The Study On Cloud Type Auto-classification Based On Millimeter-wavelength Cloud Radar Observations

Cloud observation plays an important role in weather monitoring and forecasting,maintaining the balance of atmospheric radiation budget,atmospheric chemistry,and weather modification.And the millimeter-wavelength cloud radar has obvious advantages over weather radar and Lidar in cloud detection.It can not only measure the data of cloud bottom,cloud top and vertical distribution of cloud,but also obtain quantitative data such as radar reflectivity and vertical velocity.The millimeter-wavelength cloud radar also has high spatial and temporal resolution and is not affected by clear-air turbulence and ground echoes,and its strong cloud penetration capability enables it to detect echoes from multi-layer clouds.In addition,ground-based millimeter wave radar can carry out unattended all-weather continuous observation,which is very suitable for automation of cloud observation business.This work focuses on the automatic classification of cloud echoes detected by the millimeter-wavelength cloud radar.With 206 samples of the cloud echo,the value ranges of characteristic quantities and other parameters are obtained for six types of clouds,including cirrus,altostratus,altocumulus,stratus,stratocumulus and cumulus.Then the target of automatic classification of clouds has been reached by using the value range for multi-parameters threshold discrimination method in a hierarchical order.The automatic classification effect is evaluated by comparing with that of manual classification,showing 82% consistence between the two methods.The automatic classification method did not well identify stratus and cumulus,due to the limited number of samples(10 stratus samples and 9 cumulus samples).In addition,based on the influence of different seasons on various kinds of clouds,the cirrus,altostratus,altocumulus,and stratocumulus are classified according to different seasons using the above-observed cloud data and multi-parameters threshold discrimination method.The results show that the hit rates in spring,summer,autumn and winter are 95%,98%,83% and 84% respectively,and the accuracy in spring and summer is obviously improved,which shows that the classification of cloud echoes by season is feasible and effective.With processing on more samples,more reliable information on the characteristic quantities for various types of clouds will be obtained,thus the accuracy of automatic classification will be definitely improved.This work of cloud classification technique based on millimeter-wavelength cloud radar is highly expected to promote the operation of automatic cloud observation in near future.