Tests of density dependence using indices of relative abundance in a deer population

A major question in animal ecology is explaining the causes of population fluctuations. Consensus about the most reliable method to detect density dependence (DD) or environmental effects from time-series data, however, has not yet been achieved. Times series analyses have been used with indices of relative abundance in numerous studies, although these indices are rarely validated. Here, we used three different time series of relative abundance (number of deer seen per hunter per day, hunting success and proportion of males in the harvest) to explore direct and delayed DD in a white-tailed deer Odocoileus virginianus population on Anticosti Island, Quebec, Canada. Three mathematical approaches were tested: linear models, autoregressive (AR) models, and total DD in life history. Tests of DD using different indices of abundance on the same population should lead to similar results if all indices exhibit similar behaviour. Indices of relative abundance correlated with each other, although sometimes weakly, such that we obtained similar DD estimates with each index using detrended non-stationary series. In most time series, linear regression of Nt−1 and Nt and AR models did not detect DD, while we obtained strong evidence for DD from the life-history approach. This meant that contrasting conclusions about the role of density dependence within this population were reached depending on which method was used. We conclude that the method that incorporates most biological realism, the life-history approach, provided a different result than classical interpretation of autoregressive coefficients. Only the life-history interpretation supported our a priori belief that density dependence operating through competition for food regulates the Anticosti deer population. Phenomenological analysis aiming to investigate changes in abundance should be carefully conducted as the use of inappropriate indices or methods could lead to inappropriate conclusions or management strategies. Preferably, the method used should match the time scale of the population sampling regime and species life history.