Food Webs and Trophic Cascades

An ecosystem isn't a list of species. It's a network of energy flows — who eats whom, who parasitizes whom, who scavenges whose kills. The structure of that network determines which species are abundant, which are rare, and how the whole system responds when you add or remove a piece. The Yellowstone wolf reintroduction is the most dramatic natural experiment on these questions in recent memory, and it's still being argued about three decades later. Meanwhile, the parasitologists have been quietly demonstrating that we've been drawing the network wrong — leaving out the most numerous organisms in the system.

Wolves, Elk, and the Top-Down Question

When wolves were reintroduced to Yellowstone in 1995-96, the elk population was already declining from an all-time high. The wolves grew rapidly — near the maximum rate ever recorded — feeding on the abundant prey. As wolf numbers increased, elk numbers dropped further, but separating the wolf effect from drought, human hunting, and a naturally expanding grizzly bear population turned out to be fiendishly difficult. This isn't a bug in the science; it's the central challenge.¹

The strongest evidence for a top-down effect (predators shaping the system from above) is behavioral rather than numerical. Elk changed where they fed — avoiding riparian areas where they could be ambushed. This allowed aspen, cottonwood, and willows to recover in those areas for the first time in half a century, which in turn increased the abundance and diversity of riparian bird species. The vegetation recovery covered less than 2% of the park's area but had outsized ecological significance.¹

But the bottom-up camp has a point too: the same period saw prolonged drought reducing forage, a lengthened growing season from climate change favoring woody plant recovery, and post-fire conifer regeneration from the 1989 Yellowstone fires. Any of these could explain some of the vegetation changes attributed to wolves. The honest scientific position, as Dobson puts it, is that we need another decade of data and probably new mathematics to disentangle these forces.

What makes the Yellowstone case fascinating for complex systems thinking isn't the answer (we don't have one yet) but the structure of the question. You have multiple interacting forces operating at different spatial and temporal scales — predation (fast, local), climate (slow, regional), fire recovery (medium, patchy), human hunting (politically modulated) — and the system's behavior is overdetermined by their interaction. No single force explains the observations. The debate about top-down versus bottom-up is really a debate about whether ecosystems have a dominant organizing principle or whether they're irreducibly multi-causal. Dobson argues that the mathematics of food webs "may even need new mathematics to deal with these levels and layers of complexity," drawing explicit parallels to the forces that structure physical systems at atomic and astronomical scales.¹

There's a beautiful irony in the political economy of the reintroduction. Ranchers initially opposed wolf reintroduction, fearing livestock predation. But wolves suppress coyotes (the actual primary sheep predator), and may help control chronic wasting disease by removing infected elk from the wild reservoir. The ranching community has been "noticeably silent" on these benefits. Meanwhile, the conservation community risks sanctifying wolves as bringing "only benefits," which isn't supported by the data either. Ecology, like Emergence, resists simple narratives.

The Hidden Half: Parasites as Network Structure

Kevin Lafferty has spent three decades arguing that ecologists have been drawing food webs wrong. Traditional food webs include free-living species — plants, herbivores, predators — and the arrows between them. Parasites, which may constitute nearly half of all animal species, are simply omitted. When Lafferty and colleagues added parasites to the food web of an estuary in Baja California, the number of links and their complexity increased dramatically. This isn't a minor correction; it's a fundamental structural change to the network.²

The reason parasites matter isn't just completeness for its own sake. Parasites actively shape the network's topology. Trematodes in California salt marshes castrate horn snails, hijack killifish brains to make them flash at the surface (increasing predation by herons tenfold), and collectively outweigh all the birds in the ecosystem. Each of these effects alters energy flow through the food web. A trematode that forces a killifish to be eaten by a heron is redirecting energy from one trophic pathway to another — it's an active link, not a passive one.²

The brain-manipulation angle is particularly striking. Toxoplasma gondii makes rats attracted to cat urine (facilitating transmission to its primary host, the cat). It infects about a third of the world's human population, and there's suggestive evidence — not proven, but hard to dismiss — that it may subtly alter human behavior: slower reaction times, increased impulsiveness, possibly explaining some cross-national variation in personality traits. If parasites are manipulating the behavior of their hosts at population scale, then any food web model that ignores them is missing a major control mechanism.²

Lafferty's deeper point is that parasites are indicators of ecosystem health, not just pathology. They thrive where biodiversity is rich and ecosystems are robust. Their abundance tracks the abundance of their hosts, which tracks the overall health of the system. A heavily parasitized ecosystem is, counterintuitively, likely a healthy one — the parasites are there because there's enough energy flowing through enough species to support them. This inverts the usual framing of parasites as problems to be solved and reframes them as diagnostic instruments.

Networks All the Way Down

The Azimuth Project's survey of network theory reveals just how many different fields are independently developing diagrammatic notations for networks of interacting components — systems dynamics from Jay Forrester in the 1950s, Odum's energy systems language for ecology, gene regulatory networks in synthetic biology, Petri nets in chemical kinetics, and SysML for systems engineering. Each field developed its own formalism because the same structural pattern — components connected by flows, with feedback loops and nonlinear interactions — keeps appearing everywhere.³

The deep question, which John Baez and the category theory community have been chasing, is whether there's a single mathematical framework that unifies all of these. If food webs, chemical reaction networks, electrical circuits, and tensor networks are all instances of the same mathematical structure, then techniques developed in one field should transfer to others. Ecological stoichiometry already treats food webs as chemical reaction networks. Control theorists working in behavioral ecology labs already shuttle ideas about abstraction across disciplinary boundaries.³

This matters for the Yellowstone question because the debate about top-down versus bottom-up forces is structurally identical to debates in other network domains. Is an electrical circuit's behavior determined by its source (bottom-up) or its load (top-down)? Is a gene regulatory network driven by upstream signals or downstream selection? The answer, in every case, is both — and the interesting question is how the forces interact, not which one dominates. The mathematical tools to handle this properly may already exist, scattered across half a dozen fields that haven't yet realized they're working on the same problem.