Meteorological data (1993 to 2004) from 2 public weather stations in Phoenix, Arizona, and Athens, Georgia, were analyzed with test day milk yield data from herds near weather stations to identify the most appropriate temperature-humidity index (THI) to measure losses in milk production due to heat stress in the semiarid climate of Arizona and the humid climate of Georgia. Seven THI with different weightings of dry bulb temperature and humidity were compared. Test-day data were analyzed using 2 models to determine threshold of heat stress and rate of decline of milk production associated with a specific THI. Differences in thresholds of heat stress were found among indices and between regions. Indices with higher weights on humidity were best in the humid climate, whereas indices with larger weights on temperature were the best indicators of heat stress in the semiarid climate. Humidity was the limiting factor of heat stress in humid climates, whereas dry bulb temperature was the limiting factor of heat stress in dry climates.
