COS 128-6
Using multi-species occupancy models to improve inference of metacommunity structure

Friday, August 15, 2014: 9:50 AM
308, Sacramento Convention Center
Joseph R. Mihaljevic, Ecology and Evolutionary Biology, University of Colorado at Boulder
Maxwell B. Joseph, Ecology and Evolutionary Biology, University of Colorado, CO
Pieter TJ Johnson, Ecology and Evolutionary Biology, University of Colorado at Boulder
Background/Question/Methods

Pattern-based metacommunity approaches use observed species incidence data to determine how a metacommunity is structured and subsequently to infer structuring processes. However, this type of community data suffers inherently from imperfect detection, and false absences could bias the methods used to assign metacommunity structure. Multi-species occupancy models, however, utilize re-survey data to disentangle species’ site-specific probabilities of detection and occurrence to estimate ‘true’ occupancy at each sampled site. Here, we illustrate how occupancy models can be integrated with current methods used to identify metacommunity structure in order to improve inferences of structure and to better understand structuring mechanisms. First, to explore whether occupancy models can indeed reduce bias in determining structure, we simulated true occupancy states of a metacommunity and imposed imperfect detection to simulate observed species incidences. We then compared the metacommunity structure based on these observed occupancy states to the structure based on the estimated incidences from an occupancy model. Second, taking advantage of occupancy models’ ability to assign species-specific covariate effects, we simulated communities with different assumptions about how species respond to an environmental covariate to better understand the processes that can lead to distinct metacommunity structures.

Results/Conclusions

By reducing bias associated with imperfect detection, incorporation of a multi-species occupancy model substantially improved our ability to determine true metacommunity structure compared to traditional methods using raw observed data. In our second simulation, we determined that structure was sensitive to both the distribution of covariate values across local sites and the distribution of covariate effects among species. Furthermore, stochastic variation in either of these distributions often resulted in different metacommunity structures, even when the underlying biology did not change. These results help to quantitatively evaluate verbal hypotheses about how metacommunity structures arise, illustrating that species-specific responses to covariates can, but do not always, drive structure. These findings also suggest that metacommunity structure could be driven by heterogeneity in dominant environmental covariates, such that survey designs and sampling effects could influence interpretations about emergent structure. Overall, our study shows that using an occupancy model framework can improve our understanding of metacommunity structure and its underlying mechanisms, and could therefore improve predictions of when and where certain structures should emerge.