The quality and contribution of volunteer collected animal vehicle collision data in ecological research

摘要

Volunteer-collected data are increasingly important for research, and the accuracy and applications of such data needs evaluation. We utilized the California Roadkill Observation System (CROS) to assess the quality of and potential contribution that volunteer-collected animal-vehicle collision (AVC) data to understand species ranges. We classified volunteers as professional biologists or non-professional users, and compared their rates of species observation and identification accuracy. Seventeen percent of the volunteers were professionals, but they provided 35% of the observations, and they observed a more diverse set of species than non-professionals. The accuracy of species identification was similar for both groups. We compared species' ranges estimated from CROS data with ranges derived from the largest available database of animal observations (Global Biodiversity Information Facility; GBIF). CROS observations expanded GBIF data-derived ranges for 139 of 411 species. Average range expansion was 8.2% for all species, including species with no range expansion, or 24.3% for species with range expansion. We also used species distribution models to estimate range sizes. Adding CROS observations to GBIF records expanded the average predicted species-range by 5%. Professional observers expanded previously-estimated ranges by an average of 1.1%, while non-professional observers expanded ranges by an average of 5.3%, reflecting their larger spatial coverage. While non-professional users observe fewer species, their larger number of observations and occurrence over a larger area improves our knowledge of the ranges of a number of species. Volunteer-collected AVC data from both professional and non-professional users can be an important data source for augmenting conventional databases of wildlife observations to create species range maps.