The authors contrasted and analyzed the spatial distribution map of hourly air column vapor content retrieved from the 63 GPS/MET monitoring stations from May to September during 2009-2011 in Shanxi, and the corresponding meteorological observation data in 459 days, and the rainstorm falling area in 42 rainstorm days, and the corresponding flow pattern configuration map, and discovered that: (1) When the horizontal gradents of the air column vapor content spatial distribution are between 25 to 40 mm/latitude(longitude), in the next 12-36 hours, the probability of rainstorm and above is 100 percent in the big value area of horizontal gradents and its neighboring 0.5-1.0 latitude and longitude range from south to north (east to west); when the horizontal gradents of the air column vapor content spatial distribution is ≥40 mm/latitude(longitude), the probability of rainstorm is 63.6 percent in the big value area and its neighboring 0.5 latitude and longitude range from south to north (east to west); (2) The rainstorm falling area appears in the south (east) or north (west) 0.5-1.0 latitude and longitude of the horizontal gradents big value area of the air column vapor content spatial distribution, different flow pattern configuration would be a different result. Using the hourly GPS/MET data and hourly automatic weather station (AWS) maximum wind speed data, and based on the different location of rainstorm appearing in the horizontal gradents big value area in air column vapor content spatial distribution map, the authors built several rainstorm conceptual models in different flow pattern configurations; under C/S construction, the authors improved rainstorm falling area 12-36 h forecasting model with the contour recognition technology, and achieved automatic operation runs, and the quasi operation run in 2011 proved good effects.