Based on an in-depth investigation on the randomness of the time series of the atmospheric extinction coefficient, this paper uses the monitoring data of atmospheric visibility, relative humidity (RH), the mass concentrations of particulate matter (PM₁₀ and PM_2.5), and NO₂ content during autumn and winter in Chengdu from 2015 to 2017. In addition, the time series of the aerosol extinction coefficient in the corresponding period in Chengdu is obtained. The detection of nonstationarity in the aerosol extinction coefficient time series is conducted using generalized additive models for location, scale, and shape, followed by the covariate analysis of the aerosol extinction coefficient time series. The experimental results show that the aerosol extinction coefficient time series in Chengdu during autumn and winter are nonstationary, with nonlinear changes in their mean and variance. The mass concentration of fine particulate matter (PM_2.5), RH, and aerosol component structure (PM_2.5/PM₁₀) are significant covariates of nonstationarity in the aerosol extinction coefficient series. Among them, PM_2.5 contributes the most to the nonstationarity of the aerosol extinction coefficient series, followed by RH, and PM_2.5/PM₁₀ contributes the least. The explosive growth of aerosol extinction coefficient is closely related to the synergies of PM_2.5, RH, and PM_2.5/PM₁₀.