1. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Comparison of different data assimilation techniques for a convective situation during the IHOP 2002 campaign Hans-Stefan Bauer and Volker Wulfmeyer, Hohenheim University Francois Vandenberghe, NCAR RAP

2. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Outline Motivation IHOP 2002 Models and simulations First results First conclusions Outlook

3. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Motivation The correct prediction of the components of the hydrological cycle in time and space is a major goal of numerical weather prediction. Source: NOAA - 15 Not yet satisfactory fulfilled ! 

4. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe IH 2 OP 2002 Largest campaign to date dealing with the 4D representation of water vapour. It took place in May and June 2002 in the Southern Great Plains. ”An improved characterization of the water vapor field will lead to a corresponding improvement in our ability to predict convection.”

5. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Flight tracks of 24th of May 2002

6. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe LM configuration Operational configuration 7 km horizontal resolution, 35 levels, 40 s time step GME boundary and initial forcing Nudging of observations Some advances parameterisations were used which were not operational when the simulations were done. Cloud ice scheme Advection of precipitation 1.Simulation only with forcing from the global model (lateral boundaries and initial conditions). 2.Simulation with forcing as in experiment 1 and in addition with nudging against observations during the simulation. LM simulations

7. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe LM data assimilation To be able to use high-resolution asynoptic observations operationally an assimilation based on observation nudging was implemented. G ψ = 6·10 -4 s- 1 for all assimilated quantities vertical quality horizontal time Weight depending on the temporal and spatial distance as well as the quality of the observations.

8. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Situation 24 May 2002 (ETA, 12Z analysis) Source: IHOP field catalogue

9. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe GOES-8 visible (1km resolution) Radar loop (Altus, Oklahoma) Source: IHOP field catalogue

10. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe

11. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe

12. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Precipitation 25 May 2002 00Z

13. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Differences MM5 to LM Different and larger choice of parameterisation schemes MM5 supports different nesting capabilities (one-way, two- way, moving nests) to reach very high resolutions. More assimilation systems are available Nudging against observations and/or analyses 3DVAR 4DVAR Easy conversion of the output to standard graphics tools Machine dependent output files

14. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe MM5 3DVAR system A 3DVAR system calculates the best possible analysis state by iteratively minimizing the following cost function. © Barker et al., 2003 Observations: SYNOP METAR SHIP TEMP AMDAR AIREP PILOT SATEM SATOB GPSPW SSMT1/2 TOVS COLDWARM

15. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe MM5 IHOP_2002 domains Domain 1: 21 km, 120x120 points Domain 2: 7 km, 181x181 points Domain 3: 2.3 km, 232x250 points

16. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe MM5 IHOP_2002: used physics Enhanced Kain-Fritsch convection scheme Sophisticated microphysics from the Goddard Space Flight Center (including Graupel) Boundary Layer scheme from the MRF model Cloud radiation scheme (compiler problems with Open MP with the more sophisticated RRTM scheme). Domain 0 and 1: Domain 2: Same as above but with explicit convection.

17. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe First preliminary results

18. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Preliminary Conclusions LM works in this region Synoptic-scale features are represented As expected a better representation of the surface cold front occurs in the simulation with assimilation Locally forced convection is not represented reasonably in both LM simulations. MM5 shows the same problems for the precipitation even with a first try to use 3DVAR assimilation.

19. Oct. 28 th 2003 5 th SRNWP, Bad Orb H.-S. Bauer, V. Wulfmeyer and F. Vandenberghe Outlook Detailed comparisons for other quantities. Detailed comparisons with observations. Higher resolution runs with explicit and parameterised convection. Avoid the “cold start” problem. Performance of WRF in this region. MM5 using 4DVAR assimilation. Influence of different physics options.