Validation of a citizen science-based model of coyote occupancy with camera traps in suburban and urban New York, USA

We evaluated the accuracy of a previously published model of coyote (Canis latrans) sightings in suburban Westchester County, New York. This model was originally developed using citizen reports of coyote sightings to predict the probability of a human-coyote interaction based on proximity to habitat features. Because the data were obtained from surveys, researchers could not separate patterns of site occupancy by coyotes from possible patterns of detection by respondents. Nevertheless, the model could be an indicator of site occupancy within the suburban matrix. We sought to evaluate the predictive power of the human-coyote interaction model with data gathered via a more rigorous method. To build a set of validation sites, we surveyed 11 parks in Westchester County and one park in Bronx County, NY with camera traps between April and October of 2010. The probability of photographing a coyote in a single trap-night was 0.06 ± 0.12 and all sites had >0.9 probability of detecting a coyote at least once given the total trap-nights at each site. During validation, we also added four additional sites that had been surveyed by other researchers with camera traps as additional “present” sites. Predictions of coyote presence or absence based on the human-coyote interaction model for these 16 validation sites were compared to the observed survey results. The model, which contained distances to forest, grassland, and pooled medium and high development performed well in predicting the observed data (kappa = 0.75 ± 0.17, Area-Under-Curve of Receiver-Operator-Characteristic plots = 0.90). The model appears to sufficiently predict coyote occupancy in a suburban-urban landscape and will form the basis of for development of a more comprehensive model of coyote distribution in the New York City metropolitan area. Furthermore, its accuracy illustrates how citizen science can provide reliable estimates of wildlife-habitat patterns in urban areas.