1. A. Topographic radiation correction in COSMO: gridscale or subgridscale? B. COSMO-2: convection resolving or convection inhibiting model?
Matteo Buzzi
MeteoSwiss
COSMO GM, WG3,

2. A. Topographic radiation correction: gridscale or subgridscale?
Radiation in complex
terrain
Default COSMO radiation scheme (Ritter&Geleyn,1992): topographic effects on radiation are not considered
COSMO-2 (2.2km) and COSMO-7 MeteoSwiss: grid-scale radiation correction (Buzzi, 2008)
Next step: compare the grid-scale and subgrid-scale radiation correction over a longer period
Preliminary results: 2 winter weeks ( )

3. Two radiation correction procedures
model
gridcell
Grid scale (COSMO km, COSMO km) :
COSMO Topo
slope angle,slope aspect
skyview, horizon
COSMO run
correction factors
Müller and Scherrer (2005), Buzzi (2008)
Orange: gridscale topography (e.g. 2.2 km horizontal resolution)
Yellow: subgridscale topography (e.g. 200 m horizontal resolution)
Process or calculation parts in the dashed rectangle are performed before the model integration
Subgrid scale (COSMO m, COSMO-7 1 km):
DEM topo
correction factors
skyview
aggregation
COSMO
run
slope angle,slope aspect
skyview, horizon

4. Test suite 2 winter weeks: 12-25.12.2007
COSMO-2 with own assimilation cycle
COSMO-7 with own assimilation cycle
Compared the operational suite with gridscale radiation correction and the testsuite with subgridscale radiation correction

5. Results - 2m temperature
COSMO-2
COSMO-7

6. 2 m dew point
COSMO-2
COSMO-7

7. RMSE 2m temperature: summary
N stn
Subgrid
Grid
Subgrid
Grid
Subgrid
Grid
All
491
3.06
3.03
3.22
3.21
1339
2.26
0-500
283
2.64
2.69
1032
2.15
120
3.71
3.68
3.85
3.82
193
2.94 39
4.01
3.93
4.15 56
2.98
2.96 27
3.08
3.05
3.74
3.76 33
3.37
3.45 22
2.08
2.11
3.00
3.02 25
2.32
[°C]
2 correction options are virtually equivalent, no significant differences
Stations with positive and negative effects are in balance
Stations are located near gridpoints with low impact

8. Mean difference 12-25.12.2007 subgrid-grid, direct shortwave radiation
09 UTC
14 UTC
[W/m2]
COSMO-2
[min,max]=[-204,223]
COSMO-7
[min,max]=[-102,84]

9. Mean difference 12-25.12.2007 subgrid-grid, diffuse downward radiation
09 UTC
14 UTC
[W/m2]
COSMO-2
[min,max]=[-12,15]
COSMO-7
[min,max]=[-11,11]

10. Mean difference 12-25.12.2007 subgrid-grid, longwave downward radiation
09 UTC
14 UTC
[W/m2]
COSMO-2
[min,max]=[-11,18]
COSMO-7
[min,max]=[-10,14]

11. Mean difference 12-25.12.2007 subgrid-grid, shortwave balance
09 UTC
14 UTC
[W/m2]
COSMO-2
[min,max]=[-146,164]
COSMO-7
[min,max]=[-67,57]

12. Mean difference 12-25.12.2007 subgrid-grid, surface temperature
09 UTC
14 UTC
[°C]
COSMO-2
[min,max]=[-7.6,8.5]
COSMO-7
[min,max]=[-4.9,4.7]

13. Mean difference 12-25.12.2007 subgrid-grid, 2m temperature
09 UTC
14 UTC °C
COSMO-2
[min,max]=[-2.3,3.1]
COSMO-7
[min,max]=[-1.5,1.4]

14. Conclusions and outlook
Verification differences are very small
But: impact on the surface radiation components and on surface temperature are significant
Stations are probably not really representative for the entire alpine region, compensation effects
Stable conditions: upward transport of information reduced
Differences at the surface can be relevant for snow melting
Is not possible to indicate the winner!
Gridscale correction has an important practical advantage:
Easy to be operated: long preprocessing calculations for computation of external parameters can be avoided
Additional verification is necessary
2 summer weeks
Radiation verification with satellite data and ASRB stations

15. B. COSMO-2: Convection resolving or convection inhibiting model?
RADAR
mm/24h

16. But COSMO-7 is not really better…
mm/24h
COSMO-7
COSMO-2
RADAR

17. Why is COSMO-2 so bad? Some hypothesis...
Beyond of predictability issues ...
We are faced in these situations (no strong dynamical forcing) with “non-equilibrium convection cases” (Craig et al.)
Limiting factor is the trigger (CIN overcoming) and not the forcing (generation of CAPE)
PBL and soil humidity distribution, local circulations, topography effects, small scale surface moisture convergence, PBL profiles near the surface
Independent shallow convection and turbulence parameterization (communication between scales)
LHN is the rescuer, but has always a positive impact?
 Problem is probably located in the PBL

18. Parallel test-suites - July 2008
Motivation:
Try something in order to save model reputation
Past tests: COSMO model is very sensitive to changes of the turbulence scheme parameter tur_len (Koller, 2008; Seifert, 2007)
Smaller values reduce turbulent mixing, increase gridscale fluxes
Testsuites:
Modified asymptotic turbulent length scale (tur_len)
Change of settings in the statistical cloud scheme:
1. opr: clc_diag=0.75, qcrit=4.0
2. and 3. test: clc_diag=0.5, qcrit=1.6
6 days: all 3 configurations ( ): fuzzy verification for 00 runs
30 days: configuration 1 and 2 ( ): fuzzy verification for 00 runs

19. Referece: radar composite
Fuzzy Verification
Verification on coarser scales than model scale: “Do not require a point wise match!“
Referece: radar composite
Method
Raw Data
Fuzzyfication
Score
Example result
Upscaling
Average
Equitable threat score
Fraction Skill Score (Roberts and Lean, 2005)
Fractional coverage
Skill score with reference to worst forecast X x X x
good
bad

20. Fuzzy verification 2-8.07.2008 1. COSMO-2 500 2. COSMO-2 250

21. Tests vs operational 2-8.07.2008 2.COSMO-2 250 vs 1. 500
TEST better
operational better

22. July 2008 – 00 UTC RUNS 1. COSMO-2 OPR 500 2. COSMO-2 test 250
Difference
1.COSMO-2 operational better
COSMO-2 operational better
2. COSMO-2 test better
COSMO-2 test better

23. July 2008 – all runs with 3h cut-off
1. COSMO-2 OPR 500
2. COSMO-2 test 250
Difference
1.COSMO-2 operational better
2. COSMO-2 test better

24. Conclusions and outlook
Reduction of the asymptotic mixing length:
No wonder results! We still have a trigger problem.
Test 1 with 150m:
Too many gridpoints react
Small deterioration from small to medium scales and from small to medium thresholds
Test 2 with 250m:
small improvement at medium to large scales and medium to large thresholds
Optimal value should be close to 250 m
Outlook:
three case studies have been selected: further investigations (soil moisture, vertical profiles, role of the LHN, shallow convection, SBL)
Verification of other meteorological parameters

25. Thank you for your attention!

26. Wind speed and direction COSMO-2

27. Wind speed and direction COSMO-7