Northern Prairie Wildlife Research Center
|Figure 1. Three general locations where six species of diving ducks were collected during fall, winter, and spring, January 1992-April 1993.|
All food items were identified to genus or species and number of individual items counted. Frequency of occurrence (number of waterfowl with a particular food taxon divided by the total number of waterfowl) and aggregate percent (proportion of each food item in each bird averaged for all individuals; Swanson et al. 1974) were calculated for each waterfowl species. When the upper GI tract contained only animal material (n = 74), we calculated aggregate percent based on numerical data, rather than dry mass or volume, because food items were generally the same size (average length of zebra mussels = 8.1 mm, isopods = 7.8 mm, amphipods = 5.5 mm, caddisflies = 8.8 mm) and because molluscs were an important component of the diet. When upper GI tract samples contained both plant matter and animal matter (n = 8 Redheads, 2 Buffleheads), volumetric measurements of food items were taken and the volumes used to calculate aggregate percent (Bartonek and Hickey 1969, Gammonley and Heitmeyer 1990). Aggregate percent based on numbers is preferable to aggregate percent based on dry mass when molluscs are an important component of the diet because dry mass inflates the importance of molluscs in the diet due to the proportionately large mass of undigestible shell material. Additionally, numerical counts are less prone to measurement error than either the dry mass or volume methods, which is important when measuring small volumes or dry masses as in our study. Aggregate percent based on dry mass does facilitate making energy and nutritional inferences. Dry mass of zebra mussels in our study can be estimated based on their shell length (Draulans 1982). Even though methods to determine percent composition of diet differs between studies, we follow the common practice of comparing our data to other studies (Afton et al. 1991, Dirschl 1969, Gammonley and Heitmeyer 1990, Hoppe et al. 1986).
Shell length of ingested zebra mussels ≥5 mm was measured to the nearest mm and those < 5 mm were measured to the nearest 0.2 mm. All zebra mussels were measured along the longest axis of the shell (Hamilton 1992b). Average size of zebra mussels was calculated for each duck and those means used for statistical comparisons. An individual duck was the measurement unit for zebra mussel size comparisons among species, ages, sexes and locations.
Data were analyzed in a step-wise manner. We used analysis of variance (ANOVA) on aggregate percent and zebra mussel size data. Bartlett's tests were used to test for homogeneity of variances prior to each ANOVA. If variances were not equal, data were rank transformed, which in all cases equalized the variances. Untransformed means ± 1 SE are presented in text and tables. When main factors and the interaction term in 2-way ANOVAs were non-significant, these factors were combined in subsequent 2-way or 1-way analyses. Similarly, sequential chi-square and/or Fisher's Exact tests were used on frequency-of-occurrence data; sexes, ages, and seasons were combined in subsequent analyses as these factors were found to be not significantly different. Differences in size distributions of zebra mussels consumed by waterfowl were tested with Kolmogorov-Smirnov two-sample tests.