How have WOCE Observations Challenged Ocean Models

1
How have WOCE Observations Challenged Ocean
Models?

• Julie McClean
• Naval Postgraduate School

18-22 November 2002 San Antonio, Texas
2
Overview

• Introduction The state of ocean modeling for
  climate studies at the start and end of WOCE.
• The realism of ocean models as revealed by the
  WOCE data.
• Model sensitivity studies guided by WOCE data.
• Improved model parameterizations motivated by the
  WOCE data.

3
Global Change Research Priorities

• To more reliably predict changes in climate and
  the global environment over decades to centuries
  that will result from projected changes in
  population, energy use, and other factors.
• Develop model components and coupled models that
  can be used for energy policy, IPCC, and the
  National Assessments

4
Basin to Global-Scale Ocean Modeling at the Start
of the WOCE era

• Vector platforms Cray Y-MP (1Gflop/s)
• Bryan-Cox type
• Eddy-permitting resolutions (35-40 km and 20-30
  vertical levels)just sufficient to allow
  hydrodynamic instabilities responsible for eddy
  formation
• Complete representation of thermodynamic
  processes responsible for water mass formation
• Gent and McWilliams subgridscale eddy mixing
  parameterization
• Layered Models
• MICOM (Isopycnic) not eddy-permitting, slab
  mixed layer.
• NLOM hydrodynamic, eddy-resolving
• Sigma-Coordinate
• SPEM
• POM
• Climatological wind forcing
• Surface and lateral boundary temperature and
  salinity restored to Levitus climatology

5
Basin to Global-Scale Ocean Modeling at the End
of the WOCE era I

• Earth Simulator (25 TF/s) vector shared memory
• Distributed memory IBM SP 4 at NCAR (0.4 TF/s)
• Bryan-Cox type
• Eddy-resolving resolutions (10-20 km and 40-50
  vertical levels)allow hydrodynamic instabilities
  responsible for eddy formation
• Complete representation of thermodynamic
  processes responsible for water mass formation
• Ongoing improvements of planetary and bottom
  boundary layer subgridscale parameterizations
  e.g. KPP mixed layer, Beckmann and Doscher
• Improvements to eddy mixing parameterizations
  GM and anisotropic viscosity for low-resolution.
• Hybrid co-ordinate
• HYCOM (Isopycnal, z-level, sigma)
  eddy-resolving, slab mixed layer.
• Layered Models
• MICOM eddy-permitting, slab mixed layer
• NLOM hydrodynamic/thermodynamic, eddy-resolving
  1/64?

6
Basin to Global-Scale Ocean Modeling at the End
of the WOCE era II

• Realistic prescribed surface fluxes computed
  using bulk formulae and combine atmospheric
  quantities and model SST. The former obtained
  from atmospheric prediction models and subsequent
  reanalyses (ECMWF, NCEP).
• Some surface restoring of T or S to avoid drift
• Some lateral boundary T and S restoring to
  Levitus climatology
• Coupling of low-resolution ocean models to sea
  ice, air, and land components eg. PCM, CCSM,
  HadCM3

7
Goal is to have a climate model that can run
efficiently on more than 1000 processors with an
atmospheric and land resolution of 30 km, ocean
resolution of 10 km, and land resolution of 1 km
8

• The fidelity of coupled climate system dynamics
  depend on its air/sea interactions.
• The ocean model must supply realistic SST fields
  to the atmospheric model.
• To do this the ocean model must realistically
  simulate all aspects of the hydrodynamical nature
  of the ocean. These include
• Energy levels
• Mean, variability, and location of currents?
• Intrinsic scales
• Modes of variability
• Planetary waves
• Water masses
• Meridional heat and mass transports
• Representation of sea ice (Not covered here)

9
(No Transcript)
10
Gulf Stream Heat Transport 1.3 PW
AVHRR Snapshot of SST from
Comparison of SST from AVHRR and 3 global ocean
models that differ only in horizontal resolution.
Each is run for 1 model year, and all are
initialized from WOCE SAC climatology. Courtesy,
David Webb (SOC)
11
1/12o
1/4o
1o
1/4o
1/12o
1o
12
SSH Variability (cm) from three North Atlantic
POP simulations and TOPEX/POSIEDON Altimeter Data
(Bryan et.al. 1998)
0.4o
0.2o
T/P
0.1o
13
RMS SSH (cm) from TOPEX/POSIEDON Altimeter Data
and Global 0.1o, 40-level POP
T/P
0.1o POP
McClean Maltrud
14
Sea surface height variability (cm)
Maltrud McClean
15
Sea surface height variability (cm)
Maltrud McClean
16
Mean SSH for runs 14a,14b,14c during 1994-1996
14a
14b
14c
Courtesy, R. Smith
17
Mean SSH for runs 14a,14b,14c during 1998-2000
14a
14b
14c
Courtesy, R. Smith
18
SSH Variability for runs 14a,b,c during 1998-2000
14c
14a
14b
Courtesy, R.Smith
19
Large-Scale SSH Variability OCCAM 1/4?
1
2
3
0
Courtesy, D. Webb
20
SE Pacific Modes
1
2
3
0
Courtesy, D. Webb
21
Surface drifter tracks (left) and North
Atlantic 0.1?, 40-level POP numerical
trajectories for 1993-1997 (McClean et al. (2002)
22
Lagrangian velocity integral time scale in days
for a) zonal and b) meridional directions from
in-situ drifters c) d) same for MICOM drifters
(Garraffo et al. 2001, JMS)
23
Lagrangian Time Scales (days) from Drifters and
0.1?POP
McClean et al. 2002, JPO
24
McClean et al. 2002
25
Lagrangian timescale TL as a function of latitude
for the actual () and simulated (o) floats, with
standard error bars. The zonal averages are
conducted separately for surface (top) and deep
(bottom) observations and data west of 35oW
(left) and east of 30oW (right). Solid lines with
shaded error bars indicate zonally averaged TE
from highpassed model output, weigthed by the
mean density of simulated Lagrangian
observations.
Lumpkin et al. (2002), JPO
26
How Realistic is the High Frequency Signal of a
0.1 Resolution Ocean Model? R. Tokmakian J.
McClean, JGR (in press)
Wavelet decomposition of SSH anomaly signal near
Atlantic City
Correlations (above) model/tide gauge - gray
model/Topex - black,open open
0.4-0.5, black gt0.5
e) Time series, at 3 days for the model (black
line) and the tide gauge measurements (gray line)
- units of cm. f) wavelet power spectrum for
tide gauge data in log2 of normalized variance
units g) same as f, except for the model.
a-d) averaged power squared for temporal bands
30days, 45.5 days, 65.5 days, and 100.5days.
Dashed lines 95 signif. Level model gray line,
tide gauge black
Contoured lines show significant signal at the
given periods for a white noise spectra and the
dashed lines shows the confidence interval for
the time series.
27
Temporal and spatial structure of the Atlantic
eddy flow between 1980 and 2000 WOCE current
meter database versus 1/6 CLIPPER model
28
North Brazil Current Ring ExperimentWilson et
al, GRL 29, Apr 02observed rings of
Subsurface, Deep, Intermediate and
Shallow types
Subsurface
deep
shallow
29
MICOM North Atlantic 1/12 deg.
subsurface
deep
intermediate
shallow
Similar to observations, the rings vertical
structure show subsurface, deep, Intermediate and
shallow ring types. (Garraffo, Johns,
Chassignet and Goni, submitted 2002, Elsevier).
30
(No Transcript)
31
(No Transcript)
32
Model Observation
Comparison of model mean distributions of
eastward velocity, temperature, and salinity with
the observed distributions by Wyrtki and Kilonsky
(1984). (Ishida et al., 1998)
33
Equatorial Pacific Subsurface Countercurrents
Synoptic section of zonal velocity and neutral
density showing SCCs in the western Pacific from
ADCP and the JAMSTEC model (1/4?,55 levels) from
Donohue et a, 2002 (JPO)
0.1?, 40-level global POP (McClean Maltrud)
34
Equatorial Pacific Subsurface Countercurrents
from three global z-levels models Annual mean
zonal velocity for 1995
Courtesy, K. Donohue
35
Core positions (top middle) and velocities
(bottom) of the SCCs. Bars show plus/minus one
STD. Observations (left) from Rowe et al. (2000)
are shown with large solid circles squares. The
model (right) reproduces the progressive eastward
lightening but not the equatorial divergence of
the primary SSC cores. From Donohue et al. 2002
(JPO)
36
The Subsurface Countercurrents
Core positions and velocities of the SCCs. Dashed
lines connect the model means, with bars showing
plus-minus one standard deviation. Observations
from Rowe et al. (2000) are shown with large
solid circles and squares. The model reproduces
the progressive eastward lightening but not the
equatorial divergence of the primary SCC cores.
Maximum core velocity in the model NSCC is found
near 115W. (Donohue et al., 2002)
37
Pacific NSCC quantities in core neutral density
layers averaged in three longitude ranges
(Donohue et al. 2002)
38
Saunders, 1999 (JGR)
39
Saunders, 1999 (JGR)
40
Forced with high frequency climatological ECMWF
winds and thermal forcing
Courtesy, J. Metzger
41
Velocity Cross-section at WOCE PCM-1 Current
meter data (top) vs. 1/12 Pacific HYCOM (bottom)
in the upper 1000 m
PCM-1 data from September 1995 to May 1996
Current meter data from Lee et al. (2001, JGR) 6
year mean from HYCOM forced with high-frequency
ECMWF winds and thermal forcing No ocean data
assimilation in HYCOM
42
Velocity Cross-section at WOCE PCM-1 Current
meter data (top) vs. 1/12 Pacific HYCOM (bottom)
in the upper 1000 m
Courtesy, J. Metzger
PCM-1 data from September 1995 to May 1996
Current meter data from Lee et al. (2001, JGR) 6
year mean from HYCOM forced with high-frequency
HR winds and ECMWF thermal forcing No ocean data
assimilation in HYCOM
43
Velocity Cross-section South of Japan Current
meter data (top) vs. 1/12 Pacific HYCOM (bottom)
in the upper 1500 m Along the ASUKA line
Courtesy, J. Metzger
Note The HYCOM section does not exactly follow
the ASUKA line
Current meter data from the ASUKA group 6 year
mean from HYCOM forced with high-frequency ECMWF
winds and thermal forcing No ocean data
assimilation in HYCOM
44
Observed versus modeled sea level along the
Mexican coast associated with the coastally
trapped waves generated by Hurricane Juliette in
2001
1/12 Pacific HYCOM forced with FNMOC NOGAPS/HR
winds and FNMOC NOGAPS thermal forcing. No data
have been assimilated into this model. Sea level
data provided by the University of Hawaii and the
Secretaria de Marina de México.
Courtesy, J. Metzger
45
(No Transcript)
46
Observations of change in the Indo-Pacific
Zonally averaged salinity differences on density
surfaces from Wong et al. (1999)
Courtesy, H. Banks
47
Methodology

• HadCM3 B2 historical emissions and concentration
  estimates
• Take measurements from HadCM3 from the same
  years as observations were taken
• Calculate model error bars from standard
  deviation of 5 year differences
• Observed error bars reflect internal variability
  from mesoscale eddies and seasonal variations

48
Using a water mass axis an example
Observations
HadCM3
Redmodelled change
Blueobserved change
Change on density surfaces
Actual profiles bold1970 dashed1995
49
Modelled and observed changes
Redmodelled change
Blueobserved change
Earlier sections 1962 - 1970 Later sections
1985 - 1995
Banks and Bindoff (J. Climate, in press)
50
Repeat XBT lines IX1 and IX12
Courtesy, A. Schiller
51
IX12 In-Situ and ACOM2 Depth of 20? Isotherm
Anomalies.
Courtesy, A. Schiller
52
IX1 In-Situ and ACOM2 Depth of 20? C (m) Anomalies
Courtesy, A. Schiller
53
Seasonal cycle of geostrophic transport (Sv)
across IX1 - observed and ACOM2. Negative values
indicate flow from the Pacific into the Indian
Ocean
Courtesy, A. Schiller
54
McClean, Ivanova, Sprintall
55
McClean, Ivanova, Sprintall
56
Joint EOF SST, Dynamic Height and Depth of 20?
isotherm
Joint EOF Temp at 100 m, Dynamic Height and
Depth of 20? isotherm
McClean, Ivanova, Sprintall
57
(No Transcript)
58
Time-mean, normal transport in temp. bins (
1oC) in the upper ocean ( 5oC), averaged
in time (8 yr.), and integrated along the
repeated hydr. ship track in the tropical North
Pacific (a) modeled Eulerian, (Sv)
(b) modeled eddy-induced, (Sv) (c) as in (b)
exept based only on q (S35psu) (d) measured
eddy-induced transport (Roemmich and Gilson
2001). Numerical labels indicate total southward
and northward transports in the upper ocean.
McWilliams and Danabasoglu, 2002 (JPO)
59
Time-mean, normal transport in temp. bins (
1oC) in the upper ocean ( 5oC), averaged
in time (8 yr.), and integrated along the
repeated hydr. ship track in the tropical North
Pacific (a) (Sv)
from the model (b) from the
eddy-resolving measurments (Roemmich et al. 2001)
McWilliams and Danabasoglu, 2002 (JPO)
60
Tidally-driven circulation in a global
OGCMSimmons et al. (sub. Ocean Modeling)

• Parameterization incorporated into a coarse
  resolution global OGCM that explicitly accounts
  for tidal energy source for mixing. Mixing
  evolves both spatially temporally with the
  model state
• 3 cases
• Tidal mixing parameterization variable mixing
• Uniform mixing 0.9 cm2 s-1
• Arctangent mixing profile 0.3 cm2 s-1 in upper
  ocean, 1.3 cm2 s-1 below 2500 (Bryan and Lewis,
  1979)

61
Variable Mixing
Uniform Mixing
Meridional overturning for a) global b),
Pacific/Indian, c) Atlantic oceans (Simmons et
al.)
62
Meridional heat transport for the global ocean.
WOCE inversion (Ganachaud and Wunsch 2000, Nature)
Simmons et al.
63
Bias in globally-averaged temperature (left) and
salinity (right) measured as a departure from
climatology (Levitus and Boyer 1994)
Simmons et al.
64
Conclusions

• To be added