Distance sampling detection functions: 2D or not 2D?

Area of Science:

• Ecology
• Wildlife Biology
• Statistical Ecology

Background:

• Conventional distance sampling (CDS) assumes uniform animal distribution, which is often violated by animal movement responses to observers.
• Non-random placement of survey lines or points can also compromise CDS assumptions.
• Existing methods struggle with unknown, non-uniform animal distributions near survey transects.

Purpose of the Study:

• To develop a distance sampling method that accounts for non-uniform animal distributions and responsive movement.
• To estimate animal density when the assumption of uniform distribution is violated.
• To assess the reliability of time-to-detection data in distance sampling.

Main Methods:

• Formulated distance sampling models as survival models.
• Incorporated time to first detection alongside distance (perpendicular or radial) into models.
• Developed a maximum likelihood estimator for detection probability and effective strip half-width.
• Investigated estimator properties through simulations with non-uniform distributions.

Main Results:

• The new method allows density estimation even when animal distribution is non-uniform and unknown.
• Time to detection provides insights into the failure of the CDS assumption of detection probability being 1 at distance zero.
• The estimator performs well when detection probability is 1 at distance zero, yielding unbiased density estimates with prior responsive movement.
• Estimates may show bias with continued responsive movement but are likely less biased than traditional CDS.

Conclusions:

• Time to detection is a valuable metric for enhancing distance sampling accuracy.
• This approach offers a more robust alternative to CDS for wildlife density estimation in complex scenarios.
• The method was successfully illustrated with primate and dolphin survey data exhibiting avoidance and attraction, respectively.