Optical satellites (Sentinel-2 ≈5 days, Landsat ≈16 days) only image a given tile on a fixed revisit interval T. On any given pass the tile may be hidden under clouds with probability p — a real, persistent limiter for tropical forest monitoring, where cloud cover regularly exceeds 60–80%.
P(pass is clear) = 1 − p
E[passes until clear] = 1 / (1 − p) (geometric distribution)
E[latency to 1st clear pass] ≈ T / (1 − p)
Even a clear pass isn't enough: single-pixel noise and classification error mean an operational pipeline only confirms a clearing once the contiguous cleared area exceeds a minimum mapping unit (MMU) — real programs like Global Forest Watch use MMUs on the order of a Landsat/Sentinel pixel cluster, roughly 0.1–1 ha, to suppress false alarms.
- Revisit interval — days between satellite passes over this tile.
- Cloud-cover probability — chance any single pass is blocked and produces no usable image.
- Minimum mapping unit — contiguous cleared cells required before a clear pass counts as a confirmed detection.
- Simulation speed — how many simulated days elapse per real second.
This 2D top-down view shows the exact same discrete-event model as the 3D version — a clearing that spreads outward from a seed cell across a grid, monitored by a satellite ground track orbiting the tile on a fixed schedule — redrawn as a flat map so the frontier growth, cloud blackouts and detection threshold are easier to read at a glance. Drag to pan, scroll or pinch to zoom.