A screenshot of the Atlantic sturgeon encounter risk dashboard for Delaware Bay shows sections of the bay color coded by risk level.
It started in 2009 when Dr. Matthew Breece, now assistant professor of marine science at St. Mary’s College of Maryland, was conducting his master’s research with Dr. Dewayne Fox at Delaware State University. They wanted to understand how Atlantic sturgeon were using habitat in the Delaware River.
“We started with trying to determine where sturgeon are spawning in the Delaware River and that expanded rapidly with the ACT [Atlantic Cooperative Telemetry] Network and MATOS [The ACT Network database],” Breece recalls. “We were hearing sturgeon from other people; they were hearing sturgeon from us. We were seeing this broader movement, and we were able to gain information that nobody had insight on before.”
Fast forward nearly ten years, and Breece was working on a predictive model for sturgeon in their coastal habitat. With his doctoral advisor, Dr. Matthew Oliver (University of Delaware) and a team of collaborators, Breece developed a model that paired environmental data from NASA’s MODIS satellite with acoustic detections to forecast the species’ presence in the Delaware Bay.
As the work progressed, the team identified aspects of the model that could be improved in future iterations. Cloud cover on the Delaware coast frequently prevented satellites from collecting necessary environmental data and the MODIS satellite had exceeded its original mission duration, prompting the search for an alternative.
In a 2021 paper, Breece and colleagues reported that data from the Visible Infrared Imaging Radiometer Suite (VIIRS) on the Suomi NPP spacecraft could be used to develop accurate, daily predictions of sturgeon interaction risk at scales applicable to local fishery management, ensuring long-term viability. They used a DINEOF (data interpolating empirical orthogonal function) to statistically reconstruct missing satellite data, allowing the model to perform even under challenging weather conditions.
Using the updated model, the team developed three sturgeon encounter risk warning products based on stakeholder input. These included a webpage with three-day forecast, a flyer with the study area color-coded by encounter risk level, and an SMS text listing fishing regions and risk levels. They designed the latter product to be accessible to fishers without internet access on the water.
The forecasting tool and risk encounter warning products were well-received in the Atlantic Sturgeon community, Breece says. Soon, managers working in the New York Harbor and off the coast of South Carolina approached Breece to develop a similar model for their area. Funding from NASA’s Biodiversity and Ecological Conservation program, which also supported Breece’s initial predictive model, made this possible.
To build these two new models, the team is working with historical data from ACT Network receivers in the New York Harbor and off the South Carolina Coast without tagging any additional sturgeon. To Breece, this is part of what makes the project exciting.
“We’re collecting so much more data than we had in 2009, and we need to start innovating and thinking about how we can best utilize this data… I think this is a great use for this project because none of these data were collected to do this, right? They were all collected for other reasons. Those reasons have not been exhausted, but pretty much everybody’s met those grant obligations.”
This leaves him thinking: “What more can we do?”
Using preexisting data presents challenges, however. These data sometimes lack information about when receivers were active but did not detect any transmitters – what scientists call true absence data. The ASOM needs both presence and absence data to run. That’s where Dr. John Grady, a postdoctoral researcher at Saint Mary’s College of Maryland, comes in.
Grady has combed through the station records to identify the first and last dates for detected presence, using that as evidence of when the stations were turned on. When he can be confident that a receiver was operational but had no detections, Grady says this indicates a true absence.
“There’s nuance there,” he adds, “We could exclude seasons if we think they are turned off in the winter for instance; or if we thought the stations were deployed in a batch, we could look at the first and last dates for any station in that batch.” Grady calibrates this approach with confirmed absence data from Delaware receivers to ensure that his inferences are good.
The new models are still in their early stages, but there’s reason to be optimistic. The latest three-day sturgeon forecasting model for the Delaware Bay demonstrated 90% accuracy for presence data. A preliminary model Grady ran for the New York Harbor was at 95%. The team plans to deploy three receivers that will transmit sturgeon detection data in real time to corroborate predictions from the new ASOMs.
Managers and commercial fishers need access to data at temporally and geographically appropriate scales to help crews avoid areas where the risk of encountering protected species like the Atlantic Sturgeon is highest. These data allow managers to make more precise decisions about when and where to close areas to fishing to reduce incidental bycatch.
Since he started studying sturgeon 31 years ago, Fox, who is also a co-founder of the ACT Network, says that expanded data sharing and collaboration has made products like the ASOM possible. He hopes that the tool that Grady and Breece are developing will inform more effective conservation for this ancient species.
“Ultimately, I think it gets down to taking the telemetry – we all like to tag fish – and getting that into the hands of policymakers and into a tool that they can use,” says Fox.
With this product, Breece explains, “You [fishers] can keep making your money and making economic progress, and we can also conserve sturgeon at the same time.” In the future, Breece says that he hopes to develop a single ASOM for the entire Atlantic Coast.
The investigators on this project are Investigators Matt Oliver (UDel), Matt Breece (SMCM), Dewayne Fox (DSU), John Grady (SMCM). Funding for this analysis is from NASA Earth Sciences.
They collaborated with users including Andy Herndon NOAA NMFS, Amanda Higgs NYDEC, John Clark DNREC, Bill Post SCDNR, and Ellen Waldrop SCDNR.
Additional authors of the paper mentioned in this article were Edward Hale (DE Sea Grant), Danielle Haulsee (UDel), Matthew Shatley (UDel), Steven J. Bograd (NOAA SWFSC), Elliott L. Hazen (NOAA SWFSC), and Heather Welch (NOAA SWFSC).
by Molly Murphey