Bias in ring-recovery studies : causes of mortality of little owls Athene noctua and implications for population assessment

Recoveries of marked animals hold long‐term, large‐scale information on survival and causes of mortality, but are prone to bias towards dead recoveries and casualties in the range of presence of potential finders. Thus, accounting for circumstance‐related recovery probabilities is crucial in statistical approaches. For the little owl, a species of conservation concern in central Europe, raw ring recoveries suggested a strong human‐related impact on survival. We analysed the proportions of the main causes of death using a large sample of radio‐tracked birds as a reference. We compared ring recoveries in southern Germany collected 1950–2012 (n = 465 dead recoveries of 2007 recoveries of 30 623 ringed birds) with data from a radio‐tracking study in the same region 2009–2012 (n = 177 dead recoveries of 377 tagged individuals). Two assumptions of multi‐state ring recovery modelling were unrealistic. First, not all dispatched rings remained available to potential finders. Instead, 34% of tracked birds were displaced to sites where rings were irretrievable, resulting in biased estimates of recovery probability. Second, the proportions of irretrievable rings were disproportional, with 48% in predated birds and 5% in human‐induced mortality. Consequently, the sample of rings from which recoveries were drawn differed from the sample of dispatched rings. Accounting for these biases in a multi‐state model, we estimated the frequencies of main causes of mortality to 45% for predation, 20% for casualties due to traffic and at buildings and 34% for all other causes. In radio‐tracked birds, predation was even more dominant (76%). Integrating mark–recapture data and telemetry observations allowed detecting and quantifying so far unknown recovery bias and resulted in improved estimates of key population parameters. The demography of little owls likely depends mainly on predator–prey relationships rather than on human‐induced deaths.publishe