Home >  Methods >  Avian Data Analysis >  Avian Detection Bias >  Perception

'Perception' bias: distance-sampling models of the effective detection radius (EDR)

The distance of animals from surveyors (detection distance) is probably the most ubiquitous source of perception bias during animal surveys. BAM accounts for this by using distance sampling to analyse the subset of surveys where avian detections are recorded relative to two or more distance intervals. Distance sampling models the perception rate as a decreasing function of the detection distance (Fig). The resulting function is then used to calculate the proportion of birds that are missed by the survey because they were not perceived by the observers (Buckland et al. 2001) .

.

The probability of detecting available birds as a function of their distance from observers for 102 species of boreal songbirds is shown here. The marks on the x-axis denote the effective detection radius for each species.

The perception rate can be conveniently expressed as the effective detection radius (EDR). The effective detection radius results are posted on the Life History page for each species. The EDR is the distance from the sampling point at which as many birds are detected beyond EDR as remained undetected within EDR. This allows one to measure the effective area sampled by the survey. This is important because it allows researchers to estimate densities, which can then be extrapolated to population sizes. This distinguishes distance sampling from all other survey methods that are based solely on multiple observers, visits, or time intervals.          

Because the area surveyed increases with the square of radius, even small differences in EDR can result in large differences in density and population estimates. For example, a 10% decrease in EDR from non-forest habitats to closed-canopy forests in the boreal translates to a 24% difference in density, given equivalent survey counts. BAM is therefore including a tree cover variable in our distance sampling models to control for habitat effects on perception rates.

BAM has also found that the EDRs for boreal songbirds are consistently smaller than the maximum detection distances (MDD) used by Partners in Flight to estimate continental population sizes of landbirds (Rosenberg and Blancher 2005,   Partners in Flight 2007 ). This difference results in abundances estimated using EDR that average 5 times those resulting from abundances estimated using MDD. Replacing MDD with EDR should therefore increase population estimates for boreal birds.

Finally, BAM uses distance sampling to standardize density estimates relative to variation in the survey radius among the contributing projects. This is important because the raw survey counts consistently increase with the radius of the surveys. For example, raw counts of boreal birds increase on average by 31% from a 100-m, fixed-radius survey to an unlimited-distance survey. After adjusting the counts using EDR, densities differ by only 1% between these survey radii.