EEAGER: A NEURAL NETWORK MODEL FOR AUTOMATICALLY IDENTIFYING BEAVER COMPLEXES IN SATELLITE AND AERIAL IMAGERY

The North American beaver (Castor canadensis) is a keystone species which drastically alters the physical environment, creating and maintaining riparian wetland ecosystems in a variety of settings. Recent research has underscored the ability for beaver activity to sequester carbon; support sensitive, threatened, and endangered species; attenuate floods, keep vegetation green and maintain base flow during droughts; and create patches of wildfire refugia. Presently there are no systematic or widespread beaver population monitoring programs in the United States. Despite the large-scale physical, hydrologic, and ecological implications of beaver range expansion, most of the research on beaver populations and impacts have been limited to the reach or watershed scale. Relatively few landscape-scale studies have been conducted, possibly due to the significant time investments required for manual identification and mapping of beaver landforms in aerial or satellite imagery. There are currently no models that can identify the physical presence of beaver dams and ponds or associated wetland habitat resulting from that presence. We developed a model in Google Earth Engine to identify existing and/or historic beaver habitat, at a scale and efficiency that allows for landscape- and regional-scale tracking of expansion and contraction of beaver populations, as well as of the ecosystem services provided by that habitat. Due to the wide variation in the size, shape, and construction of beaver dams, machine learning methods on geospatial imagery are employed for training. Our model was highly accurate (accuracy generally greater than 90%) at identifying beaver dams in the American West. This has far-reaching implications for large-scale monitoring of beaver populations, beaver-based river restoration work, and beaver and water resource management activities.