2007 Volume 20 Issue 3 Pages 156-166
In order to understand features of a meteorological field related to a heavy rainfall event, it is important to interpret detailed information of atmospheric stratification, especially, its horizontal variability. However, the horizontal features cannot be easily interpreted because the atmospheric stratification represents multi-dimensional data consisting of a vertical profile of water vapor and air temperature. Therefore, in this study, by considering multi-dimensional atmospheric stratification as a set of data, many varieties of patterns characterizing atmospheric stratifications detected in northern Kyushu in a rainy season were classified using the Self-Organizing Map (SOM) algorithm. The result shows that the SOM could recognize inherent patterns in the rainy season by investigating warm and humid air in lower layers, and dry air in middle layers. In the next analysis, all the atmospheric stratification characterized by the heavy rainfall event that occurred in northern Kyushu on June 29, 1999, were diagnosed on the basis of the classified results by SOM, and, subsequently, all the diagnosed patterns were displayed in the target area including northern Kyushu. As a result, the horizontal patterns of the atmospheric stratification in the heavy rainfall field could be easily interpreted by using some properties of SOM. From these analyses, it is expected that detailed information on the atmospheric environment causing heavy rainfall can be systematically and easily interpreted by combining the diagnosis of atmospheric stratification by using SOM with horizontal representation of its patterns instead of a conventional analysis of atmospheric stratification.
