Abstract: Identifying critical weather factors that affect the intensity of influenza epidemics can promote early warning of influenza outbreak. Based on influenza-like illness (ILI) and meteorological observations during winter and spring in Shanghai from 2010 to 2018, a distributed lag non-linear model (DLNM) and generalized additive regression models (GAM) are used to investigate the relationship between ILI and various weather factors. Additionally, the multiple stepwise regression model is used to pinpoint significant weather factors affecting the intensity of influenza epidemics. Results indicate that there are negative correlations between ILI and temperature and relative humidity, and the relative risk (RR) of ILI increases as both factors decrease. The effects can persist until three weeks later and remain statistically significant, e.g. the cumulative RR of exposure to cold (5℃) and dry (50%) environments are up to 2.16 (95% CI: 1.18—3.95) and 2.51 (95% CI: 1.96—3.23), respectively. Meanwhile, there is also a significant interaction effect between relative humidity and temperature, with a dry environment significantly enhancing the cold impact. In addition, unstable weather elements related to air temperature are also significantly associated with ILI, with the risk rising by 1.8% (95% CI: 0.2%—3.4%) for every 1℃ increase. The frequency of cold air activity at the start of the epidemic and the peak ILI risk also feature a significant monotonically increasing linear relationship, with an increase risk of 6.8% (95% CI: 2.1%—11.7%) for one more cold air process. Among these sensitive weather factors, frequent cold air activities and dry conditions may form an environment conducive for the outbreak of serious influenza epidemic. The findings provide a theoretical basis for scientific understanding of widespread influenza epidemics that occurred in warm winters.