Humans kill with a efficiency no natural predator can match, and a new modeling study suggests that this lethality can reshape entire predator-prey communities in ways that depend sharply on which trophic level the hunters target. In an agent-based simulation published in Individual-based Ecology, researchers found that when a hypothetical “superpredator” modeled on human hunters consumed only prey, the probability that mesopredators and prey could coexist fell by roughly 31 percent, and the probability that every agent in the simulated community went extinct rose by a similar amount. That effect dwarfed the influence of natural apex predators in the same virtual ecosystems.
The study, led by researchers exploring how human hunting and fishing alter predator-prey dynamics, builds on a striking empirical backdrop. Humans exploit adult fish at a median rate approximately 14 times that of marine predators and kill large terrestrial carnivores at roughly nine times the rate at which those carnivores kill each other, according to prior work the authors cite. Overall, people prey on more than a third of Earth’s vertebrate species, around 15,000 species. This outsized impact stems not from unique hunting abilities but from technology, a vast geographic range, and large numbers, and much of the exploitation serves non-food purposes such as the pet trade, medicine, and ornamentation.
To formalize what happens when such a predator enters a food web, the team defined a lethal superpredator as an agent that kills with 100 percent per-encounter lethality and whose abundance does not depend on its prey. Natural apex predators, by contrast, were given a 35 percent hunting success rate based on empirical estimates for large carnivores, and their populations were tied to the availability of their targets. The superpredator construct deliberately mirrors a key feature of human harvest: because people can switch resources, trade, and deploy technology, their take of any single species is only weakly coupled to that species’ abundance, a feedback that tends to facilitate overexploitation.
The model itself is a spatially explicit, individual-based system built on the Mesa framework in Python, following the ODD protocol for describing agent-based models. Four agent types populate a 100-by-100 cell lattice with periodic boundaries: prey, mesopredators, apex predators, and superpredators. Prey grow logistically in the absence of predators, mesopredators hunt prey, and apex predators hunt mesopredators or both trophic levels. Every agent senses its eight-cell Moore neighborhood and responds in real time. Prey flee predators, paying a reproductive cost for each avoidance; mesopredators do the same when facing apex or superpredators. Predators move toward targets, gain energy on successful kills, and reproduce in proportion to energy. No agent learns or plans; behavior emerges from instantaneous local information and stochastic rules.
The researchers ran simulations across 2,500 combinations of prey and mesopredator birth probabilities, with 25 replicates each, discarding the first 400 of 1,000 time steps to focus on steady-state dynamics. Each run ended in one of three states: prey only, mesopredator-prey coexistence, or total extinction. In the baseline scenario with only mesopredators eating prey, coexistence was the most likely outcome at a probability of 0.779, with prey-only states at 0.132 and extinction at 0.089. Coexistence occurred at intermediate birth rates around 0.57 for mesopredators and 0.56 for prey, while extinction clustered at high birth rates near 0.79 and 0.69.
Against that baseline, the target choice of the superpredator proved decisive. When superpredators consumed only mesopredators, coexistence barely changed, declining by just 0.7 percent, while prey-only outcomes rose about 5 percent and extinction fell by 4.1 percent. Generalist superpredators eating both trophic levels produced similarly modest shifts: prey-only states increased 4.9 percent, coexistence fell 3.1 percent, and extinction dropped 1.9 percent. The authors attribute this to capping mesopredator numbers, which reduces the chance that mesopredators exhaust their prey, lowering extinction while occasionally removing mesopredators outright and nudging systems toward prey-only states.
Prey-targeting superpredators told a different story. Because their numbers and lethality stay fixed, they keep removing prey even as prey become scarce. Mesopredators then lose their food supply, and both groups disappear, driving coexistence down by roughly 31 percent and extinction up by about 31 percent. In these scenarios, coexistence occurred at lower mesopredator birth rates, averaging 0.40 versus 0.57 in the baseline. Prey populations also cycled faster under direct superpredator pressure, with a shorter dominant oscillation period, likely because strong top-down forcing prevented prey from reaching high densities. The authors note similar truncation of population cycles in human-harvested species such as snowshoe hares, moose, and cod.
Non-lethal superpredators, which elicit anti-predator responses without killing, produced only small effects regardless of whether prey, mesopredators, or both responded behaviorally. Coexistence increased slightly and extinction decreased correspondingly, with necessary conditions for coexistence and prey-only states remaining similar to baseline. The authors caution that this result depends heavily on the avoidance costs modeled, just 0.1 for prey and 0.01 for mesopredators, and on infrequent encounters with 100 superpredators across 10,000 cells. Real-world non-consumptive effects, including habitat displacement, chronic stress, and disease transmission, were not represented, and empirical studies document long-term population impacts of human disturbance in cetaceans, waterbirds, iguanas, and ungulates.
The study is explicitly exploratory theory-building, not a prediction for any real ecosystem. The model is not calibrated to any particular species, assumes homogeneous space without habitat complexity, gives predators no energetic cost for failed attacks, and includes only one species per trophic level. Apex predators could not persist without an immigration rule that replenished them when their numbers fell below 250, and the authors acknowledge that these assumptions probably overstate lethal effects and understate non-lethal ones. The outputs are qualitative and cannot be used to estimate harvest levels or support management decisions.
Still, the findings sharpen a hypothesis worth testing empirically: harvest that targets prey in communities with mesopredators should reduce mesopredator persistence more than harvest targeting the mesopredators themselves, and non-lethal human disturbance should matter mainly where avoidance costs are substantial. The work also underscores a warning the authors state plainly. Coexistence in the model is a system state, not a desirable outcome, and simplified simulations like this one could be misused to justify interference in ecosystems. The results should not be read as evidence that human harvest benefits ecological communities, even where suppressing mesopredators appears to lower extinction risk in silico.
Subject of Research: Agent-based modeling of how lethal and non-lethal human superpredators alter mesopredator-prey coexistence and population dynamics
Article Title: Multi-trophic risk from human superpredators may alter predator–prey coexistence and population dynamics
Article References: Dsouza, S. (2026). Multi-trophic risk from human superpredators may alter predator–prey coexistence and population dynamics. Individual-based Ecology, 2, Article e211852. https://doi.org/10.3897/ibe.2.211852
Image Credits: AI Generated
DOI: 10.3897/ibe.2.211852
Keywords: superpredators, agent-based model, predator-prey dynamics, mesopredator release, trophic downgrading, non-consumptive effects, coexistence, extinction, human harvest, behavioral ecology, population oscillations, Leslie-Gower model
News Source: Drew Townsend. (November 5, 2026). When Humans Hunt, Prey-Targeting Superpredators Can Tip Predator-Prey Systems Toward Collapse. Scienmag.
