Harnessing Open Data to Protect Vulnerable Marine Ecosystems

"We need more data." Pick up almost any ocean-focused research paper, report, or assessment, and you'll probably see that phrase popping up. Data, including for the seabed, is crucial for deciding what types of activities we do in the ocean and where and when we do them. Fishing, for example, "has the potential to impact ecosystems [on the seabed] that are very fragile and complex, and comprised of species that are very long-lived, rare, or endemic," says Dr Tabitha Pearman, who is based at the National Oceanography Centre, UK. "These are known as vulnerable marine ecosystems, and international and increasingly national legislation requires they are identified and mapped so we can avoid damaging them."

The Southern Patagonian Shelf hosts several fisheries that may impact vulnerable marine ecosystems, or VME as they are often called. The Falkland Islands’ Marine Stewardship Council (MSC) certified Patagonian toothfish fishery strives to minimise fishing impact on vulnerable marine ecosystems as a requirement to keep MSC certification.

Physically surveying every part of the seafloor to find these vulnerable marine ecosystems is impractical. Instead, researchers often use species distribution models to create probability maps which show where you are most and least likely to find a vulnerable marine ecosystem. "The ecosystems as a whole are really hard to map, so instead, we map indicator species," says Pearman. Indicator species are species that are frequently associated with a particular vulnerable marine ecosystem in a region. For example, on the Southern Patagonian Shelf, gorgonian corals are an indicator species for coral gardens, which often contain many different non-reef-forming coral species and act as a home for a diversity of other marine life.

The data challenge

As part of her postdoc at the Falkland Islands-based South Atlantic Environmental Research Institute (SAERI), Pearman led the Vulnerable Marine Ecosystem Project, where she created distribution models to predict the location of vulnerable marine ecosystems using indicator species. Generally, species distribution models need data about the location of the species you want to predict and information about the environment where the species is found, plus the area you want to predict the distributions across. For the indicator species, Pearman used fisheries bycatch data and legacy imagery collected by the oil and gas industry. The environmental data was a mix of oceanographic data, such as sea temperature, and seabed data, such as slope derived from depth.

The trouble is, Pearman says, for places like the Southern Patagonian Shelf, “you don’t have much data, especially about the seabed, and in the Falkland Islands, there aren't national initiatives to collect high-resolution data." Only a small area of the seabed has been mapped at high resolution. That area was mapped as part of a series of seismic surveys. For the wider region, there is only one viable option – the Nippon Foundation-GEBCO Grid. "I wanted to know what happens when you don't have high-resolution data. Can we still make predictive maps that can be used for management?" says Pearman.

To assess if the openly available GEBCO Grid could be used to predict vulnerable marine ecosystems like coral gardens as efficiently as using higher-resolution data, Pearman built two models – one using seabed attributes derived from the GEBCO grid and one using attributes derived from seismic-derived bathymetry data. On comparing the two models, "I was pleasantly surprised," says Pearman, who presented her work at the 2023 Marine Geological and Biological Habitat Mapping (GeoHab) conference.

Guiding management with open data

While the model using the higher-resolution seabed data from the seismic surveys performed better than the one built on the GEBCO data, "both picked up the same type of environmental relationships and gave similar predictions for where these ecosystems are," says Pearman. "For the stony reefs, the Bathelia candida species can be predicted really well [with the GEBCO Grid]. So can the sea pen meadows," says Pearman. However, gorgonian corals, which act as an indicator for cold-water coral gardens, didn't predict so well. The reason, Pearman says, is that while the distribution of the former correlates well with large geomorphic features that show on the GEBCO grid, the latter "are associated with the smaller glacial stone drops [rocks that fell out of icebergs during the last ice age] which can only be picked up with higher resolution data."

This doesn't make predictions for vulnerable marine ecosystems associated with gorgonian corals useless. Rather, "we need to be a bit more cautious in what we say the predictive maps can do because we aren't as confident in the predictions," says Pearman. Considering confidence isn't just an academic exercise for Pearman. The maps could guide management decisions.

When creating marine spatial plans, most countries take an area of the ocean and make it into a grid. Each cell in that grid allows different types of activities. "In the Falklands, marine planning is often implemented at the scale of the fishery. The grid cells are quite large. When we assess how sensitive vulnerable marine ecosystems are to fishing, we need to use that same scale," Pearman says. "When we bring the predictions back to that broad-scale management grid, the maps made from the GEBCO [grid] were just as accurate as those made from the higher-resolution data."

"For all the people working with low-resolution data, you can still get something out of the maps you're using," says Pearman. "They may not be as precise as if you use high-resolution data, but for management, which often works on pretty broad scales anyway, you can still create something informative."

 
This story was written for Seabed 2030