Based on horizontal visibility, relative humidity and weather phenomena obtained from 82 meteorological stations in Jiangxi Province from 1960~2016, haze days were reconstructed and the temporal variations were analysed. Additionally, the attributions of the variation were identified according analysis of the correlation of haze days with precipitation days, high wind days and breeze days. The results showed that the haze days had an obvious peak in the 1970s, followed by an increasing trend since 1980 (0.53d/a, P<0.001). Among the four seasons, increasing haze days in autumn (0.21d/a, P<0.001) contributed most to the annual increasing rate, followed by spring (0.12d/a, P<0.001). In winter, although most haze days happened, the increasing trend was insignificant (0.10d/a, P>0.05). While summer has the least increasing amplitude (0.09d/a, P<0.001). In the past decades, the reduction of atmospheric deposition ability caused by less precipitation days (-0.26d/a, P>0.05), and the reduction of atmospheric diffusion ability caused by fewer high wind days (-0.33d/a, P<0.01) but higher breeze days (1.73d/a, P<0.01), provided advantages for haze increase. However, the main attribution varied between seasons. The major reason for haze increase in spring was the reduction in high wind days (r=-0.48, P<0.01). While in summer, it had significant correlation with both high wind days (r=-0.50, P<0.01) and breeze days (r=0.37, P<0.05). In autumn, haze increase was affected by fewer high wind days, higher breeze days and less precipitation days, leading to the highest increasing rate among four seasons. In winter, haze days has a better correlation with precipitation days (r=-0.36, P<0.05) whose trend was insignificant (-0.26d/a, P>0.05). Thus, winter haze days has an insignificant trend during.