Estimates of occurrence rates for offshore oil spills are useful for analyzing potential oil-spill impacts and for oil-spill response contingency planning. With the implementation of the Oil Pollution Act of 1990 (U.S. Public Law 101-380, August 18, 1990), estimates of oil-spill occurrence became even more important to natural resource trustees and to responsible parties involved in oil and gas activities.
The Oil-Spill Risk Analysis (OSRA) model, developed in 1975 by the Department of the Interior (DOI), is a tool that evaluates offshore oil-spill risks (Smith et al., 1982; LaBelle & Anderson, 1985). This model is used to develop probabilistic estimates of oil-spill occurrence and contact. A realistic, objective methodology for estimating oil-spill occurrence rates is required for the model's application. The MMS developed oil-spill databases and the Bureau of Safety and Environmental Enforcement (BSEE) maintains these databases on U.S. OCS Spill Incidents, and tanker spills, which are used to support these estimations (Lanfear & Amstutz, 1983; Anderson & LaBelle 1990, 1994, 2012, ABS Consulting, 2016).
Oil-spill occurrence rate estimates were recently revised (ABS Consulting, 2016) based on U.S. Outer Continental Shelf (U.S. OCS) platform and pipeline spill data (1964 through 2015), worldwide tanker spill data (1974 through 2014), and barge spill data for U.S. waters (1974-2013). These spill rates are expressed and normalized in terms of number of spills per volume of crude oil handled. All estimates of spill occurrence rates were restricted to spills greater than or equal to 1,000 barrels (159 cubic meters, 159 kiloliters, 136 metric tons, 42,000 U.S. gallons) although spills as small as less than one barrel were also discussed. The cumulative frequency distributions of oil spills were examined and the trend analysis of offshore spills previously performed by Nakassis (1982) and Anderson and LaBelle (1990) was reassessed and repeated using current data. As in the previous papers in this series, this report describes the use of a Poisson process, with volume of oil handled as an exposure variable to predict the probability of spill occurrence. Anderson, Mayes & LaBelle (2012) presents a simple approach for estimating oil-spill occurrence, normalized as a function of the volume of oil handled. For this paper, volume is reported in barrels (bbl) to assist policy- and decision-makers in government and industry. Anderson, Mayes & LaBelle (2012) utilizes the same methodology as the four previous independently peer-reviewed papers (Anderson & Labelle 2000, 1994, 1990 and Lanfear & Amstutz (1983) presented in support of the oil-spill rate assumptions used for the DOI OSRA model.
As a supplement to Anderson & LaBelle (2000), 95 Percent Confidence Intervals are presented. Further statistical information supporting this approach can be found in documents identified in the Additional Statistical Background discussion below.
Additional Statistical Background
Lanfear & Amstutz (1983) examines the cumulative frequency distributions of oil spills, tests pipeline miles as an alternative exposure variable for pipeline spills, and discusses the trend analysis of offshore spills performed by Nakassis (1982). These spill rate papers tier off earlier work performed by DOI in support of the OSRA Model, and work performed by other oil-spill researchers, as referenced in the papers.
Smith et al. (1982) documents the fundamentals of the DOI OSRA Model. It describes the approach of using lambda, the unknown spill occurrence rate for a fixed class of spills, as a parameter in a Poisson process, with volume of oil handled as an exposure variable to predict the probability of spill occurrence (pages 18-24). A Bayesian methodology, described in detail in Appendix A, “Distribution Theory of Spill Incidence,” provides one way to weight the different possible values of lambda given the past frequency of spill occurrence for a fixed class of spills. Smith et al. (1982) selects volume as an exposure variable in that it is a quantity that would be more practical to estimate future exposure (a necessity for using it to forecast future spill occurrence) than the other exposure variables considered.
In support of using the Poisson process for spill occurrence and examinations of different exposure variables, Smith et al. (1982) references the works of Devanney & Stewart (1974), Stewart (1976), and Stewart & Kennedy (1978). These references, and other pertinent ones, can be found at Oil Spill Rates - Additional References.