Probability Sampling and Estimation of the Oil Remaining in 2001
from the Exxon Valdez Oil Spill in Prince William Sound, Alaska

Abstract

Disputes concerning the amount of spill oil remaining in Prince William Sound, Alaska, from the Exxon Valdez grounding in 1989 required revisiting the region in 2001. In contrast to earlier surveys which were based on purposeful selection of sampling locations, probability sampling was applied in order that unbiased estimates with measures of their precision could be computed. Beach segments and subsegments were stratified by their oiling histories and lengths, and random samples were selected for a visit. At each beach visited, the surface was grided into tidal elevation intervals and perpendicular columns, and every intersecting block was further subdivided into quadrats (0.25 m2 in area) for random and adaptive sampling. Adaptive sampling pursued oil found in initial random quadrats in order to delimit entire patches. Subsurface oiled sediments were classified to a visual scale, and oil present in selected quadrats was extracted and weighed in a calibration study. The surface and subsurface oiled areas of the sediments at the visited beaches were estimated, together with the weight of oil in their subsurface sediments from the calibrated visual scale. Conservative estimates of oiled areas and weights for the visited beaches included only oil seen at random and adaptive quadrats, but unbiased estimates were computed by expansion for quadrats not sampled. The estimates at visited beaches were expanded for unsampled beach segments of strata, and summed for the total in Prince William Sound. Precision was evaluated by analytical formulas as well as by bootstrap resampling. Unbiased estimation of oiled areas and weights from oil found at the random and adaptive quadrats with precision evaluated by bootstrap resampling determined that until 2001, Prince William Sound still had a total of 41,000 m2 surface oiled area (95% interval, 20,700 – 70,500 m2), and a total of 71,000 m2 subsurface oiled area (95% interval, 37,700 – 113,200 m2), having subsurface oil weighing 50,000 kilograms (50 metric tons (t)) (95% interval, 24.4 - 82.6 t). Unbiased estimates based on the random quadrats only were 78,000 m2 for subsurface oiled area (95% interval, 40,600 – 127,300 m2), and 56 t (95% interval, 26.1 - 94.4 t) for subsurface oil mass, agreeing well with the estimates based on combined random and adaptive quadrats when considered in light of sampling error.


View Online  (.pdf, 1.36MB).