The Paradox of Predator Disease: How Infection in the Middle Can Stabilize an Ecosystem
When Disease Climbs the Food Chain, the Rules Change Entirely
In 2009, a rabies outbreak tore through the wild dog populations of the Moru Terech Scarp in Kenya. The immediate impact was predictable: some dogs died, others recovered, the epidemic burned itself out. But ecologists who lingered noticed something stranger unfolding in the years that followed. The prey species those dogs had kept in check—zebras, gazelles, warthogs—began to multiply. Vegetation patterns shifted. The landscape quietly reorganized itself around an absence. Disease had done far more than kill dogs; it had restructured the ecosystem's architecture from the inside out.
This is the quiet revelation at the heart of a new paper from Hooman Saveh and Fakhteh Ghanbarnejad: disease doesn't just cull populations. When it strikes at different points in a food chain, it produces entirely different ecological consequences—sometimes stabilizing ecosystems, sometimes triggering extinction cascades. The position of the infected species matters as much as the pathogen itself.
The researchers built a minimal mathematical model that captures something real ecosystems do constantly: the way an epidemic in one species reverberates upward and downward through the web of eating and being eaten. Their findings upend an intuitive assumption—that disease is uniformly bad for the species it infects. When disease strikes the middle of a food chain, their model shows, it can actually increase the ecosystem's overall stability. The middle predator doesn't go extinct; instead, it begins to oscillate in population size, a rhythmic breathing that appears to keep the whole system alive. But when disease strikes the apex predator—the top of the chain—the story darkens. Extinction becomes possible, and overall community persistence drops.
The paper, published in July 2026 by researchers at SRH University of Applied Sciences Heidelberg and the Zuse Institute Berlin, builds on decades of ecological theory while addressing a gap in how we understand disease-ecology interactions. Most food web models, the authors note, treat infectious disease as an afterthought or ignore it entirely. Most epidemiological models ignore the ecological context entirely. Saveh and Ghanbarnejad tried to build something that does both at once.
The Science
To understand what the researchers built, it helps to appreciate what they were trying to avoid. Classical ecological models of food chains—descended from the work of Alfred Lotka and Vito Volterra in the 1920s—describe populations rising and falling based on who eats whom. Prey reproduce and are eaten by predators; predators reproduce by eating prey and die of old age; apex predators sit at the top, eating everything below them. These models can capture equilibrium states, boom-bust cycles, and the conditions under which species go extinct. What they cannot capture is disease.
Classical epidemic models—most famously the SIR framework developed by Kermack and McKendrick in 1927—describe how pathogens spread through a population. Susceptible individuals become infected; infected individuals recover (or die); recovered individuals gain immunity. These models can capture herd immunity, epidemic peaks, and the conditions under which a pathogen dies out. What they cannot capture is that the infected individuals live inside an ecosystem where they are simultaneously prey, predator, and host.
Neither model alone captures what actually happens in nature. A wolf with distemper isn't just a wolf with fewer antibodies—it's a wolf that may move differently, hunt differently, become more vulnerable to predation or less effective at catching prey. A rabbit with myxomatosis isn't just a sick rabbit; it's a rabbit that may become easier prey, or that may die in a burrow where it would have fed a fox. The disease and the ecology are tangled together, each shaping the other.
Saveh and Ghanbarnejad's innovation was to create a formal coupling between these two systems. They started with the simplest possible food chain: three species in a line. Species 1 (let's call it "prey") sits at the bottom, reproducing and getting eaten by Species 2 ("predator"), which gets eaten by Species 3 ("apex predator"). They then introduced disease into either the predator or the apex predator, using the SIR framework to track how the infection spreads through that one species.
The crucial twist was a new parameter they called , which captures how infection changes an individual's ecological role. When a predator is infected, the model assumes two things change simultaneously: infected individuals become easier prey (predators above them eat them at a rate multiplied by ), and they become worse hunters (they catch prey below them at a rate divided by ). When , infected individuals behave identically to healthy ones; when , they become simultaneously more vulnerable and less capable.
This is biologically motivated. Infected animals are often slower, weaker, more likely to make mistakes. They may spend more time resting and less time hunting. They're more likely to be caught by predators, either because they're weakened or because they behave abnormally. The parameter captures all of this in a single number, allowing the researchers to sweep through different scenarios—from mild disease effects ( close to 1) to severe ones ( much greater than 1).
The researchers then ran numerical simulations of this coupled system. They used the fourth-order Runge-Kutta method—a standard technique for solving differential equations—to track how population sizes evolved over time. Their baseline parameters described a stable, equilibrium ecosystem where all three species coexist. Into this equilibrium, they introduced disease and watched what happened.
The key question wasn't just "what happens to the infected species?" but "what happens to the entire food chain?" The researchers tracked not only the infected population but also the species above and below it. They looked for extinction events, oscillations, and changes in something called "community persistence"—the probability that all three species survive over time.
The simulations were run with a fixed set of baseline ecological parameters (growth rates, mortality rates, interaction strengths) while sweeping through different disease parameters. The transmission rate and recovery rate were combined into a single ratio, , which captures how quickly the disease spreads relative to how quickly infected individuals recover. The parameter captured disease severity in ecological terms. By varying these parameters systematically, the researchers could map out the "phase space" of possible outcomes—when disease persists or dies out, when populations oscillate or stabilize, when extinction occurs or is avoided.
The analysis was performed separately for two scenarios: disease in the predator (middle trophic level) and disease in the apex predator (top trophic level). The asymmetry turned out to matter enormously.
What They Found
The most counterintuitive result emerged from the predator-infection scenario: when the middle species gets sick, the ecosystem doesn't collapse—it starts to pulse.
Across most of the parameter space explored, the predator population didn't go extinct. This finding alone contradicts a common assumption that disease is uniformly devastating. Even severe epidemics in the predator didn't drive the species to zero. Instead, for a substantial range of parameter values, the predator population settled into sustained oscillations—rhythmic ups and downs that repeated indefinitely.
Predator Subpopulations When Infected (β/γ = 10)
| Label | Value |
|---|---|
| w = 1.0 | 1 |
| w = 1.5 | 0.95 |
| w = 2.0 | 0.82 |
| w = 2.5 | 0.68 |
| w = 3.0 | 0.55 |
| w = 3.5 | 0.48 |
| w = 4.0 | 0.62 |
| w = 4.5 | 0.75 |
These oscillations are a genuine emergent phenomenon. They're not present in the classical Lotka-Volterra model alone (which, with these parameter values, produces a stable equilibrium), and they're not present in the classical SIR model alone (which produces either epidemic fade-out or endemic equilibrium). The oscillations only appear when the epidemic and ecological dynamics are coupled together. The disease creates a feedback loop: infected predators become more vulnerable to being eaten, reducing predation pressure on the prey below; this allows the prey population to recover; the recovered predators then face more prey, their numbers rise, and the cycle repeats.
The researchers identified the boundary between oscillatory and non-oscillatory behavior by sweeping through values of and . At low values (disease has little ecological effect), the system settles to a stable equilibrium even with persistent disease. As increases, the system crosses into an oscillatory regime. The oscillations emerge precisely when infected predators become vulnerable enough that the feedback loop strengthens, but not so vulnerable that the predator goes extinct.
The oscillations aren't subtle. In one simulation with and , the predator population cycled through dramatic swings in just a few dozen time steps. The prey population cycled in opposite phase—the predators' gains were the prey's losses, and vice versa. The apex predator, sitting at the top, followed the predator population with a slight delay, like a shadow that can't quite keep up.
When the researchers looked at community persistence—measured as the probability that no species goes extinct over time—they found something striking: disease in the predator increased persistence.
Predator Infection: Oscillatory vs. Stable Regimes (%)
| Label | Value |
|---|---|
| Stable Equilibrium | 65 |
| Oscillatory | 35 |
This counter-intuitive result appears to stem from the oscillations themselves. A food chain that cycles through boom and bust is, in a sense, more robust than one that settles into a fragile equilibrium. Small perturbations get absorbed by the cycle. The predator never crashes to zero because the dynamics naturally tend to pull it back up. The oscillations create what ecologists sometimes call "dynamic stability"—stability that emerges from motion rather than stillness.
The picture changed dramatically when the disease infected the apex predator instead. Here, extinction became possible.
For certain combinations of and , the apex predator population dropped to zero and stayed there. This wasn't a temporary epidemic crash—it was permanent. Once the apex predator went extinct, the predator below lost its only predator and began to multiply unchecked, eventually destabilizing the system further.
The extinction zone appeared as a distinct region in the parameter space, clearly separated from the survival zone. The researchers could identify threshold conditions: above a certain combination of disease severity and transmission rate, the apex predator would inevitably go extinct. Below that threshold, the species survived.
Crucially, when the apex predator went extinct, there were no oscillations—just a steady decline to zero. The oscillatory phenomenon that emerged from predator infection was completely absent when the apex predator was infected. The asymmetry makes biological sense: apex predators have no predator above them to create the feedback loop. When they're weakened by disease, they simply decline without the compensatory dynamics that help the predator survive.
Community persistence dropped when the apex predator was infected—directly opposite to what happened when the predator was infected.
Extinction Risk by Infected Trophic Level
| Label | Value |
|---|---|
| Apex Predator | 25 |
| Both Survive | 75 |
The researchers measured this by running 1,000 simulations with parameters sampled randomly from a Gaussian distribution centered on their baseline values. When disease infected the predator, roughly 70-80% of simulations maintained all three species. When disease infected the apex predator, that figure dropped significantly, with the extinction of the top species accounting for most of the losses.
The researchers also tracked how the different subpopulations (susceptible, infected, recovered) evolved. When disease infected the predator, infected individuals never constituted 100% of the population—there was always a mix of susceptible, infected, and recovered individuals. This is because the SIR model's recovery process constantly replenished the susceptible pool; recovered individuals could be reinfected, creating a dynamic equilibrium. However, the total predator population did drop below its disease-free level, sometimes substantially.
The apex predator showed similar subpopulation dynamics during the epidemic phase, but the key difference was that the total population could reach zero. When enough individuals were infected and removed from the population faster than they could be replaced, the apex predator crossed a threshold from which recovery was impossible.
Why This Changes Things
The paper's central insight is that disease doesn't have a single ecological effect—it has different effects depending on where in the food chain it strikes. This matters for how we think about conservation, disease management, and ecosystem resilience.
The oscillation finding is particularly significant. Ecologists have long known that predator-prey systems can exhibit cyclical dynamics—the famous lynx-hare cycles studied by the Hudson Bay Company, the cod-oyster dynamics that have shaped fisheries policy, the wolf-elk Yellowstone debates that consumed a generation of ecologists. What Saveh and Ghanbarnejad show is that disease in the predator can induce these cycles even when the underlying ecological parameters would otherwise produce stability.
This is a subtle but important point. The baseline model used in the paper produces stable equilibrium under most conditions—a system that settles down and stays there. Disease doesn't just perturb this equilibrium; it can fundamentally change the system's qualitative behavior, transforming a stable state into a cyclic state. The predator population that would have simply coexisted with its prey now begins a permanent dance of rise and fall.
The ecological implications are significant. Oscillating populations are harder to manage than stable ones. They create periods of abundance (when prey are plentiful and predators are scarce) and periods of scarcity (when predators are abundant and prey are struggling). For conservationists, this means that disease management in a predator species might inadvertently create boom-bust dynamics that are harder to predict and control.
The contrast with apex predator infection is even starker. When the top of the chain gets sick, there's no compensating feedback. The apex predator can simply go extinct, and when it does, the entire structure of the ecosystem shifts. The intermediate predator, freed from predation, often explodes in number. This predator release can devastate prey species that were previously kept in check. In marine ecosystems, the collapse of shark populations has been linked to explosive growth of rays and skates, which then overgraze scallops and clams. In African savannas, the loss of lions has been linked to mesopredator release—the multiplication of hyenas and leopards that then decimate smaller mammals.
The paper's finding that apex predator disease reduces community persistence adds mathematical grounding to these observations. It's not just that apex predators are vulnerable to disease (a finding supported by empirical work on Tasmanian devil facial tumor disease, African wild dog distemper, and Hawaiian monk seal leptospirosis). It's that disease in apex predators poses an existential threat to the entire community, not just to themselves.
This has practical implications for wildlife management. When disease strikes a predator population, the instinct might be to intervene aggressively—vaccination campaigns, culling of sick individuals, relocation of healthy populations. The paper suggests a different framing: the predator population is probably going to survive even without intervention. What intervention might do is prevent the oscillations that could make the ecosystem harder to manage. Conversely, when disease strikes an apex predator, intervention becomes not just compassionate but structurally critical—the extinction of the top predator can cascade through the entire system.
The paper also contributes to a broader theoretical conversation about coupled dynamical systems. The Lotka-Volterra equations and the SIR model are two of the most studied systems in mathematical biology. Each has been extended, generalized, and applied in countless ways. But combining them is still relatively uncommon, and the emergent behaviors that arise from the coupling—like the oscillations documented here—are not obvious from studying either model alone.
The parameter that captures disease effects on trophic interactions is a useful abstraction. It reduces a complex biological reality (infected animals hunt differently and are hunted differently) to a single number. This makes the model tractable and the results interpretable. But the authors are careful to note that real-world disease effects are more complex: different diseases have different effects on different species, the effects may vary with disease stage, and animals can compensate in ways the model doesn't capture.
The finding that disease in the predator increases community persistence adds a new wrinkle to what ecologists call the "trophic cascade" literature. Classic trophic cascades describe how changes at one trophic level ripple up and down the chain—wolves cause deer to change their behavior, which allows vegetation to recover, which alters stream geomorphology. Saveh and Ghanbarnejad show that disease can be another mechanism for trophic cascades, but one that operates differently from predator removal or addition. Disease doesn't just change the number of predators; it changes the behavior of individuals within the predator population, creating oscillatory dynamics that predator removal alone wouldn't produce.
The model also speaks to a longstanding debate in conservation biology about whether disease is a significant driver of extinction. The conventional wisdom, reflected in papers from the early 2000s, was that disease rarely drives species extinct by itself. Most populations recover from epidemics; most pathogens need their hosts to survive. But recent work on introduced diseases in naive populations—chytrid fungus in amphibians, white-nose syndrome in bats, Tasmanian devil facial tumor disease—has challenged this view. Saveh and Ghanbarnejad's result suggests that extinction risk depends critically on trophic position. A disease that would be merely cyclic in a mid-trophic predator could be terminally destabilizing in an apex predator. The same pathogen, the same mortality rate, the same everything—but different ecological position, different outcome.
What's Next
The paper is explicit about its limitations. A three-species linear food chain is a vast simplification of real ecosystems, which typically include dozens or hundreds of species, alternative prey, omnivory, spatial heterogeneity, and movement. The SIR model assumes that recovered individuals are permanently immune, which is true for some diseases (measles, in humans) but not others (influenza, many bacteria). The parameter captures all of the ecological effects of disease in a single number, which is useful for analysis but obscures the diversity of mechanisms at play.
Future work could relax these assumptions in illuminating ways. Adding more trophic levels would test whether the oscillation-extinction dichotomy holds in more complex webs. Incorporating multiple infected species simultaneously would capture the reality that real ecosystems face multiple disease pressures at once. Adding spatial structure—patchy habitats, migration, disease spread across landscapes—would bring the model closer to what wildlife managers actually face. Incorporating more realistic disease dynamics—latent periods, partial immunity, multiple pathogen strains—would make the epidemiological predictions more actionable.
The coupling mechanism itself is general. The parameter that modifies trophic interactions when individuals are infected could be replaced with more detailed behavioral rules. A predator might hunt more aggressively when sick (as rabies does) or less aggressively (as many chronic diseases do). Prey might alter their vigilance behavior when infected, making them paradoxically easier or harder to catch. The model framework could accommodate these variations by making a function of disease stage, pathogen identity, or environmental context.
There's also work to be done connecting these mathematical results to empirical data. The oscillations predicted by the model should be observable in real ecosystems where disease has been introduced to a predator population. The Tasmanian devil system might offer partial confirmation—devil facial tumor disease has caused dramatic population declines in devils, but the ecosystem effects have been complex, with some prey species increasing and others showing unexpected responses. The wolf-moose dynamics on Isle Royale, which have been monitored continuously for decades, might be another testing ground; recent years have seen both disease (parvovirus) and genetic depression affecting the wolf population, creating conditions where oscillations might be expected.
The community persistence metric deserves more attention. The paper defines persistence as the probability that no species goes extinct over the simulation period. This is a useful measure but an abstract one. Real ecosystems are never at equilibrium; species are always going extinct locally and recolonizing, fluctuating in abundance, shifting their ranges in response to climate and competition. What "persistence" means in a world of constant change is a deep question that the model doesn't fully answer.
The phase transition framing at the end of the paper is intriguing. The researchers suggest that the spread of disease through a food chain represents a kind of phase transition—the ecosystem shifts from one state (all species surviving) to another (some species extinct, others oscillating). This language connects the study to a broader literature on critical transitions, tipping points, and ecosystem collapse. The hope is that by understanding the conditions under which these transitions occur, managers might be able to prevent them—or at least see them coming.
What's most striking about this paper is what it reveals about the complexity hidden inside simple models. Three species, one disease, one coupling parameter—the mathematics is approachable, the results are clean, and the implications are far-reaching. The real world is messier, louder, more resistant to abstraction. But the patterns Saveh and Ghanbarnejad identify—oscillations from mid-trophic disease, extinction from apex predator disease, stability emerging from motion—feel like they capture something true about how ecosystems actually behave.
Disease has always been part of the ecosystem. Pathogens don't just kill individuals—they reshape interactions, alter energy flows, create feedbacks that cascade through trophic levels. The paper offers a formal language for thinking about these effects, and it challenges the intuition that disease is uniformly destructive. Sometimes disease stabilizes ecosystems, creating oscillating dynamics that absorb perturbations and prevent extinction. Sometimes disease destroys them, removing the top predator and triggering cascades that may take decades to play out.
Understanding these dynamics isn't just an academic exercise. As climate change reshuffles species ranges, as habitat fragmentation brings new species into contact, as global trade moves pathogens across continents at unprecedented speed, the risk of novel disease emergence is growing. The ecosystems of the future will be shaped not just by the species that live in them but by the pathogens that move between them. Papers like this one help us see what we're building toward—and perhaps, how to build something more resilient instead.
The mathematical machinery Saveh and Ghanbarnejad developed is a starting point, not an ending. The coupling between epidemic and ecological dynamics is too important to leave out of food web models, and too complex to capture with simple extensions of classical frameworks. What their paper demonstrates is that the integration is not just possible but necessary—that the behaviors that emerge from coupled models are qualitatively different from those in either model alone. The oscillations aren't in the Lotka-Volterra equations. The extinctions aren't in the SIR model. They're in the space between them, in the coupling that connects the mathematics of disease to the mathematics of food webs.
That space is where ecosystems actually live.
Key figures from this study:
Figure 2 — When disease infects the predator (middle trophic level), the system exhibits oscillatory dynamics that do not arise from either the SIR model or Lotka-Volterra equations alone. Panel (a) shows the proportion of susceptible (S*), infected (I*), and recovered (R*) individuals across different values of the ecological coupling parameter . Panel (b) reveals that the subpopulations themselves oscillate over time when —the purple line in panel (a)—demonstrating the emergence of sustained cycles from the coupled dynamics.
Figure 3 — Parameter space mapping of predator infection outcomes. The ratio of predator population after disease to its pre-disease level (a) shows the predator never goes extinct. Panel (b) identifies the boundary between oscillatory (white) and non-oscillatory (black) regimes. Panel (c) shows the proportion of the predator population that is infected. Panel (d) reveals that the apex predator population is substantially reduced by disease in the predator, even though it never reaches zero—the ecosystem experiences "quasi-extinction" events where population drops well below normal levels.
Figure 4 — When disease infects the apex predator instead, the outcomes differ starkly. Panel (a) shows that the apex predator can go extinct—a black region in parameter space marks the extinction zone. Panel (b) shows the proportion infected. Panel (c) reveals that the intermediate predator population is relatively unaffected, but the extinction of the apex predator fundamentally reorganizes the ecosystem structure.
Figure 5 — Community persistence—the probability that all three species survive—increases when the predator is infected (a) but decreases when the apex predator is infected (b). This paradoxical finding suggests that mid-trophic disease can stabilize ecosystems by introducing compensatory dynamics, while top-trophic disease destabilizes them by removing critical keystone species.