Abstract Estimating feed intake plays a key role in measuring the efficiency of feed use in livestock production systems. However, measuring feed intake in grazing environments is challenging. While the plant-wax marker methodology can be used to measure feed intake, an estimate of diet composition is also necessary. In addition, having a measure of the composition of the diet of grazing herbivores is important for grazing management and rangeland ecology. Although accurately estimating diet composition is possible using nonnegative least squares, modeling repeated measurements or relationships among individuals is not possible. Thus, the goal of this study was to develop models that account for covariance among repeated measurements and relationships among animals. These methodologies were tested on plant-wax marker fecal data from 21 beef cattle heifers, measured at two physiological stages, offered ad libitum pure red clover (Trifolium pratense) and tall fawn fescue (Festuca arundinacea) cubed hay. Plant-wax marker concentration (C27, C29, C31, C33) from the two forage species and fecal samples were measured using gas chromatography. The performance of the proposed methods was assessed by calculating normalized mean squared error (NMSE), mean absolute differences (MAD), reconstruction error (RE), mean bias, and standard error (posterior standard deviation). In addition, the intercept and slopes from regressing observed on estimated red clover proportion in the diet were tested for equality to zero and one, respectively. While modeling repeated measurements did not consistently improve the accuracy of estimation, accounting for relationship among animals appeared to improve MAD, NMSE and RE values. In addition, estimated intercept and slope from regressing observed on estimated red clover proportion did not differ from zero or one, respectively (P ≥ 0.18). While accounting for covariance among repeated measurements only marginally improved accuracy, incorporating a matrix of relationships among animals increased the reliability of estimates of their diet composition.
Read full abstract