Shinsuke Satoh, F. Isoda, T. Sano, H. Hanado (NICT), T. Ushio (Tokyo - PowerPoint PPT Presentation

Shinsuke Satoh, F. Isoda, T. Sano, H. Hanado (NICT), T. Ushio (Tokyo Metropolitan Univ), S. Otsuka and T. Miyoshi (RIKEN) 7 th International Symposium on Data Assimilation (ISDA2019) @RIKEN-CCS, Jan. 24, 2019 1

Introduction ・ In recent years, severe weather disasters caused by localized heavy rainfalls or tornadoes have occurred frequently in various parts of Japan. ・ We developed a X-band Phased Array Weather Radar (PAWR) to watch and predict the severe weather. The Flash flood at Toga River in Kobe city (28 July 2008 ） PAWR measures 3-dim fine structure of precipitation with 100 m range resolution and 100 elevation angles every 30 seconds. ・ The first PAWR was installed at Osaka University, Suita in 2012. The second and third PAWRs were install at Tsukuba Tornado NICT Kobe and NICT Okinawa in 2014, respectively. (6 May 2012) The observation area of MLIT C-band Suita Kobe Okinawa radar and X-band MRI@ MP-PAWR Tsukuba Saitama MP radar (small in 2012 in 2014 in 2014 blue circles). in 2015 in 2017 2

Phased Array Weather Radar (PAWR) 3-dim. d ense observation every 30 sec. 3-dim measurement using 3-dim measurement using 128 slot-array antennas a parabolic antenna (150 m, with fan-beam transmitting and DBF receiving. 15 EL angles in 5 min) (100 m, 100 EL angles in 30 sec) 3

PAWR web page （ https://pawr.nict.go.jp/ ） Google maps display Rainfall Summary Retrieve archived past data Real time display （ within 1 min of obs ） 4

RIKEN real-time weather forecast http://weather.riken.jp Real-time demonstration of 3D nowcasting 30-second update nowcasting for 10 minutes started on July 3, 2017. Otsuka et al. Wea. Forecast, 2016 5

PAWR smartphone application Free app. for Android and iPhone 3D rainfall display (2 nd year ver.) - Real-time 3D rainfall display every 30 sec. (3 rd year ver.) - Heavy rainfall forecast by push notification http://pawr.life-ranger.jp 6

Request for faster QC algorithm 7th EL 2nd EL blocked by Data quality control (QC) topography such as clutter removal is essential in order to use PAWR observation 20 dBZ data for data assimilation (BDA) and 3D-nowcasting. clutter clutter map contour The Ruiz 's QC algorithm (SOLA, 2015) used for the BDA experiment requires calculation time of 40 seconds. However, it is necessary to develop a faster and general-purpose QC algorithm to perform real-time processing on the various observation data. Perform QC calculation and data transfer within 10 seconds for BDA and 3D nowcast Ruiz et al. SOLA, 2015 7

Surface clutter and interference echoes Interference echoes Ground clutter echoes Airplane echo ＋RangeSL Clutter echoes by ships Ground clutter Suita PAWR (fine weather) Kobe PAWR (fine weather) Add another data at 2016/12/01, 10:03:30 8

Contents and overview of QC flag file QC flag < 8 bit > [0]Valid data, [1]Shadow, [2]Clutter possible, [3]Clutter certain, [4]Noise, [5]RainAttn., [6]RangeSL, [7](Reserve) l A new file of 1-byte QC flag data is provided in the same format of the same polar-coordinates as Ze and Vr data. (e.g. 20150808-160021.all _pawr_qcf.dat, kobe_20150808160000_A08 _pawr_qcf.dat ) l The QC flag file will be created in NICT Koganei in real-time (within 10 sec.) < CONTENTS > [0] Valid data ： if ( Ze > -327.68 & Vr > -327.68 ) then (1) [1] Shadow ： if ( ASL(Dem) > beamHT using 4/3 equiv. earth radius) then (1) [2] Clutter possible (clutter map) ： if (statistical Ze_PD > 20%) then (1) [3] Clutter certain ： if (Ze_PD>20% & -1.5<Vr<1.5ms -1 & ZeText > 3.0) then (1) [4] Noise (Interference) ： if (rng_num > 500 & Ze_std/Ze_avg < 0.5 ) then (1) [5] Rain attenuation ： if (Ze_inetg > 50 dBZ & delta_Ze < -2 dB/km ) then (1) [6] Range Side Lobe ： if (Ze > 40 dBZ & ZeText < 1.5 & ZrTextAz < 0.8) then (1) [7] (Reserve): future use (e.g. abnormal Vr., uncorrected aliased velocity) 9

Ze and QC flag in PPIs (EL=2.0 deg) 2015/07/17,08:30:19 JST 2015/08/08,16:00:21JST 2015/12/18,10:40:34 JST Ze Ze Ze QC flag QC flag QC flag 120 km Valid >1 Shadow >2 Clutter map >4 Clutter certain >8 Interference noise >16 120 km Interference Noise Convective Rain Stratiform Rain (fine weather) determined by ZeStd/ZeAvg < 0.5 10

Range side-lobe contamination Ze EL=3.9 False echoes at the forth and back of a strong echo 2016/08/14, 14:50:16 AZ=33.6 Ze 60 km HEIGHT (km) RANGE (km) QCF EL=3.9 RangeSL flags are determined by ZeMax, ZeText, and ZeTextAZ QCF 2016/08/14, 14:50:16 AZ=33.6 HEIGHT (km) RANGE (km) 11

Record breaking heavy rainfall in July 2018 Rainfall summary for 3 days by Kobe PAWR 5 100 2018/07/07 2018/07/05 2018/07/06 4 80 3 60 2 40 1 20 0 0 00 03 06 09 12 15 18 21 00 03 06 09 12 15 18 21 00 03 06 09 12 15 18 21 JST 800 10 2018/07/05 2018/07/06 2018/07/07 8 600 6 400 4 200 2 0 0 00 03 06 09 12 15 18 21 00 03 06 09 12 15 18 21 00 03 06 09 12 15 18 21 JST Rainfall distributions observed by Kobe PAWR (CAPPI of 3 km in height) 2018/07/06 11:00 JST 12:00 JST 2018/07/06 13:00 JST 2018/07/06 Strong echoes are embedded in wide-spread stratiform rain 12

Vertical section of the heavy rainfall (a) 12:00 JST 2018/07/06 HEIGHT(km) F A Bright band o o A (stratiform) (b) convective HEIGHT(km) E B C B o D convective (f) (c) HEIGHT(km) HEIGHT(km) o F o C (e) (d) convective HEIGHT(km) HEIGHT(km) o E o D DISTANCE from PAWR (km) DISTANCE from PAWR (km) ? 13

Incorrectly determined QC flags Original Ze QC flag QCed Ze 12:00 JST 12:00 JST EL=1.0 EL=1.0 12:00 JST EL=1.0 06Jul18 06Jul18 06Jul18 A o B HEIGHT(km) o A A o o A DISTANCE from PAWR (km) HEIGHT(km) o B B o o B DISTANCE from PAWR (km) Valid >1 Rain attenuation >32 Shadow >2 Range sidelobe >64 Clutter map >4 Clutter certain >8 14

Computation time for creating QC flag << original without Interference Noise and Rng SL >> Only clutter detection (v0.8) after 19 June ## Input file: 20150717-083019.all. 10000000.dat, and .20000000.dat # Total make qc flag real time = 7.000 proc time = 7.890 # Input data read: real time = 0.000 proc time = 0.550 ７ sec. # Calc Ze_ave, rinteg: real time = 1.000 proc time = 0.500 (single CORE) # Calc Ze_texture: real time = 5.000 proc time = 5.250 # Make QC flag: real time = 1.000 proc time = 1.570 # Output QC flag: real time = 0.000 proc time = 0.020 << Single core CPU >> Current operational ver (v1.1) after 15 Sep ## Input file: 2015-0717/20150717-083019.all. 10000000.dat, and .20000000.dat # Total create qc flag real time = 34.000 proc time = 34.410 # Input data read: real time = 0.000 proc time = 0.470 34 sec. # Calc Ze_ave, rinteg: real time = 15.000 proc time = 14.820 (single CORE) # Calc Ze_texture: real time = 17.000 proc time = 17.160 # Judgement of QCF: real time = 2.000 proc time = 1.930 # Output QC flag: real time = 0.000 proc time = 0.030 openMP (8 threads) << -O3 & -fopenmp & OMP_NUM_THREADS=8 >> MOP_NUM_THREADS= 8 ## Input file: 2015-0717/20150717-083019.all. 10000000.dat, and .20000000.dat # Total create qc flag real time = 9.000 proc time = 15.490 # Input data read: real time = 1.000 proc time = 0.470 9 sec. # Calc Ze_ave, rinteg: real time = 1.000 proc time = 7.270 (4 CORE) # Calc Ze_texture: real time = 5.000 proc time = 6.390 # Judgement of QCF: real time = 2.000 proc time = 1.330 # Output QC flag: real time = 0.000 proc time = 0.030 15

Summary l The PAWR measures dense 3-dim precipitation data (100 m, 100 EL angles) every 30 seconds. l The PAWR data QC is essential for DA and 3D-nowcasting, because the observation data includes various unnecessary echoes. l We have developed real-time data QC within 10 seconds. l The correct QC flags were calculated in most of convective and stratiform echoes, but incorrect QC flags were also seen in the case of record-breaking heavy rainfall. 16

BACKUP 17

QC flag of stratiform rain echo PPI Vr Ze (EL=1.0 deg) dBZ ms -1 2015/07/17 08:30:19JST QC flag Ze Texture 1/n Σ [Ze(i,j)-Ze(i-1,j)] in 11 range x 5 az dB 120 km Valid >1 Shadow >2 Clutter map >4 Clutter certain >8 120 km 18

QC flag of convective rain echo PPI Vr Ze (EL=1.0 deg) dBZ ms -1 2015/08/08 16:00:21JST Ze Texture QC flag dB 120 km Valid >1 Shadow >2 Clutter map >4 Clutter certain >8 120 km 19

PAWR web page （ http://pawr.nict.go.jp/ ） Monthly Rainfall Summary July 2017 7 days Jul.16 Google maps display Jul.23 Aug 4 5 Jul.30 Retrieve archived 6 24 hours past data Area of Rain (Blue) Aug.6 Avg. Rain Rate (Red) Max. Rain Rate (Green) Daily Rainfall Summary 20 [ x 0.01 mm/hr ]