CHANGING HABITAT AND HABITAT USE BY BIRDS AFTER THE

Evaluations of the ecological consequences of environmental accidents can benefit from a long-term perspective. To assess the effects of the Exxon Valdez oil spill on habitat use and occupancy during midsummer by marine-oriented birds over a 12-yr period following the spill in 1989, we conducted surveys in 10 study bays in Prince William Sound, Alaska, USA. These bays varied from completely unoiled to some of the most heavily oiled locations in the spill zone. We used oiling as a quantitative variable with and without habitat measures as covariates to assess spill effects and their changes over time. We also used oiling as a categorical variable ("oiled" vs. "unoiled" bays) to conduct between-year, repeated-measures analyses that used 1984 data as a baseline and to plot abundance trends by oiling category. We combined the results of these analyses in a "weight-of-evidence" approach to determine overall impacts and recovery for each species. To assess changes in habitat condition following the oil spill, we also measured several habitat features in 1991, 1998, and 2001. There were significant increases in the proportion of shoreline covered by rockweed (Fucus) and mussels in oiled bays between 1991 and 1998-2001, suggesting recovery of oiled habitats. Of the 25 bird species whose status we evaluated, 12 (48%) exhibited no evidence of any spill impacts, 10 (40%) exhibited negative impacts that subsequently disappeared, and three (12%) exhibited positive relationships with oiling that later disap- peared. No species provided clear evidence of any delayed impacts. In comparison with our earlier analysis of a more extensive data set collected from these same study bays during 1989-1991, we detected slightly more initial spill impacts in the present study but also found that several species whose habitat occupancy had not recovered by the end of our earlier study had now recovered. Our conclusions differ from those of some other studies of the effects of the Exxon Valdez oil spill on birds, perhaps because of differences in study design, in the criteria used to assess impact and recovery, and in the definition of "recovery." These differences illustrate the difficulties of evaluating the consequences of environmental accidents, even in such a dramatic and well-studied situation as the Exxon Valdez oil spill. Long-term studies of environmental accidents create a paradox: the long-term perspec- tive is needed to see how impact and recovery dynamics are played out, yet background natural variation can make it increasingly difficult to detect clear patterns. With the passage of time after an environmental accident, more things happen, making it more difficult to attribute observed patterns to the event alone. Consequently, long-term assessments of ecological impacts may be restricted to documenting only large effects; the use of ox levels greater than the traditional 0.05 can facilitate such analyses and reduce the likelihood of committing type II errors.