Spatially Explicit Modeling of Community Occupancy using Markov Random Field Models with Imperfect Observation: Mesocarnivores in Apostle Islands National Lakeshore

How species organize spatially is one of ecologys most motivating questions. Multiple theories have been advanced and various models developed to account for the environment, interactions among species, and spatial drivers. However, relative importance comparisons of explanatory phenomena generally are neglected in these analyses. We developed a spatially explicit community occupancy model based on Markov random fields that accounts for spatial auto-correlation and interspecific interactions in occupancy while also accounting for interspecific interaction in detection. Simulations demonstrated that the model can distinguish different mechanisms of environmental sorting, such as competition and spatial-autocorrelation. We applied our model to camera trap data from a fisher (Pekania pennanti)-marten (Martes americana) and coyote (Canis latrans)-fox (Vulpes vulpes) system in Apostle Island National Lakeshore (Wisconsin, USA). Model results indicated that the observed partitioning pattern between marten and fisher distributions could be explained best by a flipped mainland-island source-sink pattern rather than by competition. For the coyote-fox system, we determined that, in addition to a mainland-island source-sink pattern, there was a positive association between fox and coyote that deserved further study. Our model could be readily applied to other landscapes (island and non-island systems), enhancing our understanding of species coexistence patterns.