Reliability of morphometric measures for determining the sex of adult and fledgling shy albatrosses, Diomedea cauta cauta, in Australia

Abstract

Dissection and measurement of the carcasses of shy albatrosses, Diomedea cauta cauta, provided morphometric data from known-sex birds from fledgling age to adulthood. These data were used to determine the efficacy of external measures for reliably identifying the sex of birds in the field. As fledglings were smaller than adults and subadults in all measures, separate age-specific equations were developed for assigning their sex. Sexual dimorphism was evident in all head, beak and leg measurements for the adult and subadult birds, with males being significantly larger than females, while there were no sex differences in measures of the wings and tail. Beak and head measurements, along with the weight of fledglings, were taken, and sex differences were evident for all parameters. A stepwise discriminant function analysis of the seven head and beak measures (head length, head width, basal bill width, culmen length, basal bill depth, minimum bill depth and upper bill depth) indicated that 98% of adult and subadult birds could be correctly sexed by measuring the upper bill depth and head width, whilst the sex of 89% of fledglings could be discerned by measuring head length and width and the minimum bill depth. Discriminant scores overlapped less for adults–subadults than for fledglings, indicating that external measurements may be used more reliably to sex this age-class. Data from three experienced observers indicated significant differences in the morphometric measures taken from the same individual albatrosses. The extent of these differences, however, varied greatly depending on the measure taken (i.e. beak v. wing measures), with head and beak measures showing the least inter-observer differences. Such differences resulted in a decrease in the success rate of the adult–subadult discriminant function analysis from 98 to 90%. When working with breeding birds, sex-allocation errors can be reduced by measuring both members of a pair and allowing the relative size of the discriminant score to identify the individuals concerned.