Based on the Pleistocene Park rewilding hypothesis (Zimov et al.): in winter, snow is an excellent insulator, so it keeps the ground beneath it warmer than the air above. Large herbivores — mammoths, bison, horses — trample and compact the snow as they graze, destroying that insulating layer and letting the ground lose heat straight to the cold winter air.
T_eq(cell,t) = T_air(t) + S(t)·I_max·(1 − c)
dT_soil/dt = (T_eq − T_soil) / τ
T_air(t) = T_mean + A·cos(2π·yearFrac) [seasonal cycle]
S(t) = snow-cover fraction, 1 in winter → 0 in summer
c = local snow compaction from trampling, 0–1
I_max = maximum insulating bonus of undisturbed snow (°C)
τ = soil thermal-inertia time constant (years)
- Population target — how many megaherbivores the landscape carries; the herd grows or shrinks toward this target over time.
- Trampling intensity — how strongly each animal compacts snow as it passes; higher values erase the insulating layer faster.
- Climate warming offset — raises the mean annual air temperature, the background pressure the herd is working against.
- Frozen fraction — the share of the grid still colder than −1°C, a proxy for how much of the ground stays as stable permafrost rather than thawing and releasing its stored carbon.
This top-down 2D map shows the exact same cell grid and herd model as the 3D version — drag to pan, scroll (or pinch) to zoom, and watch trampled patches (lighter, browner) lose their snow cover while undisturbed ground (bluer) stays insulated through the winter.
This is the same feedback loop behind real rewilding-as-climate-mitigation proposals: reintroducing grazers is proposed as a low-tech way to keep permafrost (and the carbon locked inside it) frozen for longer, buying time against a warming climate.