Genotype by environment interaction across the full temperature-humidity index range for milk yield, milk...

Thermal environment is a major nongenetic determinant of dairy performance; however, most genetic evaluations have primarily focused on heat stress within limited temperature-humidity index (THI) ranges. This nationwide observational study aimed to characterize genotype by environment interaction (G × E) for test-day milk yield, fat percentage, protein percentage, and somatic cell score (SCS) across the full THI spectrum in Korean Holstein cattle using a reaction norm framework. A total of 4,648,152 test-day records from 297,330 cows collected between 2015 and 2024 were analyzed. THI values for each farm were derived from meteorological data using inverse distance weighting and summarized as moving averages over 5 d. Test-day milk yield, fat percentage, protein percentage, and SCS were modeled across THI values from 36 to 93 using a reaction norm animal model with second-order Legendre polynomials. Milk yield showed a nonlinear response across THI, with peak performance around THI 68-82 and declines toward both low and high THI extremes. Although losses per cow were largest under severe heat stress, a frequency-weighted loss analysis, defined as the product of the mean milk yield deviation from peak yield and the empirical frequency of records at each THI level, indicated that 66.3% of cumulative annual production loss was associated with below-optimal THI, compared with 26.4% under above-optimal conditions. For milk yield, additive genetic variance declined from 7.77 at THI 36 to 4.87 at THI 93, with heritability decreasing from 0.20 to 0.13. Fat percentage and protein percentage showed similar declines, with additive genetic variance decreasing from 0.19 to 0.13 and from 0.05 to 0.02, respectively, and heritability decreasing from 0.36 to 0.28 and from 0.55 to 0.38. By contrast, SCS showed an increase in additive genetic variance from 0.19 to 0.29, whereas heritability changed only modestly from 0.09 to 0.11 across the same THI range. Genetic correlations between THI 36 and THI 93 were 0.80 for milk yield, 0.49 for fat percentage, 0.45 for protein percentage, and 0.68 for SCS. Reranking of elite animals was observed across the THI gradient, as indicated by low Spearman rank correlations between pairs of environments among the top 25 animals selected at the centered THI value used as the reference environment in the reaction norm model (THI = 64.5). Across representative THI levels (36, 55, 74, and 93), pairwise rank correlations ranged from 0.16 to 0.75 for milk yield, 0.04 to 0.89 for fat percentage, -0.16 to 0.91 for protein percentage, and 0.44 to 0.88 for SCS. These results demonstrate that genetic parameters for milk yield, milk composition traits, and somatic cell score vary continuously across the THI gradient and support the importance of modeling performance across the full thermal range when evaluating dairy cattle in thermally variable production systems.