A study on auto-classification of cloud types based on millimeter-wavelength cloud radar observations

The millimeter-wavelength cloud radar has obvious advantages over weather radar and lidar because it can provide more information on cloud particle. It becomes an effective instrument in the detection and study of cloud characteristics. This work is focused on automatic classification of cloud echoes detected by the millimeter-wavelength cloud radar. Based on 161 samples of cloud echoes, the value ranges of characteristic quantities and other parameters are obtained for six types of cloud, including cirrus, altostratus, altocumulus, stratus, stratocumulus and cumulus. Automatic classification of clouds has been realized by using the multi-parameters threshold discrimination method with these value ranges in a hierarchical order. The automatic classification results are evaluated by comparing with that of manual classification, which shows a 84% consistency between the two methods. The automatic classification method cannot well identify stratus and cumulus clouds due to the limited number of samples (6 stratus samples and 8 cumulus samples). With more samples, more reliable information of the characteristic quantities for various types of clouds will be obtained, and the accuracy of automatic classification definitely will be improved. The cloud classification technique developed in this work based on millimeter-wavelength cloud radar observations is highly expected to promote the operation of automatic cloud observations in the near future.