Figures & data
Figure 1. Illustration of the neighbourhood scale over which territory density was calculated as a dependent variable. The number of territories over a neighbourhood areas (dotted circles) was calculated, which was determined by averaged nearest neighbour distance from the randomly selected sampled grid. Grey polygons indicate individual territories, filled circles show sampled grids, and unfilled circles illustrate centres of territories.
Figure 2. Graphical summary of model calibration for the RT model to predict territory density of Dusky Orange-crowned Warblers in Bulrush Canyon, Santa Catalina Island, California, USA, in 2006–2008. (a) Cross-validated relative error rate as a function of the number of splits; the best-fitting model has two splits and exhibits the minimum misclassification rate and the smallest tree size. (b) Decision tree for the best-fit RT predicting neighbourhood density in relation to NDVI and oak coverage (%); splits are labelled with conditional cut-off values (e.g. If NDVI is less than 0.10, then territory density is 0.75 territories/neighbourhood. Otherwise, it is 2.36 territories/neighbourhood); nodes denote the predicted number of territories within a neighbourhood scale (with the converted territory density/ha in parentheses); and the number of cases assigned to parameterize the model (n).
Figure 3. Territory density of breeding Dusky Orange-crowned Warblers on Santa Catalina Island as predicted by the best-fit RT model. Black dots indicate point count survey locations. Red polygons delineate areas burned by wildfires in 1999–2007.
Figure 4. Indirect evaluation of the RT model using habitat-specific distance sampling density estimates: territory density from the RT model and individual density from the distance sampling analyses. Each bar represents mean density estimate with 95% CIs with the number of survey points (see results for the statistics).
