Wohner, P.J., P.D. Scheerer, M.H. Meeuwig, and J.T. Peterson .A comprehensive multi-state conditional occupancy model for evaluating interactions of non-native and native species. Frontiers in Ecology and Evolution
A major challenge in ecology is disentangling interactions of non-native, potentially invasive species on native species. Conditional two-species occupancy models are used to examine the effects of dominant species (e.g., non-native) on subordinate species (e.g., native) while considering the possibility that occupancy of one species may affect occupancy and/ or detection of the other. Although conditional two-species models are useful for evaluating the influence of one species on presence of another, it is possible that species interactions are density dependent. Therefore, we developed a novel two-species occupancy model that incorporates multiple abundance states (i.e., absent, present, abundant) of the native species. We showcase the utility of this model that has the capacity to incorporate random effects and covariates on both occupancy and detection and that can help disentangle species interactions given varying occupancy and detection in different abundance states. As a case study, we use snorkel survey data from the Umpqua basin, Oregon, where it is hypothesized that smallmouth bass Micropterus dolomieu, a non-native piscivore, exclude Umpqua chub Oregonichthys kalawatseti, a small endemic minnow. Our conditional two-species multi-state (2SMS) model allowed us to conclude that in general, average occupancy was low for both fishes, and that when non-native bass were present, overall native chub occupancy in the present (0.18 ± 0.05 SD) and abundant (0.19 ± 0.03) states was higher than when non-natives were absent (0.14 ± 0.02/ 0.08 ± 0.02), indicating the non-native was not excluding the native species. By incorporating a species interaction factor into the model, we found a positive association (6.75 ± 5.54 SD) between native chub and non-native bass. The covariates strongly related to occupancy were elevation, algae, and land cover type (urban and shrub). Detection probability was relatively high for both species (0.21–0.82) and was most strongly related to the covariates day of year, water temperature, gravel substrate, and stream order/ magnitude. Incorporation of detection probability and covariates enabled interpretation of interactions between the two species that may have been missed without their inclusion in the modeling process. Our new 2SMS occupancy model can be used by scientists and managers with a broad range of survey and covariate data to disentangle species interactions problems to help them inform management decisions.