Mismatched spatial scales can limit the utility of citizen science data for estimating wildlife-habitat relationships

While most carnivore populations are declining worldwide, some species are successfully living in human-modified landscapes. For example, coyotes (Canis latrans) have expanded their range across North America and into many urban areas making it important to understand factors influencing broad-scale patterns of occurrence. We used citizen science data in the form of coyote observations by archery deer hunters from throughout the state of Illinois to evaluate factors affecting coyote detection and occupancy. Our statewide participant-level occupancy estimate (0.63) was 58% greater than our naive occupancy estimate (0.40) while detection probability was <0.25, highlighting the importance of using modeling frameworks that account for imperfect detection when modeling occupancy of cryptic species with low detection rates. Time period (AM/PM) had the largest effect on detection of coyotes, with detections greater in the AM. The number of hours hunted (analogous to effort) also impacted coyote detection, with more hours hunted increasing coyote detections. In contrast, none of the landscape covariates examined had strong effects on coyote occupancy. While coyote ubiquity and generalist habitat use may at least partially explain our results, we suspect it also is because the landscape covariates were measured at the county level, as more precise participant location data were unavailable, whereas participants effectively surveyed a much smaller area. Since scale affects the strength and direction of species-habitat relationships, this scale mismatch is likely an important limitation when using many sources of citizen scientist observations to infer species-habitat relationships for widespread generalist species when precise participant location data are unavailable.

Automated summary

In Illinois, researchers found that time of day and hunting effort were key factors influencing coyote detection, while landscape covariates had minimal impact on coyote occupancy. The use of citizen science data revealed the importance of considering imperfect detection when studying cryptic species with low detection rates.