Abstract: An modified visibility parameterization scheme (S1) was proposed based on the correlation between visibility and PM_2.5 concentrations under different relative humidity, so as to improve the accuracy of visibility forecast by using the revised PM_2.5 concentrations forecast results. In order to evaluate the forecast performance of the scheme, the forecasted PM_2.5 concentrations based on a multi-model optimal integration method were used to forecast the visibility during the pollution process in Beijing for the period from February 18^th to March 5^th 2019. The scheme from Interagency Monitoring of Projected Visual Environments (IMPROVE, S2) and the a.pngicial neural network (S3) were also used to forecast the visibility in this period. The forecast results were compared and evaluated. The results showed that the forecast results of PM_2.5 concentration based on the multi-model optimal integration method and relative humidity based on WRF4.0 mode were good, and their correlation coefficient reached 0.90. All of the three schemes had good forecast for the trend of visibility in the selected period. S1 hold the best performances with the highest correlation coefficient (0.85) and lowest root mean square error (RMSE). The average deviation of homogenization in S1 was 3.6 km lower than that in S2. S1 performed better than S3 in range of the visibility greater than 10 km. The effect of PM_2.5 concentrations forecast on visibility forecast was higher than that of relative humidity forecast. The effect of relative humidity forecast could be more significant when the relative humidity exceeded 70%.