Counting Cats: The integration of expert and citizen science data for unbiased inference of population abundance

Free-roaming animal populations are hard to count, and professional experts are a limited resource. There is vast untapped potential in the data collected by nonprofessional scientists who volunteer their time to population monitoring, but citizen science (CS) raises concerns around data quality and biases. A particular concern in abundance modeling is the presence of false positives that can occur due to misidentification of nontarget species. Here, we introduce Integrated Abundance Models (IAMs) that integrate citizen and expert data to allow robust inference of population abundance meanwhile accounting for biases caused by misidentification. We used simulation experiments to confirm that IAMs successfully remove the inflation of abundance estimates caused by false-positive detections and can provide accurate estimates of both bias and abundance. We illustrate the approach with a case study on unowned domestic cats, which are commonly confused with owned, and infer their abundance by analyzing a combination of CS data and expert data. Our case study finds that relying on CS data alone, either through simple summation or via traditional modeling approaches, can vastly inflate abundance estimates. IAMs provide an adaptable framework, increasing the opportunity for further development of the approach, tailoring to specific systems and robust use of CS data.