Application of K7 Ocean DataAssimilation

1
Application of
K7 Ocean Data Assimilation
Toshiyuki Awaji Japan Agency for Marine-Earth
Science and Technology (JAMSTEC) Department of
Geophysics, Graduate School of Science, Kyoto
University

• Members
• N. Sugiura, S. Masuda, H. Igarashi, T. Toyoda
  (JAMSTEC)
• Yoichi Ishikawa (Kyoto Univ.),Masa
  Kamachi(JMA-MRI)
• Outline
• Japanese DIAS project
• Long-term ocean reanalysis by K7
• target ocean state estimation from
  1980
• 3)Summary (Outreach Future Perspective)

finished in March, 2007. So, this is the legacy
K7 is a DA project supported by MEXT exactly,
category 7 of Research Revolution 2002 called
Kyousei Program
2
The Aims of Global Earth Observation System of
Systems(GEOSS)
Construct comprehensive, coordinated, and
sustained Earth Observation System of Systems to
maintain existing observation systems, to
develop new observation systems which fill gaps,
and to integrate in-situ and satellite
observations, based on the international
cooperation and Driven by user requirements
shown in this schematic view. contribute to
Capacity Building.
Indeed, Optimal Synthesis of Models and
Observational data by Assimilation methods is a
suitable tool for GEOSS actions
User Requirement Assessment
User (Government/Research/Public/Commerce)
Observations
Satellite Systems
Data processing/exchange/dissemination
These aims functions are common to those of
CLIVAR/GODAE in ocean/climate fields.
So, GEOSS could be the challenge of
Operational Applications.
In Japan, DIAS is one of the new national
projects toward the success of GEOSS
Exchange/ dissemination systems
Data processing systems
Transform data into useful information
In-situ Systems
Archive/Data Base
Products Information
Data
In put
Network
We use advanced data assimilation approach for
DIAS (Japanese GEOSS) development
Data
User-friendly Tools
Model and Assimilation
Output
WEB
WEB
Earth Simulator JAMSTEC, Japan
6
3
Kyoto Univ. - JMSF-K7 DA system for coastal sea
forecast management
Target area a confluent region (Tsugaru
Strait and neighboring ) influenced by Oyashio,
Kuroshio and their extension, and Tsushima
currents from Japan Sea.
very complicated region
Good benchmark for assessing the current status
of assimilation skill
Tsushima cu
Use Triple nesting and 4D-VAR DA to secure
better initial and open-boundary settings.
4
horizontal resolution 1/6x1/8, 78 vertical
levels
North Pacific model 1/6x1/8 deg
Original resolution
5
Snapshot of Temperature Field
High-resolution time series for the northwestern
North Pacific will be provided by the
joint team
SST
Simulation Observation
Assimilation
Temperature at 100m
6
Move on JAMSTEC/K7
Basin-scale to global DA
Modeling
Observation
JAMSTEC/K-7
Optimal synthesis
4DVAR
Pro truth Con sparse, heterogeneous
Pro gridded, consistent Con biases
7
Control Variables

• Oceanic initial condition
• Bulk parameters controlling Air-sea fluxes of
• h
• For Smaller-scale Parameterization
• we made pre-optimization using the Green
  function approach by Menemenlis et al. (2005)

Momentum Sensible heat Latent heat
8
K7 System of Long-term Ocean Reanalysis
Platform of ODA system
Extension up to 1980 is now underway.
Target period Years 1980 2004already made
datasets from 1986
from GODAE/CLIVAR
Reanalysis dataset from 1986 at
http//www.jamstec.go.jp/frcgc/k7-base2/ Using
this reanalysis dataset
State estimation of global heat and water mass
transports in each isopycnal layer in the 1990s
using this reanalysis dataset
9
revealed
AAIW
intermediate
volume transport
subsurface
1990s
For example, robust estimate of northward volume
transport in each isopycnal layer across
24N Annual average is close to the estimate of
Talley (1999)
Subsurface layer
dominated by interannual
variation Intermediate layer
dominated by decadal variation
Diagnosis of these variabilities in the
formation and spread of important water masses
provides new insight into their physical processes
10
1990
1991
1992
Eastern subtropical mode water in the subsurface
has good memory of subtropical climate variation
identified
1993
1994
1995
Formation is associated with MLD in March
Time-series of MLD since 1990
1996
1997
1998
time
2001
1991
1999
2000
(160-120W,15-35N)
0m
100m
200m
11
Pacific water mass distribution at 47oN (Jul2002)
Presence of distinct water mass structures in the
upper layer of NP subarctic region Dichothermal
and Mesothermal
two subsurface water masses are well defined
Simulation
Assimilation
q 47ºN P1 Revisit(1999)
DS
Prof. Suga
MS
Observation
12
Reveal Origin of water masses
Subarctic region
Coming from 1) Kuroshio Extension and 2) Alaskan
Gyre regions Further, Interannual variability
close relation to ENSO by a remote effect of
atmosphere-ocean coupling (Masuda et al., 2006
GRL)
13
Dynamical analysis using a long-term oceanic
reanalysis datasetclarifies the interannual
variability of temperature inversions that
characterize water mass structures in the
subarctic North Pacific (one of the most
important regions for greenhouse gas absorption).
Dichothermal W. Mesothermal W.
Tmax depth(m)
Masuda et al.(2006), Geophys. Res. Lett.
A good correlation between the time series of
the Multivariate ENSO Index and the depth of
local maximum Tmax is visible, suggesting that
the interannual variability of Tmax water in the
subarctic region is closely related to the
equatorial climate variability via so-called
Atmospheric bridge. ? suggests that ENSO
activity could influence global warming process
in the subarctic region!
Time(year)
(a) Monthly mean time series of the depth at
which Tmax is found in the mid-subarctic North
Pacific (b) Multivariate ENSO Index time
series 1987-2004
14
Circulation of North Pacific Intermediate water
(NPIW) Efficient absorber of CO2

For intermediate layer
Decadal variation of NPIW formation and
circulation associated with the change in
subduction rate in the Okhotsk Sea
?
15
Ensure detailed analyses of El Nino and Indian
Dipole Mode evolution by dynamically
self-consistent reanalysis data
For equatorial region
time
SST snapshot in January, 1997 (upper
left) simulation result?(upper right)
observation (lower) assimilation result
west
East
Time series of heat storage down to 400 m depth
well reflects eastward travelling Kelvin waves
Nino3 SST RSMD 1.87K (forward - observation)
0.78K (assimilation
- observation)
16
(No Transcript)
17
Estimated Meridional Overturnings from K7
ODAstreamfunctions
Indian basin Pacific basin
Atlantic basin
Basically, well consistent with previous knowledge
18
Meridional Overturning in Indian Ocean Jan Jul
Lee and Marotzke (1998)
K7 products
Good correspondence
19
Decomposition of Meridional Transport Stream
Functionin Jan 1987 (South Indian Basin) gives
us more physical insight
Total
External mode
Ageostrophic component
Vertical shear
20
Contribution to CLIVAR/GSOP (2)
Coupled Data Assimilation improved state
estimation and prediction of ENSO
The long-term ocean reanalysis dataset reproduces
the climate variations, e.g., 1986/1987 and
1997/1998 El Nino including subsurface processes.
The RMSE of the estimated field is reduced by 1/2
comparing with that of simulated one.
NINO3.4 anomaly correlation scores
betweenobservation and forecast as function of
lead time
SKIP
Sea surface temperature in Jan1997. (top left)
simulation, (top right) observation, (bottom)
assimilation
time
Initialized by coupled data assimilation
Initialized by Oceanic data assimilation
imply capability of 1.5-year lead prediction of
ENSO
Persistent
The ensemble forecast from the optimized initial
condition demonstrates much better prediction
almost up to 1.5-year forecast than earlier
results.
western equatorial
eastern equatorial
Time evolution of heat content anomaly during the
1997/1998 El Nino period with realistic eastward
propagation of Kelvin wave
21

The dataset enables us to clarify the role of
WWBs on irregular occurrence of ENSO leading to
the extended recharge paradigm (submitted to GRL)
Figure 1. (a) Time-longitude distribution of the
sum of anomalous sea surface temperature (SST)
and the absolute mean SST averaged within
2oN-2oS. (b) Time change of SST averaged across
the NINO3 region. The black, red and green curves
denote Reynolds SST, simulation and assimilation,
respectively.
22
The phase diagram suggests the capability of
clarifying important roles of WWB on irregular
occurrence of ENSO
(a)Time-latitude distribution of the zonal-mean
meridional transport in the eastern equatorial
region. Units are in Sv. Positive values denote
northward transport. (b) Phase diagram of the
NINO3 SST anomalies (abscissa) and the 20oC
isotherm anomalies in the eastern equatorial
Pacific (ordinate) for the period 1991-1999. The
units are in degree Kelvin and 10m, respectively.
A 6-month running mean has been applied to both
variables. Line colors are changed for each year.
23
Important role of short term WWBs
Figure 3. (a) Time series of the 20oC isotherm
depth (solid circle) and the diagnosed 20oC
isotherm depth (open circle see text) in 5oS-3oN
and 160oE-80oW from 1991 to 1999. The units are
in m. (b) Time series of the difference of the
local gradient of two curves in (a) (e.g., local
vertical adjustment speed), and those of the wind
stress anomaly averaged in 2.5oS-2.5oN and
130oE-150oE. The right-hand axis denotes the
value of the wind stress anomaly, for which the
units are in 0.1N/m2.
More enhanced vertical adjustment speed than
expected by the Sverdrup transport convergence
often occurs in the 1990s, corresponding to the
timing of WWBs
24
Extended model is likely to explain an Irregular
behavior of ENSO
Figure 4. (a) Time series of TE (red curve) and
hE (blue curve) spanning years 10 to 25 in the
conceptual model. The solid curve denotes the
result for the case with hshort and the dashed
curve without hshort. The units are
non-dimensional. (b) Phase diagram of TE
(abscissa) and hE (ordinate) from the 14th to the
25th year. The solid and dashed curve denote the
same as in (a). The units are in degree Kelvin
and m, respectively. A 6-month running mean has
been applied to both variables. Line colors are
changed for each year.
25
Concluding remarks

• Dynamically consistent data set capable of
  offering greater information content and forecast
  potential for S-I (e.g. ENSO variability,
  subsurface water masses) is obtained. by our ODA
  and CDA experiments
• Coupled data assimilation by simply controlling
  bulk coefficients brings some improvements to
  coupled model climatology.
• We have just finished the construction of 1990s
  coupled model assimilated data and a forecast
  system ( will be open in this year).
• Assimilation datasets and observational datasets
  are available at http//www.jamstec.go.jp/frcgc/k7
  -base2/

GOOS/OOPC in Tokyo 2006/05/01
26
Enhanced prediction of monsoon onset contributes
to reduction of natural hazard
Improved El Nino prediction
CMAP
???
assimilation
NINO3SST
simulation
?????SST
1996?
1997?
1998?
Improved DM in Indian Ocean
?????SST
Drought in Indo subcontinent, 2005
Flood in Bangladesh, 1998
GOOS/OOPC in Tokyo 2006/05/01