Assistant Professor University of Southern California, California, United States
Abstract: Ectotherms are expected to be disproportionately affected by rapid climate change, resulting in both range reductions and population declines. However, the vast majority of insect species have yet to be assessed by international or federal conservation groups. When faced with potentially millions of candidate species, how should conservation groups prioritize which species should be assessed first? Citizen science data, such as iNaturalist, are an invaluable resource for documenting species’ occurrences, but exhibit taxonomic, spatial, and temporal biases. Occupancy models can address many of these issues, but typically require presence-absence data. As much of citizen science data is presence-only data, absences must be inferred. Often, observed presences of other species are interpreted as confirmation that the site was visited, and the focal species was absent. However, collector behavior is important to consider. For example, opportunistic or targeted sampling of charismatic species by casual observers provides little information on other species, and inferring absences using these observations could bias downstream results.
We used a curated database of over 1 million records that span 2000-2019 to build multi-species hierarchical occupancy maps for each of Canada’s ~300 butterfly species over the past 20 years. We reconstructed collector behavior across a number of scenarios (varying from complete community sampling to targeted observations) to examine occupancy change across these modelling assumptions. We modelled changes in occupancy both at the individual species level, as well across species to examine spatial and temporal trends in occupancy. While detection probability has increased through time, average occupancy probability has significantly decreased across species, though there is much variation between individual species. We find that how one reconstructs collector behavior can affect estimates of occupancy change through time. However, these estimates seem relatively robust overall across these scenarios. We propose that this framework can help prioritize which species should be of highest conservation priority in the future for conservation groups.
