Abstract
We investigated the precision and accuracy of an infrared burrowscope for detecting sooty shearwater (Pufffinus griseus) chicks at 13 plots from 3 islands in southern New Zealand in 2003. We partially excavated burrows systems to reveal the entire burrow contents after 2 teams of observers had prospected all burrow entrances. Accuracy was similar between islands and observer teams at approximately 85%. The majority of the inaccuracy stemmed from failure to detect some chicks. Logistic regression modeling identified 4 burrow characteristics occurring between the entrance and the nest-site that influenced detection of burrow occupants. Detection was lower at nest-sites further from burrow entrances, in burrows with a high rate of burrow division, and in burrows with a high level of curvature. There was a positive relationship between the interaction of rate of division and curvature and detection of chicks. Distance from the burrow entrance was the only parameter that could be reliably used as a predictor of detection rate, so a reduced model containing only this variable was constructed to correct for burrowscope bias. The correction factor performed well on The Snares and Bench Island where predicted bias was very similar to observed levels (within 5%), but bias was overestimated on Putauhinu by up to 19.1%. Consistent bias, lack of damage to burrows from excavation, and the successful application of a correction factor all indicate the value of further testing burrowscope accuracy on other burrow-nesting seabird species.