Fig. 5: The effect of human disturbance gradients on interaction dissimilarity (βWN).
The relationship between human disturbance distance and interaction dissimilarity, with a fitted line obtained from a Generalized Additive Model (GAM) with human disturbance distance as the only predictor variable (Supplementary Fig. 13 shows the partial effects plot for the model including all predictors). Human disturbance distance was calculated as the absolute difference in human footprint values between a pair of network sites. Each data point (pair of networks) is colored according to the mean of the human footprint values from the two networks. The histogram above the plot shows the distribution of data points across the human disturbance gradient. To explore whether disturbance distance and the mean intensity of disturbance are related, we further divided our data into three equal sized groups (top three histograms) based on their mean (of the site pair) footprint values: “Less” disturbed (low mean footprint), ‘Mix’ (medium mean footprint) and ‘More’ disturbed (high mean footprint). Dashed lines mark the 90th percentile position in each histogram. Note that data points from less disturbed site pairs are skewed towards low values of human disturbance distance, whereas pairs of more disturbed sites also had a larger average distance. Source data are provided as a Source Data file.
