PARENT SESSION
Contributed Oral Session 8: Invertebrate Ecology: Butterflies; Soil Insects
Monday, August 8, 8:00 AM - 11:30 AM, Meeting Room 518 C, Level 5, Palais des congrès de Montréal
Detection of long-term changes in an alpine butterfly community using non-parametric bootstrap methods.
O'Brien, Joshua*,1, Forister, Matthew2, Thorne, Jim1, Shapiro, Arthur, 1 University of California, Davis, Davis, California2 State University of New York at Stony Brook, Stony Brook, New York
ABSTRACT- Long-term ecological datasets can provide invaluable insights into community dynamics, but are often characterized by variation in the timing and intensity of sampling. These inconsistencies make analysis with traditional methods difficult or impossible. We present a method for inferring trends in species abundance over time in datasets with irregular or complicated sampling structure. The method relies on non-parametric bootstrap resampling of the data under a null hypothesis of no trend in abundance with time. To illustrate the use of the method, we apply it to data from a 28-year study of a high-elevation, Sierra Nevadan butterfly community. One of us (AMS) has, for the past 28 years, monitored butterfly community composition several times each year at Castle Peak (elev. 2780 m) in the central Sierra Nevada. The number and timing of visits has varied considerably, with from three to ten visits per year occurring between late May and the beginning of November, and sampling frequency has increased over the years. To test the hypothesis of no trend in a species' abundance with time, we construct a null model that incorporates sampling structure, but in which the probability of observing the species during any one part of the year does not change over time. Using the actual sequence of site visits as a template, we create 1000 bootstrap replicates (28-year sequences of presences and absences) in which the probability of observing a species at a visit is given by its average frequency during that week over the full course of the study. The slope of the regression of number of presences per year on year calculated from the actual data is compared to the distribution of slopes generated from the simulated data to produce a significance value for the observed slope. Our analyses indicate that at Castle Peak 18 of 74 species have increased significantly in abundance (two-sided p<0.05), while just one species has decreased over the past 28 years. Of the remaining species, significantly more show evidence of increasing in abundance than would be expected by chance. We discuss the relevance of these increases in light of predicted up-slope movement of butterfly populations in response to climatic warming.
Key words: butterfly, community dynamics, non-parametric bootstrap