Ocean Initialization for seasonal forecasts

1
Ocean Initialization for seasonal forecasts
ECMWF CAWCR Met Office JMASTEC
NCEP MERCATOR-Ocean MRI JPL GMAO NOAA/GFDL
University of Hamburg
Magdalena A. Balmaseda Oscar Alves Alberto
Arribas T. Awaji David Behringer Nicolas Ferry
Yosuke Fujii Tony Tee Michele Rienecker Tony
Rosati Detlef Stammer
2
Outline

• Background
• The basis of seasonal forecasts
• Standard practice
• Different operational efforts around the world
• Ocean Model initialization
• Impact of assimilation on ocean estate
• Impact on seasonal forecast skill. Overview
• Towards coupled initialization ongoing efforts
• This talk only deals with prediction of SST.
  But seasonal forecasts products go beyond SST
• Temperature, Precipitation
• Tropical cyclones and hurricanes
• Applications such as hydropower, agriculture and
  health

3
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
• Higher heat capacity (Thermodynamic argument)
• Predictable dynamics
• Oceanic point of view Initial value problem
• Prediction of tropical SST need to initialize
  the ocean subsurface.
• Examples
• A well established case is ENSO
• A more tantalizing case is the importance
  subsurface temperature in the North Subtropical
  Atlantic for seasonal forecasts of NAO and
  European Winters.
• Indian Ocean Dipole

4
Need to Initialize the subsurface of the ocean
5
Typical Seasonal Forecasting System dealing with
model error forecast uncertanty
6
Common features of the SF initialization systems

• Emphasis on upper ocean thermal structure and SST
• Climate configuration Global domain, resolution
  1 deg with equatorial refinement. 30-50 vertical
  levels.
• Usually rely on previously analyzed SST field.
• Balance relationships (T and S, density and
  velocity)
• Assimilation cycle 5-to-10 days
• Some control of the mean state
• Relaxation to climatology
• Online bias correction (T, S, prssure gradient)
• MDT either prescribed (from free model, or TS
  analysis) or estimated (corrected) online
• Reanalysis period (15-20-50 years).
• Usually 2 products
• Delayed 7-30 days
• NRT (0-7 days)
• Some have an ensemble of analyses (3-5)

7
Operational efforts routine production of
seasonal forecasts and ocean analysis

• MRI-JMA MOVE/MRI-COM.G
• http//ds.data.jma.go.jp/tcc/tcc/products/elnino/i
  ndex.html
• ECMWF ORA-S3
• http//www.ecmwf.int/ products/forecasts/d/charts
  /ocean
• http//www.ecmwf.int/products/forecasts/d/charts/
  seasonal/
• CAWCR(Melbourne) POAMA-PEODAS
• http//www.bom.gov.au/climate/coupled_model/poama.
  shtml
• NCEP (GODAS)
• http//www.cpc.ncep.noaa.gov/products/GODAS/
• Mercator/Meteo-France
• http//bulletin.mercator-ocean.fr/html/welcome_en
  .jsp
• MetOffice GLOSEA3
• http//www.metoffice.gov.uk/research/seasonal/
• GMAO ODAS-1
• http//gmao.gsfc.nasa.gov/research/oceanassim/ODA_
  vis.php
• http//gmao.gsfc.nasa.gov/cgi-bin/products/climate
  forecasts/index.cgi

8
Reducing Uncertainty

• A simple way of producing ocean initial
  conditions is to force and ocean model with
  atmospheric fluxes
• But large uncertainty in wind products lead to
  large uncertainty in the ocean subsurface
• The possibility is to use additional information
  from ocean data (temperature, others)
• Questions
• Does assimilation of ocean data constrain the
  ocean state?
• Does the assimilation of ocean data improve the
  ocean estimate?
• Does the assimilation of ocean data improve the
  seasonal forecasts

Equatorial Atlantic Taux anomalies
9
Ocean observations assimilated
The ocean observing system has slowly been
building up Its non-stationary nature is a
challenge for the estimation of interannual
variability
10
Example of potential problem
From an Old DA system
Assim of mooring data CTLNo data
Large impact of data in the mean state Shallower
thermocline
11
Impact of assimilation on the ocean state
POAMA Only T, univariate (1st generation) PEODAS
TS, multivariate (2nd generation) ORA-S3
TS, (2rd generation) CONTROL
no data assimilation Improvements was slow to
achieve. But progress is evident Alves et al 2008
Alves et al
12
Importance of Salinity
TS both temperature and salinity
corrections NOS No Salinity corrections, only
temperature
Results from MRI Fujii et al 2008
13
barrier layer and warm water content
The WWC, function of the barrier layer thickness,
plays an important role on ENSO
Fujii et al 2008
14
Impact on Seasonal Forecasts Skill

• Until very recently seasonal forecasts skill was
  considered a blunt tool to measure quality of
  ocean analysis coupled models were not
  discerning enough.
• Examples of good impact were encouraging, but
  considered a strike of good luck.
• Improvements in the coupled ocean atmosphere
  models also translate on the ability of using SF
  as evaluation of ocean initial conditions. In
  this presentation there are several examples
  showing the positive impact of data assimilation
  on the skill of seasonal forecast.
• There are even results with observing system
  experiments, where the seasonal forecasts show
  significantly different behaviour
• Need good coupled models to gauge the quality of
  initial conditions
• The initialization problem is different from the
  state estimation problem .

15
Progress is not monotonic
The quality of the initial conditions is not
always the limiting factor on the skill
ERA15/OPS S2 NOdata S2 Assim
ERA40/OPS DEM NOdata
DEM Assim
16
Impact of Initialization strategy on SFECMWF S3

• Drift and variability depend on initialization!!
• More information corrects for model error, and
  the information is retained during the fc.
• Need better (more balanced) initialization

• Relation between drift and Amplitude of
  Interannual variability.
• Upwelling area penetrating too far west leads to
  stronger IV than desired.

• Relation between drift and Amplitude of
  Interannual variability.
• Possible non linearity is the warm drift
  interacting with the amplitude of ENSO?

17
Impact on Initialiazation on SF SkillECMWF S3
Adding information about the real world improves
ENSO forecasts
18
Impact on forecast skill (ECMWF-S3)
In Central/Western Pacific, up to 50 of forecast
skill is due to atmosocean observations. Sinergy
gt Additive contribution Ocean20 Atmos
25 OCATM55
Balmaseda et al 2008
19
Impact of Different Ocean ObservationsJMA-MRI
OSEs in JM-MRI confirm the complementary nature
of the observing systems (moorings and floats) on
the skill of SF.
Fujii etal 2008
20
Impact of initialization SF skill CAWCR POAMA
In the CAWCR system, an improved data
assimilation system improves the seasonal
forecast skill.
21
Improvements in SF Mercator-MeteoFrance S3
The new Meteo-France SF system 3 shows improved
skill. A large contribution to the improvement is
likely due to better ocean initial conditions
22
The ECCO-JPL / UCLA example
RMS ERROR on SF of SST

• Improvement on SF by using ECCO-JPL. Baseline is
  a forecast from ocean initial conditions without
  data assimilation.
• From Cazes-Boezio et al 2008.

23
Initialization and non linearities
Initialization shock
24
More balance intialization

• Coupled Data Assimilation
• Assimilation of ocean data with a coupled model
• Coupled 4D-var JAMSTEC
• EnKF GMAO, GFDL
• Coupled Breeding Vectors
• generation of the ensemble by projecting the
  uncertainty of the initial conditions on the
  fastest error-growth modes of the coupled system
• Anomaly Initialization
• Depresys (Met Office)
• GECCO

25
Towards more balanced Initialization (I)Coupled
4D-var JAMSTEC
OBS First guess Analysis Control initial
conditions (IC) Control Parameters
(PRM) Control ICPRM
Sugiura et al 2008
26
Towards more balanced coupled initialization
(II) Breeding Vectors in GMAO
ACC of 9-month lead FC of SST BV can also be used
to formulate flow dependent covariances in the
ocean data assimilation
Yang et al 2008
27
Summary

• Ocean data assimilation is used operationally in
  several centres around the world to initialize
  seasonal forecasts with coupled models
• Improving the seasonal forecasts by assimilating
  ocean data has been a slow process. Limiting
  factors have been (are)
• Balance constraints between variables
• Spurious inter-annual variability due to
  non-stationary nature of observing system
• Quality of coupled models
• With the current generation of ocean data
  assimilation systems and coupled models it is
  possible to demonstrate the benefits of
  assimilating ocean data for the seasonal forecast
  skill
• The initialization shock remains a problem. There
  are currently several initiatives aiming at a
  more coupled initialization.
• Another challenge is the initialization of a
  seamless prediction system from days-months to
  decades.