UM 4D-Var Regional Reanalysis Progress

1. UM 4D-Var Regional Reanalysis Progress Richard Renshaw, Stephen Oxley, Adam Maycock, Peter Jermey, Dale Barker, DingMin Li

2. Contents Progress with technical work First attempt at reanalysis Developments for 2012/13

3. WP2.1 Building capacity for advanced regional data assimilation Set up and demonstrate an advanced regional reanalysis system Based on operational North Atlantic/Europe LAM Planning to analyse 2010 & 2011

4. MetO Regional Reanalysis: Horizontal Domain 12km grid 480 x 384

5. Technical work Build a reanalysis system • to run at ECMWF • taking observations, start file, boundary conditions from ECMWF archives • storing analyses in ECMWF archives • and producing data to validate the reanalysis.

6. Running atECMWF ECMWF “interface” Met Office

7. Technical Highlights Set up Met Office systems on ECMWF IBM Use observations from mars (ECMWF BUFR format) Preprocess satellite radiances reconstruct level 1b files from BUFR data convert level 1b counts to level 1c radiances (AAPP) map between instrument grids (e.g. AMSU+MHS to IASI) Convert and map ERA-Interim GRIB files to UM file format, for start files and boundary conditions Use ‘CYLC’ to develop and run a reanalysis suite

8. CYLC: Scheduling System CYLC, written by Hilary Oliver (NIWA) http://hjoliver.github.com/cylc/ + ROSE – adds interfaces and configuration management

9. Why CYLC ? Will replace ageing Met Office system Can build, run, monitor suites of any complexity Portable – Open source Flexible Intelligent job scheduling - define dependencies for each task - tasks will run early if possible

10. sample CYLC suite

11. Dale’s challenge Regional reanalysis - why bother ??!

12. Evidence from operational NWP 25km Global vs 12km NAE

13. ...the benefits of resolution global 25km NAE 12km screen temperature rms error (K) forecast range

14. ...and the disadvantage of boundaries! NAE global mean sea level pressure rms error (Pa) forecast range

15. ERA-Interim vs 12km EURO4M • T255 (80km), 60 levels • T159 (125km) 4D-Var • 12-hour analysis window • assimilate: • conventional obs • satellite radiances • 12km, 70 levels • 36km 4D-Var • 6-hour analysis window • assimilate: • conventional obs incl vis • satellite radiances • Ground-based GPS • Cloud • Precipitation • Initial state and boundary conditions from ERA-Interim analyses

16. model orography ERA-Interim Model T255 (80km) Var T159 (125km) Met Office Model 12km Var 36km

17. First attempt at reanalysis... • Floods in Poland, eastern Europe • Severe storms France/Spain • Russian heatwave spreading West, forest fires May 2010 June 2010 July 2010

18. Poland floods, 16th May 2010 e-obs gridded 24hr rainfall http://eca.knmi.nl/

19. Poland floods, 16th May 2010 radar 24hr rainfall Czech Hydrometeorological Institute

20. 12km reanalysis Poland floods, 16th May 2010 e-obs

21. ERA-Interim Poland floods, 16th May 2010 e-obs

22. Russian heatwave, July 2010 e-obs Tmax, 10-07-2010

23. ERA-Interim obs 12km EURO4M Tmax 10-07-10

24. Verification vs Radiosonde Temperature rms error T+0 T+6 ERA-Interim EURO4M 12km

25. Verification vs Radiosonde Vector wind rms error T+0 T+6 ERA-Interim EURO4M 12km

26. Developments for 2012/13 • Variational bias correction • ODB – obs monitoring • Extend observations dataset • Cloud and Precipitation assimilation • Validation – extreme statistics • Collaborate on cross-validation

27. Variational Bias Correction Airmass-dependent bias correction of satellite radiances (based on Harris and Kelly, 2001) Currently coeffs c are calculated off-line monthly VarBC will give smooth and automatic updating Work is underway – finish in 2012 (DingMin Li, Dale Barker, Andrew Lorenc)

28. ODB – obs monitoring database Established ECMWF database + utilities Array of tools available “for free” Metview macros (quick look) Obstat (detailed stats / graphics) Replaces several diverse monitoring tools at MO Joint development Currently ~1.5 years into 3 year joint funded post at ECMWF (Peter Kuchta) Setting up Obstat at MO Other NMSs already using ODB (CAWCR, KMA)

29. Extend observations dataset • Work with UEA (WP2.5) to extend surface • and sonde datasets • Seeking sub-daily raingauge data...

30. Cloud + Precip assimilation Operational UK models assimilate cloud products (satellite) and rainfall (radar, latent heat nudging) For EURO4M, plan to assimilate cloud reports and raingauge accumulations from surface stations Needs development work (2012/13): source sub-daily raingauge accumulations decode and interpret surface reports (Peter Francis) quality control assimilate rainfall accumulations in 4D-Var reduce effect that analysis increments trigger rain in model

31. Validation - Extremes High impact weather Should show a benefit for regional reanalysis Focus on: precipitation temperature winds Need to look at probability distributions (pdf’s), e.g. number of values outside 10th/90th percentile count consecutive days with high/low values accumulated precipitation over several days

32. Collaboration – Cross-Validation Compare our reanalysis against: SMHI ERA Obs climatologies Others ?

33. Summary • Reanalysis system ready to run • Already better than ERA-Interim in some features, but not all (yet!) • Science development and testing this year

34. Questions ? With thanks to: Tom Green, Nigel Atkinson, Sreerekha Tr, others...

35. Extra slides...

36. First attempt at reanalysis... May 2010 June 2010 July 2010 • 3 x 1 month runs, with overlap

37. How long to spin up ? rms surface pressure

38. How long to spin up ? rms screen temperature

39. 100km 4D-Var Impact of Var resolution 12km 4D-Var

40. Verification vs Radiosonde Relative Humidity rms error T+0 T+6 ERA-Interim EURO4M 12km

41. Global ODB trial output NH Surface T mean (o-b)