Model structure and fit:
We ran HMSC models using the following setup: thin = 100, transient = 12,500, samples = 250. Using four chains to traverse parameter space and sample the posterior distribution of estimates, we obtained a total of 1,000 samples from each species, which we used in downstream analysis and interpretation. Gelman’s diagnostic scores were largely within the range of acceptable values (98-99% of values were < 1.05), suggesting that we had reached convergence of the model. Examination of trace plots (not shown here) provided further support that chains were well mixed.