Using machine learning, Rutgers researchers develop a "probability map" from databases that combines whale monitoring and environmental data
Researchers at Rutgers University-New Brunswick have developed an artificial intelligence (AI) tool that will help predict endangered whale habitat, guiding ships along the Atlantic coast to avoid them. The tool is designed to prevent deadly accidents and inform conservation strategies and responsible ocean development.
Using an AI-powered computer program that learns from patterns detected between two vast databases, the researchers said their method improved upon present abilities to monitor the ocean for the distribution of important marine species, such as the critically endangered North Atlantic right whale. North Atlantic right whales have been listed as endangered under the Endangered Species Act since 1970. There are approximately 370 individuals remaining, including about 70 reproductively active females, according to the U.S. National Oceanic and Atmospheric Administration.
The researchers' report was published in Nature Scientific Reports.
The effort was led by Ahmed Aziz Ezzat, an assistant professor in the Department of Industrial and Systems Engineering at the School of Engineering, and Josh Kohut, a professor in marine sciences who in January became dean of research at the School of Environmental and Biological Sciences. Ezzat leads a research group on applied machine learning for engineering and physical sciences. Jiaxiang Ji, the paper's first author and a doctoral student in the School of Engineering, contributed significantly to the project.
This is a demonstration of the power of employing AI methodologies to advance our ability to predict or estimate where these whales are.
Josh Kohut
Marine scientist, Dean of Research at the School of Environmental and Biological Sciences
Kohut likened the output of the program to what might be learned by tracking the movements of people in a house as well as determining whether there is food in the kitchen and a television set on in the den. Such factors might determine why people are where they are at certain times of the day. Detecting certain patterns, he said, conveys predictive power.
"With this program, we're correlating the position of a whale in the ocean with environmental conditions," Kohut said. "This allows us to become much more informed on decision making about where the whales might be. We can predict the time and location that represents a higher probability for whales to be around. This will enable us to implement different mitigation strategies to protect them."
Initially, the researchers sought to develop high-resolution models of the North Atlantic right whale presence to support responsible offshore wind farm development and operation. But they said the results have far broader implications and have made the details public as an addendum to their research paper.
"These tools are valuable and would solidly benefit anyone engaged in the blue economy - including fishing, shipping and developing alternative forms of energy sustainably," Ezzat said. "This approach can support a wise and environmentally responsible use of these waters so that we achieve our economic objectives, and at the same time make sure that we cause minimal to no harm to the environmental habitat of these creatures."
Unlike typical computer programs, where instructions are explicitly written out, the machine-learning program employed by the researchers analyzed large data sets to discover patterns and relationships. As the AI program encountered more data, it adjusted its internal model to make better predictions or classifications.
"The outcome of the machine-learning model is basically a prediction of where and when you will have a higher likelihood of encountering a marine mammal," Ezzat said, describing what he characterized as a "probability map."
The information analyzed by the computer model includes all the underwater glider and satellite-based data collected by scientists at the Rutgers University Center for Ocean Observing Leadership dating back to 1992, when it was established by then assistant professor Scott Glenn, now a distinguished professor in the Department of Marine and Coastal Sciences. The analysis also included satellite data products made publicly available by the University of Delaware.
The underwater gliders are autonomous, torpedo-shaped vessels that zip along under the ocean surface of the mid-Atlantic coast. They are designed to measure many different aspects of seawater, including temperature, salinity, current strength and chlorophyll levels. The gliders also bounce sound waves off schools of fish to gauge their size and record the underwater calls of whales and other marine mammals, locating them in time and space. Satellite data includes measurements of sea surface temperature, water color, and fronts, among others.
"We've had the data but, until now, we've not been able to put the two sets - those detections of where the whales are, and what the environment is like at those places - together," Kohut said. "This is a demonstration of the power of employing AI methodologies to advance our ability to predict or estimate where these whales are."
Other Rutgers scientists on the study included: Laura Nazzaro, a lab manager in the Department of Marine and Coastal Sciences; and Jeeva Ramasamy, an undergraduate majoring in computer science.
