Once-yearly sampling for the detection of trends in biodiversity: The case of Willow Slough, California

The butterfly fauna at Willow Slough, Yolo County, California has been censused for 32 years as part of a participatory citizen-science project, the Fourth of July Butterfly Count. While the utility of a once-a-year census as a monitoring tool is potentially compromised by lack of standardization in counting protocols and variation in observer skill, at Willow Slough these issues have been minimized.We examined the Willow Slough count data for trends in both faunal diversity and the probability of presence of individual species. During the study, the number of species observed at a visit declined by 39%. Regressions of per-visit species counts against time did not detect a statistically significant decline until year 24. In contrast, Fisher’s α, a statistic designed to reduce sample-size bias, detected the decline as early as year 13. Twelve of the 24 species analyzed showed significant declines in probability of occurrence; a further nine exhibited negative but non-significant trends. Butterflies that overwinter as eggs or larvae were more likely to decline than those that overwinter as pupae or adults. Many species in decline at Willow Slough have also been observed less frequently at nearby sites which are monitored year-round, supporting the value of once-a-year monitoring. Although correlations with climatic data have been identified, they are too weak to account for the observed faunal decline. We suspect broader patterns of land use and habitat continuity are implicated in butterfly declines across the region.We conclude that once-a-year sampling, if properly and rigorously done, is in fact useful as a monitoring tool for butterfly faunas, and that Fisher’s α is well suited to early detection of trends in repeated diversity sampling.

► We examine trends in diversity in a 32-year record of a California butterfly fauna.
► We document an unexplained 39% decline in the per-visit species count at the site.
► Twelve of the 24 species showed significant declines in probability of occurrence.
► Fisher’s α detected a decline by year 13, 10 years before it was apparent from species counts.
► We recommend α’s application to similar data collected by citizen-scientists.