Stony Brook University researcher turns to AI to predict Long Island Sound water quality problems

Researchers on coastal ecology have studied for decades the conditions that can lead to plummeting oxygen levels in sensitive marine ecosystems, trying to parse how the combination of nutrient levels, water temperature, bacteria and other factors can lead to die-offs of fish, shellfish and aquatic plants.

But previous models and analyses aiming to describe how hypoxic zones emerge were missing something, according to Kamazima Lwiza, a marine physicist at Stony Brook University's marine science division.

"In the real world," Lwiza said, oxygen is usually depleted at the end of the hot season, generally in August. "But in most models it will run out sooner. So there was a process we were not accounting for — not just me and my students, but everybody else who was in this game," Lwiza said.

Lwiza has received a New York Sea Grant to develop a new model, assisted by artificial intelligence, that he expects will discover what that missing element is. He plans to use the $218,361 grant to test a machine learning tool that cannot only shed light on why and how hypoxic zones grow in Long Island Sound, but can predict where they will develop up to six months in advance.

Usually scientists need to know the relationships between various factors — temperature, salinity, the presence of various nutrients, for example — to get meaningful results. "But the beautiful thing about machine learning, you don't need to know the relationship," Lwiza said. The machine learning tool will figure it out.

"The only catch is you need lots and lots and lots of data in order to get the machine learning to do it right." With enough data, Lwiza says, he hopes to identify what the important variables are — for example, the supply of nitrogen and phosphorus — and to discover which of those can directly cause low oxygen conditions and which are more loosely related. 

Lwiza said that early notice of a developing low-oxygen zone can allow managers to make adjustments, to pause fishing in the area, for example, to reduce stress on an already challenged environment. He said if phosphorus — a main ingredient in synthetic lawn fertilizers — turns out to be a significant contributor to hypoxia, golf courses and residents could be required to switch to low-phosphorus formulas. 

Lwiza is one of 10 researchers and teams to receive a two-year award from New York Sea Grant, which is funded by the National Oceanic and Atmospheric Administration (NOAA). "This is the largest number of projects that NYSG has been able to support during any one of its biennial research cycles," the organization’s director, Becky Shuford, said in a statement. She said the projects will "respond to the environmental, social, cultural, and economic needs of the communities who live, work, play, learn, and rely on the resources and services New York’s coasts provide."

Four additional teams of researchers from Stony Brook University’s marine and atmospheric science department received funding.

Yong Chen is creating an ecosystems model of the Hudson River estuary and New York Bight, the coastal region from Cape May, New Jersey to Montauk Point. His model will evaluate the effects of global warming and other human-caused environmental changes, such as runoff and sewage releases.

Michael Doall and Christopher Gobler as well as Stephen Tomasetti of the University of Maryland Eastern Shore, will study the effects of rising temperatures and low dissolved oxygen on shellfish. The researchers hope to determine which shellfish can best adapt to rapidly warming oceans and which areas in coastal New York are most promising for various species.

Emmanuelle Pales Espinosa and Bassem Allam, along with Stephen Tettelbach and Harrison Tobi from the Cornell Cooperative Extension of Suffolk County and Emma Green-Beach of Martha’s Vineyard Shellfish Group, will study the parasite that has crashed bay scallop populations in the Peconic Estuary and elsewhere.

Donovan Finn, Kevin Reed, and Christine Gilbert will look at how experts can communicate climate risks to vulnerable coastal communities, using Jamaica Bay as a case study. "With our project we hope to uncover some of the communication barriers that exist between climate scientists and public officials on the one hand, and the messages that communities hear related to climate change induced risk on the other hand," Finn said in a statement.