Camera-Trap Occupancy Survey

Reserve Setup

Camera-trap grid 5 × 5 (25 sites)

True occupancy ψ (habitat suitability) 55%

Detection probability p (activity / sensitivity) 40%

Survey

Live Estimates

Occasions run K

Naive occupancy

—

MLE occupancy ψ̂

—

MLE detection p̂

—

True ψ (hidden)

? — click reveal

Camera trap — no detection yet Camera trap — detected ≥1 visit Wildlife (moves only on occupied sites)

How it works

Not every occupied site gets photographed on every visit — a shy animal can walk past a camera and never trigger it. Counting only "sites with a photo" therefore underestimates true occupancy. Wildlife ecologists fix this with the MacKenzie single-season occupancy model, the real method behind most camera-trap and anti-poaching monitoring programs.

Naive estimate:  ψ_naive = (sites with ≥1 detection) / (total sites)

Detection history per site (K visits, x detections):
  P(x>0) = ψ · C(K,x) p^x (1-p)^(K-x)
  P(x=0) = ψ(1-p)^K + (1-ψ)     ← occupied-but-missed, or truly absent

Maximum-likelihood estimate:
  (ψ̂, p̂) = argmax  Σ log P(x_i | ψ, p)

Grid — sets how many camera-trap sites cover the reserve.
True occupancy ψ — the real (hidden) fraction of sites the species actually uses; this drives the simulation but is not shown to the estimator.
Detection probability p — chance a present animal trips a camera on any single visit; low p (shy species, sparse cameras) is exactly why naive counts fail.
Run Occasion — simulates one more independent survey visit across every camera; Auto-Run repeats this continuously.
Watch ψ̂ climb toward the true ψ as K grows, while the naive estimate stays biased low — the more survey effort, the better rangers can target patrols and protection at the sites that truly matter.

0 confirmed occupied sites 25 total sites

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.