PP UTCS
Final Status Report
Dmitrii Mironov
German Weather Service, Offenbach am Main, Germany
dmitrii.mironov@dwd.de
13 September 2012
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
Outline
 Overview of project results
 Co-operation with external partners
 Summary and immediate plans (finalizing the
project)
 A look into the future
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
Task 2.1: Component testing of COSMO turbulence
scheme, revision of surface-layer formulations,
COSMO-SC (Matthias Raschendorfer)
Results
• COSMO-SC version tailored for COSMO-4.24 is
developed and is ready for component testing
• The new COSMO-SC version includes (i) several bug
fixes, (ii) extensions to allow forcing with measured
surface fluxes, (iii) extensions to allow forcing the soil
moisture calculation with measured evaporation, and (iv)
implicit vertical diffusion
• Some component tests are performed with COSMO-SC to
facilitate the current model development
Task 2.1 (cont’d)
Work in progress
• A revised formulation of the transfer resistance in the surface-toatmosphere transfer scheme is developed (not yet implemented)
• The code structure is modified to ease the implementation of
new formulations
• An alternative profile function for stable stratification is
implemented in the current test version as an option
See the COSMO web site for further details
Future plans (new PT, deliverables, deadlines, etc.), c/o
Matthias Raschendorfer
Task 2.2: Development and testing of non-local
turbulence length scale formulation (Veniamin
Perov, Mikhail Chumakov)
Results
• A non-local parameterization of the turbulence length scale with due
regard for the effect on phase changes on the scale of turbulence is
developed (cf. Bougeault and Lacarrere 1989)
• The new parameterization is included into the module TURBDIFF of
COSMO
• Numerical experiments show a satisfactory performance of the new
length-scale formulation; overall, turbulence and buoyancy fluxes are
more intense with (“moist”) mixing length
• Numerical experiments are performed with 3D COSMO model for
summer days, results are compared with the reference version of COSMO
and with the radiosonde and profiler data
• Use of non-local turbulence lengths scale formulation improves COSMO
performance, e.g. with respect to near-surface temperature and humidity
Task 2.2: Results
2500
Vertical profiles of Potential
Temperature for Dolgoprudnyj,
12h UTC, 17.07.09.
Vertical profiles of temperature in the
boundary layer,
Moscow, 12 h, 170711
700
2000
Measurements
from Profiler
MPT-5
600
1500
COSM
O- REF
1000
Зонд
Height, м
Height, м
500
COSMO, local
length scale L
400
300
200
500
100
0
0
298
299
300
301
302
303
Potential Temperature, К
304
17 18 19 20 21 22 23 24 25 26
Temperature, grad С
Vertical profiles of potential temperature computed with the current (blue)
and new (red) formulations for the turbulence length scale. Observational
data are shown with the black curve.
Task 2.3: Development of a unified turbulence
scheme for COSMO, COSMO-SC, and ICON
(Matthias Raschendorfer)
Results, work in progress
Numerous modifications are made to the current COSMO-model turbulence
scheme (all modifications are in the private test version c/o MR, some
modifications are already a part of the official COSMO code):
• bug fixes, stronger modularization and adaptation for use within both COSMO
and ICON;
• modified positive-definite solution of the TKE equation and the equations for
stability functions (a SUBROUTIEN called by ‘turbdiff’ and ‘turbtran’), relaxed
restrictions required to guarantee realizability;
• updated moist physics (a SUBROUTINE called by ‘turbdiff’, ‘turbtran’, and near
surface diagnostics), optional extension to allow for mixed water-ice phase;
• new generalized scheme for semi-implicit vertical diffusion (called for each
diffused variable, incl. TKE and passive tracers, without code replication),
options for explicit treatment of tendencies, gradient corrections, and lower
boundary condition.
Task 2.3 (cont’d)
Results, work in progress
• Explicit non-gradient flux corrections are introduced (e.g. SGS
condensation correction and the circulation term are now
expressed as a gradient correction within the new vertical
diffusion scheme)
• Optional stability correction of turbulent master length scale is
implemented (already in official code)
• A concept of scale separation is developed, three optional scaleseparation terms related to SSO wakes, convective plumes, and
horizontal shear eddies are introduced into the prognostic TKE
equation (already in official code, SSO term is operational at
DWD)
• Analytical work is performed on a scale-dependent mass-flux
convection scheme interacting with the turbulence scheme
including horizontal
shear – and SSOproduction
reference
pot. temperature [K]
including horizontal
shear –, SSO- and
convective
production
DWD
Matthias Raschendorfer
COSMO Lugano 2012
Task 3: Evaluation of Relative-Humidity and
Statistical Cloud Schemes against Observational
Data (Euripides Avgoustoglou)
Results
• Performance of relative-humidity (RH) and statistical cloud
schemes (incl. version with water-ice mixed phase) is assessed
using satellite and in-situ observations (taken over the Greececentered domain)
• Minimum T2m is better predicted by the statistical scheme, and
vice versa for maximum T2m
• Statistical scheme leads to stronger underestimation of low cloud
cover than RH scheme
• Medium clouds are somewhat better predicted by statistical scheme
• As the Task 3 results suggest, statistical scheme cannot replace the
default RH scheme (in radiation calculations)
Task 3: Results (cont’d)
Minimum T2m 28 April 2011.
Task 3: Results (cont’d)
Low cloud cover computed by the COSMO model with statistical
(yellow) and RH (magenta) cloud schemes vs. satellite data (blue)
Task 3: Results (cont’d)
Medium cloud cover computed by the COSMO model with statistical
(yellow) and RH (magenta) cloud schemes vs. satellite data (blue)
Task 1: Development, Testing and Implementation
into COSMO of the TKE-Scalar Variance Mixing
Scheme (Ekaterina Machulskaya and Dmitrii Mironov)
TKE-Scalar Variance (TKESV) scheme for COSMO is developed
• Apart from the TKE equation, the new scheme carries prognostic
equations for variances of the liquid water potential temperature
variance and the total water specific humidity variance and θl-qt
covariance
• Prognostic variance equations include turbulent diffusion terms
(divergence of triple correlations)
• Reynolds stress and scalar fluxes are determined through algebraic
diagnostic expressions (incl. turbulence anisotropy)
• Scalar-flux formulations include non-gradient terms
• Turbulence length (time) scale depends on static stability
Task 1: 1D Results
• TKESV scheme is favourably tested through singlecolumn numerical experiments (outperforms oneequation TKE scheme)
• Dry PBL: enhanced mixing, up-gradient heat
transfer
• Cloudy PBLs (shallow cumuli, stratocumuli): better
prediction of scalar variances and TKE, slight
improvements with respect to the vertical buoyancy
flux and the mean temperature and humidity
Task 1: 1D Results (cont’d)
Enhanced mixing, counter-gradient heat transfer
Mean Temperature
in Convective PBL
TKE and TKESV Schemes
vs. LES Data
Potential temperature minus
its minimum value within the
PBL. Black dotted curve
shows LES data (Mironov et
al. 2000), red – TKE scheme,
blue – TKESV scheme.
Task 1: 1D Results (cont’d)
TKE (left panel) and <’2> (right panel) made dimensionless
with w*2 and *2, respectively Black dotted curves show LES
data, red – TKE scheme, blue – TKESV scheme.
Task 1: 1D Results (cont’d)
Potential temperature variance (two left panels) and total water variance
(two right panels) for BOMEX (shallow cumulus test case).
Red – TKE scheme, blue – TKESV scheme.
Black solid curves in the middle figures show LES data.
Task 1: 3D Results
• TKESV scheme is implemented into COSMO
• The scheme is tested through a series of parallel
experiments, results look promising
• Verification against observations indicate
improvements as to e.g. fractional cloud cover and
2m temperature and humidity
TKESV vs. ROU
COSMO-DE, July-November 2011 (EXP, ROU)
July
2m
temperature
bias
2m
dew point
bias
August
September
Oct-Nov
TKESV vs. ROU
2m temperature RMSE
2m dew point RMSE
July 2011
July 2011
August 2011
Low clouds, July 2011
August 2011
TKESV + reduced Kmin vs. ROU
COSMO-DE, 1 July – 30 September 2011 (EXP, ROU)
2m dew point
2m temperature
Low clouds
TKESV + reduced Kmin + new aerosols vs. Kmin + NA
COSMO-DE, 1 June – 31 July 2012 (TKESV)
2m temperature
Low clouds
Task 1: Remaining Problems
Overestimation of fractional cloud cover in cumulus-topped PBL
• This error is attributed primarily to the shortcomings of quasi-Gaussian
statistical cloud scheme which is unable to describe cumulus regime. A
somewhat more sophisticated cloud scheme that accounts of nonGaussian effects is required (see below).
Skewness-dependent “diffusion + advection” parameterizations of
the third-order moments in the scalar-variance equations
• The skewness-dependent parameterizations are developed, tested through
single-column experiments, and are available as an option within the
TKESV scheme. These parameterizations reduce numerical stability of
the entire scheme (smaller time step is required) and are not
recommended for immediate implementation into COSMO (cf. threemoment SGS cloud scheme).
We Should Acknowledge
Witold Interewicz
For his on the COSMO-model cumulus convection
(e.g. attempts to slow down deep convection)
Balázs Szintai
For his work on component testing of the COSMOmodel turbulence scheme (e.g. improved
parameterization of the TKE vertical transport)
Co-operation with External Partners
DWD – NCAR (USA) Co-operation
• “Mixing in stably-stratified boundary layers over heterogeneous surfaces:
large-eddy simulation, second-moment budget analysis, and
parameterization”, Peter Sullivan (NCAR)
“Extramurale Forschung” (EMF) Program of DWD and German
Universities
• “Development and testing of a scale independent convection
parameterisation for ICON”, Richard Keane, George Craig and Christian
Keil (University of Munich)
• “High-resolution large-eddy simulation of atmospheric boundary-layer
turbulence – contribution to the improvement of turbulence
parameterisations through systematic study of higher-order statistical
moments and their budgets”, Rieke Heinze and Siegfried Raasch
(University of Hannover)
LES-Based Study of SBL over Heterogeneous
Surfaces (Mironov & Sullivan): Motivation
Models of stably stratified PBL, incl. surface layer, do
not account for many important features (e.g. gravity
waves, meanders of cold air, radiation flux divergence,
and horizontal heterogeneity of the underlying surface)
• Mixing is typically underestimated
• Models tend to quench turbulence in strongly stable
stratification
• Ad hoc tuning devices like “minimum diffusion
coefficients” do not help much (they are often
detrimental for the NWP/climate model performance)
Enhanced Mixing in Horizontally-Heterogeneous SBL:
An Explanation (Mironov and Sullivan 2010)
increased <’2> near the surface  reduced
magnitude of downward heat flux  less work
against the gravity  increased TKE 
stronger mixing
Decreased
in magnitude
Increased
Increased
How to Account for Surface Heterogeneity?
Blue – homogeneous SBL, red – heterogeneous SBL.
• Enhanced mixing is due to increased temperature variance close to the
surface (see Mironov and Sullivan 2010, for details)
• For want of a more elegant theory, use tile approach where the number of
tiles should not be large (otherwise computationally expensive) but
account for tiles with a maximum difference in terms of thermal inertia
• SGS water is crucial
TKESV Scheme and Tiled Surface Scheme
• Tile approach where different tiles have different surface
temperature (the mosaic scheme coded by Felix Ament was
used as starting point)
• Surface fluxes are computed as weighted means of fluxes over
individual tiles
• Transport (prognostic) equations for TKE and for the scalar
variances including third-order transport of scalar variance
• <’2>, <w’’2>, <q’2> and <w’q’2> determined with tiled
scheme (non-zero at the surface!) are used as lower boundary
conditions for scalar variances
(i) Off-line single-column tests
(ii) COSMO test runs using tiled surface scheme with due
regard for SGS water
TKESV + Tiled Surface Scheme:
Off-line Test
Blue – homogeneous SBL,
red – heterogeneous SBL.
COSMO Test Runs
• COSMO-EU, the lake parameterisation scheme FLake is
used to model inland water bodies
• External-parameter fields of lake fraction and lake depth are
generated using the updated lake-depth data set (Kourzeneva
2010) and the software package by Hermann Asensio
• For grid boxes with 0.05<FR_LAKE<0.5 (SGS water), the
surface fluxes are computed as weighted mean over two tiles,
water and land
• All SGS inland water bodies are 10 m deep (externalparameter software should be modified to use the actual
depth where available)
COSMO Test Runs: Results
The lake-fraction external-parameter field based on the lake-depth data from
Kourzeneva (2010) and GlobCover physiographic data. The horizontal size of the
COSMO-model grid is ca. 7 km.
COSMO Test Runs: Results (cont’d)
Difference in surface temperature between experiment with tile
approach (mean over two tiles) and control experiment (no tiles)
night, 30.04.2011 00 UTC
Warming due to SGS lakes
day, 30.04.2011 12 UTC
Cooling due to SGS lakes
COSMO Test Runs: Results (cont’d)
Control run
(no tiles)
T2m warm bias is reduced
Experiment
(two tiles)
Development and testing of a scale independent
convection parameterisation for ICON (EMF)
Motivation
• Improve representation of convection in NWP models (diurnal cycle,
meso-scale organisation, momentum transport)
• Account for a highly variable (stochastic) convective activity within
grid cells
Results
• The Plant and Craig (2008) stochastic convection scheme is
implemented into ICOSN and COSMO, the scheme is adapted to
become “GCM-independent”
• Test simulations with ICON and COSMO are performed
Outlook
• Improve the scheme computational efficiency (vectorization, etc.),
work is underway
• Perform further comprehensive tests
Precipitation Variability
Probability distribution of precipitation intensity (between
10S and 10N) for six different ICON runs.
Example rainfall snapshots for three different
convection schemes in COSMO
The rainfall variability is
closer to that of reality
with the Plant-Craig
scheme, as compared to
two different
conventional schemes.
High-Resolution Large-Eddy Simulation of
Atmospheric Boundary-Layer Turbulence (EMF)
Motivation
• In spite of numerous LES of PBL turbulence, no comprehensive analysis of
the second-moment budgets for cloudy boundary layers
• Budgets are required to (i) estimate relative importance of the various terms,
(ii) test and further develop turbulence parameterizations, (iii) determine
disposable parameters
Results
• Budgets in clear and cloudy boundary layers are analysed on the basis of
very high resolution LES
• Pressure-scrambling terms in the Reynolds-stress and scalar flux equations
are analysed (not possible with observational data!)
Outlook
• Use LES-based results to develop mixing schemes for NWP (and climate)
models
• Estimate disposable parameters of mixing schemes
BOMEX (shallow cumulus case):
TKE and its budget
BOMEX: Total Water Variance and Its Budget
BOMEX: Temperature Flux and Its Budget
!!!
BOMEX and DYCOMS-II (stratocumulus case):
pressure-scrambling
Heinze, R., D. Mironov, and S. Raasch, 2012: Budgets of scalar fluxes for cloudy boundary layers. Proc. 20th
Amer. Meteorol. Soc. Symp. on Boundary Layers and Turbulence, Boston, MA, USA, paper 11.4, 9 pp.
• Analysis of the pressurescrambling components is not
possible on the basis of
observational data
• Detailed knowledge of
pressure terms is necessary to
estimate disposable constants
of the second-order schemes
and to reduced the number of
constants to be tuned
Rieke Heinze received the 1st prize (“Outstanding Student Presentation
Award”) for her presentation at the 20th AMS Boundary Layers and
Turbulence Symposium, Boston, MA, USA, 9-13 July 2012.
Summary and Immediate Plans
 The project ends in September 2012.
 During the coming months (09-12.2012): inclusion of
modifications associated with the TKESV scheme into the official
COSMO code.
N.B. As the current version of TKESV is not coded as a separate module, new
features are implemented into the COSMO TKE scheme. Since massive
modifications to the current TKE scheme are prepared by Matthias
Raschendorfer and should soon be included into the official COSMO code,
modifications associated with the TKESV scheme will be implemented into the
COSMO version following the version with Matthias’ changes. The
implementation process will be controlled by the COSMO source code
administrator.
 Documentation of the TKESV scheme is in preparation.
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
A Look into the Future
 Development of a three-moment (mean, variance, and
skewness) statistical SGS cloud scheme that accounts
for non-Gaussian effects. Co-operation with Axel
Seifert and Ann Kristin Naumann, Hans Ertel Centre
on Cloud and Convection (HErZ), Hamburg. Work is
well underway.
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
A Look into the Future (cont’d)
 Further development and comprehensive testing of
transport equations for the skewness of scalar quantities,
coupling the skewness equations with the three-moment
statistical cloud scheme. Much work is already done.
Closure assumptions for the scalar skewness equations are formulated and tested within the
framework of PP UTCS. Furthermore, skewness-dependent “diffusion+advection”
parameterizations of the third-order moments in the scalar-variance equations are
developed and tested through single-column experiments (optional within the TKESV
scheme). These are, however, not recommended for the immediate implementation into
COSMO due to numerical stability problems. However, the scalar skewness equations
decoupled from the third-order transport of scalars but coupled to the statistical SGS
cloud scheme is a viable option.
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
A Look into the Future (cont’d)
 Improved coupling of the scalar-variance equations to the
tiled surface scheme to better account for the effect of
surface heterogeneity on the PBL structure and mixing
(Mironov and Sullivan 2010), use of PBL scalar
variances in stochastic parameterizations. Co-operation
with Peter Sullivan, NCAR.
Effort will go into the analysis of various flow regimes over heterogeneous surfaces (e.g.
temperature-heterogeneous flat surface versus temperature-homogeneous surface with
orographic features), and into the formulation of the surface boundary conditions for the
scalar variances with due regard for the surface heterogeneity.
Results from future efforts will be reported to COSMO through the WG3a.
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
Thank you for
your kind attention!
Acknowledgements: Peter Bechtold, Vittorio Canuto, George Craig, Ulrich Damrath,
Sergey Danilov, Evgeni Fedorovich, Jochen Förstner, Jean-Francois Geleyn, Federico
Grazzini, Vladimir Gryanik, Thomas Hanisch, Rieke Heinze, Richard Keane, Donald
Lenschow, Chin-Hoh Moeng, Ned Patton, Robert Plant, Siegfried Raasch, Bodo Ritter,
Harald Ruppert, Ulrich Schättler, Axel Seifert, Pier Siebesma, Alan Shapiro, Peter
Sullivan, Jeffrey Weil, Jun-Ichi Yano
14th COSMO General Meeting, Lugano, Switzerland. 10-14 September 2012.
…
Some Lessons Learnt
• Considering high complexity of the
problem, a remote mode operation is often
inefficient (e.g. the project leader cannot
keep an eye on the details of the
implementation of all tasks)
• “Priority Project/Task” is not necessarily
the best way to organise work, particularly
work of innovative character