The Mercator experience
Transcription
The Mercator experience
www.mercator.eu.org Data Assimilation and Operational Oceanography: The Mercator experience Benoît Tranchant btranchant@mercator-ocean.fr And the Mercator-Ocean Assimilation Team (Marie Drevillon, Elisabeth Remy, Mounir Benkiran, Eric Greiner, Jean-Michel Lellouche, Laurent Parent and Charles Emmanuel Testut) NEMO users meeting 2006 Mercator Ocean: An OPA user but also an OPA developer OPA configurations used in operational (2006): North Atlantic (OPA8.1/rigid lid) MNATL medium (1/3°) and PAM high (1/15°) resolution Global (OPA8.2/free surface) ORCA2 and ORCA025 Next important steps: Regional: 1/3° & 1/15° •OPA-NEMO: •ATL4, ORCA025 and ATL12 in 2007 •ORCA12 in 2008 NEMO users meeting 2006 Global: 2° & 1/4° 2006developer Mercator Ocean: A DA user and Brief overview of data assimilation schemes: SAM1-v1: Univariate OI (SLA). SAM1-v2: Multivariate OI (SLA, SST and T/S profiles) Operational using 1D vertical modes R&D SAM2-v1: Multivariate SEEK filter (SLA, SST HR and T/S profiles) using 3D modes SAM3-v0: Multivariate 3D-Var (SLA and T/S profiles) NEMO users meeting 2006 Mercator Data Assimilation Systems : SAM 2006 OCEAN MODEL: OPA-NEMO DATA ASSIMILATION SCHEMES •SAM1 : OI •SAM2 : SEEK filter •SAM3 :3D VAR SATELLITE observations IN SITU observations DATA SSALTONEMO usersCORIOLIS meeting 2006 Forcing Fluxes & RTG_SST The Mercator-Ocean forecasting systems satellite altimetry + RTG_SST + in situ data satellite altimetry PSY2G v1 PSY1 v2 Reanalysis 1992-2002 Impact studies Oceanic initial conditions for seasonal forecasts satellite altimetry + RTG_SST + in situ data satellite altimetry PSY2 v2 PSY3 v1 NEMO users meeting 2006 North Atlantic prototypes: medium (1/3°) and high resolution (1/15°) Based on: 1. OPA 8.1 model, 43 vertical levels Two configurations: North Atlantic 1/3° horizontal resolution North Atlantic + Mediterranean 5 to 7 km hor. resolution 2. PALM_research software (CERFACS) modularity of the coupling model-data assimilation system 3. SAM1v2 data assimilation system: • • • Reduced-Order Optimal Interpolation (ROOI) technique based on SOFA (LEGOS) 1D vertical multivariate Empirical Orthogonal Functions (EOFs) multivariate assimilation of SLA from JASON-1, ENVISAT and GFO, RTG_ SST, REYNAUD SSS climatology and T/S vertical profiles provided by the CORIOLIS centre. In development: ATL4 & ATL12 (OPA-NEMO), 50 vertical levels with the SEEK filter assimilation of SLA, SST HR and T/S profiles using up to 200 multivariate 3D modes for background errors NEMO users meeting 2006 OSE : Impact of PIRATA data PSY1v2 Time serie of temperature profiles In-situ measurements at SAMBA mooring (EQ-35°W) With assimilation of PIRATA data Without assimilation of PIRATA data NEMO users meeting 2006 Mercator PSY1v2 Reanalysis: MERA 11 1992-2002 Climate : temperature anomalies PSY1v2 350m NAO + 1994 Cold Subpol Gyre Warm Gulf Stream NAO - 1998 Warm Subpol Gyre Warm Subtrop Gyre NEMO users meeting 2006 PSY2v2, the Mercator High Resolution Forecasting System for North Atlantic and Mediterranean Sea An example: Eddies in the Gulf of Mexico PSY2v2 Origin, separation and westward progression of an anticyclonic eddy (mid-June to mid-October 2004) • Very good spatial and temporal agreement between model SSH and satellite Chlorophyll concentration (not assimilated) • Warm eddies are characterized by a weak chlorophyll concentration NEMO users meeting 2006 6 weeks from 25 July 2004 to 4 September 2004 Chlorophyll-A (MODIS-CATSAT) SSH (PSY2v2) PSY2V2 : T/S profiles impact The Mediterranean Water Outflow PSY2v2 Climatology Salinity at 1000m (2004 temporal mean) Salinity (11°W,40°N) on 15/08/2004 Model forecast observation NEMO users meeting 2006 PSY2v2 North Atlantic prototypes in development: ATL4 with SAM2-v1 Based on: 1. OPA-NEMO model, ice model, free surface and bulk formulae, 50 vertical levels 2. PALM_MP software (MPI-2) enhanced modularity of the coupling model-data assimilation system 3. SAM2v1 data assimilation system: • • • Based on SEEK filter (LEGI) 3D multivariate seasonal anomalies (O2), adaptative variance (P/2), control vecto (Hbar,T, S, U,V) Data assimilation of SLA, RTG_ SST, SST HR and T/S vertical profiles. First analysis increments of salinity and temperature in january 2004 NEMO users meeting 2006 GLOBAL PROTOTYPES Low resolution 1. Model: ORCA2 OPA 8.2 free surface, 31 vertical z-levels, 2° hor. resolution and ½° in the tropics 2. Palm software 3. Assimilation methods: – Operational : Assimilation of SLA data (OI, Cooper and Haines lifting/lowering method) used for seasonal forecasting & Eddy permitting resolution 1. Model: ORCA025 OPA8.2 surface, 46 vertical z-levels, ¼° hor. resolution free 2. Palm software 3. Assimilation methods: • Operational: Assimilation of SLA data (OI, Cooper and Haines lifting/lowering method) operational since october 2005 Data Assimilation schemes in development: Multivariate assimilation of SLA and T and S in situ data with 3D modes (method based on SEEK filter, LEGI) Multivariate assimilation of T/S in situ data and SLA with a 3D-var method (based on ORCAVAR, CERFACS) Model in development: •ORCA025-LIM (50 vertical zlevels, 1m-450m) •ORCA12-LIM NEMO users meeting 2006 Validation with independent in situ data TAO/TRITON moorings in the Tropical Pacific Ocean PSY3v1 110°W 5°S 140°W O°N Temperature of the global system Temperature of the global system Temperature observed by the TAO buoy Temperature observed by the TAO buoy °C NEMO users meeting 2006 http://www.pmel.noaa.gov/tao/jsdisplay/ A basis for global climate monitoring Global system heat content in 2005 (0-700m), in 1011 J/m2 PSY3v1 Climate change indicators Average depth of the 26°C isotherm in the global forecasting system in September 2005 hurricane indicators NEMO users meeting 2006 Seasonal forecasting activity with ORCA2: First results obtained with SAM2 (SEEK) •ALL1: assimilation of SLA, in situ T,S profiles and SST data (T/P, JASON, ERS2, ENVISAT, ENSEMBLE, RTG_SST) Reduced order Kalman filter (SEEK formulation), with 200 multivariate 3D modes ALL1 vs Control run ALL1 Control run Std dev. Improvement (0< means better) NEMO users meeting 2006 Equipe R&D assimilation SAM3 Based on ORCAVAR – v1.x Intercomparion experiment with SAM2 = - 3DVar : OPAVAR version with the « default options » (background error covariance matrix,…), - ENACT like experiments: corrected ERA40 fluxes, a=-200W/m2/K, … - ORCA2: free surface constant volume: daily correction of E-P. ≠ - along track T/P sla assimilated (filtered for ORCA2), MDT Rio V5.1 - observation error covariance matrix: 3D for T and S, 2D for altimetry computed using the Fu. et al. method (1993): <rr>=<yy>-<ym> Representativity SLA, zonal NEMOerror users: meeting 2006T mean Mean HC(2001) – Mean HC(1993) in °C SSH and heat content evolution from 0 to 300 m (1993-2001) Mean SSH (2001) – Mean SSH (1993) in m Data: Merged satellite Products Assim. in situ + T/P (IAO) Control Data 65°N-65°S Temperature anomalies (0– NEMO users 300m/60°N-60°S) SLA meeting 2006 global 60°N-60°S Summary 2006 OPAx and SAMx OPA8.1 OPA8.2 SAM1v1 SAM2v1 SAM1v2 OPA9 SAM3 •Analysis and 2 week forecasts every week:PSY1V2,PSY2V2,PSY2GV1 and PSY3V1 •Reanalysis: MERA11 •OSE/OSSE:Pirata impact, SSS assimilation from SMOS and Aquarius (Newsletter #21:April 2006) •Seasonal forecasting •Reanalysis with ORCA025 planned in 2007 with SAM2. NEMO users meeting 2006 Perspectives 2006 Upgrade of data assimilation schemes: SAM2-vx A better way to assimilate largest scales (H) Smoother/IAU Adaptative covariances (time dependent) Data assimilation of surface currents (drifters/Radar) and ice concentration SST HR Estimation of representativity errors (obs. errors) Numerical Optimization for huge configuration (ORCA12 400 Gb): PALM_MP using. SAM3-vx improvement of salinity assimilation Improvement of B matrix Coupling SAM2 with OPA-NEMO configurations: ATL4 and ORCA025-LIM : end of 2006/ beginning of 2007 ATL12 in 2007 ORCA12 in 2008 NEMO users meeting 2006