Methods in spatially explicit capture-recapture

Abstract

Capture-recapture (CR) methods are a ubiquitous means of estimating animal abundance from wildlife surveys. They rely on the detection and subsequent redetection of individuals over a number of sampling occasions. It is usually necessary for individuals to be recognised upon redetection. Spatially explicit capture-recapture (SECR) methods generalise those of CR by accounting for the locations at which each detection occurs. This allows spatial heterogeneity in detection probabilities to be accounted for: individuals with home-range centres near the detector array are more likely to be detected. They also permit estimation of animal density in addition to abundance.

One particular advantage of SECR methods is that they can be used when individuals are detected via the cues they produce---examples include birdsongs detected by microphones and whale surfacings detected by human observers. In such situations each cue may be detected by multiple detectors at different fixed locations. Redetections are then spatial (rather than temporal) in nature, and density can be estimated from a single survey occasion.

Existing methods, however, cannot generally be appropriately applied to the resulting cue-detection data without making assumptions that rarely hold. Additionally, they usually estimate cue density rather than animal density, which does not usually have the same biological importance. This thesis extends SECR methodology primarily for the appropriate estimation of animal density from cue-based SECR surveys. These extensions include (i) incorporation of auxiliary survey data into SECR estimators, (ii) appropriate point and variance estimators of animal density for a range of scenarios, and (iii) methods to account for both heterogeneity in detectability and cues that are directional in nature.

Moreover, a general class of methods is presented for the estimation of demographic parameters from wildlife surveys on which individuals cannot be recognised. These can variously be applied to CR and---potentially---SECR.