Ens Generation by HC12110401513

VIEWS: 1 PAGES: 46

									Initialization and Ensemble generation for
Seasonal Forecasting




                                                                                          Magdalena A.
                                                                                            Balmaseda



       Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
   Outline

• The importance of the ocean initial conditions in seasonal forecasts
     A well established case: ENSO in the Equatorial Pacific

     A tantalizing case: NAO forecasts

• Ocean Model initialization
     Ocean initialization: requirements

     Standard practice: assessment

     Other initialization strategies: assessment

     Role of ocean initialization into context

• Ensemble Generation: Sampling Uncertainty
     Seasonal forecasts versus Medium range: different problems, different
      solutions?
     The ECMWF ensemble generation system.
     Other ensemble generation strategies



            Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
   The Basis for Seasonal Forecasts

•Atmospheric point of view: Boundary condition problem
    Forcing by lower boundary conditions changes the PDF of the
     atmospheric attractor
                                           “Loaded dice”
    The lower boundary conditions (SST, land) have longer memory
       o Higher heat capacity (Thermodynamic argument)
       o Predictable dynamics



•Oceanic point of view: Initial value problem
    Prediction of tropical SST: need to initialize the ocean subsurface.
    Examples:
       o A well established case is ENSO
       o A more tantalizing case is the importance subsurface temperature in the
         North Subtropical Atlantic for seasonal forecasts of NAO and European
         Winters.


           Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                    Need to Initialize the
                    subsurface of the ocean
Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
   A tantalizing case: SF of NAO and European Winters
                        ASSIM (S3)                                                                                                                  20051016
                        Potential temperature contoured every 0.25 deg C

              SST anomaly Autumn 2005
                        Horizontal section at 5.0 metres depth
                        Plot resolution is 1.5 in x and 1 in y                                                                                Interpolated in x and y
                                                                                                                                                                         T anomaly@ 90m: Autumn 2005
                O
              80 N                                                                                                            80 N    O       2.75
                                                                                                                                              2.25
                O
              60 N                                                                                                            60 N
                                                                                                                                      O       1.75
                                                                                                                                              1.25
                                                                                                                                              0.75
   Latitude




                O                                                                                                                     O
              40 N                                                                                                            40 N
                                                                                                                                              0.25
                O
              20 N                                                                                                            20 N
                                                                                                                                      O
                                                                                                                                              -0.5
                                                                                                                                              -1
                0   O
                                                                                                                              0   O           -1.5
                                                                                                                                              -2
                O
              20 S                                                                                                            20 S
                                                                                                                                      O       -2.5
                                                                                                                                              -3
                        O           O             O              O    O       O      O      O         O     O         O
                    80 W         70 W         60 W         50 W      40 W    30 W   20 W   10 W   0       10 E      20 E

                                                                            Longitude
                 ASSIM Apr 21
MAGICS 6.10 hyrokkin - neh Fri (S3) 17:35:41 2006                                                                                                    20060115
                        Potential temperature contoured every 0.25 deg C
                        Horizontal section at 5.0 metres depth
                SST anomaly Winter 2005/6
                        Plot resolution is 1.5 in x and 1 in y                                                                                 Interpolated in x and y



                O
              80 N                                                                                                            80 N        O    2.75
                                                                                                                                               2.25
                O
              60 N                                                                                                            60 N
                                                                                                                                          O    1.75
                                                                                                                                               1.25
                                                                                                                                               0.75
                                                                                                                                                                         Anomalies below the mixed
   Latitude




                O                                                                                                                         O
              40 N                                                                                                            40 N
                                                                                                                                               0.25
                O
              20 N                                                                                                            20 N
                                                                                                                                          O
                                                                                                                                               -0.5                         layer re-emerge and
                                                                                                                                               -1
                                                                                                                                               -1.5
                0   O
                                                                                                                              0       O




                                                                                                                                               -2
                                                                                                                                                                             Do they force the
                O
              20 S                                                                                                            20 S
                                                                                                                                          O    -2.5                           Atmosphere?
                                                                                                                                               -3
                        O            O            O              O     O       O      O      O        O         O         O
                    80 W         70 W         60 W          50 W     40 W    30 W   20 W   10 W   0       10 E      20 E

                                                                            Longitude
MAGICS 6.10 hyrokkin - neh Fri Apr 21 17:36:24 2006




                                                      Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
End to End Forecasting System
   atmos
    obs




                                                                                                           Reliable probability
   atmos             atmos reanalysis            initial
     DA                                        conditions
                        land,snow…? ensemble




                                                                                                           forecasts
                                                                                     Probabilistic
                                         AGCM




                                                                                     forecast
SST analysis

                                                        OGCM
                                                         generation
                                                 initial




                                                                                                                      products
   ocean
                                               conditions
                    ocean reanalysis
    DA




                                                                                                               Tailored
                         sea-ice?

 ocean obs


           Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
        Main Objective: to provide ocean Initial conditions for coupled
                                  forecasts


       Ocean                                                                           Real time Probabilistic
       reanalysis                                                                        Coupled Forecast
time




   Coupled Hindcasts, needed to estimate climatological PDF,
   require a historical ocean reanalysis


                                                                                  Consistency between historical
                                                                                  and real-time initial initial
                                                                                  conditions is required
             Quality of reanalysis affects the
                   climatological PDF

              Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
    Creation of Ocean Initial conditions


•    Ocean model driven by surface fluxes:
     Daily fluxes of momentum, Heat (short and long wave), fresh
       water flux
     From atmospheric reanalysis ( and from NWP for the real time).
          but uncertainty in surface fluxes is large.




           Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Uncertainty in Surface Fluxes:                                         Equatorial Atlantic: Taux anomalies

Need for Data Assimilation                                                                                         ERA15/OPS
                                                                                                                      ERA40

•   Large uncertainty in wind products

    lead to large uncertainty in the

    ocean subsurface

•   The possibility is to use additional
                                                    Equatorial Atlantic upper heat content anomalies. No assimilation
    information from ocean data

    (temperature, others…)

•   Questions:

      1. Does assimilation of ocean
         data constrain the ocean
         state?                                           Equatorial Atlantic upper heat content anomalies. Assimilation
      2. Does the assimilation of
         ocean data improve the ocean
         estimate?
      3. Does the assimilation of
         ocean data improve the
         seasonal forecasts


                   Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Real Time Ocean Observations
Moorings              ARGO floats
                                                                                XBT (eXpandable
                                                                                BathiThermograph)




                                                                                  Satellite

                                                                                SST


                                                                                                     Sea Level




     Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Time evolution of the Ocean Observing
System


             1982                   1993                          2001

XBT’s 60’s       Satellite SST Moorings/Altimeter ARGO




                                                                                 TRITON

                                                                      1998-1999
                                                                      PIRATA

       Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
       Ocean Observing System
       Data coverage for June 1982
                             OBSERVATION MONITORING
                               X B T p r o b e s: 9 3 7 6 p r o f i l e s                                        Changing observing
        60°E         120°E              180°                     120°W      60°W           0°




60°N                                                                                            60°N
                                                                                                               system is a challenge for
                                                                                                                 consistent reanalysis
30°N                                                                                            30°N




  0°                                                                                            0°




30°S                                                                                            30°S




60°S
                                                                                        Data coverage for Nov 2005
                                                                                                60°S




        60°E         120°E              180°                     120°W      60°W           0°
                                                                                        60°E           120°E       180°         120°W       60°W   0°




                                                                               60°N                                                                     60°N




               Today’s Observations                                            30°N                                                                     30°N




               will be used in years to                                            0°                                                                   0°


                        come
                                                                               30°S                                                                     30°S




                                                                               60°S                                                                     60°S




       ▲Moorings: SubsurfaceTemperature
                                                                                        60°E           120°E       180°         120°W       60°W   0°




       ◊ ARGO floats: Subsurface Temperature and Salinity
                    Training Course 2009 –
       + XBT : Subsurface Temperature NWP-PR: Initialization and Ensemble Generation                              in Seasonal Forecasting
         Impact of data assimilation on the mean
                                         EQATL Depth of the 20 degrees isotherm
                                    ega8 omona.assim_an0
                  -70               edp1 omona.assim_an0



                  -75
Assim of mooring data
CTL=No data       -80



                  -85



                  -90



                  -95


                   1993      1994      1995      1996      1997          1998   1999     2000      2001       2002
                                                                  Time

                                                                                  PIRATA
       Large impact of data in the mean state: Shallower thermocline
              Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
ORA-S3
• Main Objective: Initialization of seasonal forecasts
     Historical reanalysis brought up-to-date (11 days behind real time)
     Source of climate variability

Main Features
  •ERA-40 daily fluxes (1959-2002) and NWP thereafter
  •Retrospective Ocean Reanalysis back to 1959
  •Multivariate on-line Bias Correction (pressure gradient)
  •Assimilation of temperature, salinity, altimeter sea level anomalies an
  global sea level trends.
  •3D OI, Salinity along isotherms
  •Balance constrains (T/S and geostrophy)
  •Sequential, 10 days analysis cycle, IAU
                                                                                       Balmaseda etal 2007/2008
         Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                           The Assimilation corrects the ocean mean state
                           HOPE gcm:: 0001                                                                                                                                         20020101 ( 31 day mean)
                            Mean Assimation Temperature Increment
                           ICODE=178 contoured every 0.0002 XXX
                           Zonal section at 0.00 deg N
                                                               difference from
                                                            0 ( 31 day mean)
                           Plot resolution is 1.4063 in x and 10 in y                                                                                                                                         Interpolated in y

                       0                                                                                                                                                                                                          0
                                                                                       002                                                            -0.0002
                                                                                   -0.0                                                               0.0002
                                                                                                                                                 -0.000
                                                                                                                                                                                                                                        0.0022
                                                                                                                                                        2                                       0.0002

                                                                                                                                                                                            -0.0004
                                                                                                                                                                                           -0.0006
                                                                                                                             -0.000
                                                                                                                                   6                                                                                                    0.0018
                                                                                                                                                                                                      6
                   100                                                                                                                                                                             00                             100




                                                                                                                                                                         -0.0004
                                                                                                                                       6
                                                                                                                                -0.001                                                           .0
                                                                                                                                               014                                             -0                                       0.0014
                                                                                                                                           -0.0 012
                                                                                                                                           -0.0
                                                                                                                                                -0.001
      Depth (metres)




                                                                                                                                                                  6 00
                                                                                                                                               -0.0008
                                                                                                               0.00                                                                                                                     0.001




                                                                                                                                                               -0.0
                                                                                                                   02 -0
                                                                                                                        .00
                                                                                       8
                                                                                                                           02
                                                                                     00

                                                                                                      0.0012
                                                                                   0.0




                   200                                                                                                                                                                                                            200   0.0006




                                                                                                                                                                                                       002
                                                                                                                                                                                                                                        0.0002




                                                                                                                                                                                                  -0.0
                                                                          0.00 4
                                                                          0.00
                                                                              06
                                                                              0




                                                                                                                                                                                                                   002
                                                                                                                                                                                                                                        -0.0004




                                                                                                                                                                                                              -0.0
                   300                                                                                                                                                                                                            300
                                                                         0.0002




                                                                                                                                                                                                                                        -0.0008



                                                                                                                                               -0.
                                                                                                                                                                                                                                        -0.0012
                                                                                                                                                00
                   400                                                                                                                            04 -0.0002                                                                      400   -0.0016
                                                                                                                                                                                                                                                      Equatorial Pacific   (x)
                                                                                             0.0002




                                                                                                                                                                                                                                        -0.002

                                                                                                                                                                                                                                        -0.0024
                   500               O                          O                        O                               O                             O                            O                     O
                                                                                                                                                                                                                                  500
                                  50 E                    100 E                   150 E                             160 W                    110 W                                 60 W           10 W

                                                                                                               Longitude
MAGICS 6.9.1 hyrokkin - neh Tue Jul 25 19:19:38 2006




Data assimilation corrects the slope                                                                                                                                                                                                              z
and mean depth of the equatorial
thermocline
                                                                                                                                                                                                                                                            Free model
                                                                                                                                                                                                                                                            Data Assimilation
                                                                        Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
The Assimilation corrects the ocean mean state



                                                 Analysis minus Observations
                                          Pacific
                           Western situ data
                               eq3-All in
                                   Mean(199301-200201) of Model minus Observations
                                                     S3-a S3-c
                                                                                          Equatorial Indian
                                                                                               eqind-All in situ data
                                                                                                          Mean(199301-200201) of Model minus Observations
                                                                                                                            S3-a S3-c



                                                                                                                                                    DATA ASSIM
                                                                                                                                                    NO DATA ASSIM


                     -200
                                                                                                -200
               Depth (m)




                                                                                          Depth (m)
                     -400
                                                                                                -400




                            -1.5       -1.2         -0.9          -0.6         -0.3   0
                                                       temperature                                     -0.4       -0.2          0          0.2         0.4   0.6
                                                                                                                              temperature




     Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                                                                                         Data Assimilation Improves the
                       Correl: Control                                                                                                          20020101 ( 9 year mean)
                       Sea level contoured every 0.1 m
                       Surface field
                       Plot resolution is 1 in x and 1 in y
                                                                               Interannual variability of the Analysis                                                           Interpolated in y
               O                                                                                                                                                                                                          O
             40 N                                                                                                                                                                                                40 N
                                                                                                                                                                                                                                  0.98
               O
             30 N                  CNTL:No data (1993-2001)                                                                                                                                                      30 N
                                                                                                                                                                                                                          O
                                                                                                                                                                                                                                  0.9




                                                                                                                                                                         5
                                                                                                                                                                      0.
               O                                                                                                                                                                                                          O
                                                                                                                                                                                                                                  0.8
             20 N                                                                                                                                                                                                20 N
                                                                                                                                                                                                                                  0.7
                                                         0.7




                                                                              0. 7

               O                                                                                                                                                                                                          O
             10 N                                                                                            0. 5 0. 6
                                                                                                                                                                                                                 10 N             0.6
  Latitude




                                                                                                                                        0.7
                                                                                                                                  0.8
                                                                                                                                0.9
                                                                                                                                                                                                                                  0.5
                                             0. 7




                   O                                                                                                                                                                                                  O
               0                                                                                                                                                                                                 0
                                                                                              0.8
                                                                                                       0.7
                                                                                                                         0. 9                                                                                                     -1
                                                                                                                                                                           0.6
                                                                                                                                                                         0.5
               O
             10 S                                                  0.8
                                                                                                                                         0.6                                                                     10 S
                                                                                                                                                                                                                          O
                                                                                                                                                                                                                                  -2
               O
                                            0. 50. 6
                                                                                           0. 8
                                                                                          0. 7
                                                                                          0.6
                                                                                         0.5
                                                                                                                                                                                                                          O
                                                                                                                                                                                                                                  -3
             20 S                                                                          0.6                                                                                                                   20 S
                                                                                                                                                                      Correl: Assim        -4                                                                                                                                      20020101 ( 9 year mean)
               O
             30 S                                                                                                                                                     Sea level contoured every 0.1 m
                                                                                                                                                                                      30 S -5
                                                                                                                                                                                                                          O
                                                                                                                                                                                                                                                                                                                                            difference from
                                                                                                                                                                      Surface field                                                                                                                                                       0 ( 9 year mean)
                                                                                                                                                                      Plot resolution is 1 in x and 1 in y                                                                                                                                         Interpolated in y
               O
             40 S                                                                                                                                             O                                                  40 S     O
                                                                                                                                                                                                                                  -6                                                                                                                                           O
                                                                                                                                                            40 N                                                                                                                                                                                                       40 N
                              50 E
                                  O
                                                          100 E
                                                               O
                                                                              150 E
                                                                                     O                O
                                                                                                    160 W                       110 W
                                                                                                                                        O         O
                                                                                                                                               60 W                          10 W
                                                                                                                                                                                 O

                                                                                                                                                                                                                                                                                                                                                                                   0.98
                                                                                            Longitude                                                         O
                                                                                                                                                                             Assimilation of T (1993-2001)                                                                                                                                                                     O
                                                                                                                                                                                                                                                                                                                                                                                   0.9
                                                                                                                                                            30 N                                                                                                                                                                                                       30 N
                                                                                                                                                                                                                                                                             0. 5
MAGICS 6.9 hyrokkin - neh Thu Sep 9 12:12:36 2004
                                                                                                                                                                                                                                                                                                                                                                                   0.8




                                                                                                                                                                                                                                                                                                                                           0. 5
                                                                                                                                                              O                                                                                                                                         0.7                                                                    O
                                                                                                                                                            20 N                                                                         0.6                                                                                                                           20 N
                                                                                                                                                                                                                                                   0. 8   0. 7     0.5                                                                                                             0.7




                                                                                                                                                                                                                  0.7
                                                                                                                                                              O                                                                                                                                   0.6                                                                          O
                                                                                                                                                            10 N                                                                                                                                                                                                       10 N        0.6
                                                                                                                                                 Latitude



                                                                                                                                                                                         0. 6
                                           Correlation of SL                                                                                                                                                                                                                                       0.9




                                                                                                                                                                                                0. 7
                                                                                                                                                                                                                                                                                                                                                                                   0.5



                                                                                                                                                                                                               0. 8
                                                                                                                                                                                                                                                                                                 0.98
                                                                                                                                                                  O                                                                                                                                                                                                        O
                                                                                                                                                              0                                                                                                                                                                                                        0
                                                                                                                                                                                                                                                                                                                                                                                   -1
                                           from System2 with                                                                                                  O
                                                                                                                                                            10 S                                  0. 8
                                                                                                                                                                                                                                                                       0.9



                                                                                                                                                                                                                                                                                                                      0. 8                                             10 S
                                                                                                                                                                                                                                                                                                                                                                               O
                                                                                                                                                                                                                                                                                                                                                                                   -2

                                           altimeter data (which




                                                                                                                                                                                                                                                                                    0.8
                                                                                                                                                                                                                                                                                                                                                                                   -3




                                                                                                                                                                                                                                                                                          0.
                                                                                                                                                                                                  0.6




                                                                                                                                                                                                                                                                                             6
                                                                                                                                                              O                                          0.5                                                                                                                                                                   O
                                                                                                                                                            20 S                                                                                                                                                                                                       20 S
                                                                                                                                                                                                                                                                 0.7
                                                                                                                                                                                                                                                                                                                                                                                   -4
                                           was not assimilated)




                                                                                                                                                                                                                                                                                                                             0.5
                                                                                                                                                              O                                                                                                                                                                                                                O
                                                                                                                                                            30 S                                                                                                                                                                                                       30 S        -5
                                                                                                                                                              O
                                                                                                                                                            40 S                                                                                                                                                                                                       40 S    O
                                                                                                                                                                                                                                                                                                                                                                                   -6
                                                                                                                                                                                     O                                        O                O                                O                               O                   O              O
                                                                                                                                                                                 50 E                                 100 E              150 E                          160 W                                 110 W                60 W           10 W

                                                                                                                                                                                                                                                                 Longitude
                                                                                                                                               MAGICS 6.9 hyrokkin - neh Thu Sep 9 12:12:37 2004

                                                                         Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Data Assimilation improves the interannual variability of the ocean analysis

     No Data Assimilation                                              Assimilation:T+S




                                                                     Assimilation:T+S+Alt
Correlation with OSCAR currents
Monthly means, period: 1993-2005
Seasonal cycle removed




            Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Impact of Data Assimilation
           Forecast Skill
Ocean data assimilation
also improves the forecast
skill
(Alves et al 2003)
                                                                                Data Assimilation




                                                                               Data Assimilation

        Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                              0.4




              Rm
   So far so good, but:       0.2


                                    0
•Progress is not monotonic:
               0      1                                        2         3          4          5             6
                                                               Forecast time (months)
     Need good coupled models to gauge the quality of initial conditions
              a) ERA15/OPS fluxes S2 NO Asim                            S2 Assim
                        NINO3 SST anomaly correlation
                           fluxes DEM NO Assim DEM Assim
              b) ERA40/OPS wrt NCEP adjusted OIv2 1971-2000 climatology
                                    1


                              0.9
              Anomaly correlation




                              0.8


                              0.7


                              0.6


                              0.5


                              0.4
                                        0    1                 2         3          4          5             6
                                                               Forecast time (months)
                                    Skill can be limited by the quality of the coupled model
             Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
          MAGICS 6.11 verhandi - neh Fri Jun 1 16:49:20 2007
Impact on ECMWF-S3 Seasonal Forecast Skill

                                                       NINO4 SST rms errors
                                                          76 start dates from 19870101 to 20051001
                                                          Ensemble sizes are 3 (esj6) and 3 (esj6)

                                    Fc esj6/m1
                                                          Nodata
                                                       S3Fc esj6/m0                S3 Assim
                                                                                    Persistence          Ensemble sd
               0.8
 Rms error (deg C)




               0.6



               0.4



               0.2



                     0
                         0              1             2               3              4               5        6              7
                                                             Forecast time (months)



                                              NINO4 SST anomaly correlation
                                                      wrt NCEP adjusted OIv2 1971-2000 climatology
                             Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                     1
Initialization into context
                       Ocean Initial Conditions
                                         Versus
                                 Coupled Model


                         S2       S2ic_S3model S3




     Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Some general considerations on initialization




    Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
   Initialization Problem: Production of Optimal I.C.

• Optimal Initial Conditions: those that produce the best
  forecast.
    Need of a metric: lead time, variable, region (i.e. subjective choice)
    In complex non linear systems there is no “objective searching algorithm” for
     optimality. The problem is solved by subjective choices.
• Theoretically:
    I.C. should represent accurately the state of the real world.
    I.C. should project into the model attractor, so the model is able to evolve them.
       In case of model error the above 2 statements may seem contradictory
• Practical requirements:
    If forecasts need calibration, the forecast I.C. should be “consistent” with the I.C.
     of the calibrating hindcasts. Need for historical ocean reanalysis
• Current Priorities:
       o Initialization of SST and ocean subsurface.
       o Land/ice/snow potentially important. Not much effort so far …
       o Atmospheric initial conditions play a secondary role.
                 We choose a metric, forecasts of SST from 1-6 months.


            Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
     Perceived Paradigm for initialization of coupled forecasts

    Real world                                                                                        Model attractor




     Medium range                                                                            Decadal or longer
Being close to the real world
                                                   Seasonal?                             Need to initialize the model
is perceived as advantageous.               Somewhere in the middle?                     attractor on the relevant time
Model retains information for                                                            and spatial scales.
these time scales.
                                                                                         Model attractor different from
Model attractor and real world                                                           real world.
are close?
  At first sight, this paradigm would not allow a seamless prediction system.


 •Experiments:
     •Uncoupled SST + Wind Stress + Ocean Observations (ALL)
     •Uncoupled SST + Wind Stress (NO-OCOBS)
     •Coupled           SST (SST-ONLY)                    (Keenlyside et al 2008, Luo et al 2005)


               Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
Impact of “real world” information on skill:

                             NINO3.4 RMS ERROR
                           ALL       NO-OCOBS           SST-ONLY




 Adding information about the real world improves ENSO forecasts

                                                          From Balmaseda and Anderson 2009
       Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
  Impact of “real world” information on skill:
                  Reduction (%) in SST forecast error.
              Forecast Range: 1-3 months. Period 1987-2000
Reduction (%) in SST forecast error Range 1-3 months
                                  OCOBS          ATOBS              OC+AT
                                                                                                 In Central/Western
                                                                                                 Pacific, up to 50% of
        60                                                                                       forecast skill is due
        50                                                                                       to atmos+ocean
        40
                                                                                                 observations.
        30                                                                                       Sinergy: > Additive
  (%)




        20                                                                                       contribution
        10                                                                                       Ocean~20%
         0
                                                                                                 Atmos ~25%
                NINO3


                        NINO4


                                EQ3


                                      EQPAC


                                                     EQIND


                                                             WTIO


                                                                       SETIO


                                                                               EQATL


                                                                                       NSTRATL
        -10
                                                                                                 OC+ATM~55%


                                                             NINO-W

                                              STIO                                                   EQATL
               WTIO
                                                              EQ3




                        Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
     Impact of Initialization
     Eastern Pacific                                   ALL
                                                                                               Western Pacific
                                                       NO-OCOBS
                                                       SST-ONLY

                                                            DRIFT

                                                 •Drift and Variability
                                                 depend on Initialization !!
                                                 •More information corrects
                                                 for model error, and the
                                                 information is retained
                                                 during the fc.
                                                 •Need “more balanced”
                                                 initialization methods to
                                                 prevent initialization shock
                                                 hitting non linearities

                                                        VARIABILITY



•Relation between drift and Amplitude of                                           •Relation between drift and Amplitude of
Interannual variability.                                                           Interannual variability.
     •Possible non linearity: is the warm drift                                          •Upwelling area penetrating too far
     interacting with the amplitude of ENSO?                                             west leads to stronger IV than
                                                                                         desired.

                 Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
           Initialization Shock and non linearities

                        c
                                                                                                                   non-linear
                                                                                                                   interactions
                            Initialization                                                                         important
                            shock
 Model Attractor
 (MA)

                    b
   phase space




                          a




Real World
(RW)




                                         Forecast lead time
                   Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
           Initialization shock and non linearities


                                                                                                                   non-linear
                                                                                                                   interactions
                                                                                                                   important

 Model Attractor
 (MA)
   phase space




                                                                                     Empirical Flux
                                                                                     Corrections


Real World
(RW)




                                        Forecast lead time
                   Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
      Initialization
    Uncoupled: Most common                                                  Other Strategies
•    Advantages:                                               •Full Coupled Initialization:
       It is possible                                                   No clear path for implementation
                                                                          in operational systems
       The systematic error during the
                                                                         Need of a good algorithm to treat
        initialization is small(-er)                                      systematic error. Problem with
       It can be used in a seamless                                      different time scales
        system                                                 •Simplified coupled models?
•    Disadvantages:                                                      Initialization of slow time scales
                                                                          only, limited number of modes.
       Model is different during the
                                                               •Weakly-coupled initialization?
        initialization and forecast
       Possibility of initialization shock                              Atmosphere initialization with a
                                                                          coupled model
       No synergy between ocean and
        atmospheric observations                                         Ocean initialization with a
                                                                          coupled model.
                                                                         Ocean initialization in anomaly
                                                                          mode with a coupled model
                                                                          (DePreSyS)
               Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                        ECMWF:
        Weather and Climate Dynamical Forecasts


   18-Day
                               Monthly                                                   Seasonal
Medium-Range
                               Forecasts                                                 Forecasts
  Forecasts


    Atmospheric model                                                            Atmospheric model



         Wave model                                                                   Wave model


                                                                                       Ocean model
          Ocean model


  Real Time Ocean Analysis                                                         Delayed Ocean Analysis
         ~Real time                                                                       ~12 days


  Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                      Operational Ocean Analysis Schedule

                                                                                                      D1

   BRT ocean analysis: D1-12                                        NRT ocean analysis: D1                  Time (days)

                                            1    2    3    4    5     6    7   8    9      10   11   12




                               Assimilation at D1-12


                                                                    Assimilation at D1-5



•BRT ( Behind real time ocean analysis): ~12 days delay to allow data reception
    For seasonal Forecasts.
    Continuation of the historical ocean reanalysis
•NRT (Near real time ocean analysis):~ For Var-EPS/Monthly forecasts


            Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
ENSEMBLE GENERATION
        •Representing Uncertainty without disrupting
                                       Predictability
                       •Seasonal versus Medium Range

                               •Source of Uncertainty

                                  •Different Strategies



   Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                                               Ensemble Generation
   Medium Range: Singular Vectors
                                                                  Are Singular Vectors a valid approach for
                                                                     Seasonal Forecasts?
          Tangent
          propagator
                                                x t            We need the TL& Adjoint of the full coupled
                                                                     model is required.      BUT…

                         M t ,t0 
  x  t0                                                        1.    The linear assumption would fail for the
                                                                        atmosphere at lead times relevant for
              Initial                                                   seasonal (~>1month).
                                         forecast
              pdf                        pdf                                                                            Besides

 M M  x t0    x t0                                      2.    Uncertainty in the initial conditions may
                                                                        not be the dominant source of error (See
                                                                        later)




                        Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                               Ensemble Generation

Sampling Uncertainty in Initial Conditions:
   •Random sampling of initial uncertainty (as opposed to optimal)
      •ECMWF burst mode ensemble
      •Lag ensemble (NCEP)
   •Simplified problem (Moore et al 2003) (academic, non operational)
      •Full Ocean GCM and a simplified atmosphere
      •Measure growth only on SST
   •Breeding techniques

Sampling Uncertainty in Model Formulation:
   •Stochastic physics (operational)
   •Stochastic optimals (academic)
   •Perturbed parameters (climate projections)
   •Multimodel ensemble (EUROSIP)

        Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
   Ensemble Generation In the ECMWF Seasonal
              Forecasting System
1. Uncertainty in initial conditions:
   Burst ensemble: (as opposed to lag-ensemble)
       40-member ensemble forecast first of each month
   Uncertainty in the ocean surface
       40 SST perturbations
   Uncertainty in the Ocean Subsurface
       5 different ocean analysis generated with wind perturbations
   + SV for atmospheric initial conditions
       Impact during the first month


2. Uncertainty in model formulation:
       Stochastic physics
       Multi-model ensemble (EUROSIP)



            Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
  1.1 Uncertainties in the SST

-Create data base with errors of weekly
SST anomalies,arranged by calendar
week:                                                                          SST Perturbations

Error in SST product: (differences
between OIv2/OI2dvar)
Errors in time resolution: weekly versus
daily SST
-Random draw of weekly perturbations,
applied at the beginning of the coupled
forecast. Over the mixed layer (~60m)
-A centred ensemble of 40 members



         Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
     1.2 Uncertainties in the ocean Subsurface
                                                                                                                        1-3 months
                                                                                                                        decorrelation
-Create data base with errors in the                                                                                    time in wind
monthly anomalous wind stress,                                                Wind perturbations +p1/-p1
arranged by calendar month:
                                                                                EQ3 Depth of the 20 degrees isotherm
(differences between ERA40-CORE)                             20

                                                             10

-Random draw of monthly                                       0

perturbations, applied during the ocean                     -10

analyses.                                                   -20
                                                             1986      1988      1990       1992          1994   1996   1998   2000
                                                                                                   Time

-A centered ensemble of 5 analysis is                                      on Depth of the Subsurface
                                                                    Effect EQPACOcean 20 degrees isotherm (D20)
                                                             10

constructed with:                                             5


       -p1                                                    0


       -p2                                                   -5

                                                            -10
       0                                                     1986      1988      1990       1992
                                                                                                   Time
                                                                                                          1994   1996   1998   2000


       +p1                                                                                            ~6-12 months
       +p2                                                                                            decorrelation time in
                                                                                                      the thermocline
              Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
  1.2 Uncertainty in the ocean subsurface: wind
  perturbations
    ERA40-CORE (1958-1979)                                             ERA40-CORE (1980-2000)




Ocean Subsurface: No Data assimilation                                Ocean Subsurface: Data assimilation




                                                    C.I=0.2 C

           Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
 1.3 Uncertainty in the Atmospheric initial conditions




• The atmosphere model is also perturbed using singular
  vectors (SV):
    Same as for the medium range and monthly forecasting system
    The SV affect the spread of the seasonal forecasts:
      Mainly during the first month
      Mainly in mid-latitudes


    It makes the medium-range, monthly and seasonal
     forecasting systems more integrated


       Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
2.1) Uncertainties in deterministic atmospheric
physics?
  ECMWF stochastic physics scheme:

    i) X  D  P   P                                                   Stochastic
                                                                         forcing

  is a stochastic variable, constant over time intervals
of 6hrs and over 10x10 lat/long boxes
Buizza, Miller and Palmer, 1999;
Palmer 2001

  The Stochastic Physics is a probabilistic parameterization,
  but it does not intend to sample uncertainty in the
  parameters, nor model error

            Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
                                         Ensemble Spread

Wind Perturbations (WP) Wind Perturbations No DA (WPND)                    •The spread by different methods
SST Perturbations(ST)      All(SWT)                                        converge to the same asymptotic
Stochastic Physics (SP)     Lag-averaged(LA)                               value after after 5-6 months.
                                                                           •SST and Lag-averaged
                                                                           perturbations dominate spread at
                          S         S
                          T         P
                                                                           ~1month lead time.
                                                                           •With DA, the wind perturbations
                                                                           grow slowly, and notably influence
                                                                           the SST only after 3m.
                                                                           •Without DA, the initial spread
                                                                           (<3m) is larger. The asymptotic
                                                                           value is slightly larger
                                                                           Is the level of spread sufficient?
 From Vialard et al, MWR 2005


                Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
     Is the ensemble spread sufficient? Are the forecast reliable?

                                                                                    Forecast System is not
                                                                                    reliable:
                                                                                                RMS > Spread


                                                                                         A.     Can we reduce the
                                                                                                error? How much?
                                                                                                (Predictability limit)

                                                                                          B.     Can we increase the
                                                                                                 spread by improving
                                                                                                 the ensemble
                                                                                                 generation?


To improve the ensemble generation we need to sample other sources of error:
    a) Model error: multi-model, physical parameterizations
    b) To design optimal methods: Stochastic Optima, Breeding Vectors, …
               Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
2.1) Sampling model error: The Real Time Multimodel

                                             EUROSIP
                     ECMWF-UKMO-MeteoFrance
                                 RMS error of Nino3 SST anomalies

                                                                                        Persistence




                                                                                             ECMWF
                                                                                           EUROSIP




                                                                                   ensemble spread




     Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting
2.2) Sampling model error: The Real Time Multimodel

                                             EUROSIP
                     ECMWF-UKMO-MeteoFrance
                                 RMS error of Nino3 SST anomalies

                                                                                        Persistence




                                                                                             ECMWF
                                                                                           EUROSIP


                                                                                 Bayesian Calibration


                                                                                   ensemble spread




     Training Course 2009 – NWP-PR: Initialization and Ensemble Generation in Seasonal Forecasting

								
To top