Model and covariate evaluation.

10-fold cross-validation was used with 180 training samples and 20 test samples for each fold. Prevalence is the average logistic output for all background sites and the logistic threshold is based on equal test sensitivity and specificity for test site omissions. Importance values for each covariate (normalized to percentages) indicate the contribution of each covariate to the final model by randomly permuting the covariate values and measuring the resulting decrease in AUC.