Results of the AICc-based multimodel inference procedure examining the variations in habitat preference relative to the independent variables.

The AICc-weighted relative importance (w+), the model-averaged estimate (β) and their unconditional standard error (S.E.) are reported for each parameter (main effects and interactions), as well as their respective prevalence in the candidate models (ν). The ν and w+ values range between 0 and 100%. The parameter estimates refer to the level indicated between brackets as a baseline. The interpretation of each effect is provided in case of AICc-based support.Response variable: Arrival date of males (days) = arrival dates of males in springtime, Male wing length (mm) = outermost primary wing length in males, Male mask size (mm2) = size of black eye-stripes in males.Fixed effects: F/W = farmland (F) versus woodland (W) sites, Year = 2008, 2009 or 2010 and Study area = 1 or 2, Aggregation = number of males settled within a radius of 500 metres.Random effects: territory site identity.