Understanding Long Island Sound’s New Hypoxia Forecast Model

Set to be released in late spring, the forecast will estimate dissolved oxygen concentrations and the area of hypoxia in the Sound  

Hypoxia, or low dissolved oxygen, has remained a persistent issue impacting the bottom waters of Long Island Sound and is worsened by nutrient pollution from wastewater treatment plants, stormwater runoff, atmospheric deposition, and other sources. Over the past four decades, the Long Island Sound Study and its partners have made substantial investments to reduce pollution to the Sound and reduce hypoxia. To drive public awareness and develop a scientific understanding of long-term changes in oxygen and temperature, EPA researchers with LISS are working to develop a Long Island Sound Hypoxia Forecast Model.  

Scientists have been predicting hypoxia through forecast models in the Gulf of America, Lake Erie, and Chesapeake Bay over the last decade. Inspired by these and other examples, LISS’s Science Coordinator, Jim Ammerman, PhD, pitched the idea to explore a similar effort for water quality in the Sound.   

“When compared to the actual measured hypoxic area, a hypoxic area forecast is a good tool for evaluating scientific understanding and educating people about the importance of oxygen depletion in the Sound,” said Ammerman. “People like forecasts; they like predictions, so it’s an effective way to increase an understanding around hypoxia, which is a complicated concept.”

Planning and Developing an Environmental Forecast

In 2022, EPA Region 2, in collaboration with EPA’s Office of Research and Development, formally started advancing a forecast methodology. The seasonal forecast, set to be released in late May of this year, combines efforts of researchers, outreach professionals, and science communicators and will predict the area of hypoxia for the 2025 summer season. The prediction will be accompanied by a communications toolkit, including a Story Map, model illustrations, memes, and animated videos on what has been done to address hypoxia in the Sound. 

In early stages of the project, researchers aimed to balance the technical aspects of developing a forecast while considering the utility of a prediction for communication and outreach goals. At an EPA-hosted workshop in May of 2023, federal, state, and local stakeholders discussed how a hypoxia forecast might improve the public’s awareness of low oxygen in Long Island Sound. Workshop participants included staff from the New York State Department of Environmental Conservation, the New York State Department of Environmental Protection, New York Sea Grant, Connecticut Sea Grant, the Connecticut Department of Energy and Environmental Protection (CT DEEP), NEIWPCC, Save the Sound, and the Interstate Environmental Commission. The University of Connecticut and Stony Brook University participated in the workshop virtually.  

Why Do We Forecast?

Ecological forecasting is a useful way to explain environmental changes and can help decision-makers improve their management of ecosystems. The hypoxia forecast is being developed using water quality monitoring data going back to the early 1990s. CT DEEP monitors water quality year-round on behalf of LISS. During summer hypoxia surveys, 48 stations across the Sound are sampled for dissolved oxygen, temperature, pH, and salinity. With over 30 years of data, researchers can uncover long-term water quality changes and trends in Long Island Sound, with the forecast describing the trajectory of low oxygen and warming temperatures up to the present.

How Will Researchers Forecast Hypoxia

Other coastal hypoxia forecasts usually work in one of two main ways:

• The Chesapeake Bay Hypoxia Forecast uses an operational or “real-time” simulation model to predict the present-day levels of dissolved oxygen throughout the bay, while also forecasting oxygen levels two days into the future. 
• Forecasts for the northern Gulf of America aim for a late-spring prediction of the expected extent, or area, of hypoxia that will be measured during summer surveys. The forecast relies on information about observed relationships between water flow and nutrients from the Mississippi River in May and the extent of hypoxia that occurs later that summer.

This type of delayed correlation between freshwater flow and the onset of hypoxia is called a “lag” and is a phenomenon that also occurs in Chesapeake Bay. Studies of hypoxia in Long Island Sound have not found similar “lag” relationships between hypoxia and observable late spring factors like flow from major rivers. Still, forecasting remains possible. The multi-decade data record used in the Long Island Sound hypoxia forecast is constructed to reduce the number of potential outcomes likely to occur this summer.

How Is This Possible?

Models are mathematical representations of data that provide researchers with a way to describe how water quality variables, such as dissolved oxygen, change over time and in different locations. Researchers at CT DEEP provide an analysis of completed Long Island Sound water quality surveys, applying a model to map bottom water oxygen for the days surveys are conducted.

To create the forecast, researchers use a method called Generalized Additive Models or “GAMs.” GAMs can represent complex seasonal, spatial, and other patterns in data that occur in water bodies like Long Island Sound. GAMs can be used to build models that describe how oxygen in bottom water changes from day to day or year to year at a single location. Modeling a sequence of stations allows researchers to analyze and understand changes over time and spanning across the Sound. Modeling oxygen levels from the surface to the bottom at a series of stations can provide a detailed view of spatial and temporal patterns. Ultimately, a combination of these models can provide researchers with a more complete understanding of hypoxia in the Sound and how it varies over time. 

Over the past 30 years, the maximum extent of hypoxia in Long Island Sound has spanned a wide range, defining a series of potential outcomes for the upcoming year. Data show that the area of hypoxia has decreased substantially, mirroring early decreases in nitrogen loading from wastewater plants in Connecticut and subsequent decreases in loads from New York. These changes have been accompanied by a decrease in the length of time that hypoxia persists (i.e., “duration of hypoxia”) and an increase in the minimum oxygen concentrations that occur within the hypoxic zone.