Does matrix resistance influence Red squirrel (Sciurus vulgaris L. 1758) distribution in an urban landscape?

In determining isolation effects in fragmented populations, the landscape matrix is not often considered. Usually simple distance measures are used to quantify degree of isolation. We tested the effect of the matrix on the presence of red squirrels in 354 wooded patches in the Brussels Region, by comparing several isolation measures. These were 1) distance to the nearest source patch, 2) the Hanski-measure (a combination of distance to and size of all possible sources), 3) effective distances calculated from different least cost models using the ArcView grid extension 'Cost Distance' (a combination of distance and resistance of the landscape, with different resistances for different landscape types) and 4) some combinations of the Hanski-measure and the effective distances. Size and quality of the target patches were always included in the tests of the predictive power of different isolation measures on squirrel presence/absence. All variables examined (patch size, quality and isolation) significantly influenced squirrel presence. Models using the effective distances gave the best results. Models including the Hanski-measure improved significantly when Euclidean distance was replaced by effective distance, showing that parameterisation of matrix resistance added significant additional explanatory power when modelling squirrel presence.