A polarimetric radar can be used for several kinds of applications, such as improved rainfall estimation, retrieval of rain drop size distribution (DSD) parameters, and hydrometeor classification. Among these applications, hydrometeor classification is important to understand mechanisms of severe weather phenomena, such as heavy rain, lightning stroke, blast, and hailstorm. In North American and European countries, research of hydrometeor classification have been carried out since early period, using S- or C-band polarimetric radars. In the last decade, polarimetric radars have been introduced in Japan, and in November 2007, two X-band polarimetric radars were introduced at Nagoya University. Under such a circumstance, the expectation for hydrometeor classification with X-band polarimetric radars is becoming stronger and stronger. In this study, we tried to develop a hydrometeor classification method for X-band palarimetric radar.

With a polarimetric radar, we can get three-dimensional data of several kinds of polarimeric parameters; radar reflectivity (Z_h), differential reflectivity (Z_dr), specific differential phase (K_dp), and correlation coeficient between horizontal and vertical polarization signals (ρ_hv). In this study, we used X-band polarimetric radars and tried to classify hydrometeor into 10 categories, such as rain, snowflake, graupel, hail, and so on. We adopted hydrometeor cassification method, based on fuzzy logic, for S-band polarimetric radar (Liu and Chandrasekar, 2000) as basic scheme. With fuzzy logic, we can represent the probability of each hydrometeor type being present by continuous value from 0 to 1, using membership functions. We made membership functions for each polarimetric parameter and each hydrometeor type, and we evaluate probability of each of 10 hydrometeor types being present for each parameter with membership functions. Then, we evaluate probability of each hydrometeor being present as the product of membership functions for every parameter. Finally, we regard the hydrometeor type whose probability is the largest among all hydrometeor type as the dominant hydrometeor type.

In this study, we modified membership functions of K_dp and temperature. Because the value of K_dp depends on wavelength, we modified membership functions of Kd p shown in Liu and Chandrasekar(2000) to adapt to X-band polarimetric radars. Scince some pairs of hydrometeor types, such as rain and dry snow aggregate, have similar polarimetric properties, it is very difficult to classify using only polarimetric parameters. To resolve this problem, membership functions for temperature have also used for hydrometeor classification, usually. Because the temperature where snowflake melts is dependent on relative humidity, we modified membership functions for temperature to change their forms along with relative humidity. We used the vertical profile of temperature and relative humidity obtained from ground and balloon observations.

In this study, we used ground observational data and polarimetric radar data of snowfall event to develop hydrometeor classification method. With our classification method, snowflake was classified just above the point where snowflake was observed at ground. Therefore , we can say that our classification method worked successfully. In response, we tried hydrometeor classification for another snowfall event around Nagoya City, graupel falling event over Ishikawa Prefecture, and summer thunderstorm near Nagoya City.

In the snowfall event around Nagoya, snowflake was classified near the ground; the result is reasonable. In the graupel evant over Ishikawa Prefecture, graupel was classified approximately the same area where Z_h value was high. In the thunderstorm near Nagoya, well-known hydrometeor distribution of thundercloud was classified; rain at low level, graupel around 0 °C height, and snowflake and ice crystal over graupel region. In response, we examined time evolution of volume and height range of graupel region. As a result, it began to observe lightning stroke when the volume of graupel region was maximum, and had continued until graupel region disappeared. This result is consistent with occurrence of lightning stroke.

As described above, we tried to classify hydrometeor types and got qualitatively reasonable result. The future works of this study are to modify the member ship functions based on the in-situ observation, to conduct attenuation correction of Zh and Zdr above snowflake melting level, and to classify hydrometeor types taking account for the life stage of thundercloud using three-demensional wind field.