Contents. hydrometeors mixture. a C-band polarimetric radar. based on validation by in-situ campaign observation synchronized with Video-Sonde

Transcription

1 KU-OU Symp., Uji Obaku Plaza, Kyoto Univ., Nov. 11, 2009 Quantitative estimation of rainfall intensity and hydrometeors mixture using C-band polarimetric i radar based on validation by in-situ campaign observation synchronized with Video-Sonde Eiichi NAKAKITA Research Division ision of Atmospheric and Hydrospheric Disasters Disaster Prevention Research Institute (DPRI) Kyoto University, Japan (nakakita@hmd.dpri.kyoto-u.ac.jp) Introduction Background In Japan, main weather radars for operational network has been the C- band radar. The new type polarimetric radar, which can observe K DP, has not been put into operational use in Japan. Small number of C-band polarimetric radar in the world One of the purposes p Promotion of introducing the new C-band polarimetric radars to Japanese operational radar network. Activities A A synchronized observation by the C-band polarimetric radar, COBRA, with the video-sonde sonde. Development of a new operational QPE algorithm for polarimetric with C-band radar by an improvement of existing algorithm. Classification as mixture of some types of hydrometeors. QPF with data assimilation 1 Current activities with C-band polarimetric Doppler radar Campaign Obs.+Analysis+Data Assimilation+Impact Assessment C-band polarimetric i Doppler radar (COBRA) Doppler velocity, ZHH, ZDR, KDP, ρhv, LDR Campaign direct observation of Estimation and validation of DSD hydrometeors and type of hydrometeors synchronized with the video sonde and the Hyvis Data assimilation by mesoscale atmospheric model for QPF Developing algorism for QPE Ground based observations of DSD with some kinds of Impact assessment on instruments water management Contents In-Situ campaign observation synchronized with Video-Sonde d A new operational QPE algorithm for C-band polarimetric i radar Classification of co-existing hydrometeors using a C-band polarimetric radar Operational polarimetric radars in near future by the Ministry of Land, Infrastructure, Transportation and Tourism (MLIT), and one of an important tbackground 3

2 Contents In-Situ campaign observation synchronized with Video-Sonde d (~ 5 min.) A new operational QPE algorithm for C-band polarimetric i radar (~ 10 min.) Classification of co-existing hydrometeors using a C-band polarimetric radar (~ 14 min.) Operational polarimetric radars in near future by the Ministry of Land, Infrastructure, Transportation and Tourism (MLIT), and one of an important tbackground d( (~20 min.) 4 In-Situ campaign observation synchronized with Video-Sonde (2007~2009) Eiichi NAKAKITA, Kosei YAMAGUCHI, Hidenobu TAKEHATA, Yasuhiko SUMIDA, (Kyoto University) Kenji SUZUKI (Yamaguchi University) Katuhiro NAKAGAWA, Seiji KAWAMURA (NICT) Satoru OISHI (University of Yamanashi) Kazuhisa TUBOKI, Yukari SHUSSE, Tadayasu OHIGASHI (Nagoya University) Tsutomu TAKAHASHI (University of Hawaii) C-Band Doppler Polarimetric Radar COBRA = CRL Okinawa Bistatic polarimetric RAdar COBRA is operated by the National Institute of information and Communications Technology (NICT). C-band polarimetric weather radar Peak power > 250 kw (Dual Klystron) COBRA s purpose is the >10kW(Dual TWTA) development of the meso Pulse width 0.5 μs, 1.0 μs, 2.0 μs (Klystron) hydrometeorological observation μs (TWTA) PRF 250 Hz Hz, PRT 1μs step for the next generation. (staggered PRF) Target : Typhoon, Bai-u, meso scale Antenna size 4.5m φ parabolic rainfall system Beam width 0.91deg Radome size 8m φ Polarization parameters: ZHH, ZVV, Cross pol. ratio > 36 db (Integrated value in a beam) ZDR, φdp, KDP, ρhv, LDR Antenna gain 45 dbi (including radome) Sidelobe < -27 db (one way) Ant. scan speed rpm(ppi), rpm(rhi), 0.1 rpm step Polarization H, V, +45, -45, LC, RC (pulse by pulse) 6 Ground-based observations 2-D Video Disdrometer (2DVD) 2DVD has two line scan cameras which are orthogonally set at different height. Therefore it can observe not only the size and shape but also the terminal velocity of precipitation particles. Disdrometers and optical rain gauge at Ogimi Rain gauge of AMeDAS, which stands for Automated Meteorological Data Acquision System, at Yoronjima locations of COBRA and the ground observation points. Disdrometer The momentum of raindrop particle is observed and correspond to the size of the particle. Therefore DSD can be observed.

3 Campaign Observation in Okinawa Observation Periods Preliminary IOP:Nov. 15th 28th, 2007 Main IOP (1): May 28th-June 21st, 2008 Main IOP (2): May 21th-June 21st, 2009 Collaboration Kyoto University, University of Yamanashi, Yamaguchi University, Nagoya University, it University it of ftsukuba, Utsunomiya University, it National Institute of Information and Communications Technology, Central Research Institute of Electric Power Industry, University of Hawaii (more than 30 researchers and students) Observation instruments Polarimetric Doppler radar (COBRA) Video-Sonde, Hyvis 2-D video distrometer,impact type disdrometer,micro rain radar, Laser drop-sizing gauge,optical rain gauge, etc. 8 Video-Sonde and Hydrometeor types Takahashi at el.(2001) Rain Ice crystal Graupel Snow flake Video-sonde is the radiosonde with a video camera. The video-sonde records images of particles larger than 0.5mm diameter. The video-sonde is launched with the balloon, and can directly observe hydrometeors below and in the cloud. 13mm 9 Video Zonde Example of observations 9 No.6 no data 20 (km) HEIGHT ICE CRYSTAL 氷晶 霰 GRAUPEL AGGREGATE 雪片 FROZEN PARTICLE 雹 TEM PERATU RE ( ) 2 1 雨滴 RAINDROP 20 0 Suzuki et al. (2006) DIAMETER(mm) 1220(JST) DEC Suzuki et al. (2006)

4 Hydrometeor types Rain Ice crystal Synchronized Observation Ice crystal, snow flake, graupel etc. onna COBRA ohgimi Rain drop RHI scan in the direction of the video sonde Video sonde Type, size and electric charge of hydrometeors, pressure, temperature, humidity, wind Graupel Snow flake Sonde release & tracking Radar operation Scanning of Whether Radar Scenes in the campaign observation Volume scan RHI scan before and after synchronization During synchronization 14

5 Contents In-Situ campaign observation synchronized with Video-Sonde d (~ 5 min.) A new operational QPE algorithm for C-band polarimetric i radar (~ 10 min.) Classification of co-existing hydrometeors using a C-band polarimetric radar (~ 14 min.) Operational polarimetric radars in near future by the Ministry of Land, Infrastructure, Transportation and Tourism (MLIT), and one of an important tbackground d( (~20 min.) A new operational QPE algorithm for C-band polarimetric radar (2008) Eiichi NAKAKITA Disaster Prevention Research Institute, Kyoto University Hidenobu TAKEHATA Graduate School of Engineering, Kyoto University Katsuhiro NAKAGAWA National Institute of Information and Communications Technology 16 In short Data set (Ground) Mass weighted diameter D m will be introduced into estimators R(Z( HH ) and R(K( DP), as R(Z( HH, D m m) and R(K DP,, D m m) ). D m will be used also in the selection of R(Z HH, D m ) or R(K DP, D m ). We used the DSD data observed by 2DVD and disdrometers of impact type at Ogimi and Onna. Ogimi is located 15km and Onna is located 23km from COBRA. NiCT Site AMeDAS Ogimi 2DVD & impact type disdrometer From 2008 Onna 2DVD & impact type disdrometer 18 AMeDAS=Automated A t Meteorological l Data Acquisition System 19

6 Data set No. Start Time (UTC) End Time (UTC) : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :00 Radar COBRA Z HH, Z VV, Z DR, φ DP, ρ hv, LDR, Time step every 6 minutes 14PPI(0.5,1.1,1.8,2.5,3.3,4.2,5.3,6.5, 8.1,10.0,12.3,14.8,17.4,20.5 elevation angle ) Pulse width 2μs μ (300m) Beam width 0.9 Ground data Impact type disdrometer ( ) 2DVD (4-11 June. 2006) AMeDAS ( ) AMeDAS=Automated Meteorological Data Acquisition System 20 Strategy of developing the new algorism with Dm DSD data analysis Radar data analysis DSD data Radar data Gamma type DSD model:μ Preliminary procedure Estimated Polarimetric parameters Polarimetric parameters Optimized estimator R(Z HH ),R(K DP ) Existing algorithm of rainfall estimation Mass weighted diameter (D m ) optimized equation D m (Z HH,Z DR ) New algorithm 21 Preliminary procedure (1) Preliminary procedure (2) Removing of the radar echoes by nonmeteorological factor (such as ground clutter) using ρ HV (ρ HV <0.9) Z HH ZDR K DP Calibration of the system offset of Z DR using 2DVD and impact type disdrometers measurements. Attenuation correction with KDP using the self-consistent method by Bringi et al (2001) Finally, the radar data was converted into 100m mesh data, and the mesh data was spatially averaged over 1km 1km mesh

7 Optimization of basic estimators using observed DSD RZ ( ) = HH Z Assumption Data: impact type disdrometer at Okinawa in Japan ( ) DSD is Gamma distribution Axis ratio model of raindrop rd ( ) = D D D D RK ( ) = K HH DP DP Type of basic estimators: b0 RZ ( ) az, RK ( ) = ak b1 HH = 0 HH DP 1 DP Dmax 6 ZHHg = D N( D) dd 0 Dmax 6 DNDdD ( ) 0 ZDRg = 10log10 Dmax 7/3 6 K DPg 0 rd ( ) DNDdD ( ) Dmax rddnddd ( ) ( ) CkW g 1 = Dmax λ 3 DNDdD ( ) 0 Estimation of mass-weighted diameter (Dm) This fig. shows the comparison between the observations D m and the estimations D m by radar observations. Time series of radar-estimated D m (blue line) well explain that t of observed D m (bar). mm] Dm [ radar) D m (Z HH H, Z DR ) ( D = 0.54Z 10 m HH 0.917(0.1 ZDR ) D m (disdrometer) 10 June, 2006 Time series of radar-estimated Dm. 25 Optimization of estimators depending on class of Dm (1) Optimization of estimators depending on class of Dm (2) R(Z HH ) R(K DP ) R(Z HH, D m ) R(K DP, D m ) RZ ( HH ) = Z HH RK ( DP ) = K DP R(Z HH ).VS. Rg and R(K DP ).VS. Rg depend on the size of D m. These relations can be further sophisticated using D m ZHH Dm < ZHH 1.0 Dm < RZ ( HH, Dm ) = ZHH 1.5 Dm < ZHH 2.0 Dm < ZHH 2.5 Dm < K DP 0.0 Dm < KDP 1.0 Dm < RK ( DP, Dm ) = 35.00KDP 1.5 Dm < KDP 2.0 Dm < KDP 2.5 Dm <

8 Developed algorithm for switching between R(ZDR) and R(KDP) Validation using radar (1) comparison between algorithms Basic algorithm R alg RK ( DP ) if ZHH Thz, and KDP Thk,, = RZ ( ) otherwise. D m HH = 0.54Z HH 0.917(0.1 ZDR ) R(Z HH ) R(K DP ) Mean Bias [mm] RMSE[mm] CC Ratio of rain total [%] R(Z HH ) R(K DP ) R alg (Z HH, K DP ) alg HH DP Developed algorithm RK ( DP, Dm ) if(( ZHH T hz, or D m T h, dm ) Ralg = and KDP T hkdp, ) RZ ( HH, Dm) otherwise. D m is used also for judging which should be used, R(Z HH, D m ) orr(k DP, D m ) Introducing R alg (Z HH,KDm DP,Dimproved m ) both bias 2.99 and correlation 0.915coefficient 98.20!! These figures show comparison R alg (Z HH, K DP ) between R alg (Zthe HH,K raingauge K DP,D, D m m ) ) observations and radar observations. Basic algorithm R alg (Z HH, K DP )have been composed of R(Z HH ) and R(K DP ). D m Here, D m was used also in the selection of R(Z HH, D m )orr(k DP, D m ) Validation(2) comparison with current operational QPE (Radar-AMeDAS) Contents Comparisons of 1-h rainfall between COBRA's estimates using gproposed p the new algorithm and the Radar-AMeDAS precipitation data Radar-AMeDAS precipitation data is the data which is calibrated with ground rain-gauges and synthesized using the conventional radars in Japan. Radar-AMeDAS COBRA with new alg. 30 In-Situ campaign observation synchronized with Video-Sonde d (~ 5 min.) A new operational QPE algorithm for C-band polarimetric i radar (~ 10 min.) Classification of co-existing hydrometeors using a C-band polarimetric radar (~ 14 min.) Operational polarimetric radars in near future by the Ministry of Land, Infrastructure, Transportation and Tourism (MLIT), and one of an important tbackground d( (~20 min.) 31

9 Datasets Classification of co-existing hydrometeors using a C-band polarimetric radar (2009) Eiichi NAKAKITA Disaster Prevention Research Institute (DPRI), Kyoto University Yasuhiko SUMIDA Graduate School of Engineering, Kyoto University Kosei YAMAGUCHI Institute for Sustainability Science (ISS), Kyoto University Katsuhiro NAKAGAWA National Institute of Information and Communications Technology (NICT) Kenji SUZUKI Department of Agriculture, Yamaguchi University When the typhoon 0723 approached on 27th Nov., six Video-sondes were launched. Observation time 27th Nov.(JTC) 11/27 5:30(JST) The number of particles Start time End time Observation Rain Graupel Ice Snow Sum time (min) crystal flake No.1 3:37 4: No.2 5:58 6: No.3 6:53 7: No.4 7:58 8: No.5 10:51 11: No.6 11:36 11: sum Hydrometeor Classification Hydrometeor is classified into four types; rain, graupel, ice crystal and snow flake. The video-sonde observed the mixture of some types of particles over melting layer. In previous many researches they usually classified just one type of hydrometeor at each point. We consider that two types of hydrometeors mixed-exist, when the difference in values of evaluation index between the highest and second highest hydrometeors is small. Graupel+Ice crystal, cysa, Graupel+Snow flake aeand Ice crystal+snow cysa flake, are set in addition to each types. Hydrometeor Classification Membership Function Z HH μ j Z μ DR j Z HH Membership Function Z DR Membership Function ρ HV Membership Function K DP Membership Function Rain j = 1 Graupel j = 2 Ice crystal j = 3 Snow flake j = 4 ZHH μ j ZDR μ j μ ρ HV j K DP μ j 34 35

10 Malting Layer Detection by ρ HV Hydrometeor classification Method (mixture) Melting layer height Membership function MLH μ j Membership function ZHH ZDR ρhv KDP MLH μ j, μj, μj, μj, μ j ( j : Particle type index) Rain Malting layer Icing layer (Graupel, IC, SF) Hydrometeor Evaluation index MLH Z HV ( ) ( ) ( ) ( ) ( ) HH Z DR K μ μ μ μ ρ ρ μ DP ( ) Q = h Z + Z + + K j j j HH j DR j HV j DP 0 10 It is important to distinguish between rain and icing layer. Melting layer height is detected by ρ HV. We make the Melting Layer Height (MLH) membership function 36 Decide Particle Types( j ) Mixture of hydrometeors max{ Q j ; j = 1, 4 } = Q jmax if Q jmax - Q jsec 0.11 then particle type : jmax and jsec else particle type : jmax 37 Hydrometeor Classification (mixture) (1) Hydrometeor Classification (mixture) (2) Height(m) )雨 Heig ght H m ( Rain Graupel Ice Crystal Snow Flake Ice Crystal Snow Flake Graupel + Ice Crystal Rain Graupel + Snow Flake Mass Density(g/m Density(g/m 3 ) 3 ) n el ke C Rain Graupe Snow Flak G+IC G+SF IC+SF F F 38 39

11 Contents Current Radar network by MLIT in Japan In-Situ campaign observation synchronized with Video-Sonde d (~ 5 min.) A new operational QPE algorithm for C-band polarimetric i radar (~ 10 min.) Classification of co-existing hydrometeors using a C-band polarimetric radar (~ 14 min.) Operational polarimetric radars in near future by the Ministry of Land, Infrastructure, Transportation and Tourism (MLIT), and one of an important tbackground d( (~20 min.) Old type dual polarization C-band Conventional C-band full Polarimetric from April, Observed example by the new C-band full polerimetric radar (July 24,2009) Observed example by the new C-band full polerimetric radar (July 24,2009) Z H Z V φ DP ρ HV Z DR 偏波間位相差 φ DP 偏波間相関係数 ρ HV 42 Kyushu River Bureau, MLIT, Kyushu River Bureau, MLIT, 2009

12 Radar networks by MLIT in Japan in near future One of the important background for installing operational X-band polarimetric radar (2009) C-band Conventional C-band full polarimetric from April, 2009 C-band Doppler full polarimetric within 2-3 years X-band Doppler full Polarimetric from Apri, 2010 Eiichi NAKAKITA Disaster Prevention Research Institute (DPRI), Kyoto University Hiroyuki YAMABE Graduate School of Engineering, Kyoto University Kosei YAMAGUCHI Institute for Sustainability Science (ISS), Kyoto University Background Background On July 28 (Toga River, Kobe) About 50 people were washed away by the flush flood in Toga River, Kobe, Japan, without any overpass from embankment. Five people were died. In this case, many people were playing in the river side. This is the place where public people enjoy the water front. The local people and Kobe government have been making any efforts to develop such the water front. There is a risk that same disaster could occur in the any urban small rivers. 14:36 10 min. later 14:46 On August 5 (Zoshigaya, Tokyo) Six people, working in an underground sewage pipe system in Toshima ward, Tokyo, were swept away and five people were died. These disasters were occurred by isolated cumulonimbus (isolated convective rainfall)

13 Images from monitored movie managed by Kobe City Operationally distributed radar image 14:20 Became 14:36Ground was 14:38 Large size dark begun to Rain drop be wet Water level: -0.37m 14:40 View was blinded by rain 14:20 About 5~10 minute 14:25is necessary for the 14:30 14:35 14:40 information transmission. You can get 14:20 radar data(radar image) at 14:30 15:00 Water level: 1.05m 14: m rise in 10 min. 水位 :1.01m 14:46 15:10 15:40 Water level: 0.52m Water level: 0.28m 14:4444 Side works were covered by water Water level:-0.33m 14:42 Water level rising began -9- Radar observation at 14:20 14:42 Time 14:20 14:30 14:40 The rain area covered all the small river at 14:35: Flush flood occurred at 14:42. flash flood Concentration time is very short. In this case, you may think that this 7 minutes is enough for people to escape. However you should not forget that we need time for making and transmitting radar image. People can verify the radar image which shows us that the 14:42 rain area covered all the small river base is after flash occurred. The advantage of operational 3D scan Height 20km The baby rain cell 5km Cloud droplet cm-radar can not detect catch catch catch Low elevation scan detect raindrops after the cumulonimbus growing up. The baby rain cell cannot be detected by a low elevation observation. The volume scan radar monitoring can detect the baby stage of the rainfall.

14 Baby rain-cell generated at the height of 5~6 km 30 min. before flash flood 14:20 There is no precipitation at the ground. There is no precipitation at the ground. 14:22 The rainfall reached to the ground surface. 14:28 View from Osaka

15 The baby rain cell is generated in the upper atmospheric layers. 14:20 14:20 Became dark Water level: -0.37m 14:22 14:28

16 14:36 Ground was 14:38 begun to be wet Large size Rain drop 14:42

17 Radar networks by MLIT in Japan in near future Plan of installing operational X-band Doppler Polarimetric radars over majors urban areas by River Bureau, MLIT C-band Conventional C-band full polarimetric from April, 2009 C-band Doppler full polarimetric within 2-3 years X-band Doppler full Polarimetric from Apri, 2010 雲仙 大隅 国土交通省河川局 富士 From April, 2010 From April, 2011 (proposal) 既存 X バンドレーダ New operational network by X-band radars 60 km range 30 km range Kyoto Higher sensitivity by : X band radar Higher g spatial resolution by : X band radar (250~500 m) Dense network Free from attenuation by : Polarimetric function (KDP) Dense network together with C-band Higher accuracy by : Polarimetric function (ZDR, KDP) Shorter scan interval with low elevation : 1 minutes Shorter transmission time : 2 minute 3D image : Volume scan Kobe Osaka Higher accuracy by : By the River Bureau in MLIT New operational networks by X band polarimetric Doppler radars are scheduled to be in operation in various urban areas in Japan before next April. Concluding remarks In-Situ campaign observations synchronized with Video-Sonde d have been carried out. A new operational QPE algorithm for C-band polarimetric i radar was developed. d Algorithm for classifying the co-exising hydrometeors using a C-band polarimetric Radar was developed Plan for introducing operational polarimetric radars by MLIT is introduced with an important tbackground. 67