ECMWF Report

1. ECMWF Report • Operational Implementation of NEMO/NEMOVAR • ORAS4: Ocean ReAnalysis System 4 • Some lessons learnt during preparation • Evaluation process • Plans • Magdalena Alonso Balmaseda • Kristian Mogensen

2. Atmospheric model Atmospheric model Wave model Wave model Ocean model Ocean model Real Time Ocean Analysis Delayed Ocean Re-Analysis ~ORAS4 (NEMOVAR) ECMWF: Forecasting Systems Medium-Range (10-day) Partial coupling Seasonal Forecasts Fully coupled Extended + Monthly Fully coupled Ocean Initial Conditions

3. ECMWF: Operational Ocean Changes in 2011. ORAS4 • ECMWF has a implemented new operational ocean re-analysis system. • http://www.ecmwf.int/products/forecasts/d/charts/oras4/reanalysis/ • It implies the transition to NEMO/NEMOVAR from HOPE/OI • It consists of 5 ensemble members, covering the period 1958-Present, continuously updated. • It is used for the initialization of the operational monthly and seasonal forecasts. • It is also used to initialize the CMIP5 decadal forecasts (EC-Earth …) • It is a valuable resource for climate variability studies. • Documentation in preparation: Mogensen et al 2011, Balmaseda et al 2011

4. ORAS4 Main Ingredients Ocean Model: NEMO V3.0 ORCA1 and 42 levels (ocean) Data Assimilation: NEMOVAR (3D-var FGAT). Data: Temperature and Salinity Profiles (EN3-XBT corrected and GTS), SST (HADISST/ OIv21x1 /OSTIA), along track Altimeter Sea Level (AVISO). See figure below Forcing: ERA40/ERA-INTERIM/ECMWF NWP (see figure below) Bias Correction: In T/S and P gradient. Seasonal prescribed (from Argo+Alti) + Adaptive on line Ensemble Generation: wind perturbations, observation coverage, spin-up. 5 ensemble members

5. What have we learned in the preparation process? • Which SST product to use? • Which products are available? Criteria for evaluation • Assimilation of altimeter: variational implementation of Cooper and Haines in 3Dvar. Non trivial. Sorted. • Coastal Covariances: Impact of Assimilation in the Atlantic MOC. • Bias correction scheme: estimation of the offline term from Argo period. (Not a problem, a success; It affects the results) • How to evaluate ocean reanalyses?

6. Which SST product to use? OIV2_025_AVHRR: bias cold in the global mean (regional differences) Bias decreases with time. OSTIA in beween (not shown) Weaker interannual variability Fit to insitu Temperature: bias cold in tropics, better in mid latitudes,. Not clear impact on Seasonal Forecasts DECISION: OIV2_1x1 until 2010, OSTIA thereafter. Options for Re-analysis OIV2_1x1: (weekly)~1982 onwards OIV2_025_AVHRR(daily)~1982 onwards OIV2_O25_AVHR+AMSR:~2002 onwards Options for Real-Time: As before + OSTIA (from 2008 onwards): Consistency with atmospheric analysis

7. Multivariate balance for Altimeter IN NEMOVAR the balance is between sea level and steric height Original formulation of NEMOVAR αref and βrefare calculated by linearizing the equation of estate using the background T/S values as reference. Comments: i) zref=1500m is arbitrary. An attempt to take into account that baroclinicity is low below this level. Can we account for the vertical stratification more universally? ii) this can lead to increments in model levels with large dz

8. But Impact on Steric Height not realistic: This problem not so apparent if assimilating T/S and altimeter, but it is still there. Why? Single Obs Experiments: T increment The temperature increment is applied to the thickest model levels

9. Modifications A):Weighting based on stratification. Use BV frequency to calculate αN and βN instead of equation of state B) Do not double-count balance-salinity corrections

10. New Balance formulation: Problem with Steric Height Solved Sea Level Altimeter (AVISO) CONTROL ASSIM: TS CONTROL+ALTI ASSIM: TS + ALTI Problem with deep T increments Solved New Old

11. RMSE of 10 days forecast EQ Central Pacific EQ Indian Ocean CONTROL ASSIM: T+S ASSIM: T+S+Alti TROPICAL Pacific GLOBAL Altimeter Improves the fit to InSitu Temperature Data

12. Assessment of the ORA-S4 re-analysis • Choose a baseline: the CONTROL (e.i., no data assim) • Assim Intrinsic Metrics • Fit to obs (first-guess minus obs): Bias, RMS • Error growth (An-obs versus FG-obs) • Consistency: Prescribed/Diagnosed B and R • This is insufficient to assess a Reanalysis product • Spatial/temporal consistency: long time series and spatial maps • Time correlation with Mooring currents • Correlation with altimeter/Oscar currents • Transports (MOC and RAPID): short time series • Quite limited records. Not always independent data • Skill of Seasonal Forecasts • Expensive. Model error can be a problem. • Observing System Experiments

13. Assimilation Statistics: Incremental Analysis Update

14. Fit To Obs • Bias thin lines • RMSE thick lines • Assimilation improves over the control everywhere. A large part of the improvements comes from the reduction of bias. • Note large errors in Extratropics come from WBC and coastal areas, where obs are given little weight

15. Fit to Obs ORAS4shows reduced RMSE and bias respect the CNTL, in both T and S The bias is ORAS4 is more stable in time Fit improves with time, both ORAS4 and CNTL :Not only more subsurface obs, but better surface forcing and SST data?.

16. Fit to ADCP mooring data • Some improvement of the Pacific and Atlantic undercurrents, which are still on the weak side.

17. NEMOVAR re-an: verif. against altimeter data CNTL NEMO NoObs NEMOVAR T+S ORA-S4: NEMOVAR T+S+Alti

18. Comparison with RAPID derived transports Atlantic MOC at 26N Short time series ORAS4 underestimates the MOC Note the large minima in 2010 and 2011!!

19. More MOC diagnostics MOC profile RAPID ORAS4 CNTL • In low res model the Florida Strait transport is not so well defined. • Assimilation reduction of the FST is proportional to the weight is given to the obs (not shown) • Ocean model tends to produce too strong and shallow AABW cell

20. Impact on Of ORAS4 in SST Seasonal Forecasts Anomaly correlation: ORAS4 CNTL Persistence

21. Global Surface Heat Fluxes from Reanalysis Pinatubo El Chichon

22. Climate Applications The time integral of the ERA+ASM surface heat flux results in the evolution of the total ocean heat content ERA+ASSIM heat flux integral ORAS4 total (whole depth) heat content Mean and time variability of ORAS4 oceanic heat transport. (GW2000: Ganachaud and Wunsch 2000)

23. Summary • Operational implementation of NEMO/NEMOVAR in forecasting systems and ocean reanalysis • Transition from HOPE/OI • ORAS4: new Ocean Reanalysis with NEMOVAR • Still climate resolution: approx 1x1 degree. • Some lessons learns in the preparation process • Choice of SST product for reanalysis not trivial. Next is to try OSTIA reanalysis • Balance relationship between altimeter and T/S not solved problem. Not trivial. Still room for improvement • Improved covariance needed for the assimilation of observations near the coast. • How to evaluate an ocean assimilation system and ocean reanalyses product? • Evaluation of the ocean reanalysis is a pre-requisite for the interpretation of climate signals • Standard assimilation statistics needed but not sufficient for the reanalyses • Need information about time variability: sustained time series are very important: • (altimeter, moorings, RAPID, other?) • Impact on seasonal forecast is a test and a result. • What is next? • Document system, web pages, papers • Higher resolution ocean model and reanalysis(ORCA 025) • Sea-Ice model in monthly, seasonal forecast and reanalyses • Improved coupling (bulk formula, wave effects, ocean mixed layer) • Increased coupling (forecast and analysis)