godae oceanview intercomparison task team fabrice hernandez and matt martin
DESCRIPTION
Historical background: 10 years of cal/val activities in the framework of GODAE MERSEA IP 6 months TOP1 6 months TOP2 MERSEA Strand 1 6 months MERSEA Interc. GODAEGODAE OceanView North Atlantic Med European Seas Global 3 months GODAE Interc. Ocean basins Global FOAM TOPAZ MFS MERCATOR HYCOM MyOcean FOAM TOPAZ MFS MERCATOR DMI-Baltic FOAM TOPAZ MERCATOR HYCOM MOVE/MRI BLUElink> C-NOOFS European Seas Global NWS- FOAM ARC- TOPAZ MED- MFS GLO- MERCATOR BAL- DMI Black Sea IBI NetCDF COARDS-CF OPENDAP/LAS NetCDF3 Class 1 ATL Class 2 ATL Class 3 ATL Class 1 glo Class 2 glo Class 3 glo Class 1 sea-ice Class 2 sea-ice Class 4 T/S Class 4 sea-ice Class 1 new Class 2 newTRANSCRIPT
GODAE OceanView Intercomparison Task Team
Fabrice Hernandez and Matt Martin
Contents
Historical background: Cal/Val activities during GODAE
Brief overview of existing metrics definitions
Results from final GODAE intercomparison
Issues coming out of that intercomparison
DISCUSSION:
Objectives of the Intercomparison Task Team
Main areas to focus future effort
Historical background: 10 years of cal/val activities in the framework of GODAE
2001 2003 2004 2008 2009
MERSEA IP
6 months
TOP1
6 months
TOP2
MERSEA Strand 1
6 months
MERSEA Interc.GODAE GODAE OceanView
North AtlanticMed
European SeasGlobal
3 months
GODAE Interc.
Ocean basinsGlobal
FOAMTOPAZMFSMERCATORHYCOM
MyOcean
FOAMTOPAZMFSMERCATORDMI-Baltic
FOAMTOPAZMERCATORHYCOMMOVE/MRIBLUElink>C-NOOFS
European SeasGlobal
NWS-FOAMARC-TOPAZMED-MFSGLO-MERCATORBAL-DMIBlack SeaIBI
NetCDFCOARDS-CFOPENDAP/LAS
NetCDF3
Class 1 ATLClass 2 ATLClass 3 ATL
Class 1 gloClass 2 gloClass 3 glo
Class 1 sea-iceClass 2 sea-iceClass 4 T/SClass 4 sea-ice
Class 1 newClass 2 new
GODAE metrics definitions Class 1 – daily average model fields interpolated onto pre-defined grids (eddy-
permitting view) on specified levels
Class 2 – model fields interpolated to pre-defined mooring locations and sections.
Class 3 – transports through sections and other integrated quantities such as Meridional Overturning Streamfunction and heat transports.
Class 4 – assessment of forecasting capabilities through comparison of model with assimilated and independent observations
3000250020001500100070040020010050300
Final GODAE intercomparison Design of the Intercomparison experiment (~2006-2007):
Extending MERSEA Class 1 and Class 2 metrics to global scale Definition of intercomparison objectives:
a) Demonstrate GODAE operational systems in operationsb) Share expertise and design validation tools and metrics endorsed by GODAE operational centersc) Evaluate the overall scientific quality of the GODAE operational systems
Implementation of metrics computation Demonstration phase: 3 month period (Feb, Mar, Apr 2008). Phase of synthesis:
June 2008 to GODAE Final Meeting in Nov. 2008 Additional synthesis: until January 2009
GODAE metrics produced from various groups: BlueLink, FOAM, HYCOM, Mercator, MOVE/MRI, ...
Data put on ftp servers and/or OpenDAP servers. Most of the intercomparison results obtained so far have looked at monthly means and
standard deviations against climatology or some other processed data (e.g. SST analyses)
No sea-ice results Some intercomparisons (e.g. done by the Australians) have focussed more on
comparison with assimilated and independent observations, e.g. SST, SLA, Argo and surface drifters, in the Indian Ocean and South Pacific regions.
GODAE systems in comparison
Mercator NEMO ECMWFSEEK – RkF
T,S, SLA, SST
HYCOM HYCOM FNOC-NOGAPSNCODA – MvOI
T,S,(SLA), SST, ice
FOAM NEMO UK-Met T, S, SLA, SST, ice
BLUElink ind/spa MOM4 BoMBODAS – EnOIT,S,SLA,TG, SST
TOPAZ nat/arc HYCOM ECMWFEN-kF
T,S, SLAmaps, SST, ice
MOVE/MRI npa MOVE JMAMRI - 3Dvar
T,S, SLA, MG-SST
C-NOOFS nw-nat NEMO Env. Canada no
FOAM
HYCOM
Mercator
Intercomparison in the TATO
STIA
SST
OST
IA S
ST E
rror
HYC
OM
(0.7
3 / 0
.35)
FOA
M (0
.30
/ 0.4
1)
Mer
cato
r Glo
bal P
SY3V
2 (0
.41
/ 0.5
0)
Snapshot SST comparison the 15th of February 2008 with respect to OSTIA. Numbers in brackets correspond to RMS differences in the box limited area in the Gulf of Guinée (15°W-5°E and 5°S-5°N), and the box limited area for the Northern Tropical Atlantic (55-15°W and 5-25°N), plotted for the OSTIA figure. Units +3/-3 in Kelvin.
HY
CO
M (0
.73
/ 0.3
5)P
SY
2V3
(0.3
4 / 0
.39)
0.34 0.41
0.73 0.30
0.35 0.41
0.500.39
TNA
SAT
OSTIA SST Std Feb-April 2008
HYCOM
FOAM
PSY3PSY2
OSTIA
HYCOM
FOAM
PSY3 PSY2
OSTIA
0.5°C
Box averaged SST
0.5°C
Assessment of EKE in NAT
TOPAZ FOAM
HYCOM
C-NOOFS
SURCOUFMercator
Monthly comparison in April’08
Final GODAE intercomparison: general scientific outcomes
GODAE eddy-permitting systems are consistent (i.e. match qualitatively the climatology, general patterns of the ocean circulation) and there is no “bad” surprise
Accuracy assessment reveals differences, biases, possible errors in model or assimilation schemes… These evidence are a first step for targeted corrections and improvements
Impact of horizontal resolution is evidenced on kinetic energy levels.
Further work need to be done to identify the causes of differences between the systems (e.g., impact of forcing, data assimilation schemes….)
Final GODAE intercomparison
Successes of the intercomparison:
The hindcasts/forecasts were made available and easily accessible (and people were responsive if there were problems accessing the data).
Most/all of the work was done using the Class 1 fields. These fields were generally produced using the agreed definitions (at least close enough to make it relatively easy to use them).
The comparison of the Class 1 fields highlighted some interesting differences between the systems.
Visibility of this work through scientific communication and publication Observations were made available by the observing community,
involved now in operational oceanography, and supporting the ocean forecasting centers
Final GODAE intercomparison Shortcomings of the intercomparison:
Some systems produced the metrics in their normal operational setting, whereas others were re-run in hindcast mode.
Some groups upgrade their systems in the meantime of the Intercomparison synthesis
Some systems produced forecasts and others just analysis fields. Class 1 still need some homogenisation Class 2 and 3 metrics were produced by some systems but not all, and no
comprehensive assessment of them was carried out. Very little/no work done on Class 4 metrics intercomparison Most GODAE partners used FTP rather than OPENDAP (not technically efficient) A demonstration rather than a routine intercomparison. A three month period is a really short period of time to overview ocean
forecasting system behaviour and performance It would have been useful to meet to present and discuss results: the calendar
was very tight Several groups had problems to fully contribute to the exercise, and human
resources dedicated to intercomparison synthesis were not available in all forecasting centres.
Discussion and outlook
1. What is the role of a validation/intercomparison Task Team?
2. Strategy of the validation/intercomparison Task Team
3. Clarify workplan of Intercomp/Val TT and interactions with:
OSE/OSSE TT Coastal and Shelf Seas TT Biogeochemical TT ET-OOFS WCRP-CAS WGNE
1. What is the role of a validation/intercomparison Task Team?
• Core scientific activity: develop metrics, share experience, evidence differences, cross-fertilized ideas among GODAE centers
• Be consistent as a group in our monitoring policy: provide tool for monitoring routinely the systems, and controlling inputs (e.g. link with data providers), and outputs (mandatory link with users)
Main benefit: improvements of the systems, and the quality of products
• Provide visibility as GODAE community: demonstration, publications…
2. Strategy of the validation/intercomparison Task Team
Rely on new targeted intercomparison exercices ?
Establish permanent monitoring among the OOFS?
Expect outcomes from regional activities (e.g. MyOcean, US Navy)?
2. Strategy of the validation/intercomparison Task Team Suggested scientific aspects that should be addressed
Extend comparison with other set of independent observations, e.g. ocean colour, surface drifters
Assessing the performance of the data assimilation using observation-minus-background and observation-minus-analysis statistics:
Useful to show accuracy of short-range forecasts and the performance of the assimilation.
Different analysis time-windows and operational schedules make it difficult to intercompare (o-b) between systems.
Assessing the performance of the model through estimates of forecast skill: Anomaly correlations and RMS differences between forecasts and analyses (and
between forecasts and observations). Multi-model ensemble statistics
provide error levels and monitoring tools Design user oriented metrics for targeted applications (ocean climate
monitoring, oil spill, S&R…) Demonstration of routine validation activity
Inte
rnal
Ext
erna
l
3. Clarify workplan of Intercomp/Val TT and interactions
Take into account new OOFS (NCEP, MFS, China…) Diagnostics that allow the characterisation of biases, long term
changes (link with GSOP) Link with coastal validation
Assessing the accuracy/impact of IC/BC (downscaling) Share scientific assessment methodology
Link with biogeochemistry validation Assessing the accuracy/impact of the physical variables (vertical diffusion,
coupling) Share scientific assessment methodology
Link with OSE/OSSE TT (characterize the impact of incoming data, feedbacks to relevant data providers) :
Develop common metrics for both validation and data impact assessment
Suggested plan for the coming months
Use of existing Feb-Mar-April 2008 dataset: Extended scientific validation ? Inform which metrics should be integrating a possible routine monitoring (daily
NRT production)
Discuss future implementation depending on chosen strategy Prepare workplan : roadmap document
By end of september 2009 Review by OOFS Beginning of implementation in 2010 Prepare calendar for meetings/discussions
Topics to be addressed in the roadmap: Discuss technical aspect of NRT production (storage and exchange): possible link with
ET-OOFS focus on a sub-set of useful metrics