Nutritional Condition Indices for Elk : The Good ( and Less Good ) , The Bad , and The Ugly

Research on captive ruminants has clearly established the role of nutrition on virtually all aspects of individual and herd productivity, but assessment of nutritional effects on population dynamics of freeranging ungulates is rare. Understanding influences of nutrition on wild ungulate herd demographics has been limited by a lack of practical, reliable, and cost-effective techniques for monitoring elk condition and nutrition (Cook 2002). Assessing nutritional quality of forage is difficult, unreliable, and expensive, whereas assessing nutritional condition of animals has been impractical in the field, inaccurate, or inadequately tested (Robbins 1983, Harder and Kirkpatrick 1994, Saltz et al. 1995). The most rigorous approach to test the value of condition indices involves comparing various indices to actual fat and protein levels of the homogenized carcass. Statistical analysis of indices generally involved correlation between the indices and a body component, usually ingesta-free body fat. Non-linear relationships often were transformed to facilitate analysis using general linear models (e.g., Finger et al. 1981, Watkins et al. 1991). The final value of indices usually was determined via comparison of correlation coefficients or coefficients of determination. However, these methods of analysis often are incomplete, leaving unanswered questions related to reliability, sensitivity, and applicability of such indices across space and time. Moreover, past studies often failed to clarify between indices appropriate for nutritional assessment versus those appropriate for evaluating nutritional condition. Nutrition is defined as the rate of ingestion of assimilable energy and nutrients, and nutritional condition is the state of body components (e.g., fat, protein) which in turn influence an animal’s future fitness (Harder and Kirkpatrick 1994). Cook et al. (2001a) used captive-raised cow elk (Cervus elaphus nelsoni) fed a variety of diets to induce a wide range of body conditions to develop predictive models of body fat. For live animals, they assessed serum and urine chemistry, a body condition score, thickness of subcutaneous rump fat, and bioelectrical impedance analysis. For dead animals, they assessed femur and mandible fat, two carcass scoring techniques, and three different kidney fat indices. They assessed relations between indices and percent fat and developed models to predict nutritional condition. Cook et al (2001b) also evaluated range of usefulness, bias, precision, and sensitivity to small changes in body condition for each model deemed to be most useful by standard methods. Herein we summarize these findings and report major results.