Modeling heat stress effects on dairy cattle milk production in a tropical environment using test-day records and...

In tropical environments, dairy cattle production is constrained by several factors, including climate. The seasonal loss of milk due to heat stress is a recurring challenge for many dairy producers. The objective of this study was to detect heat stress thresholds, milk yield loss and individual animal variations using random regression models for dairy cattle from test-day milk records. Data were obtained from the Kenya Livestock Breeders Organization for the years 2000-2017 and merged with weather data. The weather parameters were grid-interpolated solar and meteorological data obtained from the National Aeronautics and Space Administration/Prediction Of Worldwide Energy Resources (NASA/POWER). After editing, the records comprised 49 993, 45 251 and 36 136 test-day records for first, second, and third lactations, respectively, for the four main dairy breeds: Friesian (68.0%), Ayrshire (21.1%), Jersey (7.6%) and Guernsey (3.3%). Variance components were estimated using Restricted Maximum Likelihood in ASReml software. Random regression models with third-order Legendre polynomials were fitted to the average and individual lactation curves and the reaction norms. An extended factor analytic variance structure for the random cow effects was used to estimate (co)variances between days in milk and thermal load. The daily average temperature (TA) and temperature humidity index (THI) were identified as the most suitable thermal load indicators for assessing milk yield losses. Considering a one day lag, the estimated heat stress thresholds were about 22 °C and 69 index units for TA and THI, respectively. Almost no differences were observed for estimated residual variances between the thermal load indicators, indicating there was no better model fit by TA or THI. The heat stress thresholds and milk loss patterns are important for management of dairy production systems in the tropics with climatic conditions similar to this study. Data recording should be improved as a tool to monitor the expected impacts of climate change and mitigation measures.