OCS Study MMS 2000-007 Estimation of Oil Spill Risk From Alaska North Slope, Trans- Alaska Pipeline, and Arctic Canada Oil Spill Data Sets April 2000 Prepared by: Hart Crowser Inc. 2550 Denali Street, Suite 705 Anchorage, Alaska 99503 This study was funded by the U.S. Department of the Interior, Minerals Management Service (MMS) Alaska Outer Continental Shelf Region, Anchorage, Alaska, under Contract No. 01-99- PO-16128, as part of the MMS Alaska Environmental Studies Program The opinions, findings, conclusions, or recommendations expressed in this report or product are those of the authors and do not necessarily reflect the views of the U.S. Department of the Interior, nor does mention of trade names or commercial products constitute endorsement or recommendation for use by the Federal Government. U.S. Department of the Interior Minerals Management Service Alaska Outer Continental Shelf Region Estimation of Oil Spill Risk From Alaska North Slope, Trans-Alaska Pipeline, and Arctic Canada Oil Spill Data Sets Jeffery L. Mach, Robert L. Sandefur, P.E., and Jean H. Lee OCS Study MMS 2000-007 2 PROJECT ORGANIZATION Hart Crowser, Inc., study team: Thomas K. Noyes Principal in charge Jeffery L. Mach Project manager Oil spill data collection Oil production, pipeline throughput, pipeline mileage data collection Oil spill notification requirements research Database manager Primary report author Robert L. Sandefur Data analysis Calculation of oil spill rates Report author – Analysis of Data and Calculation of Oil Spill Risk Rate Jean H. Lee Oil spill data research and collection Oil spill notification requirements research and collection Report author – Canadian portion of Oil Spill Notification Requirements Michael L. Boese Oil spill data research and collection Adam D. Chawes Oil spill data research and collection i TABLE OF CONTENTS Project Organization.....................................................................................................................i Table Of Contents........................................................................................................................ii Executive Summary.....................................................................................................................1 Introduction.................................................................................................................................4 Methods.......................................................................................................................................6 Results.......................................................................................................................................11 Discussion..................................................................................................................................29 References...............................................................................................................................113 Appendix A - Collated Alaska And Canada Individual Oil Spill Data Used In Study...........117 Appendix B – Statistical Analysis And Oil Spill Rate Calculations Using Alaskan And Canadian Spill Data.................................................................................................................132 ii EXECUTIVE SUMMARY The Minerals Management Service Alaska Outer Continental Shelf Region commissioned Hart Crowser, Inc. (Hart Crowser) to collect and analyze data to provide an estimation of oil spill risk from Alaska North Slope, Trans-Alaska Pipeline (TAPS), and Arctic Canada oil industry activities. To populate the database needed to perform statistical calculations, data on oil spills of 100 barrels (4,200 gallons) and greater related to oil industry exploration, construction, development, production, transportation, and storage activities from within eight study areas including the Alaska North Slope, TAPS, and Arctic Canada were collected from industry, government, and commercial sources. Supporting data on annual crude oil production, pipeline mileages, and quantities of crude oil transported by pipeline or tank vessel in the study areas were gathered. Current and historical oil spill reporting criteria in effect in the study areas also were identified. The oil spill data were collated and evaluated for comprehensiveness and completeness. Attempts were made to validate the data with government regulatory authorities to which the responsible party for a spill was required to report. The reliability of the volumes of spills of 500 barrels (21,000 gallons) and greater was evaluated based on how the spill volume was determined and supporting documentation. The statistical robustness and appropriateness of using the collected oil spill data, and the validity of using potential estimators to evaluate oil spill risks from Beaufort Sea Outer Continental Shelf (OCS) development were evaluated. Finally, an oil spill rate was calculated using the optimum data set, which was corrected for an observed time trend in the spill rate. Overall, 126 oil spills in Alaska and Canada were identified which met the study criteria. All of these spills occurred between 1970 and September 1999. The most recent spill occurred in 1997. There are 28 spills of 500 barrels and greater and 14 spills of 1,000 barrels (42,000 gallons) or greater. Another 95 spills in the study areas also were identified, but were not included in the subsequent analysis because insufficient information existed to allow a conclusive determination as to whether the spills met the study criteria. Hart Crowser identified 126 spills of 100 barrels and greater that met the study criteria. Of these spills, 111 occurred in Alaska and 15 in Canada. The Alaskan oil spills most frequently are associated with highway tank vehicle accidents and operations support facilities, followed by spills related to construction camps, operations support facilities, and pipelines. Spills associated with oil production processing facilities, oil production wells, pipeline pump stations, and exploration activities also were identified. No spills meeting the study criteria were identified for the Alaska Onshore North Slope (ONS), National Petroleum Reserve Alaska (NPRA), or Beaufort Sea study areas. Oil spills meeting the study criteria were identified in each of the Canadian study areas. Canadian spill were most frequently associated with oil exploration activities, oil production wells, and oil production processing facilities. Spills from highway tank vehicles, pipelines, and vessels also were identified. Data on the Alaska oil spills were considered to be comprehensive and complete because more than 60 percent of the spill records appeared in two or more data sets. The Canadian data are more suspect in terms of being comprehensive and complete. Less than 15 percent of the Canadian spills appeared in the two data sets obtained. Canadian data since approximately 1980 are considered good, but anecdotal report and the lack of records provides suspicion that the data is not comprehensive. Because of the small number of Canadian oil spills and relatively small 1 amount of Canadian oil production, the Canadian data were not included in the subsequent statistical analysis. For both the Alaskan and Canadian oil spill data, the volumes assigned for the spills of 500 barrels and larger are reliable, but must be considered as general estimates in most cases. Documentation for these spills often does not describe how the spill volume was determined. The Alaska oil spill data was sufficiently comprehensive and complete to conduct statistical analyses and estimate oil spill risk rates. A series of box and cumulative frequency plots of the Alaska spill data were constructed to analyze the data and determine trends. Exploratory data analysis on relevant independent variables indicated little statistical difference in terms of spills that occurred within the Alaska ONS, East of NPRA, and TAPS study areas. A general check on the fluctuation of the data set indicated spill occurrence to be quite random. There appeared to be little difference in the size of spills associated with the various facilities, with the exception of pipelines, which had larger spills. Analysis of variance by oil type showed that, in general crude oil spills tend to be larger than other types of oil spills. A statistical analysis of individual spill volumes by study area, facility type, oil type, affected media, and spill cause combined did not indicate any particularly interesting correlation. Annualized groupings of spills, where total spill volumes by year were accumulated and plotted on a cumulative frequency plot, showed a mixture of several populations. Re-plotted on a logarithmic scale, a single lognormal population emerged. A count of the number of spills per year in the database is showed a possible Poisson distribution, but that hypothesis was not tested. When spill size was plotted by year to see if regulatory or reporting requirements had a significant affect, it appeared that in the period from 1975 to 1979 there were a considerable number of large spills, and then the number of spills dropped to a more or less constant rate. The year of 1977 is significant because crude oil production on North Slope and operation of TAPS began in the middle of that year. However, the years of 1978 and 1979 visually fit with years of 1975 and 1977 better than breaking the data at 1977. The 1975 to 1979 period appears to have the most number of spills. When Alaskan spill data were plotted on a yearly basis, it appears that prior to 1977, spill rates were considerably greater than in the subsequent years. When re-plotted on a logarithmic scale, it is apparent that prior to 1980 spill rates were considerably greater than after 1980. Hart Crowser calculated oil spill risk rates based on the number of spills and on volume. Hart Crowser calculated a rate based on volume because of the greater visual variability in the data. The statistical significance of this visual analysis showed a highly statistically significant correlation with spill rate and year if all of the Alaska spill data is included. If data earlier than 1980 is excluded, then there is still a correlation between spill rate and year that is significant at the 1 percent level of confidence. However, if data earlier than 1985 is excluded, then there is a correlation between spill rate and year, which is significant at the 17 percent level of confidence. Hart Crowser concluded that spill rate is the best variable to use in predicting the volume of further oil spills and that a rate of approximately 52 gallons of oil spilled per million barrels of crude oil produced will be the average, if trends that started in 1980 continue. This rate is subject to considerable uncertainty in the mean (± 50% at the 95% level of confidence) and the value derived from the logarithmic distribution is 66 gallons of oil spilled per million barrels of crude oil produced as opposed to 52 gallons of oil spilled per million barrels of crude oil produced. These two values agree within the standard deviation of the means. The 95 percent 2 logarithmic confidence limits on spills for a given year are ± 465 percent at the 95 percent level of confidence. Hart Crowser is more inclined to believe the logarithmic values than the untransformed values, because the cumulative frequency of the data is more lognormal than normal. These very wide confidence limits and individual yearly values are consistent with the small number of data points available for this prediction. Hart Crowser also calculated oil spill risk rates based on the number of spills of a given volume per million barrels of crude oil produced (spills/MMBbl), using data from 1978 through 1999. Hart Crowser found these rates to be: • 0.0053 spills/MMBbl, ±24 percent, for spills of 100 barrels and greater; • 0.00093 spills/MMBbl, ±58 percent, for spills of 500 barrels and greater; • 0.00039 spills/MMBbl, ±89 percent, for spills of 1,000 barrels and greater; and • 0.000078 spills/MMBbl, ±200 percent, for spills of 10,000 barrels and greater. 3 INTRODUCTION In the development of environmental analyses for proposed Outer Continental Shelf (OCS) Beaufort Sea oil exploration and development off of Alaska’s North Slope, the U.S. Department of the Interior Minerals Management Service (MMS) OCS Region uses national OCS statistics to estimate the likelihood that large oil spills of 1,000 barrels (42,000 gallons) or larger will occur as a result of oil exploration, construction and development, production, and transportation activities. These national statistics primarily are from the Gulf of Mexico and do not include pipeline spills inshore of the OCS, in state waters, or on land. The MMS Alaska OCS Region desires to estimate oil spill frequency based on Alaska North Slope and Canadian Arctic, rather than Gulf of Mexico oil exploration, construction and development, production, and transportation experience. The MMS contracted Hart Crowser to gather data and provide oil spill risk occurrence estimators for OCS Beaufort Sea oil exploration and development based on Alaska North Slope and Canadian Arctic statistics. More specifically, the scope of work directed Hart Crowser to: • Identify, obtain relevant supporting information, and collate data for crude oil and diesel oil spills of 100 barrels and greater related to oil and gas exploration, construction, development, production, transportation, and storage from within the following study areas: o U.S. Beaufort Sea; o Canadian Beaufort Sea; o National Petroleum Reserve Alaska (NPRA); o Alaska Onshore North Slope (ONS) East of NPRA; o Trans-Alaska Pipeline System, not including the Valdez Marine Terminal; o Onshore McKenzie River Delta; o Canadian High Arctic Islands; and o Norman Wells. • Compare data sets from different sources for the same area to increase the comprehensiveness and completeness of the data. • Identify the oil spill reporting criteria in effect for oil spills in the study areas and validate the oil spill data with the regulatory authority to which the responsible party for the spill was required to report. • Describe the overall comprehensiveness and completeness of the collected oil spill data. • Evaluate the reliability of the volumes of spills of 500 barrels and greater, based on how the spill volume was determined and supporting documentation. • Obtain and collate data on crude oil production, pipeline throughput, tanker shipments, and pipeline mileage by year for the Alaska and Canada study areas. • Examine the appropriateness of using the collected oil spill data to evaluate oil spill risks from Beaufort Sea OCS development, partly in the context of prior MMS uses and statistical evaluations of oil spill rates for OCS use. 4 • Consider the statistical robustness and validity of potential oil spill risk estimators, including the: o Effect of one or more spills on the estimators; o Size of the data set for spills of 100 barrels or more and 500 barrels or more; o Rationale for including or excluding intentional spills; o Effect of incomplete pipeline life cycles in the data set; o Correlation of pipeline mileage and/or oil production or throughput volumes with spillage; o Differences in size of onshore and offshore oil spill data subsets; o Magnitude of the record (i.e., number and volumes of spills) used to calculate oil spill risk estimators versus that used by Anderson and LaBelle (1994); and o Postulated differences, or lack thereof, in onshore and offshore oil spill risk factors. • Calculate onshore and offshore oil spill rates using the optimum data sets and including corrections for time trends in spill rates, if statistically appropriate. • Prepare draft and final reports and technical summaries concerning the study, and appendices containing the oil spill and supporting data sets. 5