PatchOccupancyModelle

PatchOccupancyModelle, often referred to in English as patch occupancy models, are statistical tools used in ecology to estimate whether a species occupies discrete habitat patches while accounting for imperfect detection. The core idea is to separate the true occupancy state of a patch from the observed detections gathered during surveys.

In a basic patch occupancy model, each patch i has a latent occupancy state z_i that equals 1 if the patch is occupied and 0 otherwise, with occupancy probability psi_i. The data consist of detections y_i,t across survey occasions t. Detections are modeled as Bernoulli trials with probability p_i,t conditional on occupancy, i.e., y_i,t ~ Bernoulli(p_i,t * z_i). If a patch is unoccupied, detections are impossible; if occupied, detections depend on detectability.

Dynamic occupancy models extend the framework over multiple time periods by introducing colonization and extinction processes. In these models, occupancy can change across occasions, governed by colonization probability gamma and extinction probability epsilon. This allows researchers to study persistence, range expansion, and responses to management or environmental change.

Data and covariates play a central role: psi, p, gamma, and epsilon can be linked to patch-level and temporal covariates such as patch size, habitat quality, or landscape configuration. Assumptions typically include closure within survey occasions, independence of detections given occupancy, and limited false positives.

Estimation is commonly performed through likelihood-based methods or Bayesian approaches, using software such as unmarked in R, PRESENCE, JAGS, or Stan. PatchOccupancyModelle are widely used in wildlife monitoring, conservation planning, and habitat management to quantify occupancy while accounting for imperfect detection and to inform decision making. Variants include multi-species occupancy and more complex hierarchical structures.