Understanding implications of detection heterogeneity in wildlife abundance estimation

Abstract

Negative bias in mark‐recapture abundance estimators due to heterogeneity in detection (capture) probability is a well‐known problem, but we believe most biologists do not understand why heterogeneity causes bias and how bias can be reduced. We demonstrate how heterogeneity creates dependence and bias in mark‐recapture approaches to abundance estimation. In comparison, heterogeneity, and hence estimator bias, is not as problematic for distance sampling and mark‐resight methods because both techniques estimate detection probabilities based on a known quantity. We show how the introduction of a known number of individuals planted into a study population prior to a mark‐recapture survey can reduce bias from heterogeneity in detection probability. We provide examples with simulation and an analysis of motion‐sensitive camera data from a study population of introduced eastern wild turkeys (Meleagris gallopavo silvestris) of known size with a subset of telemetered birds. In choosing a method for abundance estimation, careful consideration should be given to assumptions and how heterogeneity in detection probability can be accommodated for each application.

Comparing methods to estimate feral burro abundance

“…Heterogeneity in detection probability is a consistent problem in mark‐recapture that must be appropriately accounted for to remove positive bias in abundance estimates (Laake and Collier 2024). Both approaches we used to accommodate residual heterogeneity in our DOS dataset resulted in more accurate estimates compared to the standard DOS data, but at the expense of survey precision.…”

“…Rather than change the survey type, accuracy can be improved within a DOS sampling framework by accounting for residual heterogeneity in detection probability through detections of radio‐collared individuals (Griffin et al 2013, Hennig et al 2022) or by analyzing the dataset with a mark‐recapture distance sampling (MRDS) approach (Borchers et al 2006, Laake et al 2011, Burt et al 2014). Heterogeneity in detection probability is problematic because the sample used to estimate the observation process includes the most detectable individuals or groups; thus, estimates of detection will be biased high (Laake and Collier 2024). If the sample population contains a subset of radio‐collared individuals, one can add another capture history in the detection models that corresponds to detection of radio‐marked animals (Griffin et al 2013, Hennig et al 2022).…”

Abundance estimates of Gunnison's prairie dogs compared to the number of active burrows

“…The negative bias in the CMR estimates may have been a product of unmodeled heterogeneity in detection probabilities (Laake and Collier 2023). Abundance estimates based on mark-resight approaches are typically less biased because detection probabilities are based on a known number of marked animals in the population (Laake and Collier 2023). We have also found that CMR was biased low relative to mark-resight for black-tailed prairie dogs (Facka et al 2008).…”

“…We found that CMR underestimated the MNKA, although it was correlated with both MAGC and mark-resight estimates. The negative bias in the CMR estimates may have been a product of unmodeled heterogeneity in detection probabilities (Laake and Collier 2023). Abundance estimates based on mark-resight approaches are typically less biased because detection probabilities are based on a known number of marked animals in the population (Laake and Collier 2023).…”

“…We captured an estimated 26% of the populations occupying our plots. Population estimates based on CMR might be improved if a known number of marked animals can PRAIRIE DOG ABUNDANCE ESTIMATION | 13 of 17 be introduced prior to initiating a CMR effort (Laake and Collier 2023), or if a greater proportion of the population was captured, but both approaches would require increased time and effort. In contrast, MAGC can be used as an index of abundance especially if sighting probabilities can be corrected by directly estimating them at each site (as we have done here) or modeled using relevant factors (e.g., vegetation height, shrub density, and slope) expected to influence sightability (Powell et al 1994, Anderson 1996, Craig and Reynolds 2004.…”