The use of volunteer-collected data is at the forefront of scientific inquiry, but the accuracy, bias, and applicability of such data is generally unknown. We used the California Roadkill Observation System (CROS) to assess the accuracy of volunteer-collected animal-vehicle collision (AVC) data. We classified volunteers as frequent or infrequent, and professional or non-professional, and compared species observations for each. We examined commonly misidentified species. We compared range maps based on CROS data with ranges maps based on Biodiversity Information Serving Our Nation (BISON) data for the top ten reported species in the CROS database.
We found that both frequent and professional users observe more species than infrequent and non-professional users. However, we found no difference (p >0.40) in accuracy of species identification, roughly 93%, between each pair of groups when a photo of the observation was included and verified. The ranges of the top ten observed species in the CROS database had an average of 67 percent overlap with ranges from BISON observations. This suggests that volunteer-collected AVC data may be used for augmenting conventional data for wildlife management.