Abstract:Dairy cows in cold regions are in non-heat stress state for a long time. Aiming to understand the main environmental factors affecting milk yield, the relationship between temperature, humidity, wind speed, carbon dioxide concentration, ammonia concentration, light intensity and average daily milk yield of dairy cows was studied from the perspective of daily mean and percentile values based on continuous monitoring environmental data. Meanwhile, a random forest regression model based on environmental factors was established to predict milk yield. The results showed that light intensity and carbon dioxide concentration were two important environmental factors affecting milk yield, especially at low temperature. The average daily light intensity of 250lx and the average daily carbon dioxide concentration of 8×10-4 could obviously classify high and low milk yield. The sample sites with high milk yield were also concentrated in the areas where the 90th percentile of light intensity was higher than 800lx and the 10th percentile of carbon dioxide concentration was lower than 6×10-4. The mean determination coefficient and mean root mean square error used to evaluate the generalization ability of regression model were 0.7316 and 1.0370 kg, respectively. According to the results, it was suggested that during the non-heat stress period of dairy barns in cold areas, at least 2.5h of light should be guaranteed for no less than 800lx per day, and the time when carbon dioxide concentration was higher than 6×10-4 should be controlled for no more than 2.33h.