1. Implementation of the quasi-normal scale elimination (QNSE) theory of turbulence in a weather prediction model HIRLAM Veniamin Perov¹, Boris Galperin² and Semion Sukoriansky³ 1.Swedish Meteorological and Hydrological Institute (SMHI), Norrköping, Sweden 2. College of Marine Science, University of South Florida, St. Petersburg, Florida, USA 3. Ben-Gurion University of the Negev, Beer-Sheva, Israel

2. Numerical weather prediction (NWP) models involve eddy viscosity and eddy diffusivity, Km and Kh, that account for unresolved turbulent mixing and diffusion. The most sophisticated turbulent closure models used today for NWP belong to the family of Reynolds stress models. These models are formulated for the physical space variables; they consider a hierarchy of turbulent correlations and employ a rational way of its truncation. In the process, unknown correlation are related to the known ones via “closure assumptions” that are based upon preservation of tensorial properties and the principle of invariant modelling

3. according to which the constants in the closure relationships are universal Although a great deal of progress has been achieved with Reynolds stress closure models over the years, these are still situations in which these models fail. The most difficult flows for the Reynolds stress modelling are those with anisotropy and waves because these processes are scale-dependent and cannot be included in the closure assumptions that pertain to ensemble-averages quantities. Here we employ an alternative approach of deriving expressions for Km and Kh using the spectral space presentation. The spectral model produces expressions for Km and Kh based upon a self-consistent procedure of small-scale modes elimination.

4. This procedure is based upon the quasi-Gaussian mapping of the velocity and temperature using the Langevun equations. Turbulence and waves are treated as one entity and the effect of the internal waves is easily identifiable. When averaging is extended to all scales, the method yields a Reynolds-averaged, Navier-Stokes based model. The details can be found in the paper “A quasinormal scale elimination model of turbulent flows with stable stratification”, Phys. Fluids, 17,085107-1-28, 2005, by Sukoriansky S, Galperin B. and Staroselsky I.

5. Results from the theory The turbulent coefficients are recast in terms of gradient Richardson number Ri = N 2 / S 2 or Froude number Fr =  / NK Normalized turbulent exchange coefficients as functions of Ri and Fr. For Ri>0.1, both vertical viscosity and diffusivity decrease, with the diffusivity decreasing faster than the viscosity (“residual” mixing due to effect of IGW?)  Horizontal mixing increases with Ri. The model accounts for flow anisotropy.  The crossover from neutral to stratified flow regime is replicated. No critical Ri.

6. Stability functions from the QNSE model and from the Mellor-Yamada model modified by Galperin et al. (1988)

7. Comparison with experimental data Vertical turbulent Prandtl number as a function of Ri. Data points are laboratory measurements by Huq and Stewart (2004) ; solid line represents our model’s results. Inverse Prandtl number k z / z as a function of Ri. Experimental data points are from Monti et al. (2002).

8. New K-  model : In our model Detering & Etlin, 1985 introduced correction to C 1 due to the Earth rotation We generalized it to include stratification: The constants are: The model is implemented in the 1D version of the weather forecast model HIRLAM where

9. Neutral ABL Comparison with Leipzig wind profile Comparison with Leipzig wind profile

10. Comparison with CASES-99

12. Comparison with BASE Comparison with BASE

14. Testing in the numerical weather prediction system HIRLAM  NWP system HIRLAM: High Resolution Limited Area Model  Covers the North-East Atlantic, Europe, and Greenland  Hydrostatic model; 438x336 points; 22km x 22km resolution  40 vertical levels  Lateral boundary conditions are from ECMWF operations  Massive data assimilation: over 1000 stations all over Europe  Data assimilation cycle is 6 hours  From each 00, 06, 12, 18 UTC, a +48 hours forecast is run  Total: 120, +48 h forecasts in one month (January 2005)  We replace K m and K h for stable stratification only; run parallel experiment; analyze the difference (new-reference)  Region of interest: Scandinavia

15. HIRLAM turbulence K-l scheme

16. Results (versions before 6.2) : -Positive bias in the wind direction, accompanied by too strong near surface winds - Too fast deepening and too slow filling of cyclones, making HIRLAM too active towards the end of the forecast period

17. HIRLAM turbulence K-l scheme Increasing of the vertical mixing of momentum under stable stratif. Verification score becames better, but Intercomparison in GABLS shows very deep BL and a wind profile has its maximum at the wrong place (too high)

18. Modification of HIRLAM K-l scheme

22. Surface sensible heat flux: New-Reference Red – positive Blue - negative

23. Surface latent heat flux: New-Reference

24. Surface U-momentum flux: New-Reference

25. Surface V-momentum flux: New-Reference

26. Cross-section: Difference in TKE

27. Cross-section: Difference in temperature

28. Conclusions Anisotropic turbulent viscosities and diffusivities are in good agreement with experimental data Anisotropic turbulent viscosities and diffusivities are in good agreement with experimental data The model recognizes the horizontal-vertical anisotropy introduces by stable stratification and provides expressions for the horizontal and vertical turbulent viscosities and diffusivities. This is a real possibility to include 3-D turbulence in NWP and mesoscale models The model recognizes the horizontal-vertical anisotropy introduces by stable stratification and provides expressions for the horizontal and vertical turbulent viscosities and diffusivities. This is a real possibility to include 3-D turbulence in NWP and mesoscale models Theory has been implemented in 1-D K-ε and K-l models of stratified ABL Theory has been implemented in 1-D K-ε and K-l models of stratified ABL Good agreement with BASE, SHEBA and CASES99 data sets has been found in 1-D model for cases of moderate and strong stratification Good agreement with BASE, SHEBA and CASES99 data sets has been found in 1-D model for cases of moderate and strong stratification Theory has been implemented in K-l scheme of 3-D NWP model HIRLAM Theory has been implemented in K-l scheme of 3-D NWP model HIRLAM The new K-l scheme improves predictive skills of mean sea level pressure and 2M temperature for +48h weather forecasts over Scandinavia (stable BL) for January 2005 The new K-l scheme improves predictive skills of mean sea level pressure and 2M temperature for +48h weather forecasts over Scandinavia (stable BL) for January 2005