Docstoc

JMA Numerical Analysis Systems

Document Sample
JMA Numerical Analysis Systems Powered By Docstoc
					                           International Symposium on Radar and Modeling Studies of the Atmosphere
                                        12 November 2009, Kyoto University Uji Campus



    Current and future studies at MRI/JMA toward the
   dynamical and probabilistic prediction of local heavy rain


                                   Kazuo SAITO
                    Second Laboratory, Forecast Research Department,
              Meteorological Research Institute, JAPAN, ksaito@mri-jma.go.jp


Contents:
1. Development of the JMA nonhydrostatic model
2. Operational mesoscale QPF at JMA
3. Studies for dynamical or probabilistic precipitation forecasts at MRI
4. Plans toward the dynamical and probabilistic prediction of local heavy rain



                           Acknowledgements to members of mesoscale data assimilation group of MRI
                           (K. Aonashi, Y. Shoji, H. Seko, M. Hara, T. Kawabata, M. Kunii and T. Kuroda)
                           and Mesomodel Group of Numerical Prediction Division of JMA
           1. Development of the JMA nonhydrostatic model
    Ikawa (1988; JMSJ, 66, 753-776)
    Ikawa and Saito (1991; Tech. Rep. MRI, 28, 238pp)
                                            First description of MRI-NHM
                                             ○ Numerics on the dynamical cores
                                             ○ Bulk cloud microphysics
                                             ○ Turbulent closure model and surface
                                               boundary layer physics
                                            Airflow over a 3-
                                            dimensional bell-shaped
                                            mountain
                                                                     hm
                                            Zs ( x, y ) 
                                                            {( x  y ) / a 2  1}3 / 2
                                                                2     2




                                            with hm=100 m,
                                            U=8m/s, N=0.01s-1
                                            Left: a=6km, Right: a=1.2 km

Top: hydrostatic analytic solution
Middle: non-hydrostatic analytic solution
Bottom: Numerical solution by the model
                                                            http://www.mri-jma.go.jp/Publish/Technical/DATA/VOL_28/28.html
             Development of a nested model (Saito, 1994; JMSJ)
                 Numerical simulation of a local downslope wind in Japan

          u   v   w*
  DIVT                  0
          x   x    z *


J   {1 (U  U 0 )2  1 (V  V0 )2  2 (W *  W *0 )2  DIVT}d xdydz*
         2               2                2

    V




                                                   Wind and potential
                                                   temperature with a 2.5
                                                   km nonhydrostatic
                                                   model




        http://www.journalarchive.jst.go.jp/jnlpdf.php?cdjournal=jmsj1965&cdvol=72&noissue=2&startpage=301&lang=ja&from=jnltoc
        Fully compressible version with a map factor
Saito (1997; Geophys. Mag. Ser. 2, 2, 109-137.)
○Fully compressible, no linearization by the reference atmosphere
○HI-VI (3D implicit) scheme with non-iterative direct elliptic solver
○Direct evaluation of the buoyancy term
○Full evaluation of the diabatic heating term in pressure equation
○Map factor and curvature terms
○Mass-virtual potential temperature

  m   (1  0.61qv )(1 qc  qr  qi  qs  qg )
       p0 p Cv / Cp
       ( )
      Rm p0
○Fall-out of precipitable water substances in
continuity equation
        u  v      w
     m2{ ( )  ( )} m ( )  Prc,
 t      x m  y m    z m
 Prc   (Vr qr  Vs qs  Vg qg ) .
       z
Application to tropical island convection observed in MCTEX
Numerical simulation of diurnal evolution of tropical island convection
                                              (Saito et al, 2001: Mon. Wea. Rev.)
MCTEX: The Maritime Continent Thunderstorm Experiment in 1995 by BMRC,
  NCAR, CSU, etc., .. (Keenan et al., 2000: Bull. AMS, 81, 2433-2455)




                                   http://ams.allenpress.com/archive/1520-0493/129/3/pdf/i1520-0493-129-3-378.pdf
            Unified nonhydrostatic model between MRI and NPD
Documentation of the Meteorological Research Institute/Numerical
prediction Division unified nonhydrostatic model. Technical Reports of
the MRI, 42, 133pp. (Saito et al., 2001)

Flux correction scheme for advection
Parallelization for distributed memory computers
Re-implementation of the HE-VI method

   1   2P 1      g                   2
  ( )   2
                 (      P )  {                }2 P   FP.HE
       z * G z * mCm            Cm  (1   )
    1      2     1     2
   G2            2




                                     Numerical simulation of
                                     convective clouds around
                                     Japan in the winter cold air
                                     outbreak.

                                                http://www.mri-jma.go.jp/Publish/Technical/DATA/VOL_42/42.html
             Activities on the Earth Simulator
 A super high resolution (1 km) simulation of winter monsoon
 cloud bands over the Sea of Japan




GMS-5 :                               1km-NHM :        Eito et al. (2004)
                 15JST 14 JAN. 2001   Total Water Path
Visible-Image
             2. Operational mesoscale NWP and the short
                 range precipitation forecast at JMA
Operational run with a horizontal resolution of 10 km was
started on 1 September 2004
Saito et al., 2006: Mon. Wea. Rev., 134, 1266-1298.
  •Kain-Fritsch convective parameterization scheme
  •Time split of gravity waves and advection
  •Non-local boundary layer scheme
               Mesoscale model (MSM) at JMA
                                          • Purpose: to provide information to
                 3600 km
                                            prevent disaster, such as heavy
                                            rain, wind.
     2900 km




                                          • Output is used for:
                                             – Short-term weather forecast
                                             – Very short range forecast of
                                               precipitation
                                             – Forecast for aviation (TAF)
                                             – Forecast of a storm surge
                                             – Forecast of air pollutant distribution
         Computational domain
•Initial condition: given by 4DVAR with
3-hour data assimilation window.
•Boundary condition: provided from
GSM (TL959L60) forecasts.
                Upgrade of MSM
2006 March Horizontal resolution enhancement to 5 km
2007 May 3 hourly 33 hour forecast
(Saito et al., 2007; JMSJ)
・hybrid vertical coordinates
・Sedimentation of cloud ice
・MYNN3 turbulent closure model
・New trigger function in the Kain-Fritsch scheme




                  UL)Satellite image
                  UR) MSM(10km)
                  LF) MSM(5km)
                  00UTC 20 January 2006


                                          http://www.jstage.jst.go.jp/article/jmsj/85B/0/271/_pdf/-char/ja/
            Implementation of MYNN-3
• The world first implementation of the level 3 closure
  model to the operational NWP model
                                         Satellite IR image
Winter monsoon case
Boundary layer is well developed
with MY3, which is more realistic
than that with Deardroff scheme.



 TKE
  MY3                        Deardroff
    Nonhydrostatic 4DVAR (Apr. 2009-)

   Radar-AMeDAS                JNoVA
                                                Meso 4DVAR




FT=24 from 2006 Aug 17 15UTC
                                       Sawada and Honda (2009)
MSM QPF performance at JMA    1mm/3hr 20km




       Threat score (TS)     Bias score (BS)




    False Alarm rate (FAR)     Miss rate (MS)
False Alarm rate (FAR) in MSM 10mm/3hr 10km verf.



                                                     Nonhydrostatic
                                                     4DVAR




  Meso4DVAR

              Nonhydrostatic
              model          10km to 5km Upgrade 20km Global
                                         physical model
                                         process
 GPS TPW in MSM (28 Oct. 2009-)




 Radar-AMeDAS      with GPS TPW         without GPS TPW

FT=3 from 21 UTC 20 July 2009

                                Ishikawa (2009)
                                     Very Short Range Forecast of
                                         Precipitation at JMA
                      · Forecast of 1 hour accumulated precipitation up to FT=6.
                      · Covering whole over Japan with 1 km grid
                      · Operated at 30 min. intervals
                      · Calculated by merging extrapolation of calibrated radar echo
                      intensity and prediction from MSM.
Quality of forecast




                             Merging

                                                              MSM
                                          Extrapolation
                                         Persisten
                                         ce
                      0      3       6        9     12     15      18    Forecast time
                                                                         (hour)
 20km verification
10mm/h June 2009




           Ebihara (2009)
 Development of a cloud resolving
 operational model (LFM) at JMA (Dx=2km)


                                                  North region
Operation : Mar. 2012 –                           1100x1300 km2
Domain ... 1 domain (Mar. 2012-)
          3 domains (Mar. 2013-)




                                   South region
                                   1300x1100km2



                                                                  Centre region
                                                                  1600x1100km2
  Test of data assimilation at JMA
 AMeDAS: 290 wind, temperature                   3DVAR(5km)
 Wind Profiler: 8
 Doppler radar velocity :7
 ACARS wind, temperature

                                                         LFM(2km)


               18    21         22     23      00       03   06     09   12UTC
                                      MSM

                    JNoVA   JNoVA      JNoVA   JNoVA



                          MSM        MSM    MSM              LFM


                                                  MSM

                                               Ujiie (2008)
Quantitative Precipitation Forecast 06 UTC on 12th Sep. 2009
Initial time is 00 UTC on 12th 2009


        OBS                            2km




        5km                           20km




                                                 Ishida (2009)
   3. Studies for dynamical or probabilistic
        precipitation forecasts at MRI
FY 2005 – 2008: Grant-in-Aid for Scientific Research study
 “Study on data assimilation and evaluation of forecast
reliabilities for dynamical prediction of heavy rainfall”
    by MRI, Kyoto University (RISH), Tsukuba University and NPD/JMA
        •Data assimilation studies on heavy rainfall events
        •WWRP Beijing Olympic 2008 RDP


FY 2009-: “Study on advanced data assimilation and cloud
resolving ensemble forecast for prediction of local heavy rain”.
 by MRI, RISH, National Institute of Information and Communications Technology,
                      Tohoku University and NPD/JMA
                 Development of a cloud resolving 4DVAR
                                                                                                    Assimilated forecast (16JST)




                                                                                                  1.0   10   20   40
                                                                                             60
                        Rain fall amount during 10 munutes at Nerima             Amedas
mm                                                                               Forecast
     30
                                                                                            Kawabata et al. (2007; JMSJ)
     25


     20


     15


     10


      5


      0
          1510   1520    1530   1540   1550   1600   1610   1620   1630   1630   1650 JST
4DVAR assimilation of the Suginami Heavy rain
on 4 Sep 2005
   4DVAR解析
                                観測
                      降水強度 (mm/h)




                               Kawabata et al. (2009; MWR, submitted)
   Assimilation of radar
   reflectivity 2030-2100JST
Development of Cloud resolving 4DVAR




  2130 JST   4DVAR with           4DVAR without        1st guess
  Obs        assimilation of      radar reflectivity
             radar refectivity




                                 Kawabata et al. (2009; MWR, submitted)
Flashflood 28 July 2008 in Kobe City
GPS TPW impact using Meso 4DVAR
CNTL                           GEONET+IGS




   Shoji et al. (2009; SOLA)
WWRP Beijing Olympic 2008 RDP Project
Beijing Olympic 2008 (B08) FDP/RDP
is a WWRP’s research project
conducted with the Beijing Olympic
Games of August 2008.

The project is divided into 2
components:
FDP: Forecast demonstration up to
FT=6 based on nowcasting.
RDP: Research and development up
to FT=36 based on the mesoscale (15
km) ensemble prediction.

NCEP, MSC, ZAMG&Meteo Fr.,
MRI/JMA, NMC/CMA, CAMS/CMA
participated in RDP.
      WWRP Beijing Olympic 2008 RDP

          Common verification area




                                       Forecast domain (2006)




     MA      Forecast domain (2007 and 2008)


RF/mfboundary
Model domain: 232 x 200 grids with 15 km L40 NHM
   Initial perturbation method

Following 5 methods were compared;

1) Downscale of one-week EPS of JMA (WEP)
2) Global targeted SV (GSV)
3) Mesoscale model SV (MSV)
4) Mesoscale model BGM (MBD)
5) Mesoscale model LETKF (LET)


Horizontal resolution of inner models of MSV, MBD and
LET to compute perturbation was unified to 40kmL40.
                           Tier-1 MEP systems 2008
 Centers           Model             Initial      Initial       Lateral       Lateral           Physical
(Nation)     Vertical levels and   condition   perturbation    boundary      boundary         perturbation
            number of members                                  condition    perturbation
 NCEP       WRF-ARW (L60M5)          NCEP       Breeding      NCEP Global     NCEP            Multi-model
 (USA)      WRF-NMM (L60M5)         3DVAR                        EPS        Global EPS
             GEFS-Downscaled
              (T284L60M5)
MRI/JMA           NHM                 Meso      Targeted      JMA Global       Global              No
(Japan)         (L40M11)             4DVAR     Global SV        Forecast      forecast
                                   (20kmL40)   (T63L40)       (TL959L60)     (T63L40)
                                                                            initiated by
                                                                            targeted SV
  MSC             GEM               MSC        MSC Global     MSC Global    MSC Global      Physical tendency
(Canada)        (L28M20)            Global       EnKF           EPS           EPS           perturbation with
                                    EnKF                                                      Markov chain,
                                                                                           surface perturbation
ZAMG &           ALADIN            ECMWF         Blending      ECMWF         ECMWF            Multi-physics
Meteo-Fr.       (L37M17)            Global     ECMWF SV         Global      EPS forecast
(Austria                           4DVAR           with        Forecast
   and                                           ALADIN
 France)                                        Bred Mode
 NMC            WRF-ARW              WRF-       Breeding      CMA Global    CMA Global        Multi-physics
(China)         (L31M15)            3DVAR                       EPS           EPS
 CAMS            GRAPES            GRAPES-      Breeding      CMA Global    CMA Global        Multi-physics
(China)          (L31M9)            3DVAR                       EPS           EPS
  2008 Final experiment
  24 July -24 August 2008


B08RDP website by CMA
http://www.b08rdp.org

Probability of precipitation (1 mm / 3
hours)
12UTC 10 August 2008 (FT=18).

Observation is 6 hour rain from 00 UTC to 06
UTC
Threat Scores for 6 hour precipitation of Control run
(7/25 – 8/23)
Bias Scores for 6 hour precipitation of Control run
(7/25 – 8/23)
Surface (2m) temperature of Control run
(7/25 – 8/23)


   FT=18 (14LT)




   FT=18 (14LT)
          Ensemble spreads of surface variables
          (7/25 – 8/23)

                             JMA/MRI                   Psea                                                        NCEP                   Psea                                                                 MSC                    Psea
                                                                                                                                                                                                                                      Us
                                                       Us                                                                                 Us
                                                       Vs                                                                                 Vs                                                                                          Vs
                Ensemble Spread(MRI/JMA)               Ts                                          Ensemble Spread(NCEP)                  Ts                                                       Ensemble Spread(MSC)               Ts
                                                       RHs                                                                                RHs                                                                                         RHs
                                                       RR3H                                                                               RR3H                                                                                        RR3H

           2                                                10                                2                                                10                                         2                                                10


          1.5                                               7.5                              1.5                                               7.5                                       1.5                                               7.5




                                                                                                                                                                                                                                                       Spread(RH,RR3H)
                                                                                                                                                           Spread(RH,RR3H)
                                                                  Spread(RH,RR3H)




                                                                                                                                                                             Spread
Spread




                                                                                    Spread
           1                                                5                                 1                                                5                                          1                                                5


          0.5                                               2.5                              0.5                                               2.5                                       0.5                                               2.5


           0                                                0                                 0                                                0                                          0                                                0
                0    3   6   9 12 15 18 21 24 27 30 33 36                                          0    3   6   9 12 15 18 21 24 27 30 33 36                                                   0    3   6   9 12 15 18 21 24 27 30 33 36
                                  FT(hours)                                                                          FT(hours)                                                                                   FT(hours)

                                ZAMG                   Psea
                                                       Us
                                                                                                                   NMC                    Psea
                                                                                                                                          Us
                                                                                                                                                                                                             CAMS                     Psea
                                                                                                                                                                                                                                      Us
                                                       Vs                                                                                 Vs                                                                                          Vs
                    Ensemble Spread(ZAMG)              Ts                                              Ensemble Spread(NMC)               Ts                                                       Ensemble Spread(CAMS)              Ts
                                                       RHs                                                                                RHs                                                                                         RHs
                                                       RR3H                                                                               RR3H                                                                                        RR3H

            2                                               10                                 2                                               10                                          2                                               10


          1.5                                               7.5                              1.5                                               7.5                                       1.5                                               7.5




                                                                                                                                                                                                                                                 Spread(RH,RR3H)
                                                                  Spread(RH,RR3H)




                                                                                                                                                     Spread(RH,RR3H)




                                                                                                                                                                                Spread
 Spread




                                                                                    Spread




            1                                               5                                  1                                               5                                           1                                               5


          0.5                                               2.5                              0.5                                               2.5                                       0.5                                               2.5


            0                                               0                                  0                                               0                                           0                                               0
                0    3   6   9 12 15 18 21 24 27 30 33 36                                          0    3   6   9 12 15 18 21 24 27 30 33 36                                                   0    3   6   9 12 15 18 21 24 27 30 33 36
                                  FT(hours)                                                                          FT(hours)                                                                                   FT(hours)
                       Brier scores and ROC curve for 6 hour precipitation
                       of ensemble mean (7/25 – 8/23)
                                           Probability of Precipitation
                                           Brier scores
                       0.2
                                                                                                          ROC curve(1mm/6hrs)
                0.15                                                            MRI/JMA 1
Brier score




                                                                                NCEP
                                                                                ZAMG
                                                                                       0.9
                       0.1
                                                                                MSC    0.8
                                                                                NMC
                0.05                                                            CAMS   0.7
                                                                                                0.6




                                                                                     Hit rate
                        0
                                                                                                0.5
                             0                5               10          15                                                      Hr=Fr
                                           Threshold(mm/6hrs)                                   0.4                               MRI/JMA
                                         Probability of Precipitation                                                             NCEP
                                           Threshold 1mm/6hrs                                   0.3                               ZAMG
                         0.3
                                     Brier scores for 1mm/6hrs                                  0.2                               MSC
                       0.25                                                                                                       NMC
                                                                               MRI/JMA          0.1                               CAMS
         Brier score




                         0.2                                                   NCEP                                               系列2
                                                                               ZAMG              0
                       0.15
                                                                               MSC                    0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
                         0.1                                                   NMC
                                                                               CAMS                            False alarm rate
                       0.05

                             0
                                 6   9   12 15 18 21 24 27 30 33 36
                                                  FT(hours)
CNTL      RAM           SPREAD




p01       p02           p03            p04          p05




m01       m02           m03           m04           m05




       Ensemble forecast using Mesoscale SV for heavy
               rain event in Japan Dx=15km
                  FT=6 from 12UTC28Aug 2008
                                              Kunii et al. (2009)
       RAM                CNTL         SPREAD




                                        Pr02 : >=5.0mm/3hour Prob.
             Probability of rainfall    Pr04 : >=20.0mm/3hour Prob.
                                        Pr05 : >=50.0mm/3hour Prob.


Pr02                      Pr04
                                       Pr05
                           NHM-LETKF
                                    observation
       analysis
                                       R


                          Pa

                                                                   T=t2

                                                   Xif (Xif )T
                                            Pi 
                                             f

                                                       m 1
                                      Pf
  Pa
analysis                    Ensemble mean
       T=t0                      T=t1
                        Miyoshi and Aranami (2006; SOLA)
                  #MLEF has also been developed at MRI (Aonashi,
  Data assimilation of GPS PWV with NHM-
    LETKF for the 2008 Kobe flashflood
                                    Seko et al (2009)




           12JST:観測           14JST:観測
                Forecast of 1.6km-NHM




1100JST    1130JST        1200JST            1230JST
International Research for Prevention and Mitigation of
Meteorological Disasters in Southeast Asia
          Ensemble Forecast for
         Myanmar Cyclone Nargis
≈4m   Storm Surge at Irrawaddy river point
               (16.10N, 95.07E)

      Kuroda et al (2009; JMSJ, submitted)
       Saito et al (2009; JMSJ, submitted)
       Kunii et al (2009; JMSJ, submitted)
       Shoji et al (2009; JMSJ, submitted)   47
         LETKF application is underway
     4. Plans toward the dynamical and probabilistic
               prediction of local heavy rain
                                    GPS
                                   satellite

                                                            Cloud resolving data
                                         rainwater          assimilation for deep
                                                     Cb
                                                            convection
                       Mosit air


Doppler radar, lidar     GPSreceiver           Microwave
                                               radiometer
High density observation campaign around Tokyo
 metropolitan areas by MRI and JMA 2010-2014

                                                                氷
                                                               雷晶
                                         降水コア           ひょう      あられ
                                                                                               Polarmetric C band
                                                                                               Doppler radar (MRI)
             Doppler
              Lidar


                                                                                               2D scan
                                    Surface rain gauge                                       X band radar




                                         GPS receiver                            Lightning
                                                                3D scan
                                                              X band radar       Detection


                                sondes


 Moist air                MW                                                 Collaboration with MILT,
                       radiometer
                                                                             NIED, NICT, OU, ..
                                                                                 約10km
                                                 初期状態       微小時間             予報時刻
                                                  x(t0)      x(ts)            P(t=tl)


Cloud resolving ensemble                                                                 ば
                                         り程解                                             ら
          NWP                             度析
                                          の誤
                                          広差
                                                                                         つ
                                                                                         き
                                                                                         (
                                                                                         ス
                                          が                                              プ
                                                                                         レ
                                                                                         ッ
                                                                                         ド
                                               解析値                                       )
                                                     線型的な            非線型的な
                                                     時間発展            時間発展
                                                                             アンサン       単独予
                                                                             ブル平均        報




                               SV
                               LETKF
                               MLEF




Goal:
Dynamical and probabilistic prediction of
local heavy rain with a lead time in the
scale of local municipalities by near real
time cloud resolving ensemble NWP
(hourly, 1-2 km, 40-100 members)
    The Next Generation Super Computer




The Next-Generation Supercomputer is under construction by RIKEN at
Kobe, which becomes available for the study of cloud resolving NWP in 2010.
Collaborations with universities (Tokyo, Kyoto, Nagoya, Tohoku, ..)

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:6
posted:10/15/2012
language:English
pages:46