Modeling and estimating co-occurrence between the invasive Shiny Cowbird and its Puerto Rican hosts

Invasive species threaten island biodiversity globally. For example, the Shiny Cowbird (Molothrus bonariensis) parasitizes many of Puerto Rico’s endemic species, particularly in the open forests in the island’s southwest. Less is known, however, about cowbird parasitism in the agro-ecological highlands, which contain a patchwork of forests, shaded-coffee plantations, and coffee farms without shade. In this paper, we estimated co-occurrence rates, a potential indicator of parasitism rates, between the cowbird and four host species across these three land uses, hypothesizing that cowbirds would most likely co-occur with their hosts in shaded-coffee farms. We also hypothesized that the presence of host species would increase the probability of cowbird occurrence. To investigate these hypotheses, we developed three Bayesian hierarchical occupancy models: one where the hosts and parasite occurred independently, one that used the latent host species richness as a predictor of cowbird occurrence, and one that used each latent host occurrence state as predictors. These methods addressed observation errors and appropriately propagated error to our predictions of co-occurrence rates. We selected the best performing model using WAIC, then used it to predict co-occurrence rates. While there was some evidence that host species richness increased the probability of cowbirds, the parsimonious model assumed no interaction. With this model, we found that cowbirds were more likely to overlap with certain hosts in shaded-coffee plantations. This may suggest increased parasitism at these plantations, potentially presenting challenges for managers who advocate for shade restoration to gain ecological services such as biodiversity conservation.