A Two-Species Occupancy Model with a Continuous-Time Detection Process Reveals Spatial and Temporal Interactions

Detection/non-detection data are increasingly collected in continuous time, e.g., via camera traps or acoustic sensors. Application of occupancy modeling approaches to these datasets typically requires discretizing the dataset to detections over individual days or weeks, which precludes analysis of temporal interactions between species or covariate relationships that change over fine temporal scales. To address this limitation, we developed a two-species occupancy model that assumes a temporal point process detection model. This model permits simultaneous analysis of species interactions in space (i.e., site occupancy) and time (i.e., activity patterns). The model is also capable of estimating the amount of time animals are available for detection, i.e., availability. We applied the model to detections of white-tailed deer (Odocoileus virginianus) and coyote (Canis latrans) collected via camera trapping. We found evidence of both temporal and spatial interactions between deer and coyote. Detection intensity of deer was greater and proportionally more diurnal where coyotes were present. At hunted sites, coyotes were more likely to occur at sites where deer were also present (and vice versa). These results highlight how two-species occupancy models with a continuous-time detection process can be used to infer temporal interactions between species. Our approach broadens the set of questions ecologists can ask regarding both spatial and temporal interactions between species, as well as fine-scale temporal covariates (e.g., weather). Our model should be increasingly applicable given the increasing availability of ecological data collected in continuous time.