Habitat-density associations are not geographically transferable in Swiss farmland birds


M. Schaub (michael.schaub@vogelwarte.ch), M. Kéry, S. Birrer, M. Rudin and L. Jenni, Swiss Ornithological Inst., CH-6204 Sempach, Switzerland. Present address of MS: Conservation Biology, Inst. of Ecology and Evolution, Univ. of Bern, Baltzerstrasse 6, CH-3012 Bern, Switzerland.


Knowledge about associations between habitat and population density is essential for ecological studies and management alike, for instance, to identify high-density areas worthy of protection. Furthermore, habitat-density associations are important for setting targets, e.g. in agri-environmental schemes. However, habitat associations need to be constant across management units, which may be entire countries; otherwise management should be targeted at smaller, internally homogenous regions. We use data on 13 Swiss farmland bird species from 128 study sites to test for the generality of the association between territory density and six key habitat descriptors: the density of hedges and woodland edges, and the proportion of meadows, pastures, crops and fallow land at each study site. We modelled observed territory counts in overdispersed Poisson regressions, with and without zero-inflation, and including the interactions between four geographical regions and six habitat descriptors. To test for spatial transferability of habitat-density associations, we also fitted the corresponding main-effects models. As a continuous alternative to discrete regions, we fitted models with altitude and its interactions with habitat, to test for the generality of habitat-density associations across the 1885 m altitudinal gradient in our study (365–2250 m a.s.l.). Finally, we tested for nonlinearity as a possible reason for geographical non-transferability of habitat-density associations. The interactions between either region or altitude and habitat-density associations were important in 10 of 13 species, showing that habitat-density associations were not transferable. Discrete regions were the best descriptor of the spatial variation in habitat-density associations in seven species, and altitude in six. Simple overdispersed Poisson regressions were superior to zero-inflated models for 12 of 13 species. For 6 species nonlinearity was a sufficient explanation for the observed non-transferability. Non-transferability of habitat-density relationships might stem from biological reasons or from methodological inconsistencies. To distinguish between them, to suggest better management actions and to improve species habitat models, focused habitat-density studies ought to be replicated at various sites. Our study emphasizes the need for spatial stratification of conservation measures (e.g. agri-environmental schemes), otherwise, they may be inefficient or even harmful. Ideally, the spatial extent of internally homogenous regions ought to be investigated for efficient conservation management.