The citizen science project eBird has generated a database of over 750 million bird observations, with broad spatial and taxonomic coverage. Over the past 10 years, the Cornell Lab of Ornithology has developed machine-learning models using eBird and remote-sensing data to produce high resolution, weekly estimates of range boundaries, occurrence rate, and relative abundance while accounting for many of the biases inherent in citizen science datasets, including variation in observer behavior and effort. Visualizations and modeled data products for 600 North American breeding birds, including resident and non-breeding grounds in South America, are currently available on the eBird website.This workshop will introduce attendees to the modeled data products (weekly estimates of range boundaries, occurrence rate, and relative abundance) and the ?ebirdst? R package developed specifically for working with these data. This will include a discussion of how eBird data is collected and how we prepare it for modeling by addressing spatial bias, temporal bias, class imbalance, and observer behavior and effort biases. We will also provide an introduction to the modeling process used to generate the eBird Status and Trends data products. The workshop will also include a demonstration of how to access and manipulate these data products for specific combinations of species, seasons, and regions using the ebirdst package. After the workshop, attendees will have an understanding of how and when to use these data products for applied research and conservation efforts, including within-year dynamics. Some experience with R will be helpful in following along with the demonstration. Please note, this workshop will not cover the analysis of trends or trend data.
Organizers: Orin Robinson, Cornell Lab of Ornithology, Ithaca, NY; Tom Auer, Cornell Lab of Ornithology, Ithaca, NY; Erica Stuber, Cornell Lab of Ornithology, Ithaca, NY;
Supported by: Biometrics Working Group