Pooling robustness in distance sampling: Avoiding bias when there is unmodelled heterogeneity

Area of Science:

• Ecology
• Wildlife Biology
• Statistical Ecology

Background:

• Distance sampling provides unbiased abundance estimates for total populations, even with unmodeled detection probability variations.
• However, this pooling robustness does not extend to subpopulation abundance estimates when subpopulations are ignored.
• Differences in detectability between subpopulations can introduce bias in abundance estimates.

Purpose of the Study:

• To investigate the bias in subpopulation abundance estimates caused by differing subpopulation detectability.
• To compare abundance estimates using a single pooled detection function versus a model incorporating a subpopulation covariate.
• To evaluate the impact of pooling robustness on distance sampling in ecological studies.

Main Methods:

• Simulation studies were conducted to assess the effect of detectability differences on bias.
• Analysis of multispecies songbird point transect survey data was performed.
• Species-specific abundance estimates were compared using pooled detection functions and a species-covariate detection function model.

Main Results:

• Simulations confirmed bias in subpopulation abundance estimates when using a pooled detection function, positively related to detectability disparities.
• Overall population abundance estimates remained unbiased, barring extreme heterogeneity in detection functions.
• Including a subpopulation covariate in the detection function model effectively removed subpopulation abundance estimate bias.

Conclusions:

• Pooling robustness in distance sampling ensures unbiased total population estimates but not subpopulation estimates.
• Subpopulation-specific detection functions are crucial for accurate subpopulation abundance estimation.
• Using subpopulation as a covariate in detection function models is recommended to mitigate bias in data-poor subpopulations.