Anomaly-Based Classification Study of Synoptic Patterns Associated with Hourly Heavy Rainfall over Beijing by SANDRA Method

Using SANDRA (Simulated ANnealing and Diversified RAndomization), the hourly heavy rainfall days in Beijing during the warm seasons (May-September) of 2007-2014 are classified into four types based on 500 hPa anomalous geopotential height fields. There is an anomalous low in Mongolia, Northeast and North China, Hetao region, and Russian Far East, respectively, in the four types. All the days during the warm seasons of 2007-2014 are then assigned to one of the four types according to the centroids of the types. The result shows that the type in which an anomalous low is located in Mongolia has the largest probability of hourly heavy rainfall. The statistical analysis of sounding on hourly heavy rainfall days reveals that the median specific humidity values of 925 hPa and 850 hPa are 13.01 g kg^-1 and 10.64 g kg^-1, respectively, and that the wind direction in the range of 180°-225° is dominant at these two levels.