Comparison of Two Industrial Quantitative Risk Analyses Using the OECD Risk Assessment Dictionary/Thesaurus

Comparison of Two Industrial Quantitative Risk Analyses Using the OECD Risk Assessment Dictionary/Thesaurus Dennis C. Hendershot Rohm and Haas Company PO Box 584 Bristol, PA 19007 EMail: nagdh@rohmhaas.com Prepared for Presentation at the Fourteenth International Hazardous Materials Spills Conference Chicago, Illinois April 5-9, 1998 Wednesday Morning April 8, 1998 8:30-10:00 Session Methodology for Improving International Communication of Risk Assessment Terminology Related to Chemical Accidents Copyright Rohm and Haas Company February 25, 1998 UNPUBLISHED

Comparison of Two Industrial Quantitative Risk Analyses Using the OECD Risk Assessment Dictionary/Thesaurus Dennis C. Hendershot Rohm and Haas Company Bristol, PA 19007 Abstract The OECD has developed a dictionary/thesaurus for describing risk assessment processes, laws, and regulations. The thesaurus/dictionary provides a generic structure which can be applied to a specific quantitative risk analysis to understand how it relates to other analyses, and also to risk management laws or regulations. The proposed OECD risk structure will be used to compare two industrial quantitative risk analyses. Introduction The Organization for Economic Co-operation and Development (OECD) has developed a Risk Assessment Thesaurus/Dictionary with the help of a group organized by the Wharton Risk and Decision Processes Center of the University of Pennsylvania and supported by the United States Environmental Protection Agency (EPA) (Ross and Ignatowski, 1997; Ignatowski, et. al., 1997). The Dictionary/Thesaurus is intended for use in describing and comparing various risk management laws, regulations, policies, and risk analysis methodologies, with particular attention to catastrophic industrial risks. The OECD Dictionary/Thesaurus is intended to help people in different countries, companies, governments, and other organizations to understand the meaning of the various terminologies they use to describe risk regulations, assessments, and other documents. It is intended to overcome the two communications theorems proposed by Kaplan (1997): C 50% of the problems in the world result from people using the same words with different meanings C The other 50% come from people using different words with the same meaning The Dictionary/Thesaurus is intended to provide a common base line which can be used to understand the meaning of the language in various risk assessment documents and regulations. The OECD Dictionary/Thesaurus was used to characterize two industrial quantitative risk analyses (QRAs) as a test of its utility. This discussion will briefly describe the QRAs and summarize the description using the OECD Dictionary/Thesaurus. Both QRAs were done a number of years ago,

and the studies are characterized in the context of the legal requirements, regulations, and corporate policies which were in place at the time the QRAs were actually done. The Quantitative Risk Analyses Glacial Methacrylic Acid Storage The first QRA which was characterized using the OECD Risk Assessment Dictionary/Thesaurus was for a proposed storage facility for glacial methacrylic acid (GMAA). The QRA was done to aid design engineers in choosing between two alternate locations for the storage facility. GMAA is a combustible organic liquid which can polymerize exothermically if it is not stored and handled properly. For example, Anderson and Skloss (1992) describe a railroad tank car explosion which resulted from polymerization of GMAA. Potential causes of uncontrolled polymerization include contamination with a polymerization initiator, insufficient or incorrect polymerization inhibitor, insufficient dissolved oxygen in the GMAA (dissolved oxygen is required for the polymerization inhibitor to work), or exposure to external heat. The scope of the QRA was limited to estimation of the risk of uncontrolled polymerization. The results were used to compare two alternate proposed locations for the storage facility within the same manufacturing plant. The analysis focused on estimating the likelihood of a runaway polymerization, and it used simple, extremely conservative consequence and exposure models to place an upper bound on the risk to workers and local population. Upper bound individual and societal risk measures were estimated. Although the main purpose of this risk analysis was as an engineering design tool, the results were compared to Rohm and Haas corporate risk guidelines (Renshaw, 1990). The conclusion of the study was that there was little difference in the risk of polymerization between the two proposed locations, and that the decision could be made based on other factors such as operability and economics. The facility was also determined to meet the risk guidelines. Chlorine Storage and Distribution The second QRA used to evaluate the OECD Risk Assessment Dictionary/Thesaurus was for a chlorine handling facility, done in the mid 1980s. A summary of the risk analysis and its use as a risk management and decision making tool was published by Hendershot (1991). A description of this QRA using the OECD Risk Assessment Dictionary/Thesaurus was also published by Hendershot and Schechter (1997). The QRA was done as part of an expansion project for a plant, which would result in a doubling of chlorine usage, hence increasing the potential risk. The purpose of the QRA was to assist engineers in the selection of an appropriate design for the new and expanded chlorine unloading facility. The scope of this QRA was broader than the GMAA QRA, and it considered a variety of potential chlorine release scenarios. The QRA estimated both individual and societal risk using several risk measures described by CCPS (1989). The results were used to identify the major contributors to risk, and, from that information, to identify design and operational changes in the facility which significantly reduced risks. The tolerability of the estimated risk was evaluated by comparison to the existing facility, and by comparison of the facility design to industry and regulatory standards. 2

Dictionary/Thesaurus Application to the GMAA Storage QRA Although the OECD Dictionary/Thesaurus was intended as a tool to aid in understanding risk management policies, risk regulations, and risk assessment methodologies, it can be applied to a specific QRA study. This can help in understanding the methodology of the study, and might be useful in understanding how well a particular study meets the requirements of a risk management policy or regulation. The Dictionary/Thesaurus postulates that all risk assessments include four basic elements: Element I: Element II: Element III: Element IV: Sources with the potential to cause undesired outcomes to subjects of concern Basis for generating possible sequences of events Mode of contact between the potential to cause the undesired outcome and the subjects of concern Basis for estimating the likelihood that specified undesired effects will occur The details of the structure of the Dictionary/Thesaurus are described by Ignatowski, et. al. (1997). Tables 1 through 4 summarize the characterization of the GMAA QRA, including the tools used to fulfill the various elements and subelements used to describe risk analysis processes in the Dictionary/Thesaurus. The descriptions can be summarized as follows: C Element I: Sources with the potential to cause undesired outcomes to subjects of concern: The QRA considered the risk of uncontrolled polymerization, which could cause fatality due to acute toxicity of released vapors or from the potential for explosion of the storage tank. Risk to people was estimated, including on-site and off-site population. C Element II: Basis for generating possible sequences of events Potential accident scenarios were identified based on past experience in the company and industry, and using analytical tools such as Hazard and Operability studies (HAZOP), What If, fault trees, and event trees. C Element III: Mode of contact between the potential to cause the undesired outcome and the subjects of concern The modeling of the mode of contact was done for a worst case bounding case for this QRA. Essentially the assumption was made that any uncontrolled polymerization would result in multiple fatalities in the population of concern. This is an extremely conservative assumption in the incident reported by Anderson and Skloss (1992), for example, there 3

were no injuries or fatalities. The undesired outcome was measured in terms of the (worst case) likelihood of fatality at a particular location, and as the worst case estimated number of fatalities. C Element IV: Basis for estimating the likelihood that specified undesired effects will occur Quantitative fault tree analysis and historical data were used to estimate the likelihood of occurrence of the undesired outcome. For this QRA, best estimate data were used, and the uncertainty of the estimates was not evaluated. The analysis was done for the total population, as well as for several sub-groups of interest (for example, on-site employees, nearby residents). Several different individual and societal risk measures were used (CCPS, 1989), and two design options (different locations) were considered to help the process designers select the best overall system. Dictionary/Thesaurus Application to the Chlorine Handling QRA The application of the Dictionary/Thesaurus to the chlorine handling QRA was described in detail by Hendershot and Schechter (1997). The description of the chlorine handling QRA is summarized in Tables 5 through 8 for the four elements of the Dictionary/Thesaurus. Comparison of the Two QRAs Version 2 of the Dictionary/Thesaurus provides some simple reports which allow comparison of one study entry to another. This is potentially valuable in comparing two different regulations or laws, or in comparing a particular study to the requirements of a regulation or law. This reporting facility was used to compare the two QRA studies described. Figures 1 and 2 summarize the results. Basically, the comparison reports show that both studies included consideration of all four elements of a QRA as defined by the Dictionary/Thesaurus (Figure 1). Figure 2 compares the two analyses in more detail and shows that the GMAA QRA did not include a consideration of some subelements. For example, the GMAA QRA did not evaluate the units of the potential which causes the undesired outcome in detail, but, instead, used a simple and conservative (overestimate risk) assumption to provide a worst case estimate of the impact of a runaway polymerization. Also, the GMAA QRA did not evaluate the impact on the results of alternative assumptions (for example, evacuation and escape of the surrounding population) which were considered in the chlorine QRA. Conclusions The OECD Dictionary/Thesaurus provides a logical description of the example QRAs, and documents the assumptions used to do the analyses. The description of the QRA illustrates a sound and logical risk management process and describes the basis for decision making. Work continues on improving the Dictionary/Thesaurus. Data entry in Version 2.0 is less repetitive than in earlier versions through the use of a reference database. Reporting capability continues to 4

improve, with the simple comparison reports representing a significant improvement. More detailed comparison reports would be even more useful, and such reports would show more significant differences between the two QRAs described above. Other efforts to improve the utility of the database should focus on further improvements in searching capabilities, to it easier to find specific information. Overall, the Dictionary/Thesaurus is a useful tool for understanding a quantitative risk analysis. Tables 1 through 8 represent a substantial condensation of the actual Dictionary/Thesaurus reports, but they also allow a quick understanding of the basis and assumptions used for the example QRAs. The Dictionary/Thesaurus has potential usefulness in the following areas: C For documenting the basis, tools, and assumptions used for a particular QRA C For understanding the requirements of risk management laws, regulations, and policies C For comparing a particular QRA to the requirements of laws, regulations, or policies C For comparing the methodologies of QRAs done by various organizations or agencies References Anderson, S. E., and R. W. Skloss (1992). More Bang for the Buck: Getting the Most From Accident Investigations. Plant/Operations Progress 11, 3 (July), 151-156. Center for Chemical Process Safety (CCPS) (1989). Guidelines for Chemical Process Quantitative Risk Analysis. New York: American Institute of Chemical Engineers. Center for Chemical Process Safety (CCPS) (1985). Guidelines for Hazard Evaluation Procedures. New York: American Institute of Chemical Engineers. Hendershot, D. C. (1991). The Use of Quantitative Risk Assessment in the Continuing Risk Management of a Chlorine Handling Facility. The Analysis, Communication, and Perception of Risk, ed. B. J. Garrick, and W. C. Gekler, 555-565. New York: Plenum Press. Hendershot, D. C. and S. J. Schechter (1997). Evaluation of a Proposed Thesaurus/Dictionary for Risk Assessment Using an Industrial Quantitative Risk Analysis International Conference and Workshop on Risk Analysis in Process Safety. October 21-24, 1997, Atlanta, GA. 581-591. New York: American Institute of Chemical Engineers. Ignatowski, A. J., I. Rosenthal, and L. D. Helsing (1997). An Internet Thesaurus/Dictionary for Analyzing Risk Assessment Processes, Laws, and Regulations. International Conference and Workshop on Risk Analysis in Process Safety. October 21-24, 1997, Atlanta, GA. 567-579. New York: American Institute of Chemical Engineers. Kaplan, S. (1997). The Words of Risk Analysis. Risk Analysis 17, 4, 407-417. Renshaw, F. M. (1990). A Major Accident Prevention Program. Plant/Operations Progress 9, 3 (July), 194-197. Ross, W. C., and A. J. Ignatowski (1997). OECD Risk Assessment Dictionary/Thesaurus System Instructions and Workbook. Version 2.0. Wharton Risk and Decision Processes Center, University of Pennsylvania, Philadelphia, PA. (July 1). 5

Withers, R. M. J., and F. P. Lees (1985). The Assessment of Major Hazards: The Lethal Toxicity of Chlorine. Parts 1 and 2. Journal of Hazardous Materials 12, 3 (December), 231-282 and 283-302. 6

GMAA QRA Element I: Sources with the potential to cause undesired outcomes to subjects of concern Subelement Descriptors I Sources Systems containing reactive chemicals GMAA can self-react (polymerize) exothermically, generating pressure in a closed system such as a storage tank. For example, see Anderson and Skloss (1992). ii Subjects of concern People Nearby residents, employees of neighboring factories and businesses, and on-site workers were considered. iii Undesired outcomes considered People The QRA was done using immediate fatality as the measured outcome. Other potential outcomes (for example, injury, delayed fatality, legal sanctions, economic loss, anxiety) were not considered explicitly in the QRA, although management was certainly aware of them in the decision making process. Table 1: Subelements and Descriptors for Element I for the GMAA QRA 7

GMAA QRA Element II: Basis for generating possible sequences of events Subelement Descriptor I Basis for generating accident scenarios Past experience Industry records, company records, professional judgment Technical analysis Fault tree analysis, HAZOP analysis ii Basis for estimating distribution of potential for undesired outcome over space and time Other approaches Assume that any uncontrolled polymerization reaction causes multiple fatalities in each population of concern - worst case conservative assumption. iii Basis for estimating distribution of subjects of concern Other approaches Assume that the subjects of concern are always in a location where they will be impacted by an uncontrolled reaction - worst case conservative assumption. iv Basis for boundaries or endpoints Other approaches Assume that any person in the population of interest will be impacted in case of an uncontrolled reaction - worst case conservative assumption. Table 2: Subelements and Descriptors for Element II for the GMAA QRA 8

GMAA QRA Element III: Mode of contact between the potential to cause the undesired outcome and the subjects of concern Subelement Descriptor I Type of contact People Inhalation; explosion overpressure; fireball radiation ii Basis for relationship used to predict undesired outcome People Assume any uncontrolled reaction results in sufficient exposure to impact all members of the population of interest - worst case conservative assumption. iii Measurement units of the potential specified Not specified Conservative worst case assumption results in no need to quantify the potential. iv Measurement units of undesired outcome Total number of outcome events in the specified population Worst case estimated number of deaths; Worst case likelihood of fatality at a particular location Table 3: Subelements and Descriptors for Element III for the Example QRA 9

GMAA QRA Element IV: Basis for estimating the likelihood that specified undesired effects will occur Subelement Descriptor I Basis for estimating the likelihood that the undesired events will occur Quantitative assessment Fault tree analysis Historical data Equipment failure rate data and human error rates from various chemical industry databases for use in fault trees Expressions of likelihood of undesired event Quantitative Frequency of uncontrolled polymerization reaction; probability of various sequences of events resulting in various specific outcomes Quality/uncertainty Characterization of estimate Best estimate Undesired effects Specified for a single member of the population Member closest to potential Table 4: Subelements and Descriptors for Element IV for the GMAA QRA (continued on next page) 10

GMAA QRA Element IV: Basis for estimating the likelihood that specified undesired effects will occur (continued) Subelement Descriptor ii Comparison with relevant standards and guidelines Type of standard or guideline Comparison of one proposed location to another; comparison to risk guidelines (Renshaw, 1990) Comparison metrics Worst case individual risk Consideration of all practical measures to reduce likelihood of undesired events iii Impact of alternative assumptions Not considered Table 4: Subelements and Descriptors for Element IV for the GMAA QRA (continued from previous page) 11

Chlorine QRA Element I: Sources with the potential to cause undesired outcomes to subjects of concern Subelement Descriptors I Sources Substances Toxic to humans The human toxicity of chlorine is well known (Withers and Lees, 1985). Legally defined Although the chlorine inventory in the facility is below the threshold quantity for coverage by the relevant national and local regulations, the fact that chlorine was a listed substance in those regulations identified this facility to Rohm and Haas management as one requiring special attention. Energy Pressure A Rohm and Haas internal prioritization, since published in a modified form (Renshaw, 1990), highlighted the contribution of elevated pressure to the system hazard. ii Subjects of concern People Nearby residents, employees of neighboring factories and businesses, and on-site workers were considered. iii Undesired outcomes considered People The QRA was done using immediate fatality as the measured outcome. Other potential outcomes (for example, injury, delayed fatality, legal sanctions, economic loss, anxiety) were not considered explicitly in the QRA, although management was certainly aware of them in the decision making process. Facilities Sensitive population locations (schools, hospitals, a pedestrian tunnel) were considered in the QRA Table 5: Subelements and Descriptors for Element I for the Chlorine QRA (from Hendershot and Schechter, 1997) 12

Chlorine QRA Element II: Basis for generating possible sequences of events Subelement Descriptor I Basis for generating accident scenarios Past experience Insurance and industry records, company records, professional judgment Technical analysis Fault tree analysis, event tree analysis, HAZOP, What If? analyses as defined in the original edition of the CCPS Guidelines for Hazard Evaluation Procedures (CCPS, 1985) ii Basis for estimating distribution of potential for undesired outcome over space and time Technical analysis Event tree analysis PHAST consequence analysis program and SAFETI risk analysis program (from Technica, Inc., now DNV-Technica) iii Basis for estimating distribution of subjects of concern Historical data National census data, plant population data, contact with neighboring plants and businesses to obtain population data iv Basis for boundaries or endpoints Other toxicologically derived endpoints Probit relationship for inhalation toxicity of chlorine (Withers and Lees, 1985) Table 6: Subelements and Descriptors for Element II for the Chlorine QRA (from Hendershot and Schechter, 1997) 13

Chlorine QRA Element III: Mode of contact between the potential to cause the undesired outcome and the subjects of concern Subelement Descriptor I Type of contact People Inhalation ii Basis for relationship used to predict undesired outcome People Human epidemiological data, animal data, weighted combination of data (Withers and Lees, 1985) iii Measurement units of the potential specified Concentration of substance over specified time Dose-response relationship over 0-60 minute time period based on Withers and Lees (1985) probit iv Measurement units of undesired outcome Total number of outcome events in the specified population Estimated number of deaths; likelihood of fatality at a particular location Table 7: Subelements and Descriptors for Element III for the Chlorine QRA (from Hendershot and Schechter, 1997) 14

Chlorine QRA Element IV: Basis for estimating the likelihood that specified undesired effects will occur Subelement Descriptor I Basis for estimating the likelihood that the undesired events will occur Quantitative assessment Event tree analysis, fault tree analysis, SAFETI risk analysis program, consultant and in-house vapor cloud dispersion models Historical data Equipment failure rate data and human error rates from various chemical industry databases for use in fault tree and event tree analyses Expressions of likelihood of undesired event Quantitative Frequency of releases; probability of various sequences of events resulting in various specific outcomes; probability of specific undesired outcomes at a particular location in a specified time period Quality/uncertainty Characterization of estimate Best estimate Undesired effects Specified for a single member of the population Average member Member closest to potential Members at specified locations (individual risk contour map) Specified for a group of N or more people in the specified population F-N Curve for entire population and specified sub-groups of the population (on-site workers, nearby residents) Table 8: Subelements and Descriptors for Element IV for the Chlorine QRA (from Hendershot and Schechter, 1997) (continued on next page) 15

Chlorine QRA Element IV: Basis for estimating the likelihood that specified undesired effects will occur (continued) Subelement Descriptor ii Comparison with relevant standards and guidelines Type of standard or guideline Other standards or guidelines - internal company guideline that the risk of the new facility would not exceed the risk of the existing plant. Comparison metrics Individual risk at specific locations F-N Curves Average individual risk Rate of death Individual risk contours Consideration of all practical measures to reduce magnitude and likelihood of undesired events (case studies for various design options) iii Impact of alternative assumptions Alternative assumptions Impact of sheltering in place was considered Table 8: Subelements and Descriptors for Element IV for the Example QRA (From Hendershot and Schechter, 1997) (continued from previous page) 16

Figure 1: Example high level comparison report for the GMAA and chlorine QRAs from the OECD Risk Dictionary/Thesaurus 17

Figure 2: More detailed comparison of the GMAA and chlorine QRAs using the comparison report feature of the OECD Risk Assessment Dictionary/Thesaurus 18