Dairy records from the Dairy Recording Service of Kenya were classified into low, medium and high production systems based on mean 305‐day milk yield using the K‐means clustering method. Milk and fertility records were then analysed to develop genetic evaluation systems accounting for genotype‐by‐environment interaction between the production systems. Data comprised 26,638 lactation yield, 3,505 fat yield, 9,235 age at first calving and 17,870 calving interval records from 12,631 cows which were descendants of 2,554 sires and 8,433 dams. An animal model was used to estimate variance components, genetic correlations and breeding values for the production systems. Variance components increased with production means, apart from genetic group variances, which decreased from the low to the high production system. Moderate heritabilities were estimated for milk traits (0.21–0.27) and fat traits (0.11–0.38). Low heritabilities were estimated for lactation length (0.04–0.10) and calving interval (0.03–0.06). Moderate heritabilities (0.25–0.26) were estimated for age at first calving, except under the high production system (0.05). Within production systems, lactation milk yield, 305‐day milk yield and lactation length had high positive genetic correlations (0.52–0.96), while lactation milk yield and lactation length with age at first calving had negative genetic correlations. Milk yield and calving interval were positively correlated except under the low production system. The genetic correlations for lactation milk yield and 305‐day milk yield between low and medium (0.48 ± 0.20 and 0.46 ± 0.21) and low and high production systems’ (0.74 ± 0.15 and 0.62 ± 0.17) were significantly lower than one. Milk yield in the low production system is, therefore, a genetically different trait. The low genetic correlations between the three production systems for most milk production and fertility traits suggested that sires should be selected based on progeny performance in the targeted production system.

Economic and biological values for milk yield (MY), milk butter fat (FY), daily gain (DG), weaning weight (WWT), mature live weight (MLW), calving interval (CI), pre-weaning survival rate (PreSR), post-weaning survival rate (PostSR), age at first calving (AFC), and productive life time (PLT) were estimated under fixed herd (FH) and pasture (FP) production circumstances assuming milk marketing based on volume, and volume and butter fat. Further, economic values were estimated involving risk using the Arrow Pratt coefficients at two levels. For the former economic values for the traits ranged from KSh.-17.246 to 100.536 while the biological values ranged between-1.29 to 0.791. Economic values with higher Arrow-Prat coefficient of absolute risk aversion (λ=0.02) were lower than those reported under λ=0.0001 indicating that the uncertainty of the future market is important and should be considered during the estimation of economic values. Genetic improvements targeting MY and growth traits would be recommended to production system with unlimited feed supply for profit maximization. However, since dairy production systems in the tropics are characterised by feed scarcity, fixing the herd and concentrating on genetically improved animals would result to more profitability than increasing animal populations.

Genetic parameters for test-day milk yield, lactation persistency, and age at first calving (as a fertility trait) were estimated for the first 4 lactations in multiplebreed dairy cows in low-, medium-, and high-production systems in Kenya. Data included 223,285 test-day milk yield records from 11,450 cows calving from 1990 to 2015 in 148 herds. A multivariate random regression model was used to estimate variance and covariance components. The fixed effects in the model included herd, year, and test month, and age as a covariate. The lactation profile over days in milk (DIM) was fitted as a cubic smoothing spline. Random effects included herd, year, and test month interaction effects, genetic group effects, and additive genetic and permanent environmental effects modeled with a cubic Legendre polynomial function. The residual variance was heterogeneous with 11 classes. Consequently, the variance components were varied over the lactation and with the production system. The estimated heritability for milk yield was lower in the low-production system (0.04-0.48) than in the medium-(0.22-0.59) and high-production (0.21-0 60) systems. The genetic correlations estimated between different DIM within lactations decreased as the time interval increased, becoming negative between the ends of the lactations in the low-and medium-production systems. Low (0.05) to medium (0.60) genetic correlations were estimated among first lactation test-day milk yields across the 3 production systems. Genetic correlations between the first lactation test-day milk yield and age at first calving ranged from 0.27 to 0.49, 0 to 0.81, and −0.08 to 0.27 in the low-, medium-, and high-production systems, respectively. Medium to high heritabilities (0.17-0.44) were estimated for persistency, with moderate to high (0.30-0.87) genetic correlations between 305-d milk yield and persistency. This indicates that genetic improvement in persistency would lead to increased milk yield. The low to medium genetic correlations between test-day milk yield between production systems indicate that sires may be re-ranked between production systems. Therefore, we conclude that sires should be selected based on a genetic evaluation within the target production system.