The purpose of this study was to design and test a monitoring protocol for marine waterfowl in the central Alaskan Beaufort Sea. The study provides an important case-study of how a long-term monitoring program may be affected by unanticipated human disturbances. Because of its overwhelming and widespread abundance, relatively sedentary behavior, ease in counting, and the extensive historical database, the long-tailed duck (Clangula hyemalis) was selected as the focal species. Two null hypotheses were formulated concerning potential changes in the numbers and distribution of long-tailed ducks in relation to disturbance in an industrial study area, compared to a reference study area located about 50 km to the east. A 9-year historical database (1977-1984, 1989) of long-tailed duck densities and other important data recorded during systematic aerial surveys was analyzed retrospectively using multiple regression techniques. The retrospective analyses determined which of several predictor variables recorded were significantly related to long-tailed duck density. Separate analyses were conducted for two periods: (1) the overall period when long-tailed ducks were present in the lagoon study areas, and (2) the shorter adult male molt period. The results of the two analyses indicated that 57% and 68%, respectively, of the total variation in long-tailed duck density during the two periods could be explained by variables recorded during the surveys. Predictor variables representing habitat, day of the year, time of day, amount of ice, and wave height recorded on-transect during surveys were most closely associated with long-tailed duck density. Measurement error during the surveys, and influences outside the study area such as nesting success in tundra habitats and mortality during migration and in over-wintering areas likely also had strong influences on the results, but these factors were not measurable in our study. Based on results of the retrospective analyses, a long-term monitoring protocol consisting of a program of systematic aerial surveys and an analyses of variance and covariance (ANOVA and ANCOVA) statistical procedure was designed and initially tested in 1990 and 1991. This 2-year testing phase resulted in several revisions to the monitoring protocol. Refinements were made to the original sampling procedures, to the survey schedule, and to the recommended statistical analysis procedures. Results of the ANOVA and ANCOVA indicated that there was no evidence of a change in long-tailed duck densities that could be attributable to disturbance (from any source) in the industrial study area relative to a reference area with no industrial development. Other analyses indicated that the sampling and analysis procedures would be adequate to detect long-term trends in long-tailed duck density and localized disturbance effects, but that the monitoring program should be continued well beyond two years to detect statistically significant changes. As a result, additional aerial surveys of both study areas were conducted again during 1999-2001. Results of the revised ANOVA and ANCOVA of the 1990-1991 and 1999-2001 survey data indicated that the density of long-tailed ducks had significantly declined in coastal lagoons along the central Alaskan Beaufort Sea coast during the study period. In addition, disturbances throughout the barrier island-lagoon systems used by these ducks, including both the industrial and the reference study areas, had significantly increased over the same period. However, because unanticipated disturbances from a variety of anthropogenic sources, and not just industry sources, increased in both study areas, the reference study area was not an effective statistical control. As a result, the decline in long-tailed duck density in both study areas was not attributable to industry-related activities. Although the monitoring protocol described here is an effective method to detect statistically significant changes in long-tailed duck distribution and abundance in the nearshore Alaskan Beaufort Sea, many more years of sampling would be necessary to attribute observed changes to industry-related disturbances.
Read full abstract