Sampling and Analysis Methods for Snow-Urine Nutritional Assays

Andrew Charles Pils

Investigations with captive and free-ranging cow elk have demonstrated that urinary allantoin:creatinine (A:C) ratios can be used as an index of dietary intake. Practical management applications of the technique, however, would require random collections of snow-urine, resulting in unknown proportions of samples from various sex-age classes. The goal of this research was to develop sampling protocols and statistical methods for analyzing A:C ratios derived from random collections of snow-urine. Snow-urine samples (n=2,051) were collected from cow, calf, and bull elk in the Madison-Firehole drainage of Yellowstone National Park, during five consecutive winters from 1991-92 through 1995-96. Significant differences were detected in mean A:C ratios among sex-age classes. Computer simulations using these data demonstrated that mean A:C ratios derived from random sample collections were highly variable and consistently overestimated the true mean A:C ratio of cow elk. These problems were somewhat alleviated by the use of trimmed means. Trimming 15% of the right tail and 20% of the left tail of the ordered sample distribution resulted in mean A:C ratios with consistently less bias, lower variability of sample means, and smaller confidence interval widths than did untrimmed means. Additional simulations and field experience suggested that a realistic sampling goal would be the collection of 20 samples at two-week intervals throughout the winter. This sampling protocol and the use of trimmed means was tested during the winter of 1996-97 by collecting a total of 1,008 samples from six elk herds that were managed for three different goals and were expected to have different nutritional levels. More differences in average overwinter A:C ratios were detected with trimmed means than with untrimmed means. Differences in average overwinter A:C ratios that were consistent with expected differences in the nutritional levels among herds were also detected among the six herds, indicating that when used with the sampling protocol and statistical methods detailed in this paper, A:C ratios were sensitive enough to detect important differences in the nutritional levels among herds.