THE CHARACTERISTICS OF DROP SIZE DISTRIBUTIONS AND CLASSIFICATIONS OF CLOUD TYPES USING GUDUCK WEATHER RADAR, BUSAN, KOREA Dong-In Lee 1, Min Jang 1, Cheol-Hwan You 2, Byung-Sun Kim 2, Jae-Chul Nam 3 Dept. of Environmental Atmospheric Sciences, Pukyong National University, Korea 1 Remote Sensing Division, Korea Meteorological Administration, Korea 2 Meteorological Research Institute, Korea Meteorological Administration, Korea 3 Abstract A weather radar and POSS observations were carried out to classify the cloud types and investigate the characteristics of drop size distributions (Here after DSDs) at Busan, Korea. The DSDs of POSS and those of Marshall Palmer(1948) were compared to know the characteristics of DSDs with respect to total rain drop number at Busan. There were significantly discrepancies especially in the smaller and larger drop size than 2 mm of diameter. To calculate its own characteristics of raindrop size distributions the least square method was used. And the gamma distribution was also used to know the features of DSDs according to the rain rate. To classify the cloud types using radar echoes, the algorithm of Steiner et al. (1995) was applied to weather radar at Busan, Korea. Key Words: DSDs, POSS, Marshall Palmer, slope, intercept, cloud types Introduction Measurements of rainfall by radar are based on the relationship between reflectivity factor and rain rate at the ground. The relations can be obtained through a theoretical or an empirical approach. Experimentally measured DSDs have been extensively used to calculate both radar reflectivity and rain rate (Campos and Zawadzki, 22). By plotting rain rate against reflectivity or by correlating these statistically, the relationship between these two parameters can be determined (Rinehart, 1997). Many researchers have been studied to estimate rainfall using a radar with different methods. It is considered that there are several essential things to estimate more accurate rainfall using a radar. One of them is to obtain optimum Z-R relationships with geographic features and precipitation system. DSDs is one of the most important parameters to calculate Z-R relationship since both of reflectivity and rain rate are dependent on DSDs. Most of precipitation systems in Korea have convective and stratiform rainfall. Therefore, in this paper the characteristics of DSDs in Busan were examined and classifications of cloud types were executed using Steiner et al. (1995) algorithm. Descriptions of methods Numerous measurements under a variety of conditions showed that for diameters more than 1 mm, the average size distribution of raindrops near the ground may be well represented by the following simple equation proposed by Marshall and Palmer(here after M-P) (1948); N( D) = N exp( ΛD), (cm -4 ) (1) Melbourne, Australia 2-4 February 24
4.21 1 Where, N =.8( cm ) and Λ = 41R ( cm ). And the widely used gamma distribution function (Ulbrich 1985) was also employed for modelling the DSDs in this study : m N( D) = N D exp( ΛD) (2) A method of moments approach was used to calculate the intercept, slope and shape parameters (Kozu and Nakamura 1991). For gamma DSD model, the x th moment of the DSDs, is expressed as Γ( m + x + 1) Mx = N. m+ x+ 1 Λ To classify the cloud types, Steiner et al (1995) algorithm were tested. The criteria for identifying convective precipitation of this are as follows; Any grid in the radar reflectivity field with at least 4dBZ is automatically labelled as a convective centre, since rain of intensity more than 4 dbz could practically never be stratiform. Any grid point in the reflectivity field not identified as a convective centre in the above step, but which exceeds the average intensity taken over the surrounding background by at least the reflectivity difference. The background difference is linear average within a radius of 11km around the grid point. For each grid point identified as a convective centre by one of the above two criteria, all surrounding grid within an intensitydependent convective radius around that grid point are also included as convective area. In this study the classification was applied at 1km height of 121*121*59 dimensions in considering of Busan radar scan strategy. M x Results To know the characteristics of DSD in Busan, the monthly N(D) were examined. The DSDs obtained by POSS were compared with Marshall-Palmer distribution (M-P), gamma distribution, and a new fitted distribution of M-P using least square method. Fig. 1 shows the DSDs of observed and M-P at April, July, and September in 21. The average rain rate at April is.6mm/h and total spectra is 1,552 minutes. The black line indicates the DSDs observed by POSS and blue line shows that of M-P. (a) (b) (c) Fig. 1. The comparison of the DSDs characteristics between POSS and M-P (a) April, (b) July, and (c) September in 21. Melbourne, Australia 2-4 February 24
Drops smaller than 2 mm were underestimated and larger ones than 2 mm were overestimated at POSS observation. It may be concerned that the variability of DSDs with locality is significant. The discrepancy of drop number concentrations near 2 mm of diameter is dominated. It is considered that the effects of topography, improvement of observing technology and so on were mainly 3.37 dominated. Harimaya and Okazaki (1993) classified cloud types with N = 7 1 R and.14 Λ = 3.8R using optical spectrometer. They also found that small raindrops below 1.5 mm in diameter are fewer than those of corresponding to M-P, however larger raindrops above 1.5 mm are more than them in stratiform clouds. A new method with well-fitted DSDs should be required for calculating adequate DSDs in Busan. Therefore, the least square method was used to obtain it in this study. Table 1 summarizes the intercept and slope of DSDs at each month in 21. Table 1. The slope and intercept values of DSDs using least square method at Busan in 21 Month N Λ Month N Λ April 956 2.3 August 5839 2.5 May 129 2.4 September 124 2.5 June 968 2. October 1626 2.2 July 93 1.9 November 129 2.4 Until now, the parameters of DSDs were monthly focused and explained. In 1985 Ulbrich used gamma model to calculate more reasonable parameters of DSDs with rainrate. Six categories (Very light, light, moderate, heavy, very heavy, and extreme) with respect to rainrate were classified from April to November in 21. The observed DSDs (solid line) and calculated ones (dotted line) in 21 were shown in Fig. 2. Fitted DSDs were overestimated at the whole of diameter in the category of less than 1 mm/h and the contribution to the total rain was 55.9%. The parameters of DSDs were 59. of intercept and.1 of slope(fig. 2.(a)). (a) (b) (c) Melbourne, Australia 2-4 February 24
(e) (f) Fig. 2. The comparison of DSDs between observed DSDs and fitted ones using gamma model with respect to average rainrates from April to November, 21 Fig. 2(b) is in the class of 1 to 2 mm/h. The contribution to the total rainfall was 18.9% and fitted line was underestimated less than about 3. mm and overestimated more than 3. mm. The fitted and observed DSDs were good agreement in Fig. 2(c) and (d) but the fitted ones were overestimated in Fig. 2(e). Both weak rainrate (<1mm/h) and strong one (>1mm/h) were a little deviated. Steiner algorithm was used to classify the cloud types by three steps with mean background reflectivity. Fig. 3 shows the result of Steiner algorithm on 8 LST 5 th July in 22. Convective area is indicated by blue color and stratiform cloud is red one. It is just examination whether Steiner algorithm could apply for middle latitude region or not. (a) (b) (c) Cloud types 8 LST 5 July 22(ST, Small) Cloud types 8 LST 5 July 22(ST, Midium) Cloud types 8 LST 5 July 22 (ST, Large) -6-6 -6-3 3-3 3-3 3 6-6 -3 3 6 6-6 -3 3 6 6-6 -3 3 6 Melbourne, Australia 2-4 February 24
(d) (e) -6 Reflectivity 8 LST 5 July 22 7 6-3 3 A B Height (km) 5 4 3 2 6-6 -3 3 6 1 A 1 2 3 4 5 6 Distance (km) B -2-1 1 2 3 4 5 6 dbz -2-1 1 2 3 4 5 dbz Fig. 3. The classifications of cloud types used by Steiner algorithm 8 LST 5 th July in Busan. (a) is small area, (b) for medium, (c) for large, (d) indicates CAPPI, and (e) is the cross section along with line AB. Conclusion In this research, the characteristics of DSDs were found out according to the total drop number (Power law regression) and the rainrate (Gamma model) at the first time at Busan in Korea. The possibility of Steiner algorithm was also examined in the middle latitude region. Firstly, there were significantly discrepancies especially at the smaller and larger size than 2 mm diameter in Busan. To remove these differences between both and obtain new parameters in Busan, the least squares method and gamma model were used. And the classifications of cloud types were just demonstration but verifications in this study. Therefore, the verification of Steiner algorithm should be done at the middle latitude and then apply it to the estimates of radar rainfall in the near future. Acknowledgements This study was performed for the project, Technical Development for Remote Sensing Meteorology, one of the Meteorology and Earthquake R&D programs funded by the Korea Meteorological Administration. References Campos, E., and Zawadzki I. (2). Instrumental uncertainties in Z-R relations. J. Appl. Meteor. 39, 188-112. Harimaya, T., and Okazaki, K. (1993). A comparison of the raindrop size distributions from stratiform clouds with those from convective clouds. Jour. Fac. Sci. Hokkaido Univ. Ser. 7. 9, 341-353. Kozu, T., and Nakamura, K. (1991). Rainfall parameter estimation from dual-radar measurements combining reflectivity profile and path-integrated attenuation. J. Atmos. Oceanic Technol. 8, 259-271. Melbourne, Australia 2-4 February 24
Marshall, T. S., and Palmer, W. M. K. (1948). The distribution of raindrops with size. J. Meteor. 5, 165-166. Ronald, E. R. (1997). Radar for meteorologists. Rinehart publishing. 428. Steiner, M., Houze Jr. R. A., Yuter, S. E. (1995). Climatological characterization of threedimensional storm structure from operational radar rain gauge data. J. Appl. Meteor. 34, 1978-17. Ulbrich, C. W. (1983). Natural variations in the analytical form of raindrop size distribution. J. Climate Appl. Meteor. 22, 1764-1775. Melbourne, Australia 2-4 February 24