Abstract: Hourly observations of visibility, relative humidity (RH), and particulate mass concentration in Wuhan for the period from September 2014 to March 2015 have been analyzed in this study to investigate the relationship among these variables. Nonlinear prediction of visibility in Wuhan is explored preliminarily. It is found that the frequent occurrence of haze in Wuhan is largely responsible for the severe reduction in visibility. The formation and accumulation of fine particulates are two important factors inducing haze and low visibility. Both the RH and the particulate mass concentration affect the variation of atmospheric visibility. High RH and large fine particulate mass concentration can significantly reduce the atmospheric visibility. Under wet conditions (RH ≥ 40%), the visibility deteriorates because the hygroscopic growth of the fine particulate can efficiently enhance light absorption and scattering. When the RH is higher than 90%, the visibility decreases linearly with the increase in RH. Averagely, the visibility decreases by 0.568 km as the RH increases by 1%. Under dry conditions (RH2.5 concentration becomes a critical factor for the rapid decrease in visibility. In urban areas where fine particulates in the atmosphere are primary pollutants, the visibility has a nonlinear relationship with RH. This is partly attributed to the influence of PM2.5 on the visibility and partly attributed to light scattering effects of hygroscopic particles. Results also indicate that there exists a nonlinear relationship between the PM2.5 concentration and the visibility,which can be described by a power function. The correlation between the PM2.5 concentration and the visibility is most significant when the RH is less than 90% but larger than 80%. The sensitivity threshold of PM2.5 concentration for the atmospheric visibility decreases with increasing RH. Under dry conditions, the visibility of 10 km corresponds to a PM2.5 concentration threshold of 70 μg/m3, whereas the value is 18-55 μg/m3 under wet conditions. Decreases in the PM2.5 concentration can lead to significant improvement in visibility when the PM2.5 concentration is less than 40 μg/m3. In addition, results of preliminary experimentshave shown that the visibility prediction model, which is developed based on the neural network method, performs well in prediction of visibility in Wuhan. The correlation coefficient between observations and predictions can be up to 0.86, and the Root Mean Square Error (RMSE) is 1.9 km. The TS score is 0.92 for the visibility that is less than 10 km. These results indicate that the model has a crucial skill for haze prediction.