Theoretical ecology · Spatial simulation
Food webs under habitat loss
Theoretical-ecology research: a spatially-explicit simulation of multitrophic communities — food webs living on a landscape — used to show that the pattern of habitat loss, not just how much is lost, decides whether an ecosystem stays stable. Published in Nature Communications.
A run of the model. The four panels track individuals of each trophic level — plants, herbivores, omnivores and predators — across a landscape; the graph below follows each population over time as the food web settles. Removing habitat in different spatial patterns changes whether it holds together.
The question
Habitat loss is one of the biggest drivers of biodiversity decline — but ecosystems aren't collections of independent species; they're food webs, where nearly everything eats, or is eaten by, something else. This work asked about the whole web at once: when habitat disappears, what happens to the stability of the entire multitrophic community — and does it matter how the habitat goes, scattered at random or taken in one contiguous block?
The model
The approach was a spatially-explicit simulation — the run in the video above. A community of many species across four trophic levels — plants, herbivores, omnivores and predators — lives on a landscape; individuals move, feed and reproduce according to who-eats-whom, and the food web assembles and persists (or collapses) out of those local interactions. Habitat is then removed under different spatial patterns and the community's response measured.
It's complex-systems modelling in its purest form: no single equation for ‘stability’, just many simple agents interacting, with the system-level behaviour emerging from the bottom up — and only visible once you run it at scale.
The finding
The pattern of loss mattered as much as the amount. Contiguous habitat loss — carving the landscape away in a block — was consistently more destabilising than the same fraction removed at random, because it acted differently on species' mobility, on diversity and network structure, and on the strength of the interactions holding the web together.
The work was published in Nature Communications (McWilliams et al., 2019). The through-line to everything I do now is already here: understand the system as a system, model it honestly from the mechanisms up, and let the scale of computation show what closed-form maths can't.