Heterogeneity of variance for milk production traits between the low and high input dairy production systems of South Africa

Abstract

South African dairy cattle industry is characterized by a dual production system, comprising of a high input commercial production system and low input smallholder and emerging dairy herds. Performance data from both systems are included in national genetic evaluations, with models that assume homogeneous variances. If variances are heterogeneous, above average animals in more variable herds will be favoured over high performing animals in the less variable herds. This may result in biased selection and inaccurate estimation of breeding values. With intensified selection, genetically inferior animals could be chosen, thereby decreasing the realised genetic gain, resulting in lower productivity. The aim of the study was to investigate the extent of heterogeneity of variance between the two dairy production systems South African Holstein cattle. Milk production data were obtained from the Integrated Registration and Genetic Information System of South Africa (INTERGIS).High input production system data set consisted of 68 000 lactation records from 741 herds, recorded between 2006 and 2018.Pedigree file comprised of 38 126 daughters of 2 472 sires and 4 305 dams. Data for the low input production system comprised of 32 388 lactation records of 3 325 daughters of 134 sires and 253 dams from 59 herds recorded from 2006 to 2018. Hartley’s Fmax test was used to test for heterogeneity of variances for 305 day yields of milk, fat and protein between the high and low input production systems. Non-genetic factors affecting these traits were then determined by the Proc GLM procedure of SAS. Genetic and phenotypic parameters among these traits were estimated, for each production system, by the Restricted Maximum Likelihood (REML) procedure in the ASREML software.vHeritability estimates for milk, fat and protein yield, respectively, were 0.70 ± 0.027, 0.55 ± 0.35 and 0.64 ± 0.03 for the low input production system compared to 0.16 ± 0.014, 0.11 ± 0.012 and 0.145 ± 0.013 for the high input production system. Estimates for genetic correlation between milk and fat, milk and protein and fat and protein were 0.68 (0.03), 0.81(0.01) and 0.81(0.02) in the high input production system and 0.80(0.34), 0.90(0.02) and 0.91(0.01) in the low input production system respectively. Phenotypic correlations in the high input dairy production system were 0.85(0.00), 0.92(0.00) and 0.88(0.00) for milk and fat, milk and protein and protein and fat and 0.82(0.08), 0.91(0.01) and 0.91(0.34) in the low input production system, respectively. Genetic prediction models for milk production traits, in South African Holstein cattle, shouldaccount for heterogeneous variances between the high and low production systems. Herd year-season of calving, parity and linear and quadratic effects of age at calving should beincluded on the models. There is a need to increase selection pressure in the low input production system, in order to improve genetic merit for milk production traits