Oil Spill Risk Analysis for Assessing Environmental Contact Probabilities in the Gulf of Mexico

The Federal Government plans to offer for oil and gas leasing in the Gulf of Mexico (GOM; Figure 1): the Eastern Planning Area (EPA), Central Planning Area (CPA), and Western Planning Area (WPA). Because oil spills may occur from activities associated with these lease sales—such as offshore oil exploration, development, production, transportation, and decommissioning—the U.S. Department of the Interior (USDOI) Bureau of Ocean Energy Management (BOEM) conducted an oil spill risk analysis (OSRA) to support environmental impact statements (EISs) that are completed before proposing to lease these areas.

The occurrence of oil spills is fundamentally a matter of probability. There is no certainty regarding the amount of oil that would be produced, or the size or likelihood of a spill that could occur during the estimated life of a given lease sale. Neither can the winds and ocean currents that transport oil spills be known for certain. A probabilistic event, such as an oil spill occurrence or oil spill contact to an environmentally sensitive area, cannot be predicted, but an estimate of its likelihood (its probability) can be quantified.

The OSRA was conducted in three parts which correspond to different aspects of the overall problem: 1) the probability of oil spill occurrence, which is based on estimated volumes of oil produced and transported and on spill rates derived from historical data; 2) calculated trajectories of oil spills from hypothetical spill locations to locations of various resources, which are simulated using the OSRA model (Smith et al. 1982; Ji et al. 2002); and 3) the combination of results of the first two parts to estimate the overall oil spill risk if there is oil development. Note that the analysis estimates spill contacts, not spill impacts. Further measures that should be evaluated to determine impacts, such as the natural weathering of oil spills and the effects of cleanup activities, are not directly factored into the analysis but should be added to the interpretation of its results.

For this analysis, both benefits and risks are functions of the volume of oil produced and are mutually dependent. Due to the inherent uncertainties associated with an assessment of undiscovered resources, probabilistic techniques were employed, and the results were reported as a range of values corresponding to different probabilities of occurrence. A wealth of historical data and information derived from more than 50 years of oil and gas exploration, development, and production activities were used extensively by BOEM (formerly the Minerals Management Service [MMS]). The projected oil production in billion barrels (Bbbl) for a typical proposed lease sale and for the cumulative existing oil and gas exploration and production activities in the GOM OCS region (OCS Program) is shown in Table 1 below.

The study area includes the entire GOM and surrounding area (Figure 1). The geographic boundaries that encompass the resources at risk from a hypothetical oil spill from OCS operations in the lease areas are shown in Figure 1. The resources considered in this analysis were selected by BOEM analysts in the GOM OCS region with supplementary input from the National Marine Fisheries Service and the U.S. Fish and Wildlife Service. BOEM analysts also used information from its Environmental Studies Program studies results, literature reviews, and professional exchanges with other scientists. To create maps of resource locations vulnerable to oil spill impact, the analysts used geographic digital information on the biological, physical, and socioeconomic resources that could be exposed to contact from OCS oil spills. These maps (e.g., Figure 2) depict locations that were analyzed by the OSRA model, which included onshore resources and the surface waters overlying or surrounding offshore resources. Detailed discussions of risks to all considered resources can be found in Ji et al. (2016).

Spills less than 1,000 barrels (bbls) are addressed in the EIS for each proposed action without the use of trajectory modeling, because smaller spills may not persist in the environment long enough to be simulated by trajectory modeling. Spills greater than or equal to 1,000 bbl persist in the environment long enough to be modeled and are addressed in this OSRA study. Furthermore, larger spills are likely to be identified and reported; therefore, these records are more comprehensive than those of smaller spills. For potential oil spills exceeding 1 million bbl, Ji et al. (2014) used the extreme value theory to analyze the probability of catastrophic spills.

Anderson et al. (2012) analyzed platform and pipeline spills in Federal waters from 1964 through 2010 and crude oil tanker spills that occurred in U.S. waters from 1974 through 2008. Spill rates (in spills per Bbbl of oil) were estimated for platforms, pipelines, and tankers on the OCS, as shown in Table 2.

Oil spill occurrence estimates for spills greater than or equal to 1,000 bbl were calculated for production and transportation of oil during the 50-year analysis period associated with the proposed action in the OCS Program (2017–2066). These probabilities are based on the volume of oil estimated to be found, produced, and transported over the life of a typical lease sale and on the rates that have been calculated for oil spills from OCS platforms, pipelines, and tankers by Anderson et al. (2012). The probabilities of one or more oil spills greater than or equal to 1,000 bbl occurring as a result of OCS exploration, development, and production and transportation resulting from a typical lease sale or resulting from the OCS Program are found in Table 3.

At distance intervals of one-tenth of a degree (1/10°) of latitude (about 11 kilometers) and intervals of 1/10° of longitude (about 10 kilometers), 6045 launch points for hypothetical spills were identified in the study area. The spatial resolution of the spill simulations (1/10° north-south and 1/10° east-west) was selected to reflect the spatial resolution of the input data. We applied the OSRA model to estimate the trajectories of one spill per launch point per day for 15 years of available wind and current data in the GOM from 1993 to 2007. In total, over 40 million hypothetical oil spills were modeled.

The OSRA model was designed to track the movements of potential oil spills before they happen. The model was originally developed by Smith et al. (1982) and later enhanced by BOEM over the years (LaBelle et al. 1985; Ji et al. 2002, 2011, 2014). The model calculates the movement of hypothetical oil spills by successively integrating the time sequences of two spatially-gridded input fields: surface ocean currents and sea-level winds. With this method, the OSRA model generates time sequences of hypothetical oil spill locations—essentially, oil spill trajectories.

Conducting OSRA requires detailed information on ocean currents and wind fields (Ji 2004). Ocean current inputs were numerically computed from an ocean circulation model of the GOM based on meteorological forces (near-surface winds and total heat fluxes) and observed river inflow into the GOM (Oey 2005). For surface wind data, the OSRA model incorporated concurrent wind fields (6-hourly surface wind speed and direction data analyzed by the European Center for Medium Range Weather Forecasting [ECMWF]). The OSRA model used the same wind field data to calculate the empirical wind drift of the simulated spills.

The OSRA model also tabulates the simulated oil spill contacts to specific locations or resources. The model uses geographical boundaries of a variety of identified resources, such as shoreline (counties/parishes), offshore topographical features, and Habitat Areas of Particular Concern (HAPC). At each successive time step, the OSRA model compares the location of the hypothetical spills against the geographic boundaries of shoreline and designated offshore resources. The OSRA model then counts the number of “contacts,” which is comprised of the number of times the oil spill contacts the shoreline plus the number of times the oil spill contacts an offshore resource during a time when it is known to be in use.

Each modeled trajectory was allowed to continue for as long as 30 days. However, if the hypothetical spill contacted shoreline sooner than 30 days after the start of the spill, the spill trajectory was terminated, and the contact was recorded. A contact to an offshore resource that is not a shoreline (such as a wildlife refuge area in the middle of the ocean) did not stop the calculation of the trajectory.

The probability that a hypothetical oil spill will contact a specific resource within a given time of travel from a certain location or spill point is termed a “conditional probability.” A critical difference exists between conditional probabilities and combined probabilities (Smith et al 1982; Ji et al. 2002). Conditional probabilities depend only on the winds and currents in the study area. Combined probabilities, on the other hand, depend not only on the winds and currents, but also on the chance of spill occurrence, the estimated volume of oil to be produced or transported, and the oil transportation scenario.

More than 300 onshore resources and offshore resources were considered in this study. Detailed combined probabilities are presented in statistical tables (Ji et al. 2016). As one might expect, resource locations closest to the spill sites have the greatest risk of contact. Also, as the model run duration increases, more of the identified resources and shoreline segments could have meaningful probabilities of contact (≥0.5%). The longer transit times allowed by the model (up to 30 days) enabled more hypothetical spills to reach the resources and the shoreline from more distant spill locations. With increased travel time, the complex patterns of wind and ocean currents produced eddy-like motions of the oil spills and multiple opportunities for a spill to make contact with any given resource or shoreline segment.

For example, Table 4 gives the combined probabilities of the East Flower Garden Bank and the West Flower Garden Bank, which are the “crown jewels” of the GOM (Figure 2). Table 4 provides the probabilities (expressed as percent chance) of one or more offshore spills greater than or equal to 1,000 bbl and the number of spills (mean) that could occur and could contact offshore resources within 10 days and within 30 days given the estimated volume of oil produced from a proposed action for the Alternative A leasing scenario (Table 3). This table shows that the East Flower Garden Bank, which is illustrated on Figure 2, has a probability of less than 0.5% of being contacted by spilled oil within 10 days, if there is an oil spill from a proposed action with an estimated oil production volume of 0.210 Bbbl (Table 1). Its probability of being contacted by the spilled oil within 30 days increases to 1%.

Using historical current and wind data, the OSRA model was also applied to analyze the contact probabilities of the Ixtoc I Oil Spill, which happened on June 3, 1979 in the Bay of Campeche of the GOM and lasted for 10 months. The Ixtoc I Oil Well suffered a blowout, resulting in one of the largest oil spills in history and 3 million barrels of oil spilled. There are very limited numerical simulations on Ixtoc oil spill using ocean current and wind data (e.g., Hooper 1982).

This analysis used the historical GOM wind and current data from 1993 to 2007—the same data discussed above—to determine if the data would produce a pattern of probabilities similar to what was observed from the Ixtoc spill. Special OSRA runs were conducted in order to estimate the probability results of spills occurring at the Ixtoc I oil well (lat 19°24′N, long 92°13′W). Trajectories of hypothetical spills were initiated from the launch point every 1.0 day over the 15-year simulation period. Each trajectory was allowed to continue for as long as 30 days. However, if the hypothetical spill contacted the shoreline after the start of the spill, the spill trajectory was terminated, and the contact was recorded. The simulation period was from 1 January 1993 to 31 December 2007, rather than 3 June 1979 to 23 March 1980 (the actual dates of the spill).

The patterns of the grid cell probabilities of oil location for the 15-year average are shown in the upper panel of Figure 3. The color bar shows the probability of contact in percentage. For instance, the red color area is near the Ixtoc I well location. On average, this area has more than 11% probability of contact with the oil spilled at the launch point within 30 days. The area of the figure indicating 0.5% or higher probability of contact is to the west of the release point and crosses the Mexico-Texas border at lat 25°N.

The lower panel of Figure 3 depicts estimated oil spill travel time averaged over the 15 years and shows that oil from the Ixtoc well could reach the Mexico-Texas border (at lat 25°N) within 30 days. Both the probability of contact and the probability of travel time in Figure 3 have patterns similar to the ones described by Hooper (1982).

The winds and currents in the GOM exhibit strong seasonal variability. Consequently, the oil spill trajectories also exhibit strong seasonal variability. For the purpose of comparison, the patterns of the grid cell probabilities of oil location for the four seasons are also analyzed. It is found that during the four seasons, the oil spilled is predominantly transported in this order: westward, northward, along the coastline, and then into U.S. water after 30+ days. The fall season (July, August, and September) showed the highest probability of oil entering the U.S. water as the result of fall current and wind patterns. These model results are also consistent with the observations described by Hooper (1982).

Because the OSRA model does not take into account response actions, such as oil skimming, burning, or use of dispersants or oil-absorbent booms, the affected area in OSRA model results should be larger than the actual oiled area if response efforts were effective. In the case of the Ixtoc spill, a major response effort was launched, limiting the size of the oiled area. Comparison between OSRA model results and the maximum extent of the oiled area did show that the OSRA model can provide a reasonable representation of the area affected by an oil spill for which no response activities were conducted.

BOEM uses the OSRA model (Smith et al. 1982; Ji et al. 2002) for the analysis of possible oil spill contact from offshore oil and gas operations. The OSRA model provides a useful tool for estimating the oil spill trajectories, which is used for analyzing the environmental impacts of potential spills.

The OSRA model was applied to estimate the risk patterns due to projected oil and gas development in the GOM. Using the currents from the ocean model and the wind data from the ECMWF, the OSRA model simulated hypothetical oil spill trajectories in the GOM. The simulation period was 15 years, from 1 January 1993 to 31 December 2007. The OSRA model results exhibited seasonal variability during the simulation period.

The model results for the 1979 Ixtoc oil spill simulation using the same historical current and wind data from 1993 through 2007 were consistent with the descriptions of the actual oil spill trajectories by Hooper (1982) even when considering factors such as seasonal variability and response actions.