Quantifying the richness of ecological communities in patchy greenspaces is a key to conserving urban biodiversity and managing human-wildlife interactions. However, the diversity of any urban habitat patch is influenced by multiple factors, e.g., the match between species-specific habitat requirements and ecosystem attributes, interspecies interactions, and human use. Understanding the importance of these factors to the diversity of urban wildlife is achieved through large-scale, long-term data collection and multi-species patch occupancy modeling. This talk highlights the development of such a model, and its application to a five-plus-year study of Chicagoland mesocarnivores using camera traps. Camera traps are commonly used to study urban biodiversity, but are susceptible to malfunctions and failed detections. To address sampling- and process-associated errors while allowing inferences about community dynamics, our model is hierarchical, mechanistic, and dynamic. We base our model on a combination of single-species, multi-season patch occupancy models and multivariate autoregressive models. Our model includes species interactions, movement, and covariate effects, and can be parameterized using both prior data and observations through space and time. Specifically, we apply the model to a study of the mesocarnivore guild of Chicagoland (red fox, skunk, opossum, coyote, and raccoon). Since 2010, we have been sampling this guild using camera traps at >100 sites located along three transects radiating from downtown Chicago. The model allowed all pairwise first-order interactions among species and included the potential impact of housing density, patch area, and canopy cover on colonization and persistence.
Our model shows that community composition is highly dynamic, with turnover rates ~33% each season. Although habitat patches most commonly (26%) contain two of the five species, 6% of the patches contain four or all five members of the guild simultaneously, and 21% of patches are empty at any given time. In general, species richness increased with increasing canopy cover. Our model indicates that raccoon, opossum, and skunk are more likely to persist in the same patches, as evidenced by positive pairwise interactions (posterior 95% credible intervals do not overlap with 0). Conversely, there was evidence that red fox and coyote were less likely to persist in the same habitat patches. Using model-based rarefaction, we determined that >10 days of camera time is necessary to robustly estimate “true” species richness of a patch. We are using the results from this Chicagoland-focused modeling effort in a Bayesian framework to design future studies in additional cities to inform inter-city patterns of urban biodiversity.