Abstract: The paper aims to find the abnormal change reasons of the haze days and to analyze the effects of the change of criteria, the use of automatic observation of visibility and human activities on haze days. The author analyzed the data of haze days from 1981 to 2017 by the methods of linear tendency, accumulative anomaly, M-K test, and slide t test. The results showed that: there were 5 obvious mutations of haze days in Pukou since 1981, which were 1988, 2000, 2007, 2011 and 2016; except 2016, all the other mutation points were related to the change of influence factors of haze day. Before 2007, the main factors affecting mutation were human activities and the change of observed environment; thereinto, the data of haze days had no use value due to the influence of human factors during 1988-1999; the change of haze criterion and the automatic visibility observation had the most obvious effect on haze days at the beginning of 2007; before 2016, the mutation points of haze days coincided with the change time of the influence factors, indicating that the effects of different factors on haze days in different time periods were more significant than that of meteorological conditions and air quality.