THE RISE OF MACHINE LEARNING IN MARINE ECOLOGY RESEARCH
As more species are listed under the Endangered Species Act (ESA) because of factors such as over-exploitation, damaged habitats, and resource competition leading to a decline in their population size, the countries in which these species are found must do their part in protecting, monitoring, and recovering them. One new innovative way to monitor such species is through the use of machine learning (ML). This involves using pattern recognition methods based on Artificial Intelligence (AI) techniques to identify animals to allow monitoring of populations and migrations. This project with the National Oceanic and Atmospheric Administration (NOAA) aims to study the reliability of ML for Steller Sea Lions (SSLs) in the Aleutian Islands, which are classified as the western distinct population under ESA. The reliability of AI will be evaluated by comparing human-monitored observations of each sea lion with the ML algorithm’s outcomes. To accomplish this, I participated as an observer of SSLs that would act as comparative data to multifunction on how accurate NOAA’s current AI model is improving, and to feed in new observed data for the AI to learn. With the final comparative data of my work to the AI model, I found an increased miss rate of SSLs by the AI model, but additionally increased at catching SSLs that I the observer missed. This project will serve as one of several baseline tests to assess whether it is possible to efficiently monitor the impact of humans and global warming on animal health, distribution patterns, and population size.
