<b>unmarked</b>: An<i>R</i>Package for Fitting Hierarchical Models of Wildlife Occurrence and Abundance

TLDR

Ecological studies of unmarked wildlife rely on survey methods that are prone to measurement error, and hierarchical models have been developed to separately model latent abundance or occurrence and detection processes, gaining popularity for their interpretable structure. The R package unmarked provides a unified modeling framework for hierarchical models of wildlife occurrence and abundance. It offers tools for data exploration, model fitting, criticism, post‑hoc analysis, and model comparison.

Abstract

Ecological research uses data collection techniques that are prone to substantial and unique types of measurement error to address scientific questions about species abundance and distribution. These data collection schemes include a number of survey methods in which unmarked individuals are counted, or determined to be present, at spatially- referenced sites. Examples include site occupancy sampling, repeated counts, distance sampling, removal sampling, and double observer sampling. To appropriately analyze these data, hierarchical models have been developed to separately model explanatory variables of both a latent abundance or occurrence process and a conditional detection process. Because these models have a straightforward interpretation paralleling mechanisms under which the data arose, they have recently gained immense popularity. The common hierarchical structure of these models is well-suited for a unified modeling interface. The R package <b>unmarked</b> provides such a unified modeling framework, including tools for data exploration, model fitting, model criticism, post-hoc analysis, and model comparison.