Coupling in goshawk and grouse population dynamics in Finland

Abstract

Different prey species can vary in their significance to a particular predator. In the simplest case, the total available density or biomass of a guild of several prey species might be most relevant to the predator, but behavioural and ecological traits of different prey species can alter the picture. We studied the population dynamics of a predator-prey setting in Finland by fitting first-order log-linear vector autoregressive models to long-term count data from active breeding sites of the northern goshawk (Accipiter gentilis; 1986-2009), and to three of its main prey species (1983-2010): hazel grouse (Bonasa bonasia), black grouse (Tetrao tetrix) and capercaillie (T. urogallus), which belong to the same forest grouse guild and show synchronous fluctuations. Our focus was on modelling the relative significance of prey species and estimating the tightness of predator-prey coupling in order to explain the observed population dynamics, simultaneously accounting for effects of density dependence, winter severity and spatial correlation. We established nine competing candidate models, where different combinations of grouse species affect goshawk dynamics with lags of 1-3 years. Effects of goshawk on grouse were investigated using one model for each grouse species. The most parsimonious model for goshawk indicated separate density effects of hazel grouse and black grouse, and different effects with lags of 1 and 3 years. Capercaillie showed no effects on goshawk populations, while the effect of goshawk on grouse was clearly negative only in capercaillie. Winter severity had significant adverse effects on goshawk and hazel grouse populations. In combination, large-scale goshawk-grouse population dynamics are coupled, but there are no clear mutual effects for any of the individual guild members. In a broader context, our study suggests that pooling data on closely related, synchronously fluctuating prey species can result in the loss of relevant information, rather than increased model parsimony.