Detecting detectability: identifying and correcting bias in binary wildlife surveys demonstrates their potential impact on conservation assessments

The European Commission Habitats Directive requires that changes in the conservation status of designated species are monitored. Nocturnal and elusive species are difficult to count directly and thus population trajectories are inferred by variation in the incidence of field signs. Presence/absence techniques are, however, vulnerable to Type II errors (false negatives). The Eurasian otter (Lutra lutra), listed by the IUCN as ‘near threatened’, is monitored throughout Europe using the ‘Standard Otter Survey’ method. We explored the reliability of this approach by analysing species incidence at 1,229 sites throughout Ireland. Naïve species incidence was 72 % 95 % confidence interval (CI), 69–75 %] with variation affected significantly by survey team and, at running freshwater sites, the number of bridges present and rainfall during the month, and most notably during the 7 days, prior to survey. Rainfall had no effect on static freshwater sites or the coast. Marginal estimated mean species incidence derived from a GLM assuming the β coefficient of the survey team associated with the highest prevalence, no rainfall in the week prior to survey and sites that had multiple bridges, was 94 % 95 %CI 78–97 %]. We demonstrate that bias and error in binary wildlife surveys can have a major impact on a conservation assessment even when conducted on an apparently well-known species in a developed country with good infrastructure and a long history of similar ecological studies. Our results provide empirical evidence for further criticisms of the Standard Otter Survey method calling into question its value in monitoring changes in otter populations throughout Europe.