Separation

1. Along the aim of COSMO-SP to consider missing interactions: • Mass fraction and number concentration of • cloud constituents, • precipitation and • passive tracers (aerosols) • Optical properties: • optical thickness, • single scattering albedo, • asymmetry factor and • delta-transmission function Increased set of diagnostic or prognostic variables: modified by SGS-processes feed-back with SGS-processes PP T2(RC)2 Radiation transport Cloud-microphysics Local parameterizations: Parameterizations of source terms integrated in GS parameterizations: Parameterizations of SGS processes interaction STIC Turbulence Separation Circulations PT ConSAT Surface and Soil-processes

2. The coarse resolution extension of TURBDIFF: STIC: Separated Turbulence Interacting with non-turbulent Circulations • Application of turbulence approximations only to small SGS scales • separationof the sub grid scale flow in different classes with specific closure assumptions • by application of associated filter scales • turbulent budgets with additional production terms due to shear terms with respect to the separated sub gird scale circulation flow of • wake vortices by SSO (sub grid scale orography) blocking or gravity wave breaking • [operational in COSMO and ICON] • large separated horizontal shear vortices [operational in ICON] • surface induced density flow patterns [operationalin COSMO and ICON] • shallow and deep convection patterns [not yet operational active] production terms dependent on: specificlength scales and a specificvelocity scale(= ) production terms depend on: single turbulentlength scale and the turbulentvelocity scale (= ) and other moments of and other moments of circulation-scale turbulence-scale is the scale interaction term shifting SKE form the circulation part of the spectrum (CKE) to the turbulent part (TKE) by virtue of shear generated by the circulation flow patterns.

3. Priority Task “Consolidation of the Surface-to-Atmosphere Transfer-Phase 4: (ConSAT4) according to a dynamically adapted list of actions being the base also for future PTs Geneve 2017 COSMO Matthias Raschendorfer

4. General Task: • Implementing the next steps of consolidating The combined schemes for • surface-to-atmosphere transfer SAT including the calculation of near surface variables • model for moist turbulent statisticsinteracting with the statistics of non-turbulent SGS statistics <-> STIC • the generalized code for vertical diffusion • the heat budget of the roughness layer and its obstacles including the description of evapotranspiration • Improving the overall model performance and particularly the daily and annual cycles of near surface model variables by an iterative cycle of the following general actions: • Diagnostics by component test in order to detect potentials of improvement • Activation of already implemented options • Implementation of further extensions and corrections along a dynamically adapted action list, which is according to the strategy of the current COSMO science plan content of module TURBTRAN content of module TURBDIFF (small) part of TERRA- code

5. Current state with the common TURBDIFF code: • Common blocked COSMO/ICON TURBDIFF code implemented into COSMO • Including a couple of extensions and generalizations with regard to SAT and the turbulence model • Including a generalized formulation of vertical diffusion applicable • for all full- and half-level variables • for gradient- and non-gradient fluxes • Verified in several parallel experiments with COSMO-DE (including also the blocked version of vertical diffusion • called (with the exception of statistical moments) at the end of the physics section • called for each staggered horizontal wind component (in contrast to ICON) including wind components, all active and passive scalars and tracers, as well as statistical moments

6. Revision of the combined turbulence- and SAT-scheme TURBDIFF for ICON and COSMO: Matthias Raschendorfer, Günther Zängl (DWD) • In block-data structure and with stronger modularization • Generalized semi-implicit vertical diffusion (also for non-gradient fluxes) • One additional STIC-term active (due to separated SGS horizontal shear circulations) • Increased shear –production of TKE by STIC-terms also considered definition of Ri-number • Less restrictive prevention of possible singularities • Application of some first empirical hyper-parameterizations (by Günther Zängl): Some parameters of the scheme formulated as a function of model-state variables: e.g. ”minimal diff-coeff” = fnc ( “Ri-number”, “height above ground” ) • Complete moist physics applied to surface level (including turbulent cloud diagnostics) • Near-surface interpolation of vertical profiles in conserved variables • Zero-concentration condition for qi and qc at the surface (e.g. impaction of fog) Geneve 2017 Matthias Raschendorfer

7. Upper air verification with COSMO-DE for June 2016: revised TURBDIFF Reference = Routine geopotential [m] geopotential [m] rel. humidity [%] rel. humidity [%] -------------------- Mean Error ------------------- ------------------------- RMSE ----------------------- temperature[K] temperature[K]

8. Related new Tasks: Regarded to the general design of the common turbulence code the following actions need to be transferred from the previous phase: WP I: Shifting all remaining surface layer calculations from the atmospheric part into the surface layer part[in order to allow for a better application of surface tiling, makes the code more clean]. Ra: 0.05 FTE WP II: Implementing optional metric correction terms into common blocked routine for vertical diffusion [only required for running 3D-turbulence with this code, not really urgent]Ra: 0.05 FTE Ba: 0.05 FTE Further, some development related to the moist extension is in the queue (since long), promising a more consistent design of turbulent diffusion for the different sets of model equations of COSMO and ICON and of the cloudy boundary layer in general: WP III: Preparing diffusion of the conserved variables (total water content and total water potential temperature) and transforming the tendencies into those needed in the respective models (ICON or COSMO) by means of saturation adjustment [desired first of all for ICON, in preparation]Ra: 0.05 FTE WP IV: Including ice-processes by introducing a mixed water- and ice-cloud into the turbulence code [desired also for T2(RC)2] Ra: 0.05 FTE Due to the expressed urgent requirements, we also include: WP V: documentation of the whole package [management requirement, almost ready for TURBTRAN]Ra: 0.1 FTE

9. Current state with the diagnostics of SAT for stable stratification: • SC component tests for investigation of the stability dependency of the Prandtl-layer part of atmospheric transfer resistances for scalars : • Combined approach for SAT(employing the STIC turbulence model rather than MOS) offers the possibility to include additional processes of the stable boundaryinto the SAT formulation • Currently some interaction termsare missing and substituted either by crude approximations (current thermal circulation term) or by simple numerical security constraints (minimal diffusion coefficients) • too much nocturnal near surface mixing for grid cells with a homogeneous surface • the consideration of a special interpolation function of the transfer layer turbulent velocity profile (TTVP) for stable stratification (necessary for being consistent with the utilized turbulence scheme) has almost no effect due to the additional mixing • The combined SAT-scheme of TURBDIFF is reducible to the well established solution for a homogeneous surface, if • these so far non-adaptive (on average beneficial but for a homogeneous surface unphysical) constraints are switched off • and the special TTVP for stable stratification is applied • and a so far implemented restriction of a transfer layer stability parameter is reduced

10. Current state with the diagnostics of SAT for stable stratification: Surface Layer formulation for the models COSMO and ICON: Intercomparison with Monin-Obukhov based schemes for stably stratified conditions being submitted to JAMC • starting point for full documentation Ines Cerenzia, Matthias Raschendorfer • homogeneous surface conditions disabling all additional mixing in TURBDIFF lin. TTVP enabled branch hyp. TTVP atmospheric height transfer layer turbulent velocity profile (TTVP) and the profgile parameter

11. Already implemented first consequences: • Switchabaloptional stable branch of the TTVP (with hyperbolic profile) • Ri-number dependent modulation of the crucial minimal turbulent diffusion coefficients ‘tkhmin’ and ‘tkmmin’ [has already been done within ICON] • Modulation of the previously constant parameter ‘pat_len’ of the current thermal circulation term in terms of the standard deviation of SSO height[has already been done within ICON]

12. Related new Tasks: WP VI: testing and varying the new extensions together with an activation of the since long available Deardorff-limit of the turbulent length scale in combination with the activation of the stable branch of TTVP and the relaxed restrictions for the transfer layer profile parameter [promising natural testing step as a result of previous phase]. NN: 0.05 FTE WP VII: trying to find an even better adapted statistical parameterization for the minimal diffusion coefficients[natural and promising consequence but possibly a more elaborate methodology of setting up statistical parameterizations necessary <-> CALMO] NN: 0.15 FTE WP VIII: trying to reformulate the ‘pat_len’-dependent STIC-term (the near surface “circulation term”) to be better adapted to density flows along SGS slopes [natural and promising consequence form last phase, in preparation]Ra: 0.10 FTE WP IX: investigating the effect of thermal surface patterns that would results fromimplemented surface tiles by means of idealized LES simulations with the aim of later describing a related SGS thermal circulations by an adapted parameterization [basic investigation of the significance of not yet properly represented SGS process; part of Ines PhD]Ce: 0.20 FTE The success of performed SC component tests (with regard to the stability dependency of the Prandtl-layer resistances) and recent diagnostics of the diurnal cycle of T2m/Td2m encourages us to the following next step: WP X: completion of the vertically not resolved roughness layer representation (based on a generalized BLA) which has resulted in a SAI-dependent resistance formulation (substituting a specific roughness length for scalars): Introduction of the roughness layer heat budget by linearization of LRF and LHF and investigation of related (global and local) parameters (rlam_heat, rat_sea, c_land, c_soil, LAI e_surf, dep_d, dep_v) [important and very promising step of consolidation, first version implemented in non-blocked TERRA]Ra: 0.15

13. In future also supported by Darius in Turino

14. Idealized simulation of the stable BL above thermal surface patterns: decreasing surface temperatures temperature profiles after 10 hours warm tile +3K Height [m] T_S [K] -3K cold tile Run-time [h] Temperature [K] Results: Experimental setup: • patterns of T_S even sharpen decoupling of surface • already represented by tiling • non-linearity of “flux=K*grad(T)” dominates against extra mixing by thermal circulations • Tcirc is a thermal SSO-term and (in contrast) produces MORE mixing. It should not be active at flat surfaces! • simulation with COSMO-Single-Column • representing column above a 400X400m2 idealized flat surface • with 100X100m2 checker board elements (of ∆T_S=6K after 8h) • using TURBDIFF (with): • surface tiling (2Tiles), • a STIC-term for thermal driven near surface circulations (Tcirc) • progn. scalar variances (SV), (with) Var(T_S) as lower BC (SBC) • COSMO-LES: 3.125 m horizontal resolution

15. Towards a new prognostic equilibrium surface temperature in combination with SAT and the soil model: • Completion of the roughness layer model (TERRA-part) • Thermal decoupling of a cover built by roughness elements (canopy) above the dense soil (shading effect) • Representation of the thermal energy storage of a semi-permeable roughness layer with a compromise of complexity • Linearized energy budget of cover implicitly coupled with heat conductance equation of the soil • stronger numerical stability • possibly without a flux limiter • To be implemented into blocked ICON-version of TERRA Matthias Raschendorfer Geneve 2017 Matthias Raschendorfer

16. Case study: 23.06.2016 COSMO-DE with lateral boundaries from ICON-EU • only for rather smooth surfaces; applied filter • almost saturated soil due to long standing rain period before • almost no clouds due to high pressure situation; + applied filter TD_2m T_2m still much too moist in the afternoon still much too cold during day-time nocturnal warm bias removed but perhaps a new nocturnal dry bias direct analysis of T_2m and TD_2m operational configuration revised TURBDIFF imported from ICON Geneve 2017 Matthias Raschendorfer

17. Case study: 23.06.2016 COSMO-DE with lateral boundaries from ICON-EU • only for rather smooth surfaces; applied filter • almost saturated soil due to long standing rain period before • almost no clouds due to high pressure situation; + applied filter TD_2m T_2m removed cold bias in the afternoon! almost perfect double wave! amplitude slightly overestimated still some problems during heating of the cover (stable stratification) revised TURBDIFF imported from ICON + new decoupled surface cover: revised TURBDIFF imported from ICON direct analysis of T_2m and TD_2m Geneve 2017 Matthias Raschendorfer

18. ncover layers including the surface of the dense soil (n=0) are connected by long-wave radiation interaction and sensible heat exchange thermally decoupled roughness elements (shading) • Only a part of the inner surfaces is connected to A by the resistance chain for SHF and LHF, the other part is for the inter- surface exchange strongly effects the LAI-impact of transpiration! functions of SAI only Geneve 2017 Matthias Raschendorfer

19. Related work: Jan-Peter Schulz: Testing the general effect using observations at MOL and an easy extension of TERRA (only a few lines of code): • Virtual skin covering the whole surface • No discrimination between TB and TC and no adaptation of LAI-impact • No permeability and no heat storage of cover • No implicit coupling to heat conduction of soil • Similar to a previous SC test-implementation already discussed in an early CUS by MR Jürgen Helmert: Implementing an alternative approach from HIRLAM for comparison: • No heat storage of cover • No implicit coupling to heat conduction of soil • No adaptation of LAI-impact • Resistances not easy to substitute by TURBTRAN concept Matthias Raschendorfer: Developing the full roughness layer effect as an extension of TURBRAN and TERRA • Includes the functionality of the test/alternative approaches as a special case

20. Testing & Tuning of Revised Cloud Radiation Coupling T2(RC)2PP: Status & Highlights from Workshop in Israel Project leader: Harel Muskatel (IMS) Ulrich Blahak (DWD), Matthias Raschendorfer (DWD), Pavel Khain (IMS), Alon Shtivelman (IMS), Oliver Fuhrer (MCH),Xavier Lapillonne (MCH), Gdaly Rivin (RHM), Natalia Chubarova (RHM), Marina Shatunova (RHM), Alexey Poliukhov (RHM)

21. Revised parameterization of optical ice-cloud properties: • In the COSMO radiation scheme (Ritter & Geleyn 1992) • Optical properties of ice-clouds are described crudely and don’t include precipitation products extinction coeff. , single scattering albedo, asymmetry factorg , delta-transmission factor fd • Effect of inhomogeneity is taken into account by means of a constant reduction factor radqcfact=0.5applied to the mass fractions. • New parameterizations of optical propertiesbased on idealized calculations according to Fu • Visible-bands: Ray-tracing for randomly orientated hexagonal ice particles (Fu 2007) • IR-bands: weighted average of Mie-scattering and related methods (Fu et al. 1998) • Optical propertiesare treated as functions ofeffective arguments • effective radius Reff • aspect ratioAR and wave-length • Argumentsdeduced from inherent assumptions in terms of particle size distribution N(L), mass size relationm(L) and particle shapes (expressed by L and D): • Suitable also for complicated ice particles (bullet rosettes, aggregates with rough surfaces, fractional crystals) • Extending particle size range (5μm - 300 μm)by using 7000 size-modified Generalized Gamma-DistributionsN(L) • Fitting the calculated relations between optical properties and effective arguments • In terms of rational functions using spectral averaging for 8 distinctive spectral bands

22. Extended effective -Radius Reff calculation for water clouds based on an aerosol-climatology: Currently: grid scale clouds: Reffis a function mainly of cloud mass fractionqC sub-grid clouds: using a fixed Reff as a tuning parameter New:using … • Tegen-climatology: (Tegen et al., 1997), later: prognostic aerosols (COSMO-ART) • Optical thickness for 5 aerosol categories: • sea-salt, mineral dust, black carbon, organics • Assumed specific extinction coefficients • Assumed mean particle radius and density • Assumed exponential vertical decrease => grid-column-integrated aerosol-mass per m-2 => aerosol number concentration NCN(z) in m-3 convective velocity scale • cloud-activation parameterization according to Segal/Khain (2006): • aerosol number concentration NCN • effective updraft wind speed: =>effective radius: • height of cloud base • characteristic cloud depth • c1 and c2: tuning parameters • valid for cloud water qc composed by all scales • (activ.) cloud number concentration

23. Problem: New cloud-radiation scheme depends on 30 parameters! • Perform idealized COSMO simulations • for many parameter combinations • and special cloud types • Replace attenuation of radiation (in %) by an analytic function (quadratic form) of the parameters (meta-model): scaled dimensionless parameters ranging from -1 to 1 • most sensitive parameters: automatic tuning by CALMO • Calculate sensitivity of each parameter : • less sensitive parameters : “expert-tuning” • List of 8 most sensitive parameters : Real numbers: Selectors: • Assumed number concentration of cloud droplets • Properties of sub-grid water clouds • Representation of sub-grid variability • Operational / new scheme • Include rain, snow & graupel

24. Task 1 - Testing and tuning of the new cloud-radiation scheme: Done: Sensitivity analysis of ~30 new parameters of the radiation scheme has been performed using the COSMO idealized frame work for six types of clouds. A new CALMO-type methodology was developed to perform a fast sensitivity check (PK) Next steps: Real case studies, expert tuning – reducing number of user-defined parameters, upgrading new radiation code to blocking structure and to GPU capable format Task 2- Analysis/Revision of SGS cloudiness in the radiation scheme: Done: Learning the current schemes and discussions (MR, UB, HM, PK), discussion with Prof. Alexander Khain about different possibilities in using SBM+SAM 3D and HUCM-2D codes to simulate different cloud types for a more accurate parametrization of SGS clouds Next steps: Adaptation of the parametrization of effective radius from grid-scale clouds scheme to the SGS clouds (currently constant), revision of the parametrization of LWC & IWC in the alternative SGS clouds schemes, including mixed phase parametrization (RHg) in the so-called “statistical scheme” of SGS clouds (MR, UB) Task 3- MACC aerosol fields instead of climatology in the radiation scheme: Done: Revision of INT2LM code that now digest the MACC files and linearly interpolates the 11 aerosols 3D fields. The new code is written in a way that more tracers could be easily added in the future (UB) Next steps: Preparing a new look-up table of aerosols optical properties adapted from ECMWF and adjusted to COSMO wavelength bands (HM), implementation of the new diagnostic aerosols fields in the radiation scheme (UB, HM, PK). Summary

25. Task 4a - Adapting switchable single/double precision to the radiation scheme: Done: AS, PK, UB, OF and XL met in a workshop in Zurich for an introduction on GPU computing as well as on single/double precision in the radiation scheme. The perturbation module was successfully implemented at IMS and first results regarding SP/DP were obtained Next steps: Figuring out the problematic parts of the radiation schemes when switching to SP Task 4b- Testing the temporal resolution call time of radiation scheme: Done: COSMO 7-km experiments with partial CLC and high WS performed using different temporal resolutions of radiation call-time. The results are compared with a benchmark run for which the radiation is called every COSMO time step (AS), The T2M, CLC, fluxes RMSE was compared. Next steps: Repeat with other weather conditions, repeat for COSMO 2.8km, checking localized biases, checking CPU costs (AS). Task 5- Implementation and testing of the MCSI method: Preliminary studies were done including learning about the work previously done by Bodo Ritter (HM) Summary

26. Task 6- Testing the new radiation code against observation including new MACC/Kinne aerosol: Done: • Clear-sky studies comparisons performed including: observational data (from Moscow State university), CLIRAD model COSMO with both Tanre and Tegenclimatologies (NC) • Comparison of Kinne and Tegenclimatolgies against AERONET observations (NC) • New Kinne climatology (MACv2 model) data file made ready for implementation in ExtPar, comparison (NC, MS) • Implementation of Kinne climatology aerosols fields in INT2LM code (UB) Next steps: • Implementation of Kinne climatology in radiation code (UB, MS) • Making of the ExtPar data files (MS) • Clear-skies studies of all four aerosols schemes on one frame work: Tanre, Tegen, Kinne& MACC against observational data and CLIRAD model (MC, MS) • Checking the possibilities of other observational locations i.e. Israel or Germany (Lindenberg obs. ?) (MC, MS) Summary

27. Mixed water/ice phase in the SGS condensation scheme and the moist turbulence scheme • The moist turbulence scheme is combined with a statistical SGS condensation scheme for non precipitating clouds heat release of SGS freezing does not effect turbulence (not a dominating effect) excluding the ice phase • We use a different scheme of SGS condensation for radiation and general diagnosis of fractional cloud cover - based on relative humidity This should be substituted by statistical SGS condensation scheme Common tuning of parameters in radiation and the statistical SGS condensation Cloud ice needs to be included Solution: Introducing a mixed water/ice phase - into the statistical condensation scheme similar to the procedure in the scheme based on relative humidity - into the moist turbulence scheme for consistency and for a higher level of generalization cloud water (liquid and ice) mixed phase saturation humidity icing factor mixed phase condensation heat

28. turbulent kinetic energy [m^2/s^2] Lon -5 5.5 Lat -5 6.5 Effect of SGS release of icing heat DWD COSMO Geneve 2017 Matthias Raschendorfer

29. Boundary layer perturbations for convection triggering in COSMO-DE Ulrich Blahak2 (DWD), Kirstin Kober1 (LMU) 1 Original inventor 2 Implementation and testing at DWD Continued by a PhD-student at LMU DWD EWGLAM/SRNWP Rome 2016 COSMO Geneve 2017 Matthias Raschendorfer Matthias Raschendorfer

30. The method • perturbed physical tendencies within ABL dependent on turbulent variances • aiming to reflect unresolved effects in variability of surface fluxes on convection initiation tuning-parameter • from TRUBDIFF • modif. By U. Blahak: • 0, above BLH • maximal BL-value below • 2D-field of Gussian random numbers, folded by   a Gaussian smoothing kernel to generate spatially coherent patterns. • It is updated every 10 minutes (approximate eddy turnover time in shallow  convection). • The spatial width of the kernel is specified as a multiple of horizontal grid length . • According to Teixeira and Reynolds (2004), described in Kober, Craig et al. (2015) • For : • temperature • specific humidity • vertical velocity DD COSMO Geneve 2017 Matthias Raschendorfer

31. T2m – TD2m COSMO-DE experiments with EDA driven by ICON-EU for August 2015: with stoch. BL-Pert reference • strong dry BIAS • due to overdrawn precipitation already during assimilation? 1 Tot. Prec. Rate + stoch. BL-Pert COSMO-DE-KENDA experiment (Axel Seifert): Tot. Prec. Rate (reference) domain averaged precipitation-rate vs. obs. ens. mean fisrt guess RMSE ens. spread domain averaged precipitation rate vs. obs. • strong spin-down for precipitation • improved spread/RMSE – ratio for 1h assimilation cycle • not yet effective for triggering more convective precipitation!