Landscape, cropping and field boundary influences on bird abundance

Given likely dramatic land use changes in the UK, there is a need to predict their impact upon biodiversity. Such predictions require an understanding of the habitat factors that determine species' abundances. We relate British farmland bird abundance at the 1 km square scale to three variable sets: landscape composition (for example, areas of woodland, urban and arable), field boundary characteristics and cropping. We then use multi-model inference to summarize model fits for each variable set and compare their explanatory power to identify the relative importance of the three forms of description of the farm landscape. The important variables across species within each set show the habitat features that are most significant for the farmland bird community. Models comprising landscape variables provided the best fits for 21 of 31 bird species, while field boundary variables were most important for seven species and cropping variables for one. After controlling for landscape influences, we then found that field boundary variation was slightly more influential than cropping. These patterns remained when farmland-specialist species were considered alone. This shows that landscapes are the most important determinant of bird abundance and that cropping patterns and field boundary characteristics are influential only within the constraints of landscape structure. However, boundaries and, especially, crops remain important from a management perspective because they are more volatile, and can still drive population change. Important single variables that affected multiple species included woodland (predominantly negative even within farmland-dominated landscapes), hedges with trees (providing nesting habitat and song-posts) and heterogeneity both of landscapes and of cropping (predominantly positive effects). The results indicate which habitat features offer the most scope for induced population changes via management action and potentially provide relationships for input into predictive models with respect to land-use change