Habitat Percolation: Patch Map and Size Distribution (2D)

Remove habitat cell by cell and find the sharp threshold near 59 % where the connected corridor breaks into isolated patches; a fox walks inside the largest patch.

Switch to 3D
Controls

Presets

Parameters

Model checks

The science

Percolation. A landscape is a square grid in which each cell is habitat with probability p (the rest is matrix: fields, roads, water). Two habitat cells belong to the same patch if they touch (4 neighbours, or 8 if the animal can cut corners). For site percolation on the square lattice there is a sharp threshold pc = 0.5927 (4 neighbours; a numerical constant, not a closed form) and 1 - pc = 0.4073 for 8 neighbours. Below it habitat is a dust of small patches; above it one giant patch spans the whole landscape and an animal can walk from one side to the other. Near pc the patch-size distribution is a power law, so a small loss of habitat can cut a corridor abruptly.

Nested landscapes. One random number u per cell is drawn once; a cell is habitat when u < p. Moving the slider therefore adds or removes cells without reshuffling, and the whole curve of ‘largest patch vs p’ is computed in a single pass that inserts cells in order of u and merges patches with union-find. The threshold of this one landscape is the p at which a corridor first appears; it fluctuates around the infinite-lattice value by a finite-size amount.

Exact facts used as checks. A single isolated habitat cell has density p(1-p)z (z = 4 or 8 empty neighbours). On a finite rectangle either habitat spans top to bottom, or the matrix spans left to right with the complementary neighbourhood (8 for 4 and vice versa) — never both and never neither (planar duality). Union-find is compared with an independent flood fill. At pc the crossing probability is close to one half.

Simplifications. A neutral random landscape has no spatial autocorrelation; real habitat is clustered, which moves the threshold. The fox walks randomly inside the largest patch and is only an illustration of connectivity; real dispersal has distance kernels, mortality in the matrix and behaviour. The site-percolation threshold is finite-size blurred on grids this small.

-
-