Résumé
Overall mortality rates often are based upon a variety of mortality sources such as predation, disease, and accidents, and each of these sources may influence population dynamics differently. To better understand population dynamics or to derive effective conservation plans, it is thus crucial to know the frequency of specific mortality causes as well as their variation over time. However, although the mortality cause of retrieved marked animals is often known, this information cannot be used directly to estimate the frequency of a mortality cause. By calculating the ratio of the number of animals reported dead from a specific cause to the total number of retrieved animals, one does not consider the fact that the probability of finding a dead individual depends on the cause of its death. Although frequently used, such ad hoc estimates can be heavily biased. Here we present a new way of estimating the frequency of a mortality cause from ring-recovery (band-recovery) data without bias. We consider the states ''alive,'' ''dead because of mortality cause A,'' and ''dead due to all other causes'' and estimate within a multistate capture-recapture framework the transition probabilities as well as the state-specific resighting probabilities. Among the transition probabilities are the overall survival probability and the proportion of animals dying because of A. From these, the probability that an animal dies during a year due to the specific cause of interest (cause A kill rate) can easily be calculated. We illustrate this model using data from White Storks Ciconia ciconia ringed in Switzerland to estimate the proportion of storks that died due to power line collision. Average unbiased estimates of this proportion were 0.37 0.08 (mean 1 SE) for juveniles, about 25% lower than ad hoc estimates, and 0.35 0.09 for adults. The annual survival rate of juveniles was 0.33 0.05 and of adults, 0.83 0.02. Power line mortality is thus important for White Storks, with about one in four juveniles and one in 17 adults dying each year because of power line collision. We discuss advantages and disadvantages of the new model and how the results could be used to explore the link between a specific mortality cause and population dynamics.