Background/Question/Methods PRBO Conservation Science has coordinated citizen science projects to improve ecosystem conservation and management since 1971. Recently we created the California Avian Data Center (CADC) as the first regional node of the Avian Knowledge Network. CADC serves as a repository of data, analysis, tools and information for studying, managing and conserving birds and their habitats throughout California. CADC enables a wide variety of observation types to be archived and accessed via the internet, including systematic surveys conducted by professional ornithologists (~1.7M records) and more casual records via California eBird. In partnership with the Cornell Lab of Ornithology and Audubon California, eBird enables thousands of citizen scientists to contribute data (~1.6M records). Drawing from these large datasets, and using a distribution modeling approach supported by a machine-learning algorithm (Maxent), we investigated the potential impacts of climate and land use change on bird distributions. California's terrestrial ecosystems are particularly vulnerable to future changes in the global climate, including increased temperatures, changing precipitation patterns, and changes in human infrastructure and development. Information on the potential effects of climate change on bird communities can help guide effective conservation and inform land management decisions. Modeling the impact of climate change on rare species is important, but challenging because rarer species usually have sparser data from which to derive models. Unlike systematic monitoring surveys, citizen scientists can readily cover vast geographic areas including places seldom visited by professionals. We used systematically collected records collected by professional ornithologists to create detailed current and future distribution models for several common and rare species. We then compared model performance results after including eBird citizen science data.

Results/Conclusions Most models of current distribution showed very high concordance with expert-generated distribution maps, and most species were projected to have significant changes in their distributions due primarily to climate change effects on their habitat. Here we focused on comparing the performance of the current models with and without eBird data and discuss the implications of the differences for building and validating distribution models for rare species. Since eBird data significantly increased the accuracy of predictions of current distributions, its inclusion for predictive models for future distributions is important since it can help in the discovery of critical management information for the most vulnerable species. Our results show that citizen science data substantially improved models for rare species and highlight the value of a new generation of internet-based distributed observational databases.