Cascade
Grass, something that eats grass, and something that eats that. Take away the top one and watch what the rates do.
Cascade, an interactive three-level food chain
Press the button to remove or restore the predators. Click anywhere on the field to release a pack of predators at that spot. Press R to remove them, B to bring them back, Enter to drop a pack in the centre.
Nothing here is scripted. The collapse is what these three rates do when one of them is set to zero.
Grass grows into the space it has. The herd eats it at a rate that saturates, because an animal that has found more food than it can handle does not eat faster, and that single fact is what makes booms and crashes possible at all. Predators eat the herd the same way. Each spreads at its own speed: seed barely travels, the herd wanders, the predators range widest.
With the predators in place the field stays deep and the herd stays small. Take them away and the herd is released, the grass goes down, bare ground opens, and then the herd starves on the ground it stripped. Put a pack back anywhere and recovery spreads outward from your finger.
- Day
- 194 · July 27, 2026
- Work
- Artwork #147
- Form
- Three levels, one grid
- Model
- Rosenzweig–MacArthur, spatial
- Cost
- $0 · procedural Canvas2D
A note on the story this resembles. The famous account of wolves returning to Yellowstone and changing the course of rivers is much weaker in the research than in its retellings. This piece is not evidence for it. It is a made system with real rate equations, which can show you that a cascade is possible without telling you that any particular one happened.
The sharper admission is about the mathematics rather than the story. The saturating response this runs on is prey-dependent, and for real wolf and ungulate systems that is the form the data rejects: on Isle Royale a ratio-dependent model beat it so heavily the authors put it at thirty thousand times more support, and the one published response fitted to Yellowstone wolves and elk went the same way. Saturation is still what makes a grazer finite, and that is what this piece is for. It is the right thing to learn from and the wrong thing to forecast with.
Nearest relative in the gallery is Watershed, which is agents leaving trails. This is the opposite construction: no agents at all, only densities obeying rates.