1. Measurements and (preliminary) Modeling of Turbulent properties in the Adriatic Sea Sandro Carniel – Mauro Sclavo (CNR-ISMAR, Venice) Lakshmi Kantha (Univ. of Boulder, CO, USA) Hartmut Prandke (ISW, Germany) Jacopo Chiggiato (ARPA-SIM, Bologna) ROMS-TOMS European Workshop, November 6-8, 2006

2. Sub-grid Scale parameterization …WHY?... important to air-sea exchange, weather, climate, biolog. productivity, oil-spill tracking, counter-mine warfare, S&R etc. Mixing in the upper ocean is primarily surface-driven: 1. Momentum flux from winds and waves 2. Negative buoyancy flux due to cooling and evap. Below the active ML: Shear instabilities, Double-diffusion…HOW?... 1) Acquiring microstructure measurements via turbulence profiler 2) Starting from this picture of turbulent parameters, to test and accordingly modify SCM employed in 3-D hydrodynamical models ONR NICOP Grant (N00014-05-1-0759)

3. DART region 2006: measuring turbulent properties, during the period of DART06 A & B (March and August) 2007: - assessment Turbulence parameterization in 3-D ocean models - other measurements (June) 2008: refinement of parameterization and delivery ONR NICOP and DART Project

4. Location and Instruments

5. Free-falling profiler 2 velocity microstr. Shear sensors 1 microstructure Temp. sensor standard CTD sensors for prec. measurements turbidity (light scattering) sensor vibration control sensor (ACC) surface detection sensor 1024 s.p.s., 16 bit

6. MSS Profiler 1.The microstructure sensors are placed at the tip of a slim shaft, about 150 mm in front of the CTD sensors. 2.Shear sensor is a piezoceramic beam: if V~sinking velocity, G is sensor gain, E s ~signal, du/dz ~ (  GV 2 ) -1 dE s /dt TKE dissipation rate 3.TKE Dissipation Rate 1.The microstructure sensors are placed at the tip of a slim shaft, about 150 mm in front of the CTD sensors. 2.Shear sensor is a piezoceramic beam: if V~sinking velocity, G is sensor gain, E s ~signal, du/dz ~ (  GV 2 ) -1 dE s /dt TKE dissipation rate 3.TKE Dissipation Rate

7. Dissipation rate for isotropic turbulence: ( water kinemat. viscosity) i.o.  =   u i /  x j  (  u i /  x j +  u j /  x i ) Dissipation rate of microstruct. Temperature variance: (K T molecular diff. for heat in water) Eddy Diffusivity from Diss. Rate: Eddy Diffusivity of Heat: Useful Info from Measurements…

8. MSS profiler - Example of measurements

12. Meteo Conditions during DART06-B (August 2006) Red= air T Blue= SST March 2006: 252 profiles August 2006: More than 300 profiles, 160 of them at B90 site in 5 days, divided into 5 O.P.

13. Heat & Buoyancy Fluxes, DART06-B Estimated errors Bulk formulae VS measured turbulent fluxes are easily up to 40%

14. OP-1 (39 casts, 13.09-15.26 UTC) Weak surface forcings case Weak w, 4.9 m/s Wstar= 1.2 cm/s D=13m > MO=5m L T =0.5 m (RMS of Thorpe displ., length scale of turb. oveturns, i.e. weak mixing) N freq. indicates strong pycnocline T variance diss. rate U * =(  /  ) 1/2 W * =(J b0  D) 1/3 MO=U * 3 /(  J b0 ) R=(W * /U * ) 3

15. OP-2 (24 casts, 17.03-18.48 UTC) Wind-driven case Stronger w, 11 m/s Cooling, -150 W/m 2 Wstar= 1. cm/s Ustar=1.3 cm/s (R=0.4, i.e. wind driven) D=13m < MO=60m (large, indicating wind driven case) L T =up to 2 m d U * =(  /  ) 1/2 W * =(J b0  D) 1/3 MO=U * 3 /(  J b0 ) R=(W * /U * ) 3

16. OP-3 (40 casts, 23.29-02.17 UTC) Convection-driven case Weak w, 3.4 m/s Clear sky, high cooling, -200 W/m 2 Wstar=1.1 cm/s Ustar=0.4 cm/s (R=20, convection) D=20m > MO=2.6m (buoyancy dominated ) L T =3 m (strong convection mixing) d U * =(  /  ) 1/2 W * =(J b0  D) 1/3 MO=U * 3 /(  J b0 ) R=(W * /U * ) 3

17. March 2006 – late winter (DART06-A) Weak wind (4 m/s) Layered density structure. Double diffusivities convection from cold fresh water masses over warm salty ones?

18. Turbulence Scaling: Dissipation rates  =  c +  s,z  D =  i,z >D …better scaling in the interior good scaling in the ML...  c =0.58 J b0  s =1.76 u * 3 /  z  =  c +  s Below the ML,  i =0.03 L 2 T N 3

19. Original resolution: 15 arc seconds (1/240°). Source: data collected during the project ADRIA 02-03 with various contributions from CNR-ISMAR Bologna, CNR-ISMAR Venice, HHI Split, IIM Genova, IRB Zagreb, NIB Piran, NURC La Spezia. Grid 160 x 60 w/ variable resolution ~ 3 Km (north) ~ 10 Km (south) 20 levels

20. LAMI wind 10 m mean sea level pressure air temperature 2 m dew temperature 2 m total cloud cover net short-wave radiation Mediterranean GCM (OPA-MFSTEP) daily forecasted temperature and salinity + tidal elevation and currents (M2, S2, O1, K1) from QUODDY model 48 rivers and springs

21. -Resolution 6x6 km or 2x2 km (running) 2x2 km (running) - Restart file from operational version ARPA-SIM 6km -TCMs:  GLS as “gen”  GLS as “gen”  GLS as “k-kl”  GLS as “k-kl” -Wave-breaking on (6km)/off (2km) on (6km)/off (2km) -Radiation Stresses on (one-way, 6km) via SWAN run ROMS turbulence modeling: March

22. Validation (CTDs Urania) RMSE suggest: Similar behaviour as for CTD-Alliance in the SAd and MAd (actually the temperature is a little bit colder compared to observations) Low errors for temperature in the NAd, salinity fresher by 0.5 PSU Major errors are located along the WACC

23. RMSE suggest: Southern Adriatic: AdriaROMS is colder by 1°C and fresher by 0.3 PSU. The two bias cancel each other, and resulting sigma-t is nearly unbiased. Middle Adriatic: AdriaROMS has a negligible temperature bias but is still fresher by 0.2÷0.3 PSU; the resulting sigma-t is now slightly biased Validation (CTDs Alliance)

24. ROMS turbulence modeling: March 2006 6x6 km 2x2 km CB wave- breaking onRadiationstresses on via SWAN run NO CB wave- breaking NO radiation stresses

25. ROMS turbulence modeling: March 2006 5x5 km 2x2 km 6x6 km 2x2 km NO CB Wave- breaking NO Radiation stresses CB wave- breaking onRadiationstresses on via SWAN run

26. ROMS turbulence modeling: March 2006 6x6 km CB wave- breaking onRadiationstresses on via SWAN run

27. ROMS turbulence modeling: March 2006 2x2 km NO CB wave- breaking NO Radiation stresses

28. Considerations 1.Dissipation measurements and data processing have to be carried out carefully to avoid falsification resulting from pseudo shear, sensor bottom hits, high particle conc., strong shear layers and pycnoclines (change of profiler sinking leads to falsified dissipation rates) 2.Despite this, modern turbulence profilers enable routinely TKE dissipation measurements in marine environment and useful diffusivity estimations 3.Need of longer „time-series“, i.e. repeated measurements in the same spot (intermittency…) combined with other info (shear, meteo…)

29. Considerations 4. 2-eqs TCMs are now integral part of ocean models (POM, ROMS, NCOM, ICOM?), but experience gained over past 2 decades indicates that these models can be made more skillful. OF COURSE, influence of W-B, radiation stress, horiz. resolution, proper initialization to be investigated!!! We need to have a correct vertical structure, first. 5. However, outstanding issues are: a) Better performance under free convection b) Inclusion of surface wave effects on mixing in the OML

30. Considerations 6.Surface Wave Effects have not been included properly in OML models until recently. Two kinds of effects: a) TKE injection at the surface into the water column by wave breaking – has a surface effect (Umlauf, JSR 2003; Kantha, OM 2004). ROMS is following one approach (Warner GLS), but… b) Stokes production of TKE by the interaction of waves and turb. in the OML, can enhance turb. in the interior 7. Langmuir cells (wind driven shear+Srokes drift) can also produce strong vertical velocities in the OML 8. simple non-local models (counter-gradient term) need to be constructed for free convection situations. 9. Can we possibly think of “assimilating” these typology of measurements?

31. TKE prod. by LC andwhere Adding additional production terms…

32. #define craig_banner #define tke_wavediss Surface TKE fluxes Two formulations to account for surface injection of TKE due to breaking waves. For GLS each formulation requires boundary conditions for k and . ~ 100; = surface stress …how get Zos ? #define charnok #define zo_hsig  ~ 0.25 = wave energy dissipation a = 1400 a = 0.5; Hs = significant wave height

33. End of Presentation

34. Conclusions 1.Modern turbulence profilers enable routinely TKE dissipation measurements in marine environment 2.Dissipation measurements and data processing have to be carried out carefully to avoid falsification resulting from pseudo shear 3.Dissipation measurements in coastal waters require special attention due to sensor bottom hits, high particle conc. (falsified shears), strong shear layers and pycnoclines (change of profiler sinking leads to falsified dissipation rates) 4.Need of longer „time-series“, i.e. repeated measurements in the same spot (intermittency…) combined with other info (shear, meteo…)

35. Theoretical Dissipation rates In the convective layer:In the wind-stress driven ML: Assuming L T proportional to Ozmidov scale For turbulence generated by momentum flux + destabilizing buoyancy flux:  =  c +  s,z  D =  i,z >D Within ML Below ML A length scale for the description of turbulent flows under stable stratification, defined as… In flows where turbulence and wave motion are simultaneously present, the inverse of the Ozmidov scale defines the buoyancy wavenumber, which separates the buoyancy subrange from the inertial subrange.

36. Friction & Convective Velocity scale U * =(  /  ) 1/2 Wstar=(J b0  D) 1/3 MO=U * 3 /(  J b0 ) R=(Wstar/U * ) 3

37. ADCP velocities

38. OP-6 (6 casts, 14.45-15.13 UTC) -Stronger wind (7 m/s) -Insolation decreased to 300 W/m 2 Ustar=0.95 cm/s D=18m, MO=21m, L T =3 m (strong mixing above pycnocline)

39. Shear probe data processing From the lift force at the airfoil caused by potential flow (Allen and Perkins, 1952): F = 1/2   U 2 A  sin2  (   15  ) Definition of S (F  E)  E =  U 2 S  sin2  rms. measurement for S  E =  2  U 2 S rms  sin2  sin 2  = 2sin  cos   E = 2  2  Vu S rms Differentiation, gain  du/dt = (2  2  VG S rms ) -1  dE/dt du/dt = V  du/dz  du/dz = (2  2  V 2 G S rms ) -1  dE/dt TKE dissipation rate definition:  =   u i /  x j  (  u i /  x j +  u j /  x i ) … and assuming isotropic turbulence:  = 7.5   (du/dz) 2 …from the voltage output of the probe we then obtain an estimate of the TKE diss. rate

40. Where to improve/What to include 1.Wave breaking effects were ignored on Epsilon scaling (upper 2-3 m are not covered by measurements) 2. Langmuir circulation, Stokes production 3. Wave observation (Wave rider, buoy) 4. Measurements of turbulent fluxes at sea VS bulk fluxes 5. Measurements of shear (ADCP) and of the broad context around OP