New mathematical tool reveals who eats whom in nature

Scientists have developed a mathematical method that can reconstruct the hidden architecture of food webs with more than 80 percent accuracy, offering a new way to understand what eats what in marine ecosystems. The technique, known as the Superposition Method, uses chemical clues found in animal tissues and Bayesian statistics to estimate predator–prey relationships that researchers may never observe directly. The study, published in Methods in Ecology and Evolution, could help conservationists predict how ecosystems respond when fishing, climate change, pollution, or species loss disrupts the delicate pathways through which energy moves across the ocean.

Food webs are far more complicated than simple food chains. A food chain suggests a straight sequence in which one organism eats another, but real ecosystems are dense networks containing hundreds or thousands of overlapping interactions. A single fish may consume several species, while being hunted by multiple predators. Some prey are important sources of energy for many animals at once, meaning that removing one apparently minor species can affect organisms far higher in the network. If a heavily fished species disappears, its predators may lose a major food source. If a small forage fish declines, the consequences can spread through the ecosystem and eventually reach human fisheries.

Mapping these relationships has traditionally required painstaking fieldwork. Researchers may examine stomach contents, observe feeding behavior, analyze animal remains, or use cameras and sensors to record interactions in the wild. Such approaches can provide valuable information, but they are difficult to apply across large ocean regions, especially where animals live at great depths or move constantly. Stomach contents also offer only a snapshot of a recent meal, while many species are too elusive to collect in sufficient numbers. As a result, scientists often have detailed information about a limited number of species but only an incomplete picture of the entire network.

The new approach began with a conversation between Ettore Barbieri, a senior researcher at the Japan Agency for Marine-Earth Science and Technology and a researcher at the Advanced Institute for Marine Ecosystem Change, and Naoto F. Ishikawa, leader of JAMSTEC’s Organic Molecule Research Group. Ishikawa was describing stable isotope analysis, a technique that allows scientists to infer an animal’s position in a food web from the chemical composition of its tissues. Barbieri recognized that the information could be connected to a problem from engineering mathematics: how to work backward from an observed outcome and estimate the hidden structure that produced it.

Stable isotopes are versions of elements that differ slightly in mass. As energy passes from prey to predator, the proportions of certain isotopes, particularly nitrogen isotopes, often change in predictable ways. By measuring these chemical signatures, researchers can estimate an animal’s trophic position, or its relative location in the food web. Plants and algae occupy low trophic positions, herbivores sit above them, and predators occupy progressively higher positions. The method can reveal whether an animal is feeding near the base or the top of an ecosystem, but it normally cannot identify the exact species it consumes. Many different food-web arrangements can produce the same trophic ranking, creating a mathematical problem with multiple possible solutions.

Instead of attempting to identify one definitive food web, Barbieri and Ishikawa treated each possible interaction as a probability. Their method breaks a complex predator’s diet into smaller two-prey relationships. In principle, estimating the position of a predator that feeds on dozens of species at once is extremely difficult. But the researchers showed that the problem becomes more manageable when the network is decomposed into simple pairs. The model then eliminates biologically implausible combinations, such as a sardine consuming a shark, and evaluates the remaining possibilities using known biological constraints and chemical data.

The calculations rely on Bayes’ theorem, a framework for updating the probability of a hypothesis as new evidence becomes available. In this case, the hypothesis is that a particular predator consumes a particular prey species. The algorithm begins with possible interactions and then adjusts their probabilities according to the animals’ trophic positions, the expected transfer of isotope signatures, and the biological plausibility of each relationship. Thousands of these pairwise estimates are then superimposed, allowing the model to assemble a broader picture of the ecosystem without requiring researchers to observe every feeding event directly.

The team tested the algorithm using 158 fully documented ecosystems from a global database. Importantly, the researchers withheld information about the actual food webs before running the reconstruction, creating a test of whether the method could recover known relationships from trophic information alone. The resulting networks matched the documented ecosystems with more than 80 percent accuracy. The model also estimated the proportion of prey consumed by predators with an error rate below 5 percent, suggesting that it can provide not only a list of likely interactions but also an approximation of their relative importance.

The researchers report that the method remained stable even when the input data contained noise, a critical feature for real-world ecological research. Field measurements are rarely perfect: isotope values can vary, samples may be limited, and biological communities can change between seasons or locations. A model that works only with flawless data would have little practical value in the ocean. By showing that the Superposition Method can tolerate imperfect measurements, the team has opened the possibility of applying it to ecosystems that are too large, deep, remote, or rapidly changing to map through conventional observation alone.

The potential implications extend from marine conservation to food security. Managers could use reconstructed food webs to identify species whose loss would trigger the greatest ecological disruption, distinguish critical prey from less important ones, and anticipate how fishing pressure might alter predator populations. The method could also help scientists monitor ecosystems undergoing warming, acidification, habitat loss, or invasive-species expansion. Barbieri and Ishikawa now aim to apply the technique to real marine systems and improve its accuracy beyond 80 percent. By turning chemical signatures into probabilistic maps of ecological relationships, the researchers may have created a powerful new way to see the invisible machinery that keeps the ocean—and the human communities that depend on it—alive.

News Publication Date: 6-Aug-2026

Web References: https://doi.org/10.1111/2041-210x.70376

References: Methods in Ecology and Evolution, article published 5-Aug-2026; World Premier International Research Center Initiative; Advanced Institute for Marine Ecosystem Change, Tohoku University and JAMSTEC.