Home >  Methods >  Avian Data Analysis >  Avian Detection Bias


Detection bias occurs during avian surveys when not all of the birds present are detected by surveyors. This common bias results in raw survey counts that underestimate the number of birds present in survey areas.

The ability to detect a bird is affected by factors that can be categorized as the 'availability' of cues to detect a species, and 'perceptibility' of the cues by the observer.

Availability bias.-The birds present at a survey location may vary in the rate that they are giving detectable cues which affects their availability to be detected during a survey. For example, a female bird quietly sitting on a hidden nest is not available for detection because she cannot be heard or easily seen, while her brightly-colored mate singing from a conspicuous perch is available for detection. BAM deals with availability bias by estimating avian singing rates using a removal model (Availability Bias).

Perception bias.-Birds may be singing and therefore available for detection. However, not all of the available individuals will be detected by the surveyors. For example, a bird singing from a perch 10 m from a survey point is more easily perceived by a surveyor than a bird singing from a perch 100 m away. Similarly, a bird with a loud song will be perceived more often than a softly-singing bird at the same distance. BAM deals with perception bias by estimating the effective detection radius (EDR) using distance sampling (Perception Bias).

Overall detection rates.-The overall detection rate can be viewed as the product of availability and perception rates.  Each of the available estimators of avian abundance do not deal adequately with both of these components of bias inherent in the detection process. Therefore, BAM is using a combination of removal models and distance sampling to minimize bias in our estimates of bird densities. Density estimates are summarized by species under RESULTS.