October 22, 2018

PFC/RSGAL Guest Speaker: Opportunities and challenges for Deep Learning in the Environmental Sciences

Talk Title: Opportunities and challenges for Deep Learning in the Environmental Sciences – Dr. Tony Chang

Monday, 11:00 am 10/22/18

Bloedel 357/ 389

Bio: Tony is an ecological data scientist at Conservation Science Partners (CSP) with a background in forest modeling, species distribution modeling, deep learning, data visualization, and big data analysis. He specializes in applying machine learning techniques to ecological issues, including climate change and forest disturbances. His current work focuses primarily on forest structure modeling utilizing deep convolutional neural networks to interpret high resolution imagery in the western U.S. Tony has 13 years of combined field- and research-based conservation experience with numerous federal agencies and NGOs. A 2013–2017 NASA Earth and Space Science Fellow, he is a 2017 recipient of the David H. Smith Conservation Research Fellowship from the Society for Conservation Biology and 2018 recipient of the Microsoft AI for Earth Grant.

Abstract: Deep learning, a class of machine learning algorithms, has recently showed impressive results across a variety of domains. The ecological and earth sciences are data rich, but the data are large, complex, and difficult to utilize for many applications. This talk will provide a broad overview of deep learning and its potential in the fields of environmental science, ecology, and remote sensing. I also provide a case study of a new architecture for forest structure modeling. Deep learning methods may enhance and speed up investigations, but must be used appropriately. Challenges include, limited amounts of standardized labeled datasets for training models and a lack of adoption among environmental scientists to provide a baseline of code to develop upon. Nonetheless, the influence of deep learning will likely transform how we model and map patterns of natural systems in the coming years.