Annual DRAKKAR Meeting, Jan , 2007 Greg Smith, Keith Haines, Dan Lea and Ben McDonald Environmental Systems Science Centre (ESSC), Reading University Evaluation of 47yr ORCA1 Simulations using a Generalized Observation Operator
ORCA1 Model Runs: All runs use CORE Interannual forcing, with namelist settings as in G70. Run for years No relaxation is used under sea ice. Runs: 36 day SSS relaxation 219 day SSS relaxation 180 SSS relaxation, with 3D relax at high latitudes (as in KAB001) 36 SSS relax, DS3 Precipitation field 36 SSS relax, DS3 Precipitation field, ERA40 winds (in progress)
Globally-averaged mean fields
Model/Levitus differences for January year 2004 Salinity
Model/Levitus differences for January year 2004 Temperature
Comparison with Levitus01 Mean salinity: 0-300m ORCA1-R04ORCA025- G70
Comparison with Levitus01 Mean salinity: m ORCA1-R04ORCA025- G70
Comparison with Levitus01 Mean salinity: m ORCA025- G70 ORCA1-R04
Comparison with Levitus01 Mean temperature: 0-300m ORCA025- G70 ORCA1-R04
Comparison with Levitus01 Mean temperature: m ORCA025- G70 ORCA1-R04
Comparison with Levitus01 Mean temperature: m ORCA025- G70 ORCA1-R04
Generalized observation operator Collocate observations on depth, isotherm and isopycnal levels By evaluating model-data difference on isotherm or isopycnal levels can better assess errors in water mass properties. For isotherms: Given T, S -> Calculate Z(T), S(T) for model and observations Z’(T) = Z b (T) + ΔZ(T) S’(T) = S b (T) + ΔS(T) For isopycnals: Given T, S -> Calculate Z(ρ), π(ρ) for model and observations Z’(ρ) = Z b (ρ) + ΔZ(ρ) π’(ρ) = π b (ρ) + Δπ(ρ) π = Spiciness
What is spiciness? Spice = c*(S/S o + T/T o ) S o = reference salinity T o = reference temperature Spice = ρ(T o,S) + ρ(T,S o ) High spice warm + salty Low spice cold + fresh ρ(T,S o ) ρ(T o,S) T S
# obs at 500m
Perform a series of experiments over the period 2003-present. Test Assimilation scheme with high number of Argo observations Study impact of Argo through data-witholding experiments Develop spice(rho) assimilation scheme and investigate possibility of using this method in the absence of salinity observations. Look at best way to assimilate Rapid Array data and its influence Development Plan:
Assimilation on density levels Density /kg m -3 Spiciness increment π (ρ) Density level depth z(ρ) before and after assimilation Can spread spiciness over larger spatial scales on ρ levels T(z), S(z) → z(ρ), π (ρ) → Assimilate → T(z), S(z) Spiciness /kg m -3 Results from FOAM – UK MetOffice
NEMO in GoogleEarth
Outline Variance along isotherms and isopycnals NEMO : plans and progress… ORCA1 47yr runs Generalized observation operator An example of assimilation on isopycnals NEMO in Google Earth
Relevant questions: Can we recover water mass variability and THC changes using data assimilation? How can we best use the available observations? What does the data assimilation tell us about model errors? Main aim: Produce high-resolution multi-decadal global ocean reanalysis using physically-based assimilation techniques. Use this reanalysis to study water mass variability and the effect on ocean climate signals.
How can we best make use of the available observations? Two types of variability: dynamic and that due to water mass changes Dynamic: high frequency short correlation scales Watermass: low frequency long correlation scales
One point correlation maps from HadCEM S(z)S(T)
NEMO: Plans and progress... Strategy: Use 1 o model for testing the assimilation scheme and to perform multiple sensitivity studies. Then transfer assimilation scheme to 1/4 o global model to produce high-resolution reanalysis. Current Status: 2°, 1 ° and 1/4 ° resolution global NEMO ice-ocean models set-up and running, both at ECMWF and ESSC cluster. Several 45 year hindcasts of 1° NEMO complete. Generalized observation operator has been implemented in NEMO Can now begin implementing assimilation system