Hiroyasu HASUMI

1
6-8 September 2007 in Delmenhorst GEOTRACES
Data-Model Synergy Workshop
An overview of global scale ocean modeling from
low to high resolution
Hiroyasu HASUMI (Center for Climate System
Research, The University of Tokyo)
2
Aim of this talk

• Introduce a hierarchy of global-scale ocean
  modeling studies, by use of OGCMs, conducted at
  CCSR (with collaborating partners)
• Give estimates of required computational resource
  for each of the above
• Serve as a basis for discuss about future
  direction of GEOTRACES-related modeling
• Introduce our recent attempt on
  GEOTRACES-oriented modeling

3
Model

• CCSR Ocean Component Model (COCO)
• Free-surface, z-coordinate OGCM developed at
  Center for Climate System Research in
  collaboration with FRCGC/JAMSTEC
• Ocean component of the coupled climate model
  MIROC (contributor to IPCC AR4)

4
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies

5
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies

6
Low resolution modeling
Model grid and bathymetry
7
Low resolution modeling
Model performance Atlantic (30W) T/S
Model T
WOA T
WOA S
Model S
8
Low resolution modeling
Model performance Pacific (180E) T/S
Model T
WOA T
WOA S
Model S
9
Low resolution modeling
Modeled THC/MOC and its sensitivity Atlantic
0.1x10-4 m2/s
0.5x10-4 m2/s
0.13.0x10-4 m2/s
1.0x10-4 m2/s
10
Low resolution modeling
Modeled THC/MOC and its sensitivity Pacific
0.1x10-4 m2/s
0.5x10-4 m2/s
0.13.0x10-4 m2/s
1.0x10-4 m2/s
11
Low resolution modeling

• Merits and demerits
• Global scale distribution of water masses and
  associated (deep) circulation are reasonably
  captured
• Representation of intermediate waters and
  associated circulation is relatively poor
• Representation of transient response and/or
  variability might also be poor

12
Low resolution modeling

• Required resource
• O(1) days real-time for O(103) years model
  integration with a small number of CPUs of usual
  super computers (e.g., with 1 node 8 CPUs of
  NEC SX6)
• O(1) MB output data for a single time shot of a
  single variable
• (O(10) MB for climatological monthly data of a
  single variables)

13
Low resolution modeling

• Merits and demerits
• Global scale distribution of water masses and
  associated (deep) circulation are reasonably
  captured
• Representation of intermediate waters and
  associated circulation is relatively poor
• Representation of transient response and/or
  variability might also be poor
• Easy to run and easy to examine
• ...good for studying sensitivity of (steady/
• climatological state of)THC and associated
  biogeochemical cycles to model parameters over
  their wide ranges

14
Low resolution modeling
NO3 distribution sensitivity to ballast
parameters (Oka et al., 2007)
Atlantic
WOA
Modeled
Pacific
WOA
Modeled
15
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies

16
IPCC-class resolution modeling
Model grid and bathymetry
17
Low resolution modeling
Model grid and bathymetry
18
IPCC-class resolution modeling
Deep water formation process
Modeled wintertime MLD and surface velocity
19
IPCC-class resolution modeling
THC Variability (coupled modeling)
Wintertime mixed layer depth
long-term mean
time series for 600 years
Deep convection see-saw
Oka et al. (2006)
20
IPCC-class resolution modeling

• Merits and demerits
• Formation processes and distribution of deep
  water masses (and associated circulation) are
  significantly improved
• Intermediate waters and circulation are little
  improved
• Western boundary currents (and other narrow,
  swift currents) are insufficient
• Interannual to decadal climate variability (ENSO,
  PDO, NAO, ...) is reasonably captured when
  coupled to atmospheric models

21
IPCC-class resolution modeling

• Required resource
• O(1) days real-time for O(10) years model
  integration with a small number of CPUs of usual
  super computers (e.g., again with 1 node 8 CPUs
  of NEC SX6)
• O(10) MB output data for a single time shot of a
  single variable
• (O(10) GB for 100-year-long monthly data of a
  single variables)

22
IPCC-class resolution modeling

• Merits and demerits
• Formation processes and distribution of deep
  water masses (and associated circulation) are
  significantly improved
• Intermediate waters and circulation are little
  improved
• Western boundary currents (and other narrow,
  swift currents) are insufficient
• Interannual to decadal climate variability (ENSO,
  PDO, NAO, ...) is reasonably captured when
  coupled to atmospheric models
• Not very difficult to run the model for O(1000)
  years and to handle its output

23
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies

24
Eddy-permitting resolution modeling
CCSR/NIES/FRCGC MIROC 3.2 (a coupled GCM
contributing to IPCC AR4)
25
Eddy-permitting resolution modeling
Coupled modeling (IPCC AR4 climate projection)
26
Eddy-permitting resolution modeling
Ocean hindcast (CORE forcing 1958- without data
assimilation)
27
Eddy-permitting resolution modeling

• Merits and demerits
• Representation of western boundary currents and
  other narrow, swift currents becomes realistic

28
Eddy-permitting resolution modeling
Western boundary current (Kuroshio)
Eddy-permitting
IPCC-class
realistic path and separation point
weak, broad, and overshooting
29
Eddy-permitting resolution modeling
Change of the Kuroshio under global warming
Surface current of the control climate high-res.
Difference (2071-2100) (1971-2000)
Accelerated under global warming 30 cm/s
Sakamoto et al. (2005)
30
Comparison of horizontal resolutionbetween
previous and present models
?
?
?
?
(1 1 )
(1/4 1/6 )
IPCC AR3(Hashioka Yamanaka)
IPCC AR4(Hashioka developing NW Pacific)
Kuroshio Separation latitude
31
3-D regional NEMURO in the Western North
Pacificforced by CCSR/NIES/FRCGC climate model
with 1/4x1/6 degrees Just obtained two month ago
after two years efforts
Using off-line method (giving dairy interval data
of u, v, T, S and insolation instead of no their
calculation) We have a plan model
intercomparison among NPZD, NEMURO, eNEMURO, and
PlankTOM5, using the same physical fields.
Courtesy of Dr. Yamanaka
32
Eddy-permitting resolution modeling

• Merits and demerits
• Representation of western boundary currents and
  other narrow, swift currents becomes realistic
• Mesoscale variability is reasonably captured

33
Eddy-permitting resolution modeling
Mesoscale variability (SSH std. dev.)
Eddy-permitting
IPCC-class
34
Eddy-permitting resolution modeling

• Merits and demerits
• Representation of western boundary currents and
  other narrow, swift currents becomes realistic
• Mesoscale variability is reasonably captured
• Intermediate waters and associated circulation
  are somewhat improved, but not very much

35
Eddy-permitting resolution modeling

• Required resource
• 1 day real-time for 5 years model integration
  with 100 nodes ( 800 CPUs) of Earth Simulator
  (whose single node/CPU is slightly faster than
  NEC SX6)
• ...gt x10 time on usual supercomputers
• O(100) MB output data for a single time shot of a
  single variable
• (O(100) GB for 10-year-long 5-day-interval data
  of a single variable)

36
Eddy-permitting resolution modeling

• Merits and demerits
• Representation of western boundary currents and
  other narrow, swift currents becomes realistic
• Mesoscale variability is reasonably captured
• Intermediate waters and associated circulation
  are somewhat improved
• However, these features are still to be improved
• Integration for O(100) years is feasible, but
  only for limited cases and on limited
  computational platforms, and data storage/
• handling/transfer accompanies difficulty

37
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving lt 10 km
• for process studies

38
Eddy-resolving resolution modeling
Southern Ocean modeling
Surface velocity in 8 km grid modeling
39
Eddy-resolving resolution modeling
Southern Ocean modeling
Subduction velocity (TRM-w)
AAIW (30W salinity)
SSH variability
IPCC-class
Eddy- permitting
Eddy- resolving
40
Eddy-resolving resolution modeling

• Merits and demerits
• Intermediate waters and circulation are
  reasonably captured
• (subduction is governed by sub-mesoscale
  processes)
• Short-term and/or small-scale variability is
  sufficiently simulated
• Computational cost is very high
• ...available only for limited process studies

41
Eddy-resolving resolution modeling
Required resource ...No general remark
should/can be given here
42
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies

43
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies
• ...Case study on global (deep) TEI distribution

44
REE modeling
Distribution of elements in the ocean
Can we obtain its 3-D, global version via
modeling?
45
REE modeling
Governing process for Nd (REE) distribution
Internal recycling
Source and sink
primary factor for basic vertical profile
46
REE modeling
Model Result
smaller K (weaker scavenging)
K5.0e5
K4.0e5
K3.0e5
K2.0e5
K1.0e6
K6.0e5
Observed REEs
heavier REE
lighter REE
520yr
570yr
2440yr
2890yr
2430yr
2420yr
1820yr
740yr
longer residence time
47
REE modeling
Governing process for Nd (REE) distribution
Modeled only by internal recycling (no
source/ sink)
Observed concentration Southern lt Atlantic lt
Pacific
Modeled with source distribution
48
Modeling study hierarchy

• Several standard setups of COCO global modeling
• Low resolution 300 km
• for global THC and related biogeochem. cycles
• IPCC-class resolution 100 km
• for decadal to centennial climate variability
  (including coupled modeling)
• Eddy-permitting 20 km
• for mid-term (30-100 yr) ocean hindcast and
  climate projection
• Eddy-resolving 10 km
• for process studies
• ...Case study on global (deep) TEI distribution