Comparing piecewise regression and hysteresis models in assessing beef cattle heat stress

Climate change may generate more frequent heat waves, resulting in substantial cattle production losses through increased heat stress. Time lags between air temperature and an animal's body temperature have been recognized as valuable measures of heat stress, and developing methods for detecting time lags is important. Existing hysteresis models are useful for estimating air-body temperature time lags, especially when the air temperature follows a consistent diurnal sinusoidal function, such as when animals are housed in a controlled environment. However, in cattle feedlot or pasture operations, consistent sinusoidal air temperature patterns are not realistic, and a more flexible approach would be useful. In this article, piecewise regression models (linear and quadratic) are developed to estimate time lags under more general temperature trend conditions. Both piecewise regression and hysteresis models were fit to heat stress data of feedlot cattle. Simulations were conducted to compare the estimated time lags using both types of models. In the simulations, the asymmetric harmonic hysteresis model estimated time lags best, followed by the piecewise linear regression model, while the piecewise regression models were generally more efficient for both simulated and actual data. It was concluded that piecewise regression models are more appropriate than hysteresis models when applied to heat-stressed cattle in production environments.