“This paper describes a new methodology that has broad applicability to animal sampling problems involving many taxa and sampling methods.”

Conventional wisdom holds that it is not possible to estimate abundance or density from simple counts in the presence of imperfect detection. I think this paper is receiving attention because it contradicts this view, and because it provides a general inference framework for data that are widely collected in animal sampling problems.

This paper describes a new methodology that has broad applicability to animal sampling problems involving many taxa and sampling methods.

One of the major concepts in animal population sampling is that of imperfect detection, that not all animals in a sampled population are detected. Much of contemporary animal population sampling is concerned with this problem, its effects, and solutions for estimating or otherwise accommodating imperfect detection either by sampling design, or use of statistical models. One of the most common types of data collected in animal sampling are simple counts of individuals, typically for a number of spatially referenced sample units. The utility of simple count data, referred to as "point counts" in bird sampling, has been debated extensively, in part because of the perception that the data cannot be used to draw inferences about abundance in the presence of imperfect detection. This paper describes a framework for modeling and estimating both abundance and detectability from simple point count data.

I have had the privilege of working with many extremely talented biologists and other scientists in the USGS and the USFWS, and my research has always been strongly influenced by these collaborations. This specific paper arose out of a collaborative effort, several years ago, to develop models and inference methods for species occurrence, which spawned a USGS/FWS collaborative field project to evaluate methods of sampling waterfowl broods in the Prairie Pothole Region. Attempting to understand the relationship between occurrence and density led directly to the N-mixture model for simple counts.