects of changes in this activity, particularly those which might inadvertently increase the likelihood of a catastrophe, before such accidents have occurred.
The analysis framework and methods described in this study are based on the National Transportation Safety Board's Special Study 'Risk Concepts in Dangerous Goods Transportation" released in 1971. Within that framework, relating particularly to tank car transportation of hazardous materials, the analysis of past data and future problems is presented
Some of the data needed for illustrating the methodology are incomplete or wholly absent, requiring the use of estimates. Therefore the analysis and the results should be viewed as an indication of the risks rather than absolute values. It is intended that when appropriate data become available, they can be substituted for the estimates used herein to arrive at more representative results.
The number of changes considered has been limited to two risk ‚Äëreduction alternatives, and the change in risk has been examined for only two changes because the study focuses on the method of analysis rather than the undertaking of an analytical program.
The study assumes that catastrophic accidents in hazardous materials transportation are unacceptable, but does not examine the risk acceptance decision process. Various risks are described but the issue of the acceptability of such risks is beyond the scope of this study.
The general approach to the problems described underlines a risk‚Äëoriented concept, described in the Safety Board's prior risk concepts special study. 11/ "Risk'' is the expression used to identify the forecasted injury to persons or property at risk during an activity, in this instance, rail transportation of hazardous materials. Implementation of this general approach is facilitated by adopting the concepts outlined below and described in detail in the following sections. This approach assumes that accident injury is possible during the activity being considered, and that the elimination of all accidents is not possible.
The overall approach considers that the transportation of hazardous materials is an activity which either exposes people to injury while they are engaged in their normal pursuits, or impacts the normal activities of people who might be exposed to accidents involving this activity.
Risk is considered to be a function of the likelihood of an accidental transportation occurrence during the conduct of a hazardous materials/activity ( HMTA) the consequences of such a mishap, and the exposure relationships between the HMTA and the parties at risk.
A " modular'' approach has been adopted, in which the railroad pathway is segmented for analysis. Each module can be described by the characteristics and conditions of the module, and can then be examined to ascertain the "risk" of injury within that module. This permits comparison among the modules analyzed and among commodities moving within a module. effects
A comparison between the predicted effects of certain actions or changes, and the comparison of subsequent results of such changes is a principal consideration under the "risk" approach. This addresses one of the major problem areas with present safety regulations and control efforts.
A rationale for identifying needed safety changes is provided with a risk "peak shaving" approach where the likelihood of an accidental occurrence and/or the magnitude of an injury in such an occurrence are substantial, a "risk peak" exists. These risk peaks can provide the focal point for safety control efforts, and a reasonable basis for establishing safety action priorities.
The approach lends itself to the reduction of risks experienced by individuals, or aggregate risks experienced in a geographic location, such as a module, or are different activities. The latter / particularly significant when catastrophic risks are being addressed.
Finally, analyses of the risk peaks should strive to identify the "risk raising" hazards which may be common to several risk peaks. The criteria for any technical analysis effort should include this consideration.
The risk of accidental injury or death associated with any activity is contingent upon the nature of the activity. The activity of principal interest in this study is the movement of hazardous materials in the rail transportation system. Because of the threat to public safety during the transportation of such materials, this activity might properly be considered a "hazardous activity.
Three types of relationships exist between the "hazardous" activity and "normal" activities of the variety of parties at risk. First, there is exposure to the hazardous activity during the normal activity, as occurs when residents occupying a dwelling along a railroad pathway in the "normal" activity of resting, recreation, or homemaking, are exposed to the ''hazardous" shipments in a passing train. A second relationship consists of the convergence of the normal activity into a hazardous activity, as occurs when firemen respond to a wreck track crews are reassigned to clean up a rail wreck. The third relationship arises when the hazardous activity is concurrently the normal activity as in the case operating of train crews operating a train containing hazardous materials. These relationships are shown in graphic form in Figure 2
The significance of these relationships is evident when the undertaking of the risks is examined. In the first instance, exposure to the risks associated with hazardous materials transportation usually occurs without the awareness of the person engaged in the normal. activity. Such personsexposed to the hazardous transportation materials activity (HMTA) usually are not in a position to exercise effective personal control over their destiny during the period such exposure occurs.
Insert Fig 2 here
In the second instance, when a normal activity converges with a hazardous activity, a limited measure of self‚Äëdetermination in the risk-taking decision arises. For example, in an accident, firemen might withdraw from the wreckage if the threat to their own safety is discernible.
In the third instance, participation in the hazardous activity occurs willingly, as with the train crews, and the participant relies on precautions which enable him to keep his personal risks at a reasonable level.
The objective of regulatory safety controls is to attempt to assure that the risk levels associated with the transportation of hazardous materials, both on an individual and an aggregate basis, is within acceptable limits. This approach contemplates that the risk levels are discernible for both the individual and aggregate populations exposed to, converging with or participating in the HMTA. These distinctions are essential for identifying the perspective from which accident risks must be considered, and for the risk level comparisons ‚Äë on either a relative (incremental) or absolute basis ‚Äë which are weighed in the risk acceptance decision process. Evidence exists 18/ that the level of risk which is acceptable to a willing participant and direct beneficiary in a hazardous activity (Case 3) is likely to be much higher than the level of risk acceptable to the bystander with no perceived benefits from its movement (Case 1). Thus, individual and aggregate risks must be considered.
An example of a framework for risk analysis is presented in the Safety Board's risk concepts special study. That framework provides the basis for the analysis methods in this study. To that framework has been added an additional "exposure" element of risk. A condensed modified framework is described in Figure 3.
Insert Fig 3 here
Within this framework, risk is the forecasted accidental injury per unit of exposure to the hazardous materials transportation activity (HMTA ). Risk will be expressed as forecasted fatalities per unit of personal exposure time, although other measures such as lost time, dollars, reduced life expectancy, etc., might also be used. At this time, fatalities are used because the data appear to be most reliable, and avoid the complicating difficulty of dealing with the degree of injury..
Risk is a function of the likelihood of an undesired injury producing incident during the HMTA, of the likely consequences of attributable to hazardous materials in such occurrences, and of the "exposure" to the HMTA.
The method for calculating risks varies according to the viewpoint being considered, which in turn depends upon the relationship of the normal activity and the HMTA, and upon the need to be addressed by the risk calculation. If the regulatory needs described in the Safety Board's prior study_11/ are considered risks to individuals, classes of individuals aggregate and catastrophic risks require identification. Methods for calculating each are possible.
For example, consider the risk to an individual (ri) from the perspective of a resident along a railroad pathway.
Insert Fig 4 here
The likelihood of an occurrence of a loss producing event affecting this individual is dependent upon the probability of a mishap during the HMTA, the probability of hazardous materials involvement in an HMTA mishap, and the probability of a release of a hazardous material in a hazardous materials mishap. The likely loss, or severity of the consequences, is dependent upon the probable proportion of the total population exposed in the hazardous materials release, the likely proportion of the exposed population which is injured, and the likely proportion of the injured population which experiences fatal injury. The exposure is dependent upon the likely proportion of the HMTA to the duration of the normal activity being considered.
The risk for a given individual in a pathway module can then be expressed as follows:
(1) ri = Po x Ph x e where
Po = the estimated probability of the occurrence of an accident at the location of the individual in which injury is attributable to hazardous materials.
Ph = the probability of harm attributable to such a mishap, and
e = the exposure of the party at risk to such a mishap.
Po is a function of
Pa = the probability of an accident during the HMTA
Pb = the probability of the involvement of hazardous materials in an accident, and
Pc = the probability of a release of hazardous materials when they are involved in an accident.
Ph is a function of
Px = the probable proportion of the population which will be exposed to a hazardous material, release.
Py = the probable proportion of the exposed population which will suffer injury upon exposure.
Pz = the probable proportion of the population which will suffer fatal. injury.
e is a function of the probable duration of the exposure time to HMTA for the period being considered.
Each of these factors, in turn, is a function of other variables, and can be conveniently analyzed for incremental change in such variables to assess the effects of proposed safety controls.
The risk (rm) for the total exposed population (Pm) in a given analytic module (m) is the sum of the individual risks for each of the exposed module occupants, and can be expressed:
(1) rm = ri x P m or
(3) rm = Po x Ph x (e x Pm) or
(4) rm = Po x Ph x E where E represents the aggregate exposure.
The risk level (R) associated with the activity of transporting hazardous materials by rail in N total modules between two terminals on a rail line is the sum of risks which exist in all affected modules, and can be expressed:
(5) R = rm1 + rm2 + rm3 + . . . rmN = ÖN rmi
i = 1
To estimate the risk level associated with rail transport of specific shipments (Rs ), the factors constituting R can be adjusted in = Po , Ph and e to reflect the specific characteristics of the shipment and its movement. This provides a method for comparing risks associated with alternative handling of specific shipments or for making comparisons among shipments..
Two considerations impose limits on the risks associated with hazardous materials. The first is the maximum acceptable risk level for the activity, and the second is the need to avoid catastrophes.
Assume Kmax is the maximum acceptable risk level, or
(6) Po x Ph x E = Kmax
The risk Kmax provides an upper limit against which both present conditions and proposed changes can be assessed or compared.
The second constraint, against catastrophes, reflects the consideration that some accidents are unacceptable because of the magnitude of the injury, regardless of the probability of their occurrence (Po = 1). To analyze for such catastrophes, the identification of types of accidents in which catastrophic injuries can be expected would focus on the expected harm and the exposures. This could be expressed:
(7) Ph x E = Kmax
which represents the maximum casualties or losses which constitute the threshold of an unacceptable catastrophic accident, various types of accidents in different modules can be examined to determine whether such losses can be expected to occur.
Note that the probable casualties, rather than the maximum possible casualties (maximum credible accident) is the basis for this determination, because this more nearly reflects what is likely to occur. It is acknowledged that in specific incidents casualties may be greater or less than this value, because chance deviations in the locations of persons exposed can be expected in specific cases.
A description of the system to be analyzed is a necessary prerequisite to its analysis. For this study, the system to he analyzed is described on the basis of the pathway in which the HMTA is conducted. The System Description is shown in Figure 4. The description incorporates modules to facilitate the analysis tasks.
Insert Figure 4 here
Modules with one half mile square dimensions were adopted because of the availability of existing mileage markers on many railroad lines and corresponding compatible engineering records for such lines. Additionally, a module of this size extends approximately 1, 300 feet the limit of hazardous materials danger zones, noted in some past accidents, on each side of the mainline track.
The module would be described in terms of the pathway, operational and geographical characteristics or conditions. The characteristics or conditions which are of principal interest are those that contribute to risk peaks within a module. These factors would include, for example, the track condition, the terrain, the population density, the traffic density, the traffic makeup, etc.
The factors to be considered are dependent upon the elements described in the risk equations. As a better understanding of these factors is achieved, the individual elements of the risk equations might be expanded.
A modular approach is presently utilized in the regulation of pipeline operations (49 CFR 192. 5). This usage of the modular concept provides four general classes of modules. It is anticipated that the railroad pathway modules might be classified similarly into a small number of classes with common characteristics. This would permit the determination of risk by class of module and characteristics of operations, and reduce the number of risk calculations required to analyze the rail or other transport systems.
The module size and classes ultimately selected should be compatible for all modes of transportation, if comparative risk determinations are to be developed in the future. Therefore, the module definition or sizes selected for this study should not be considered at this time.
The modules permit analytical consideration of the associated risks from numerous perspectives which may be needed by the analyst. These perspectives include risks to individual occupants of a module, risks to the total population in a module, risks in adjacent modules, overall risks for movements through a series of modules between origin and destination points, relative risks for various types of shipments moving through modules under varying conditions, and risks associated with different activities being conducted by parties at risk in modules. These perspectives are related to regulatory needs discussed in the Safety Board's Risk Concepts Study11/ in the following table:
Insert Table 1 here
The need to validate the risk levels predicted for an activity requires that the risk levels be stated in terms that are compatible with the expression of casualty rates. For example, the sum of the risks associated with the movement of hazardous materials through a railroad system should approximate the accident experience for that system over a statistically.meaningful period of time. It is important to recognize that the risk based approach will not identify the time and place of the next accident, but will rather describe the expected overall experience over a period of time nor will it predict the time and place of the next catastrophe. However, the approach does provide a means for addressing the problem on a broad basis for national safety policy and program purposes
It may be necessary to improve both accident and exposure data collection to meet this need.
The general approach contemplates the identification of segments of the activity which contain 'risk peaks. " This applies equally to the identification of catastrophic risks, and of high risks to parties exposed to the activity. (See Figure 5.)
Insert figure 5 here
Catastrophic risks may be viewed as one time risk peaks in the sense that when an accident occurs the losses would be unacceptable. Isolating the conditions and events during which such results could occur is a major thrust of this "risk peak" analysis.
Risks to individuals may peak during the conduct of certain types of activities, during the transportation of certain types of shipments, at certain locations along a pathway, or in other circumstances. Such peaks may involve both the high likelihood of the occurrence of an accident, or severe consequences if it does occur, as contrasted to catastrophic risk which is concerned solely with the consequences.
By focusing on the "risk peak, " a method is provided to establish priorities for efforts to control unsafe conditions with the resources available, or to identify the need for allocating additional resources to achieve an acceptable level of safety.
One of the underlying constraints during the preparation of this study was the effort to identify hazardous conditions which could have had significance in not just railroading but in other modes of transportation of hazardous materials. This consideration should be a criterion for all safety analysis efforts, if cross modal comparisons are to be developed and made available to the parties introducing such risks or responsible for their control.
Within the preceding framework, methods for analysis of risk of accidental harm associated with the rail transportation of hazardous materials have been prepared and sample calculations made. This section describes the results and examines the effects of these efforts. Risks have been estimated from the perspectives described in Table 1. Calculations utilized available data wherever available, and ''best guesses" where no sources are indicated. Derivation of best guess" values is also described. A. Estimated Historical Risk Levels Casualty data shown in Figures 1 A and 1 B indicate that during the past 25 years, the transportation of hazardous materials by railroad has been responsible for about two fatalities and at least 50 injuries per year in the United States. This casualty rate might be interpreted to suggest that there is no cause for concern. Examination of the casualties on a risk basis suggests different interpretations.
The overall casualty rate attributable to the railroad HMTA for U. S. civilian residents during a recent 7 year period (1965 through 71) has been about .000002. fatalities per million person hours of exposure. This represents the risk to all civilian resident members of this country's population, whether or not they might personally be exposed to the HMTA from time to time. Adjusting the population exposed to reflect average occupancy of all modules constituting the 206, 000 miles of rail line in the United States, the estimated casualty rate was about . 00003 fatalities per million person hours of exposure.
Insert Tabloe 1 here
Approximate Fatalities per Million Person Hours Exposure
U. S. Civilian Resident HMTA .000002
U. S. Population 21 Disease 1.0
U. S. Population 21 Natural Disaster .0001
U. S. Population 21 Electric Power .002
On this basis, the casualty rate (historical risk level) for rail transportation of hazardous materials suggests that other national safety problem areas are relatively more pressing.
2. Classes of Parties at Risk
As the overall casualties are segregated into classes of parties at risk, a different picture emerges. Known casualties, attributable to railroad hazardous materials transportation are shown by class in the following table.
Injuries attributable to IIMTA by class from 8 Selected Accident Reports 6 to 10, 14, 15
Class Fatal Nonfatal Total
1. Carrier Employees
a. Train crew 0 1 1
b. Wreck crews 8 75 83
a. Residents 9 569 578
b. Newsmen 0 9 9
. Nonresident Bystanders 23 19 42
3. Emergency Personnel
a. Firemen 1 100 101
b. Other 0 1
Totals 41 774 813
By Class, Total:
1. Carrier Employee 8 76
2. Bystander 32 597
3. Emergency Personnel 1 101
Assuming these data reflect the distribution of the total injuries attributable to railroad HMTA, the risk of fatal injury varied among the classes of parties at risk as follows:
Extimated Fatality Rates by Class
Fatalities per Million Person Hours Exposure
Carrier Employess, Total
a. Train crews
b. Wreck crews
c. Other bystanders
Emergency personnel (not determined)
This table indicates that, based on a small number of fatal injuries, significantly different levels of risk exist for firemen and wreck crews, as contrasted with other classes.
This is not surprising, considering the phase of the accident in which their respective activities are conducted.
The casualties in Table 3 were attributable to only four of the roughly 1, 100 commodities regulated in 49 CFR 170-179. The distribution of the casualties among the commodities is shown in the following table:
Injuries Attributable to HNTA by Commodity
from 8 Selected Accident Reports
Commodity Fatal Nonfatal Total
A 6 53 59
B 26 348 374
C 8 322 330
D 1 51 52
Totals 41 774 815
From these data, the historical casualty rate for a U. S. civilian resident due to the activity of transporting the respective commodities, in the quantities which moved is directly proportional to the fatal injuries. Thus the rate for the activity of transporting commodity B is 26 times greater than D, in the aggregate. This does not take into account differences in the scope of the respective activities, so it is useful primarily as a measure of the activity as it has existed. If the scope changes, the effects will change the above historical relationships.
The identification of changes in historical risks utilizes the risk framework and the modular approach previously described. The statistical bases for determining the probabilities associated with the risk equations are most advanced in the area of forecasting accident occurrences, and weakest in the consequences area. Therefore, Po Ph and e are treated separately in the following sections.
In the calculation of r, (Equation 1), data for estimating the probability of the occurrence of an accidental release of a hazardous material in an accident for the 3 years 1968 to 1970 can be broken down as follows:
Pa = . 00038 train accidents per train hour
Pb = .047 hazardous materials (tank car) accidents per train accident *
Pc = .33 hazardous materials releases per hazardous materials accident
or about .000006 releases per train hour. This works out to about 128 releases per year. From 1965 to1971, the actual count was 129 per year.
* Not adjusted to exclude Canadian occurrences.
If accidents were distributed uniformly over the 206, 000 miles of rail line, the p in any module would be about 129 releases per year divided by (206, 000 mi X 2 modules/mile), or about .0003 releases per module year. However, distributions of accidents among the modules is more likely to be determined by the conditions pertinent to a given module. By analysis of conditions determining Pa , Pb and Pc above, predictions of the effects of proposed changes in these probabilities for specific modules can be made. For example, if the tank car head shields being considered in the RPI AAR research project are adopted and prove 100 percent effective, the value of Pc would decline to about _________
It can be seen that new Federal Railroad Administration regulations which address track and car conditions will impact Pa above. The Safety Board is aware of no actions which are directed toward Pb at this time. The work of the RPI AAR Safety Research is directed toward improving Pc .
No known data analyses of casualties are available to identify historical casualty rates from which casualty projections might be developed. Therefore, selected accidents for which historical data are available were examined to develop a basis for such estimates (Tables 3 5).
To forecast Ph ' the factors px, py and pz in equation (1) need to be developed.
The magnitude of the expected harm can be shown to vary with identifiable conditions. For example, the probable proportion of the population exposed (p ) in an accident varies with the area impacted by a hazardous material release in the accident. The affected area, in turn, is contingent on released quantities, emission characteristics and velocities, atmospheric and topographical conditions, and other factors. Methods for predicting the dispersion range are receiving extensive and continuing attention, 22,/23,/24/ and it is reasonable to expect that they can be made for most injury producing releases with increasing confidence in the future.
The population exposed varies widely, with the response options available to the population. For example, if the dispersion of the hazard is delayed and the threat to public safety is perceived, the time available to evacuate the impacted area may be sufficient to reduce p to nearly zero. Where the dispersion occurs during the crash phase of an accident, this option is not available. Thus a catastrophe is more likely when the dispersion is rapid or the threat is not adequately understood. Documentation of these response options should be encouraged. By combining the impacted area data with the population data in the module being analyzed, reasonable estimates of Px appear possible.. Development of predictive values for Px in given analysis modules with specified topographic and population characteristics can probably best be undertaken by comparing 1) the likely dispersion characteristics of the hazardous material released and the resultant danger zone size and exposure intensity distribution with 2) the population within the danger zone. This would be accomplished with an "overlay" of the danger zone onto the population distribution
way to identify the potential population exposed to the effects of the release under varying conditions. The approach is illustrated in the charts in Figure 5 attached.
This chart illustrates various dispersion modes. A fireball after an abrupt rupture of a tank in a fire projects injurious radiant energy hemispherically, as might a radioactive material spill. Lethal contents of a leaking anhydrous ammonia car would spread in a plume shaped dispersion pattern over level ground, or along stream beds in hilly terrain. LPG vapor clouds from leaking cars might spread first in an engulfing plume like fashion, then ignite and spread their lethality in hemispherical shock waves and thermal radiation.
The expected exposures could utilize a pollution 'rosette" 23/ approach to accommodate the variations in ambient weather conditions or directional dispersion experience to arrive at an "average" Px for a given module and shipment. Other predictive approaches also merit consideration.28/ From these charts, Px in a module might range from less than . 01 to 1. 0 for the various types of accidents described. Other injury producing accident types can be similarly analyzed for Px
Conditions affecting the probable injuries (Py) among exposed persons depend upon the nature of the injury mechanisms associated with the dispersed hazard. These have been roughly delineated in Table 6 for the 8 selected accidents used previously.
To probe the relationship between these injury mechanisms and the principal dispersion mode, the injury mechanisms in the accidents analyzed were ranked according to the number of injuries and linked with the principal dispersion mechanism in Table 7.
Insert table 7 here
This table suggests that hemispherical dispersion mechanisms are the most likely to produce injury. It would appear from the number of injuries (Table 6) that hemispherical dispersion mechanisms introduce higher Py values than the linear, plume or circular mechanisms, by a considerable magnitude. This may be partially attributable to the timing involved in encroaching on the exposed persons. Each dispersion mechanism affects differently the probable proportion of the population exposed in the accident. For example, the likelihood of injury would be greater for anyone in the path of a rocketing tank car section than someone in range of a nonflammable, nonpoisonous "compressed gas" vapor plume.
Past accidents might provide useful data for this purpose. In two accidents involving different commodities, the injuries in spills of approximately the same quantities and release rates varied by a factor of 4. Comparing two accidents involving similar materials in different quantities but with similar dispersion mechanisms, the injuries were similar. Further analyses would appear promising.
The value of Py is also affected by the timing of the dispersion mode. If an exposed person has an opportunity to respond to the exposure with effective countermeasures, the injuries might be reduced. For example, if an individual engulfed in a vapor cloud ignition can adequately shield exposed portions of his body before ignition, he might escape thermal injury.
Timing of the dispersion mode also affects Px . Refinement of this approach and expansion of the data base upon which the evaluation is based should permit a better predictive approximation of Py to be developed.
Conditions affecting the lethality of the release appear to be dependent principally upon the nature of the commodity, although post injury treatment and other circumstances can not be overlooked.
Once exposed persons suffer injury in an accident, the probability of survival depends upon the injury mechanism in the circumstances. If the exposure to continuing injury can he expected to be terminated and timely treatment begun, survival is more likely. Some materials, such as respiratory paralyzers, are more rapidly lethal than materials which might act through skin absorption, and survival may be relatively unlikely. Other injury, such as by mechanical puncture of vital organs by shrapnel, is less dependent on duration of the exposure than timely treatment.
Data from past accident casualties should be useful for predicting The relative fatality rates from Tables 8 and 9 below suggest large differences among commodities and injury mechanisms which additional data may or may not support.
These data suggest that the probable fatal injury rate in hazardous materials accidents varies significantly among the classes of persons atrisk, the shipment and the injury mechanism which might be involved.
These kinds of data, which might be derived from both transportation and non transportation accident histories, would appear to be applicable for preparing informed estimates of fatalities which might reasonably be anticipated for various classes of accidents.
Fatal Injuries to Total Injuries by Class
From 8 Selected Accident Reports
Class Injuries fatal
1. Carrier Employees
a. Train Crew 1 0
b. Wreck Crew 83 9.6
a. Residents 578 1.6
b. Newsmen 9 0
c. Bystander/Gawkers 42 54.8
3. Emergency Personnel
a. Firemen 101 1.0
b. Other 1 0
By Class, Total 815 Injuries
1. Carrier Employee 84 9.5 10
2. Bystander 629 5.1 77
3. Emergency Personnel 101 1.0 13
Fatal Injuries to Total Injuries, by Commodity From 8 Selected Accident Reports
Commodity Total Injuries Percent Fatal
A 59 10
B 374 7
C 330 2.4
D 52 2.0
The interrelationship of the activities described earlier provides the basis for establishing the probable exposure factor for the parties at risk. In Figure 2, Case 1, the exposure of an individual in a module to the HMTA going on in the module is directly proportional to the duration of HMTA compared to the total duration of his presence in the module. Thus, and if an individual resides adjacent to a railroad pathway, and never leaves his home, he will be exposed to the HMTA for a national average time of about 82 hours during the 8, 760 hours of the year he is in the home. This could be estimated more precisely for specific modules using actual train operating and traffic data. For emergency personnel, like firemen, their activity merges with the HMTA when they go to a wreck scene (Figure 2, Case 2). The duration of their exposure is the time they are on scene engaged in handling or coping with the emergency. Carrier employees' exposure to HMTA as a proportion of their total exposure can be estimated on a similar basis under Figure 2, Case 3 concepts.
In the absence of directly applicable predictive data, analysis for catastrophe is based on identification of all the sets of conditions which must he present for its occurrence, and the likelihood that the elements of any set might exist in the combination required. From equation 7, Kmax in a single accident is considered to be the undesired event which is to be addressed by the safety analysis 11/ effort. For Kmax to occur, one of a limited number of hazardous material releases must occur, persons must be exposed to the effects of the occurrence, and at least Kmax persons must be injured. For large exposure to occur, the hazardous material release (or its effects) probably * must migrate beyond the boundaries of the railroad track involved. This migration might take the form of any one of the dispersion modes in Figure 5. For a catastrophe to occur, the shipments involved must be capable of one of these dispersion modes Thus the presence of such shipments constitutes one of the ingredients which must be present for a catastrophe to occur. The following table shows the essential relationships for various Kmax values which might be selected.
* A wreck involving a passenger train and hazardous materials would be an exception from this generalization.
Effects of Module Population on Ph for Various Kmax
Maximum If population Ph for HMTA
Tolerable Kmax of module (s) is must be
10 10 1
10 1,000 .01
10 5,000 .002
1 5, 000 .0002
The effects of the value of Kmax selected for given analytical modules, on the controls necessary to achieve the required Ph are apparent. What controls might be instituted to limit
From Figure 5 and Tables 6, 7, 8, and 9, the conditions which must be present for h to attain the values to produce Kmax in the module (s) being analyzed can be productively approximated, for purposes of ranking the problems sets to indicate which conditions merit priority attention.
It must be recognized that such predictions can not identify when, where and how bad the next accident can be. There is a randomness connected with the existence of all the specific conditions which must be present in an accidental occurrence. The control efforts are undertaken to remove one of these "must" conditions, reduce the frequency of their existence or reduce the likelihood of their existence in the combination needed for an accidental occurrence. Safety analysis/,control efforts can not be expected to prevent all accidents, but can be expected to reduce the risks associated with an activity to acceptable levels.
Approaching the analysis from the perspective of different "shipments" in a given module, the alternative sets of "must" conditions would be derived in detail utilizing techniques such as logic trees found in reference 28, Appendix B, and the effects in an "effects analysis" such as that illustrated in Figure 6 or other documents. 28/, 31/ Rankings or probabilities (depending on the level of effort or state of knowledge) can be assigned to each set of "must" conditions in a module. The higher ranked or more likely conditions would constitute the "risk peaks " which should be given priority attention.
Insert Figure 6 here
These considerations suggest that a definition of the HMTA in the modules being analyzed is an essential first step in the risk analysis process. This, in turn, suggests that the carriers who undertake the HMTA in the modules are the most likely candidates to undertake the generation of the underlying data and analysis to identify the risks.
A simplified application of the methods discussed is presented to illustrate how the method produces information for improving safety decisions.
From Table 4, the greatest risk of injury appears to be borne by firemen fighting wreck emergencies. Wreck emergencies, in which firemen are fatally injured, can be viewed as a progression of events represented by Pa , Pb , Pc , Px , Py and Pz in equation (1).
Each successive event is contingent on the occurrence of the preceding event, and the continuation of the progression requires the existence of identifiable conditions (or hazards). Countermeasures to prevent the continuation or arrest of the progression of accident events can be proposed for each of the events involved. The most successful countermeasures usually are those which arrest the chain of events before they begin. The earlier the progression is arrested, the more likely injury can be avoided, and the greater the expected reduction in risk.
Firemen enter into the progression of accidental events after the accident involving a hazardous material tank car has begun. The hazardous materials "release'' may or may not have occurred by the time of their arrival. Based on past accidents, the initial "release" will have occurred.
or be in process when the firemen arrive. Most of the casualties have been associated with secondary "releases" after firemen are on scene. Thus the point for application of the first countermeasures to protect the firemen is during the event preceding their entry into the danger zone associated with the accident (the junction of the "normal.'' activity and HMTA, Case 2, Figure 2.
A proposed countermeasure is a change intended to reduce the probability of the occurrence of a sequential event in an accident sequence. Thus, by examining the affect of the change on the target event, and then examining the effects on the subsequent events, the change in risk level for the activity being examined can be predicted and its success measured.
In this example, the activity addressed in firefighting during a hazardous material accident, and the risk level is the forecasted fatalities per exposure hour for the individual fireman engaged in the activity.
The historical risk level has existed when Pc was equal to 0. 16. In other words, for every six fires, one secondary release could be expected. If the hazard(s) determining Pc can be reduced to zero (i.e., thermal shields being studied by the industry would be 100 percent effective) the risk of accidental injury to firemen from secondary releases could be eliminated. Achievement of 100 percent control is unlikely. However, with thermal barriers, the present Ph (approximately .007) would also decline because of the increased response time would allow firemen to set up remotely operated apparatus (assuming this time would be efficiently utilized) and to withdraw from the danger zone.
Thus assuming a 90 percent efficiency, Pc would decline to .016, and would also decline by some significant amount, probably in the range of 25 to 50 percent. Thus, the risk level for a fireman, responding to a fire, from this action would decline from 9. 9 fatalities per million exposure hours to about 0. 1, calculated as follows:
Assume total annual duration of this firefighting activity is 6, 740 hours (Appendix I) and no fireman attends more than one such fire during period. From Equation
rif = Pa x Pb x Pc x Ph x e Ä Pm
rif = 1 x 1 x .016"'x .00525' x (8/6740 ) x 1 or
rif = 0. 1 forecasted fatal injuries per million exposure hours.
This reduces the individual level of risk for this activity to a level comparable with that now existing for the activity of clearing wrecks.
With this estimated effect on the risk level, the ''safety" element of the decision making process can he quantitatively weighed against the other considerations which enter into such decisions.
1. 83rd Annual Report on Transport Statistics in the United States for the Year ended December 31, 1969, Interstate Commerce Commission, Washington, D. C.
2. Yearbook of Railroad Facts, 1971 Edition, Association of American Railroads, Washington, D. C.
3. Accident Bulletin No. 1.39, Calendar Year ].970, Federal Railroad Administration, Department of Transportation, Washington, D. C.
4. Wage Statistics of Class 1 Railroads in the United States, Statement No. A. 300, Calendar Year 1971, Interstate Commerce Commission, Washington, D. C.
5. Final Phase 01 Report on Accident Review, Report RA 01 4 l6, Railroad Tank Car Safety Research and Test Project, do Association of American Railroads, Washington, D. C.
6. National Transportation Safety Board, Washington, D. C., Railroad Accident Report, "Southern Railway Company Train 154 Derailment With Fire and Explosion, Laurel, Mississippi, January 25, 1969. 11
7. National Transportation Safety Board, Washington, D. C., Railroad Accident Report, "Chicago, Burlington, and Quincy Railroad Company Train 64 and Train 824 Derailment and Collision with Tank Car Explosion, Crete, Nebraska, February 18, 1969.
8. National Transportation Safety Board, Washington, D. C., Railroad Accident Report, "Derailment of Toledo, Peoria and Western Railroad Company's Train No. 20 with Resultant Fire and Tank Car Ruptures, Crescent City, Illinois, June 21, 1970. 11
9. National Transportation Safety Board, Washington, D. C., Railroad Accident Report, "Illinois Central Railroad Company Train Second 76 Derailment at Glendora, Mississippi, September 11, 1969.
11. National Transportation Safety Board, Washington, D. C., Special Study, "Risk Concepts in Dangerous Goods Transportation Regulations."
12. United States Life Tables by Causes of Death, 1959 1961, U. S. Department of Health, Education and Welfare, May 1968, Washington, D. C.
13. Statistical Abstract of the United States, 1970, U. S. Department of Commerce, Washington, D. C.
14. Railroad Accident Investigation, Ex Parte No. 218, Seaboard Air Line Railroad Co., Meldrin, Georgia, June 28, 1959, Interstate Commerce Commission, Washington, D. C.
15. Railroad Accident Investigation Report No. 3838, Missouri Pacific Railroad, Monroe, Louisiana, Jan. 22, 1959, Interstate Commerce Commission, Washington, D. C.
16. BE Report No. 61, August 1968, The Bureau for the Safe Transportation of Explosives and Other Dangerous Articles, Association of American Railroads, Washington, D. C.
17. Communication from Southern Railway to National Transportation Safety Board, 1972.
18., Chauncey Starr, "Social Benefit vs. Technological Risk, " Science Vol. 165, September 19, 1969.
19. Minutes of March 25, 1971, Meeting, National Research Council, Committee on Hazardous Materials Advisory to the United States Coast Guard, Washington, D. C.
20. Code of Federal Regulations, Title 49, Parts 171 173, U. S. Government Printing Office, Washington, D. C.
21. Chauncey Starr, Benefit Cost Studies in Socio Technical Systems Proceedings of the Conference on Hazard Evaluation and Risk Analysis, Houston, Texas, August 18 19, 1971, Division of Chemistry and Chemical Technology, National Research Council, National Academy of Sciences, Washington, D. C.
22. D. Bruce Turner, Workbook of _Atmospheric Dispersion Estimates, Revised 1970, Public Health Service, Environmental Health Service, U. S. Department of Health, Education and Welfare, Cincinnati, Ohio.
23. Technical Report, Washington I) C. Metropolitan Area Air Pollution Abatement Activity, Public Health Service, U. S. Department of Health, Education and Welfare, Cincinnati, Ohio, 1967.
24. Federal Register, Volume 35, No. 201, Thursday, October 15, 1970, Page 16180, Docket HM 60.
25. Comments on Recommendations from ERA on National Transportation Safety Board's Report No. NTST3 RAR 72 2, dated August 19, 1972.
26. Hazardous Materials Regulations Board Public Docket Files (Docket No. HM 60), Department of Transportation, 400 6th Street, S. W., Washington, D. C. 20590.
27. op. cit. 49 CFR, Part 213
28. B. J. Garrick, W. C. Gekler, 0. C. Baldonado, H. K. Elder, and J. E. Shapely, A Risk Model for the Transport of Hazardous Materials, prepared for Department of the Army, Fort Detrick, Maryland 21701, under Contract No DAAA13068 C 0190.