Biodiversiry monitoring is important to identify biological units in need of conservation and to check the effectiveness of conservation actions. Programs generally monitor species richness and its changes (trend). Usually, no correction is made for imperfect species detectability. Instead, it is assumed that each species present has the same probability of being recorded and that there is no difference in this detectability across space and time, e.g. among observers and habitats. Consequently, species richness is determined by enumeration as the sum of species recorded. In Switzerland, the federal government has recently launched a comprehensive program that aims at detecting changes in biodiversity at all levels of biological integration. Birds are an important part of that program. Since 1999, 23 visits per breeding season are made to each of >250 1 km2 squares to map the territories of all detected breeding bird species. Here, we analyse data from three squares to illustrate the use of capture-recapture models in monitoring to obtain detectability-corrected estimates of species richness and trend. Species detectability averaged only 85%. Hence an estimated 15% of species present remained overlooked even after three visits. Within a square, changes in detectability for different years were of the same magnitude when surveys were conducted by the same observer as when they were by different observers. Estimates of trend were usually biased and community turnover was overestimated when based on enumeration. Here we use bird data as an illustration of methods. However, species detectability for any taxon is unlikely ever to be perfect or even constant across categories to be compared. Therefore, monitoring programs should correct for species detectability.
