An error occurred while loading the PDF.
A15: City-Scale Species Distribution Modeling Through Aerial Imagery
Presenter(s)
Phoenix Jarosz
Faculty research mentor
Nathan Jacobs
Poster/exhibit session
10:00AM - 11:00AM: Poster session A
Acknowledgments
Srikumar Sastry, Nathan Jacobs, the Multimodal Vision Research Laboratory, and the Center for the Environment.
Abstract or Description
We present a high-resolution species distribution model (SDM) for the Greater St. Louis area, leveraging deep learning to map species presence by using crowdsourced presence-only species observations and 1-meter resolution aerial imagery. Prior work generates low-resolution range maps, making those models unsuitable for local conservation efforts. Aerial imagery provides fine-grained environmental features that allows the model to learn about the habitat of where species are present. To improve predictive accuracy and explore the significance of absence data, we visually compare two models: a baseline SDM using only presence data and a variant (RS-SDM) that integrates pseudo-absence data by randomly sampling locations and assuming they have no species presence. A visualization method is used to produce range maps of every species in our dataset.