Long-Running Sea Scallop Survey Diversifies for the Future

Our sea scallop survey has been providing data on population status since 1979. Over that time, the stock has recovered from very low levels to a population large enough to support one of the nation's most valuable single-species fisheries. The status of the population isn't the only thing that's changed. Increasingly sophisticated ways to track sea scallop populations have come along. We have adapted and diversified our survey methods to take advantage of them.

Adding Survey Capacity

While we started with one vessel surveying the entire area with a dredge, the area is now divided among our science center and several partners who use both dredges and optical (photographic) instruments deployed from research and commercial scallop vessels. All these data are used for stock assessments and for quota setting.

This year, for the first time, our scie nce center's survey covered Georges Bank and adjacent waters using a dredge, a towed camera and sensor system called the HabCam, and our newest sampling device-a long-range autonomous underwater vehicle nicknamed "Stella." These were deployed off three different vessels: a university-owned research ship, a commercial sea scallop vessel, and a NOAA research vessel.

Stella, the Long-Range Autonomous Underwater Vehicle

The most recent addition to our survey capacity is Stella, our long-range autonomous underwater vehicle. It carries the same imaging package as our towed sampling system, the HabCam, but can be programmed to operate without human supervision. Stella is being developed with partners at the Woods Hole Oceanographic Institution.

"Stella was developed for surveying inside the new wind energy development areas," said Peter Chase, who is in charge of our Center's resource surveys. "It's one way to capture data to replace towed HabCam sampling in those areas."

Large vessels similar to those currently used for our dredge surveys will likely not operate among the multiple turbines planned within wind energy areas. "Our focus now is on testing deployments and operations of the Stella from different platforms before formally incorporating it into our regular survey," said Chase. "We are also gathering more information on how long Stella's batteries last under different conditions. That's important for any fully autonomous operations. Ideally we'd be able to release Stella from shore for a round trip to a survey area."

Stella has the same capacity for sampling as HabCam but travels more slowly. "It has a number of advantages," said Dvora Hart, our lead sea scallop assessment scientist. "We can deploy it untended while other operations continue, and we have so far used it to further investigate areas of particular interest. For example, if the dredge encounters high densities of sea scallops, Stella can rework the area, taking photographs that can confirm what's on the bottom as well as how efficient the dredge is at capturing what's in its path."

What's Next for the Sea Scallop Survey

After this year's successes with the Stella and deployments from a variety of vessels, Chase said the survey will likely continue to use multiple vessels.

"We will be better positioned to consistently complete the survey. We can use a commercial vessel to do all the dredging and the research vessels to deploy the camera systems. The commercial vessel has a sister ship and we have two options for research vessels," said Chase. In recent years, research vessels have encountered operational issues that led to reduced or canceled surveys.

Finally, we are working on ways to handle the data we collect more quickly and efficiently. One outcome of expanded sampling is significant increase in data. In particular, approximately four million images of the ocean bottom are collected during an annual survey. Right now, humans are annotating about one in 50 of the images, identifying sea scallops and other animals. Hart and others have been working to automate that process using artificial intelligence to both speed it up and gather more data from each image.

"People will still need to review a representative sample of the images, but we are training computers to do it using artificial intelligence solutions. The current test version can process in 1 day what it takes us more than a month to do with humans," said Hart. "Our goal is to process all the images we need using machine annotation, continue to have people manually annotate many fewer than we do now, and free those people up for other tasks."

Public link

Please use the above public link if you want to share this noodl on another website.