An Evaluation of Regression Methods to Estimate Nutritional Condition of Canvasbacks and Other Water Birds

Regression equations that use mensural data to estimate body condition (i.e., the general health of a bird based on body reserves of fat or protein) have been developed for several water birds. These equations often have been based on data that represent different sexes, age classes, or seasons, without being adequately tested for intergroup differences. When compared to methods using total fat or moisture content, mensural equations frequently provide poor fits to measures of condition, or use body measurements that do not appreciably increase a model's precision. We used proximate carcass analysis of 538 adult and juvenile canvasbacks (Aythya valisineria) collected during fall migration, winter, and spring migrations in 1975-76 and 1982-85 to test regression methods for estimating body condition. We weighed, measured, and analyzed each canvasback for body fat, protein, and ash. Analyses of covariance provided estimates of total extractable fat(R2 = 0.71), a condition index (total fat divided by fat-free dry mass; R2 = 0.64), and protein (R2 = 0.74) that could be accounted for by several explanatory variables. We adjusted each regression equation for significant effects of age-sex classes and seasons. Body mass, season, and age-sex class were useful in explaining condition, but mensural characters added only 1-3% to accountable variance.