AI Model Flags Threats to 10,000 Freshwater Fish Species Before Extinction Risk

Almost one in three freshwater fish species faces possible extinction. That figure, drawn from decades of conservation assessments, represents a staggering scope of potential biodiversity loss - from redfin pickerel threading through Maine's Kennebec River to sturgeon navigating the Great Lakes. Identifying which species to prioritize, and why some survive while others decline, has long been an imprecise, resource-intensive process.

Christina Murphy, an assistant professor at the University of Maine and assistant unit leader for the U.S. Geological Survey's Maine Cooperative Fish and Wildlife Research Unit, spent five years building a different approach. The result is a computer model capable of assessing extinction risk for more than 10,000 freshwater species worldwide - and, crucially, identifying what conditions allow species to thrive before crisis sets in.

What the Model Actually Does

The tool draws on data from 12 publicly available sources, the majority compiled by the International Union for Conservation of Nature. It uses artificial intelligence trained to recognize patterns across 52 variables, including damming, water abstraction, habitat degradation, pollution, economic conditions, and invasive species. Rather than simply flagging what is going wrong for imperiled fish, the model maps the conditions associated with stability - what the researchers call signals of ecological well-being.

That distinction matters. Previous conservation models have often focused on documenting threats after populations collapse. Murphy's team flipped the analysis: by identifying the conditions that keep species safe, managers can act before a species requires emergency protection.

"This uses new metrics to identify what is working to keep species from being listed," Murphy said. "Managers may be able to protect a lot of fish."

The AI was programmed and trained to analyze millions of nonlinear connections among species - relationships that linear statistical models typically cannot capture. Users can then examine the driving conditions for any given species and ask whether those same conditions exist for closely related or geographically nearby species not yet in immediate danger.

Most Species Can Still Be Saved

One of the more encouraging findings: the majority of species in the model's dataset still have viable conservation windows. Maine's Arctic Char (Salvelinus alpinus) and certain char populations in other regions appear among those that could be safeguarded under current conditions - if managers act proactively rather than reactively.

Co-author J. Andres Olivos, a postdoctoral researcher at Oregon State University, offered a medical analogy: "Our results suggest conservation works like human health: the signals of 'well-being' are often more consistent than the many pathways to illness. For freshwater fishes, safe conditions tend to be predictable, while extinction risk can come from countless combinations of threats."

Wildlife managers overseeing dozens of species across large watersheds can set conservation programs based on what has demonstrably worked for related species. "Managers can set up new conservation programs based on what has worked in the past because a lot of species share what works," Murphy said.

The Socioeconomic Dimension

The model's 52 variables include economic and governance indicators alongside ecological ones. Countries and regions with stronger governance, better environmental regulations, and more stable economies tend to produce better outcomes for freshwater fish. Ivan Arismendi, an associate professor in Oregon State University's College of Agricultural Sciences, emphasized the practical urgency of acting on these findings. "People sometimes go in to protect species when it's already too late. With our model, decision makers can deploy resources in advance before a species becomes imperiled."

Validation and Scope

The team validated the model against existing conservation assessments, and it performed well against these benchmarks. Murphy began the project in 2020 as a postdoctoral researcher at Oregon State, collaborating with Arismendi and Olivos alongside scientists from the USGS, the U.S. Forest Service, and the University of Girona in Catalonia, Spain. The findings were published in Nature Communications. The reliance on publicly available data keeps implementation costs low - important for conservation agencies working across vast geographic areas with constrained budgets.