Comparing model with observations: methods, tools and results

1
Comparing model with observationsmethods, tools
and results

• Mélanie JUZA, Thierry Penduff, Bernard Barnier
• 
  
• LEGI-MEOM, Grenoble

DRAKKAR meeting, Grenoble, France, 11-12-13
February 2009
2
Objectives / Activities

• Assessment of DRAKKAR simulations
• - Quantitative and systematic comparisons
  model/observations
• - Intercomparison of simulations
• (impact of resolution, forcing, numerical
  scheme, parametrizations)
• Observability of the ocean dynamics (OSSE)
• - Accuracy of ARGO array
• Distribution of data and tools to the
  scientific community

• Development of tools collocation
  model/observations, statistics, vizualization
• Scientific studies. Papers in preparation

3
Hydrography collocation
4
Hydrography collocation
ARGO 1998-2004
5
Hydrography simulated and observed MLD
Mixed layer depths (MLD) (m)
ARGO
ORCA025-G70
? Realism of simulated and observed MLD
6
Hydrography method for the analysis of mixed
layer quantities

• Distribution of Mixed Layer Depth / Temperature
  / Salinity / Heat and Salt Contents
• Medians and percentiles 17 and 83

Exemple MLD in North Atlantic
MODEL BIAS
SAMPLING ERROR
September 1998-2004
-- full model -- subsampled model
(like ARGO) -- ARGO
7
Hydrography sampling errors
Monthly cycles of MLD (1998-2004)
zone MNW-ATL
-- subsampled model (ARGO) -- full model
MLD
Solid lines medians Dashed lines percentiles
17, 83
Sampling error ? well observed monthly cycle.
Sampling error in winter.
8
Hydrography sampling errors at global scale
ARGO sampling errors on the monthly MLD
(1998-2004)
Sampling error ltsubsampled model gt ltfull
modelgt
Bins 30 x 30 x 1 month (1998-2004)

• ARGO sampling errors maximum in winter (extreme
  values 100m)
• Especially in inhomogene (Southern Ocean, North
  Atl.) and coastal regions

9
Hydrography sampling errors at global scale
ARGO sampling errors on the monthly MLD
(1998-2004)
Sampling error ltsubsampled model gt ltfull
modelgt
Bins 30 x 30 x 1 month (1998-2004)

• ARGO sampling errors maximum in winter (extreme
  values 100m)
• Especially in inhomogene (Southern Ocean, North
  Atl.) and coastal regions

10
Hydrography sampling errors at global scale
ARGO sampling errors on the monthly MLD
(1998-2004)
Sampling error ltsubsampled model gt ltfull
modelgt
Bins 30 x 30 x 1 month (1998-2004)

• ARGO sampling errors maximum in winter (extreme
  values 100m)
• Especially in high variable (Southern Ocean,
  North Atl.) and coastal regions

11
Hydrography sampling errors at global scale
ARGO sampling errors on the monthly MLD
(1998-2004)
Sampling error ltsubsampled model gt ltfull
modelgt
Bins 30 x 30 x 1 month (1998-2004)

• ARGO sampling errors maximum in winter (extreme
  values 100m)
• Especially in high variable (Southern Ocean,
  North Atl.) and coastal regions

12
Hydrography conclusion

• Assessment of ARGO sampling errors
• - More dependence on spatial distribution of
  floats rather than number of floats
• - MLT, MLS, MLHC, MLSC

• Assessment of the simulations
• - Mixed layer monthly cycles
• - Impact of resolution

• Perspectives
• Extension to - recent years (maximum ARGO
  coverage)
• - the last 50 years
  (interannual cycles)
• - all instruments (ARGO
  floats CTD, XBT, moored buoys)

13
Altimetry collocation
14
Altimetry interannual SLA (statistics)
Impact of resolution on low-frequency variability
SLA standard deviation (cm)
(1993-2004)
AVISO
¼ ORCA025-G70
1 ORCA1-R70
2 ORCA246-G70
½ ORCA05-G70.113
? Global increase of interannual variability with
resolution
15
Altimetry interannual variability (EOFS)
Assess the ability of models to reproduce the
observed interannual variability in various
regions

• Data processing
• - Observed SLA EOFs (decomposition spatial mode
  temporal amplitude-PC)
• Projection of simulated SLA on observed SLA EOFS
• Comparison PC(obs)/projections variance,
  correlation

Exemple interannual SLA in North Atlantic
(1993-2004)
Associated obs. amplitude and mod. projections
Mode 1 Observed SLA var17
Lag with NAO (weeks)
obs ¼ ½ 1 2
Projections of simulated SLA reproduce main
features of the obs. variability. More explained
variance with 1/4
Simulated lags more realistic with increase of
resolution
Intergyre gyre of Marshall
? Resolution improves space-time variability
16
Altimetry interannual variability (EOFS)
Exemple large-scale (gt6) and interannual SLA in
Southern Ocean (1993-2004)
Mode 1 Observed SLA var18
Associated obs. amplitude and mod. projections
Response to ENSO
Resolution does not change variance projected on
observations
Conclusion - Global and regional (North Atl.,
Gulf Stream, Equat. Pac., Indian, Southern
Ocean) - Resolution improves space-time
variability, except in Southern Ocean (intrinsic
variability?) - Similar processing applied to SST
analysis (Reynolds, NCEP) - Response of ocean to
atmospheric variability (NAO, ENSO, SAM, AAO) -
Impact of mesoscale on low-frequency variability
17
Conclusion

• Collocate and compare model observations T,
  S, SLA, SST
• Assess simulations. Quantify model sensitivities
• Evaluate the accuracy of observing systems (ARGO
  sampling errors, paper in preparation)
• Tools are mature. Technical report users
  manual. Fields are being distributed.

• Perspectives
• Further assess the interannual variability in
  eddying models (paper in preparation)
• Evaluate every new simulation (global, regional,
  reanalyses)
• Extend to new datasets current meters (G.
  Holloway), ice field thickness (A. Worby),
• gravimetry, maregraph, SSS,
• Foster collaborations

http//www-meom.hmg.inpg.fr/Web/pages-perso/Melani
eJuza/
18
Hydrography model bias at global scale
Model bias of the monthly MLD (1998-2004)

• run ORCA025-G70
• run ORCA246-G70

MLD
too shallow
too deep
ltBiasgt ltcollocated model ARGOgt
Bins 30 x 30 x 1 month (1998-2004)

• Model biaises seasonal, regional, too deep MLD
  in winter (max50m)
• The increase of resolution improves the
  representation of MLD

19
Hydrography model bias at global scale
Model bias of the monthly MLD (1998-2004)

• run ORCA025-G70
• run ORCA246-G70

MLD
too shallow
too deep
ltBiasgt ltcollocated model ARGOgt

• Conclusion - ORCA025-G70, ORCA05-G710.113,
  ORCA1-R70, ORCA246-G70
• - MLT, MLS, MLHC, MLSC
• - Resolution improves mixed
  layer monthly cycles
• - Use of all instruments from
  1956 to present (interannual cycles)

20
Altimetry SLA standard deviation
HF (Tlt5months)
MF (5ltTlt18months)
LF (Tgt18months)
AVISO
ORCA025-G70
ORCA05-G70.113
ORCA1-R70
ORCA246-G70
(1993-2004)
21
Altimetry SLA zonal variance and correlation
SLA standard deviation (cm)
Model/obs SLA correlation
(1993-2004)
LF (interannual)
MF (annual)
HF (high freq.)
AVISO ORCA025-G70 ORCA05-G70.113
ORCA1-R70 ORCA246-G70
? Zonal variability increases with resolution

• Zonal correlation decreases with resolution in
  S.O.

gt Forced vs intrinsic variability in the
Southern Ocean
22
Biais global T/S modèle global ¼
Structure verticale moyennes temporelles
Pdf de
Pdf de
Ecart modèle ¼ (sous-échantillonné comme ARGO)
- observations (ARGO)
Structure horizontale intégration sur les
couches de surface (1998-2004)
23
Biais régional T/S modèle global ¼
Courant Nord Atlantique
Kuroshio
-3C 300-500m
2C 100-400m
-0.6 à -0.25 0-600m
0.2 100-400m
24
Conclusion - Perspectives

• Altimetry
• Resolution improve space-time variability (lag
  NAO)
• Increased resolution yields stronger local
  variances (depend on latitude), similar or
  smaller correlations (increased intrinsic
  variability), improved basin-scale space-time
  variability
• Interannual variability impact resolution
  correlation , variance increase
• In general, enhanced variance projects on
  observations (except in Southern Ocean)
• perspective Impact of mesoscale on
  low-frequency variability. Forced vs intrinsic
  variability.

• In the future
• Continue to investigate the impact of the
  resolution on the realism of the model
  (2,1,1/2,1/4)
• Systematize the assessment of simulations
  (forcing, parametrization,)
• Regional simulations (NATL4, NATL12), with
  assimilation (HYCOM),
• Scientific studies and collaborations
• Others datasets current meters (G. Holloway),
  ice field thickness (A. Worby), gravimetry,
• maregraph, SSS,

25
Altimetry interannual SLA (statistics)
SLA standard deviation (cm)
AVISO
(1993-2004)
ORCA025-G70
ORCA1-R70
ORCA246-G70
ORCA05-G70.113
? Global increase of std(SLA) with resolution