Partitioning variation in measurements of beef carcass traits using ultrasound.

Ultrasound technology provides cattle breeders with a quick, noninvasive, and inexpensive way to measure carcass data on live animals. Ultrasound data are used as indicator traits in cattle genetic evaluations for economically relevant carcass traits. Ultrasound cattle genetic evaluations assume homogeneous additive genetic and residual variance. Thus, the objective was to partition phenotypic variance in ultrasound carcass measurements into components for additive genetic effects, technicians, contemporary groups within technicians, and residual and to examine the homogeneity of these variances among image interpretation laboratories. Records of longissimus muscle area (LMA), percentage of intramuscular fat (IMF), and subcutaneous fat depth (SFD), measured using ultrasound, were provided by the American Angus Association ( = 65,967), American Hereford Association ( = 43,182), and American Simmental Association ( = 48,298). The data also included contemporary group, technician, imaging lab, and a three-generation pedigree for each animal. Variance components for ultrasound carcass measurements were first estimated with univariate animal models for each breed and imaging laboratory using derivative-free restricted maximum likelihood. Then, treating data from each imaging laboratory as separate traits, genetic correlations between laboratories for LMA, percentage of IMF, and subcutaneous fat were estimated with trivariate animal models. The technician explained 12-27%, 5-23%, and 4-26% of the variance for IMF, SFD, and LMA, respectively, across all three breeds. Variance due to technician was often greater than variance due to additive genetic effects but almost always less than that explained by the contemporary group. Within breeds, estimates of additive genetic variance for LMA, SFD, and IMF differed (range divided by mean) among laboratories by 4.5%, 21.5%, and 39.4 % (Angus); 31.6%, 15.0%, and 49.1% (Hereford); and 19.9%, 46.6%, and 55.3% (Simmental), respectively. Likewise, estimates of residual variance for LMA, SFD, and IMF differed among laboratories by 43.4%, 22.9%, and 43.3% (Angus); 24.9%, 15.2%, and 79.2% (Hereford); and 26.4%, 32.5%, and 46.2% (Simmental), respectively. Genetic correlations between labs across breeds ranged from 0.79 to 0.95 for IMF, 0.26 to 0.94 for SFD, and 0.78 to 0.98 for LMA. The impact of the observed heterogeneity of variance between labs on genetic evaluation requires further study.