How does choice of statistical method to adjust counts for imperfect detection affect inferences about animal abundance?

Abstract

There is ongoing debate in ecology about the value of the increasing complexity of analytical methods, especially as it relates to models that correct occupancy or abundance estimates for imperfect detection. While both arguments for and against increasing complexity have merit, there is a need for greater clarity on how to determine what level of complexity is necessary. We present a general approach and case study for comparing alternative detection methods that vary in their complexity. Our approach puts emphasis on the logistical costs of methods, which are often overlooked in the debate about method complexity, and on developing models that address common sources of error in ecological data sets while avoiding unwarranted complexity. We used point counts of saltmarsh sparrows (Ammodramus caudacutus) to compare estimates of abundance from three alternative protocols that vary by logistical costs: single observer, multiple observers and multiple visits. We also compared results from counts to those from captures and nest searches from the same populations to provide broader context for the evaluation of point count methods. We found that parameter estimates derived from alternative count protocols were similar and that predictions of point-level abundance were highly correlated (r = 0·96). We found lower correlation between pairwise comparisons of abundance estimated from point count data, the number of individuals captured and the number of nests (all comparisons r = 0·50 or less). Choosing point counts over alternative measures of abundance may have a greater effect on inferences than choosing among specific count protocols. For saltmarsh sparrows, there is likely little added benefit to adopting count protocols that require additional logistical costs. Determining the frequency of cases like this has broad implications for the appropriate design of studies that rely on estimates of abundance, especially when resources are limited. The general approach we present can be used to assess whether general rules of thumb can be developed to benefit people charged with implementing field studies and allocating limited resources.