Quantificação de um critério de conservação: raridade de aves de terra firme em uma floresta na Amazônia Central

Abstract

The distinction between rare and common species is fundamental in research and management decisions. Nevertheless the definition of these categories often relies on a mix of sampling methods, biological processes and subjective impressions. Most studies that used field sampling techniques to count landbirds failed to present measures of precision and do not account to imperfect detectability, in that way this counts are based in a fraction of true population. We tested if classifications of bird rarity based on ornithologist s perceptions are the same when we estimated occupancy accounting for imperfect detection. We selected ten species pairs whose members are related phylogenetically and one member is reported to be rarer than the other. On three different occasions 760 ha of primary terra-firme forest was sampled by 10 observers with point counts. Playback and autonomous recorder was used as complementary approach. All observers were trained and evaluated using memory-improvement software. It wasn t possible refute the hypothesis that rarer species occur in higher proportion of places as common species because only five pairs showed a clear difference in rarity and all were in the direction predicted by existing literature. Only two of these, Dendrocolaptes picumnus and Cyphorhinus arada, are genuinely rarer species and occupy less than 50% of sampling points. Although published rarity classifications never mistook a genuinely rarer species to be more common than its pair,ours results gather clear evidence that distinction between rare and common species must be based in quantitative measures that account for detection probabilities and that presents measures of precision. Because species rarity figures so prominently in landbird conservation, it is essential that researches and managers attempt to provide reliable information. Thus, in research and management that depend on correct rarity classification, we recommend the use of quantitative methods with calculated confidence intervals and that dealing with imperfect detections.