LACE Research and Development on Predictability Report for 2006 and Plan for 2007

1. LACE Research and Development on PredictabilityReport for 2006 and Plan for 2007 Yong WANG, ZAMG

2. Fulfillment for 2006 • Research Areas: • Dealing with uncertainties in analysis • Dealing with uncertainties in LBC • Dealing with uncertainties in model physics • Dynamical downscaling • EPS post-processing • Common EPS verification • Comparison with SRNWP-PEPS

3. Dealing with uncertainties in analysis • Breeding • Status: Effort on testing rescaling factor and breeding cycles. A one-month test which makes the verification meaningful were carried out with 10 members Breeding. Due to limited computer capacity, experiments with one side perturbation and multi-physics perturbation haven’t done.

4. Downscaling vs. Breeding

5. Dealing with uncertainties in analysis • ETKF + spherical simplex transformation: • Status: a one-month test for 10 members ETKF has been carried out. A preliminary comparison between ETKF with Breeding and ET has been done. The inflation method used in the experiment is the similar one used as in Breeding. The innovation inflation technique was not tested for this one-month test. Only some case studies were done with this innovation inflation technique.

6. Dealing with uncertainties in analysis • ET + spherical simplex transformation: • Status:Same as ETKF, a one-month test for 10 members ET has been carried out. A preliminary comparison between ET with Breeding and ETKF has been done.

7. ET, ETKF and Breeding

8. Dealing with uncertainties in analysis • ALADIN native Singular Vector (Edit and Horanyi): • Status: Research with ALADIN SV has been started in 2006 at Budapest. As a first step conf 401, 501 and finally 601 was used and tested. Speed of the convergence (during the singular vector computation), as well as CPU time and memory usage was analyzed. These are quite important issues, since the number of iterations performed determines the accuracy of the computed singular values/vectors. The first results with the ALADIN singular vectors were presented in the ALADIN HIRLAM LAMEPS workshop in Vienna in November, 2006. More details are available from Hágel and Horanyi.

9. Dealing with uncertainties in analysis • Blending ARPEGE PEARP and ALADIN Breeding: • Status: implemented, first test has been done. More tunnig with • Breeding are needed. • With LACE Funding, Martin Bellus, 1 Month

10. Blending vs. Breeding and downscaling

11. Dealing with uncertainties in LBC • Impact of inconsistent LBC and IC perturbation • ALADIN Breeding coupled with NCEP EPS system and ARPEGE • SV EPS system • Status: finished partly, more experiments are needed.

12. Coupling with SV vs. Breeding: Spread Spread of Surface Pressure (top left), Geopotential (top right) and Kinetic Energy (bottom left) as a function of forecast range. Experiments: Breeding coupled with PEARP (purple), Breeding coupled with NCEP-ENS (black), PEARP members (light blue), NCEP-ENS (yellow). Breeding coupled with NCEP-ENS perturbed LBC show much larger spread, partly due to the spread of the coupling model itself, partly due to a combination with Breeding within LAEF.

13. Dealing with uncertainties in physics • Multi-physics parameterization: • 11 combinations of different physics parameterizations and tunings • in ALADIN have been chosen for downscaling the PEARP members • schemes of Lopez, Bougeault, Kain-Fritsch etc.... • Status: finished partly

14. Results of multi-physics option

15. Results of multi-physics option

16. Dealing with uncertainties in physics • implementation of ECMWF stochastic physics • implementation of UKMO/ECMWF Stochastic Kinetic • Energy Backscatter algorithm SKEB • Status: not touched

17. Dynamical downscaling • Downscaling of ECMWF EPS • Similar as COSMO-LEPS using clustering method • Status: case studies, computational framework at ECMWF partly • implemented. • With LACE Funding, Tascu, Pistotnik, 2 Months

18. ECMWF EPS downscaling: case study, HU

19. Experiments with clustering: CRO

20. Downscaling of ARPEGE EPS • Status: Works have been done in HU,CZ and AT. • In HU, experiments with different optimisation time and target domain in ARPEGE EPS and then ALADIN downscaling are done. More details by Edit. • In CZ, to investigating the performance, in particular the more added value of the downscaling of ARPEGE EPS. Cases studies, especially those related with big problem with precipitation forecast in ALADIN, are under way to be investigated.

21. EPS post-processing • Bias correction • developing a statistical method for correcting the bias in • LAEF forecast. • Status: code implemented.

22. Common EPS verification • Verification • developing a common verification package for LACE • research and development on prdictability • Status: The common verification package has been developed for verifying the upper air parameters. The scores, like Talagrand diagram, Brier Score and skill score, ROC etc have been implemented. With LACE Funding, Edit, Mladek, 1.5 Months

23. Common EPS verification: examples

24. Comparison between ALADIN EPS and PEPS • comparison with SRNWP-PEPS • To evaluate the performance of ALADIN EPS, • understand the multi-model, multi-analysis EPS system • Status: not touched

25. Other activities: • International collaboration, GLAMEPS etc, • ALADIN-HIRLAM Workshop in Vienna, 13-14. Nov. 2006

26. Summary of means and LACE Funds

27. Working plan for 2007 • Research focues: • ALADIN native SV, 13 man/month, 1 mon. LACE funds • Blending, 5 man/month, 1 mon. LACE funds • Downscaling of ECMWF EPS, 15 man/month, 1 mon. • LACE Funds, 1 mon. ALADIN flat rate funds • Common verification package, 5 man/month, 1 mon. • LACE funds • GLAMEPS, 4 man/month, 1 mon. LACE funds

28. ALADIN native SV Action (HU and CZ): In 2007, tests and validation of perturbations generated by ALADIN singular vectors will be carried out. Work on singular vectors would be to deeply validate the singular vector computations for ALADIN (from the scientific point of view) and to inter-compare their results to global (possibly ARPEGE and IFS) counterparts with possibly different horizontal resolutions. At the next stage perturbations should be computed from the singular vectors and those perturbations should be used as initial conditions for a limited area ensemble prediction system. The proposed work can be considered as contribution to the GLAMEPS.

29. Blending Action (AT and SK): In 2007, the work will be continued for test the performance of the blending perturbation. The main efforts will be on tunning the small scale part of the uncertainty.

30. ECMWF EPS downscaling Action : In 2007, the work will be continued for set up the computational framework at ECMWF. This is strongly related with GLAMEPS project. Case studies, performance test, choice of representative members will be investigated. The subject is contributed by: CRO: 4 mon; RO: 4 mon and AT 6 mon. PT: 1 mon by ALADIN funds Coordination needed.

31. Common verification package Action : The continuation of the implementation of the verification package will be done in 2007, in particular, the work will be concentrated on the verification of surface parameters, e.g. precipitation.

32. GLAMEPS Action : Many subjects in LACE plan are strongly related with the GLAMEPS, like SV, downscaling of ECMWF etc. Coordination with HIRLAM, and participation on working week of GLAMEPS will be conducted.

33. Other issues There are other subjects: Downscaling of ARPEGE EPS, uncertainty in LBC and model physics, EPS post processing and multi-model EPS and so on. To avoid works be done repeatedly, exchange of information, coordination are needed.

34. Summary of means and LACE Funds ALADIN flat rate: 1 month, PT colleague Participation on workshops related EPS: 2 months