The race is not to the swift: long-term data reveal pervasive declines in California's low-elevation butterfly fauna

Understanding the ecology of extinction is one of the primary challenges facing ecologists in the 21st century. Much of our current understanding of extinction, particularly for invertebrates, comes from studies with large geographic coverage but less temporal resolution, such as comparisons between historical collection records and contemporary surveys for geographic regions or political entities. We present a complementary approach involving a data set that is geographically restricted but temporally intensive: we focus on three sites in the Central Valley of California, and utilize 35 years of biweekly (every two weeks) surveys at our most long-sampled site. Previous analyses of these data revealed declines in richness over recent decades. Here, we take a more detailed approach to investigate the mode of decline for this fauna. We ask if all species are in decline, or only a subset. We also investigate traits commonly found to be predictors of extinction risk in other studies, such as body size, diet breadth, habitat association, and geographic range. We find that population declines are ubiquitous: the majority of species at our three focal sites (but not at a nearby site at higher elevation) are characterized by reductions in the fraction of days that they are observed per year. These declines are not readily predicted by ecological traits, with the possible exception of ruderal/non-ruderal status. Ruderal species, in slightly less precipitous decline than non-ruderal taxa, are more dispersive and more likely to be associated with disturbed habitats and exotic hosts. We conclude that population declines and extirpation, particularly in regions severely and recently impacted by anthropogenic alteration, might not be as predictable as has been suggested by other studies on the ecology of extinction.