ASSESS INTERTIDAL IMPACTS: LESSONS LEARNED FROM 11 YEARS OF MONITORING IN PRINCE WILLIAM SOUND1

Abstract

ABSTRACT “How many samples do we need?” and “Where should we collect them?” are questions common to most field monitoring programs. They are especially compelling when an oil spill impinges on a coastline and an intertidal monitoring program must be rapidly implemented to assess impacts. Moreover, these initial sampling decisions profoundly affect the ultimate validity of the assessment study. While it is important to avoid squandering limited resources by oversampling, an undersized study is equally wasteful if it cannot reliably discern significant impacts and produce useful results. All else being equal, the power to detect change in an ecosystem is dictated by its inherent variability. Variability estimates were computed for a variety of intertidal assemblages using data collected within Prince William Sound over eleven years, following the Exxon Valdez oil spill. The sample-size recommendations that emerge from these power analyses demonstrate that spill and cleanup impacts to most intertidal taxa can be reliably detected using four to six replicate samples collected at three or more treatment sites, and at an equivalent number of reference sites. They also demonstrate that collecting more than eight replicate samples within each site does little to increase the power to detect differences in mean populations. Analogously, “parallelism” tests on multi-year data provide a reliable means of quantifying long-term recovery in intertidal populations following an oil spill. Parallelism tests examine temporal trends in mean abundance at impacted sites relative to reference sites. However, the ability to detect major recolonization events with parallelism tests varies widely depending on the taxa being tested. For example, the marked recolonization in Fucus and epifaunal invertebrates that was experienced four years after the Exxon Valdez spill could have been detected by sampling at only two reference and two impact sites. In contrast, infaunal populations exhibit greater variability, making detection of nonparallel trends difficult without sampling at least six reference and six impacted sites.