Our oceans provide over 80 million tons of seafood annually, providing a source of animal protein to nearly 3 billion people and a livelihood for 10% of the global population, but more than half of today’s fishery stocks are in decline or overexploited. Historical overfishing and climate change challenge the future productivity and abundance of the world’s fisheries, even as many of them are actively working to rebuild and ensure the recovery of previously overfished species.
This rebuilding is complicated by the lack of accurate tools to measure fish stock. New research from UC Berkeley scientists and their colleagues utilizes computational methods inspired by artificial intelligence (AI) and robotics to assess population regrowth in fisheries and to make better resource management decisions.
Traditional fisheries stock assessment is performed by scientists, who analyze the data on fish abundance to determine how much to fish the population in a sustainable way. This decision analysis takes into account several goals, such as maximizing the number of fish caught while reducing the probability that the number of fish will decline below a critical threshold—it is essential to keep a number of fish above a certain population so they can continue producing offspring.
Complications can arise during these predictions, because the parameters of time-frame, future population growth, mortality, and environmental conditions are often very uncertain. Deciding how much to fish only gets more challenging as scientists and fisheries consider more potential sources of uncertainty due to climate change and increasingly unstable weather patterns.
In the study, published this week in the Proceedings of the National Academy of Sciences, the researchers borrowed numerical techniques used in the field of autonomous vehicle navigation to create a model that calculated the optimal decision of fishery harvest under a range of uncertain future conditions.
“Fisheries management is quite an old science, and the basic management principles were laid down before electronic computers existed. This meant making a lot of simplifying assumptions, which spilled over into modern day engineering fields,” says Carl Boettiger, study co-author and an assistant professor in the Department of Environmental Science, Policy, and Management. “Researchers working on autonomous vehicle navigation found that some of those 'simplifying assumptions' meant their cars kept getting stuck, but after a few decades of research they found a way to tackle the problem of ‘imperfect information.’ Referencing the algorithms used in self-driving cars, we saw that we could return these methods to fisheries.”
Boettiger and colleagues found that their algorithm outperformed traditional approaches typically used in fishery management. By analyzing 109 fish stocks from all oceans, current practices demonstrate a 55% fish stock recovery on average, while their methods yield an 85% recovery of global stocks by 2050, higher economic returns, and greater resilience to environmental catastrophe.
“This increase occurs primarily due to the higher probability of avoiding catastrophic population collapse when accounting for the larger range of future biological and environmental uncertainty,” says study co-author Greg Britten. “Roughly speaking, our algorithm re-discovered the ‘precautionary principle’ for fisheries management, where it ultimately pays to err on the side of cautious harvesting when future conditions are uncertain—something that traditional fisheries management has often failed to do.”
The field of conservation and natural resource management is increasingly engaging computational power, like the algorithms described in this study, to help make decisions. “Ecology is complex and messy, so perhaps our approaches to managing it will need to be complex too,”Boettiger says. “Perhaps the transformative examples we see elsewhere in technology and engineering can help us do more than talk to our speakers and drive our cars. Perhaps they can help save our planet.”
Additional authors of the study include Mailad Memarzadeh in the Department of Civil and Environmental Engineering at UC Berkeley, Gregory L. Britten in the Department of Earth, Atmospheric, and Planetary Sciences at the Massachusetts Institute of Technology, and Boris Worm in the Department of Biology at Dalhousie University.