Reduction of bias when estimating bird abundance within small habitat fragments

We used the distance detection function from five-minute point counts entirely within large woody vegetation patches to derive a method of truncating counts of birds detected close to the observer to estimate their relative abundance in small habitat patches. Our method trades off loss of information by truncation of bird sightings at successively larger distances from the observer to reduce sampling bias. Truncation of counts to include detections within 10 m of the observer gave similar absolute density as distance methods for the six most abundant native and six introduced species. Distance analysis showed that introduced species were in general more conspicuous than New Zealand native species. Use of counts very close to the observer reduces detectability biases for species and habitat comparisons to give more robust measures of community structure, allows inclusion of very small habitat fragments into the analysis, and provides a density measure for infrequently encountered species. However, the counts are still best treated as relative indices rather than absolute density estimates. Much of the international literature using counts and distance sampling estimation methods to claim increased bird diversity and abundance in larger habitat patches may be unreliable because these include directional biased estimation of abundance in small patches.