Abstract: Using correlation analysis between hourly simulation results of Beijing Regional Environmental Meteorology Prediction System and the limited simultaneous observation data of PM_2.5 concentration in November, December 2015 and January, May, June, and July 2016, a method to determine the optimal observation station distribution strategy is proposed and examined. Using this method, the observations are used to select better model results. This method is useful for determining station distribution strategy in sparse observation area based on high resolution and long time series data from numerical model simulation and available observations. This method is applied to station distribution strategy of PM_2.5 observation in Beijing. Generally, the concentration of PM_2.5 in the south is higher than that in the north, and characterized by high concentration along the mountain foot. In summer, the area has the highest horizontal concentration gradient located in the southwest, and it extends toward northeast in winter. Therefore, the optimal distribution strategy of PM_2.5 monitoring stations is with high density (small distance between stations) in the southwest and gradually increasing the distance between stations towards the north and east. On the other hand, to get same quality of horizontal distribution information, more stations should be used in winter than those in summer.