That depends on a number of issues; what was the design of your survey, how may "disturbance variables" and habitats does the data set include, is the data hierarchical (nesting), how many data points (samples) you have, etc.. All these factors will interact to determine how you best should analyse the data. For example, if you have limited data you could only test quite simple models.
If each site is a sample (from which you have the beetle community as well as data on the different variables and habitats present), I would start with a simple linear additive model along these lines: