A LAYER DETECTION AND CLASSIFICATION ALGORITHM FOR SODAR FACSIMILE RECORDS

This paper describes an application of image processing and pattern re cognition techniques for the detection and classification of the layer echo structure within a window of a facsimile record of a monostatic acoustic sounder (sodar). The algorithm identifies automatically the u pper and lower boundary of every layer in a facsimile record, even if it is a complex pattern. Consequently, it uses the boundaries of the l ayers to extract the classification features. This technique provides quantitative information for a facsimile record and follows a differen t approach than previous classification schemes that classify the reco rd manually according to predetermined patterns. The algorithm was dev eloped and tested using data collected by two different types of sodar under various meteorological conditions. An overall skill better than 90 per cent of the cases was achieved. The algorithm experienced some difficulty identifying the thermal structure over transitional atmosp heric conditions, during which there is no distinct layer structure.