Predicting rectal temperature and respiration rate responses in lactating dairy cows exposed to heat stress

Milk production and time effects are considered related to heat stress but they have not yet been combined in predictive models. In two parts, this study aimed to develop new models to predict heat stress (rectal temperature and respiration rate) of lactating dairy cows by inputting predictors, including ambient temperature (T_a), relative humidity (RH), wind speed (WS), milk yield (MY), and time blocks. In the first part of the study, we built the quantitative foundation for the second part, including the regression relation between respiration rate and rectal temperature (to convert predicted respiration rate to predicted body temperature), as well as between rectal temperature and respiration rate when heat stress was triggered (to recognize whether herds were under stress). In the second part, we built models that combined the abovementioned predictors to predict respiration rate. In part I, data were obtained from 45 high-producing Holstein cows within a T_a range of 9.5 to 30.8°C. We found a very strong correlation between mean respiration rate (MRR) and mean rectal temperature (MRT), where MRT = 0.021 × MRR + 37.6 (R² = 0.925), suggesting that for each 4.8 breaths per minute (bpm) increase of MRR, MRT would be expected to increase by 0.1°C. Rectal temperature was determined to be 38.6°C when heat stress was triggered, which corresponded to a respiration rate of 48 bpm. In part II, data were obtained in 3 stalls within a T_a range of 6.9 to 33.3°C over 3 time blocks, all of which were the 90 min preceding milking (0630-0800, 1230-1400, and 1830-2000 h). We found a nonlinear response of MRR to T_a, which could be linearized by the quadratic term of T_a. The response of MRR was the highest in the 0630-0800 h block, followed by 1230-1400 h, and finally 1830-2000 h. We proposed a model combining 3 time blocks (R² = 0.836): MRR in 0630-0800 h was determined to 56.28 + (-3.40 + 0.11 × T_a + 0.02 × RH) × T_a - 0.21 × RH - 2.82 × WS + 0.62 × MY; MRR in 1230-1400 h and 1830-2000 h were 4.6 and 10.3 bpm lower than that in 0630-0800 h, respectively (reducing the intercept of the expression in 0630-0800 h). Compared with temperature-humidity index equations, the proposed model performed better at suppressing prediction error, and had better sensitivity and accuracy in recognizing whether heat stress was triggered.