Detectability

Absence versus non-detection

As you know from your field experience, no survey method provides a 100% guarantee of detecting a species on a particular visit

It’s common to hear people talk about presence/absence surveys, but it’s important to be clear that if you fail to record the species on your survey, it’s not necessarily absent from that site

Failure to record the species is not proof of absence, as it could be explained either by:

  • True absence: The site is not occupied by that species
  • Non-detection: Failure to detect the species even though it is actually present at the site

Standardised survey effort

In the past, ecologists focussed on standardising survey methods to keep detectability constant, so that it was possible to compare data from different locations, times or field teams

With distance sampling, we explicitly estimate detectability so that these comparisons are more robust

This means we can be more certain the differences in our estimates of the state variable such as density are due to genuine temporal/spatial variation in density, and not just due to differences in our ability to detect the species

For example:

  • An apparent decline in population size could actually be the result of the species becoming more nocturnal in areas of human encroachment - it is just as common, but has become more difficult to detect during daytime surveys
  • We could wrongly conclude that a species prefers open habitat simply because it’s harder to spot in dense vegetation

What is detectability?

Estimating detectability is key to distance sampling1

A formal definition of detectability:

“The probability of detecting a species, given that it is present at a site”

We will use the terms ‘detectability’ and ‘detection probability’ interchangeably during this course

Implicit assumptions

If you don’t measure detectability, you are implicitly assuming that:

  • Detection is certain (i.e. detection probability = 1), or
  • Detection is not certain (i.e. detection probability < 1), but it is constant across all distances from your transect

However, it’s not realistic to assume that detection is certain! It’s also unlikely that detectability is constant at all distances from the observer - it’s far more likely that the chance of you detecting an animal decreases the farther away it is from you

Many factors can influence detectability, including:

  • Terrain and land-cover characteristics
  • Observer experience
  • Weather
  • Activities of other species including humans
  • Time of day etc

Estimate detection probability

By estimating detection probability instead of assuming that it is 1 (or <1 but constant), you allow yourself to determine:

  • How detection varies with distance from the observer
  • How detection probabilities vary in space or time

By knowing the detection probabilities, you can make a more robust estimate of density1 because you can correct for the effects of varying detectability in your analysis

You can also figure out whether detectability is affected by a factor that also influences density, which allows you to make better inferences about density because you can separate out the detection process from the species’ actual abundance