1 / 60

Consortium Meeting October 8, 2009

Consortium Meeting October 8, 2009. Cooling. Adding two more evacuating racks to blow out air from the last two clusters. Now all clusters blow hot air to great outdoors. Absolutely no trouble dealing with the warmest days on record Need to write up our supergreen cooling approach.

lotte
Télécharger la présentation

Consortium Meeting October 8, 2009

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Consortium MeetingOctober 8, 2009

  2. Cooling • Adding two more evacuating racks to blow out air from the last two clusters. Now all clusters blow hot air to great outdoors. • Absolutely no trouble dealing with the warmest days on record • Need to write up our supergreen cooling approach.

  3. Major Policy Issue: Blocking At the request of Rob and Mike, and based on the vote at the last meeting, I have blocked access to the Oregon Dept. of Forestry. The Canadian Weather Channel is another issue. I have talked to the U.S. Weather Channel…50% owner…and they may get the funds. Decisions: who else should get blocked…and what about Brian Lamb’s side of the effort?

  4. September Stats

  5. Massive Tests This Summer • Dave Ovens has done massive testing this summer to evaluate many issues: • Is the new version of WRF (released this spring) better? • Do the new PBL schemes offer improvements for PBL heights or surface properties? • Can we improve a major system flaw-poor initialization and short-term forecasts of stratus and low clouds over the Pacific? • Do the new radiation schemes offer any hopes for improvements?

  6. Some sample current verifications

  7. WRF Version 3.1 Dave has run hundreds of tests at 36-12km for summer and winter. Very little changes in results, but V3.1 is considerably more stable…less problems losing runs. Something you may not know about…how we deal with problem runs (roughly 1%). Automated scripts will try other physics (starting with Cu parameterizations) from the save point before it bombs. Almost always works.

  8. The New RRTMG Solar Radiation Scheme • According the WRF group, this is the best solar scheme available. • Much more sophisticated than the Dudhia scheme we have been using. • But how does it verify here?

  9. Comparative Statistics Not a consistent improvement

  10. The Low Cloud Problem • Currently we don’t have any cloud at initialization time. • The model rapidly spins up orographic cloud and synoptic clouds, but marine clouds are a real problem for over a day into the forecast.

  11. Today

  12. Previous forecasts valid this morning

  13. Bottom Line Problem We don’t do cloud initialization (tried it…WRF blows up) Low stratus cloud slowly develops…takes 18-24 hr to be reasonable…sometimes longer.

  14. Fixing this problem We have tried two complementary approaches…more aggressive and extensive nudging of low level moisture. Initialization of cloud using a new facility in WRF…digital filter initialization. Digital filtering is essentially running the model backwards and forwards a few times steps while removing (filtering) higher frequency/short wavelength features. Produces a balanced physical state.

  15. Non-DFI

  16. DFI Initialization

  17. Nudging Nudging at a lower level and increasing the nudging coefficients for winds and temperatures by a factor of 3 and moisture by a factor of 300 (to the NCAR defaults) has a greater effect on low clouds

  18. Comparisons of PBL Schemes: YSU, ACM2 (Pleim), QNSE, Mellor Yamada Bottom line of our results and others presented at the WRF meeting in June….no silver bullet. Lets examine two REALLY BAD sample cases: one winter, one summer

  19. 18 Sep 2008 Marine Intrusion Case Study

  20. 18 Sep 2008 Marine Intrusion Case Study

  21. 18 July 2009 Marine Intrusion Case Study

  22. 18 July 2009 Marine Intrusion Case Study

  23. PBL Schemes Bottom line…after running several weeks…no clear winner…will be running more extensive verifications for longer periods this year.

  24. Recommended Changes/Non-Changes to the System This Fall • Change from WRF version 3.0.1.1 to 3.1. Tests show improved numerical stability. • Pending favorable verification statistics for 2 months and 2 seasons, enhanced nudging on the 36-km domain should be implemented. Should improve stratus and low clouds • Continue to use YSU PBL scheme. • Continue with RRTM longwave and Dudhia shortwave radiation schemes. • To improve cloud-water initialization, a cycling approach can be used in combination with the Digital Filter Initializaton. This approach would require running the 36-km domain separately from the 12-km domain and would change our timestep from 216 seconds, 6-delta x, to 225 seconds (an approach used by Brian Ancell). This method has not been extensively tested, but it shows promise.

  25. Real-time High Resolution Data Assimilation • The goals: • High-quality, high resolution analyses using all our data assets. • Our own data assimilation • Frequent, real-time updates • High quality short-term forecasts for various consortium needs

  26. Status • Major advances—4 km and 3h cycle • We showed you the excellent results with 12 km and 6h cycle…but we need better… • But problems with insufficient computer hardware

  27. Real-time WRF EnKF • 36km outer domain/4km nested domain D2 (4km) D1 (36km)

  28. Real-time WRF EnKF • 3-hr update cycle for both domains (analyses and 3-h forecasts 8 times daily) • Bias removal on 4km domain • Analysis and hourly forecast output saved on both domains (analysis,1,2,3-hr forecast) • 80 members on 36km, 42 members on 4km (best we could do with current computer resources) • Digital filter initialization performed on both domains (reduces noise). New technique—run model backward and forward a few times to settle it down before forecast.

  29. Real-time WRF EnKF • ~3000 surface observations available for assimilation on 4-km domain (~75% of all observations after terrain check) • ~1500 observations assimilated, ~1500 for verification • Assimilation of radiosonde, cloud-track wind, and ACARS data still occurs on both domains Terrain Height (m) 36-km 4-km

  30. Real-time WRF EnKF • Runtime for full 36km/4km 3-hr cycle: 2:58 • Improvements in near future (~month): 1) Surface background variance inflation 2) Real-time graphics on web • Analysis and forecast verification with GFS, NAM, RUC, 4km GFS-WRF coming soon!

  31. Real-time WRF EnKF • Sample early results (before bias removal): Analysis RMS errors against unassimilated observations: 4km GFS-WRF4km WRF EnKF Sfc Temperature 2.84 K 2.34 K As was the case with the 12km runs, bias removal and variance inflation is expected to further improve analyses…

  32. Real-time WRF EnKF 36km Mean 4km Mean Near-surface Windspeed, MSLP ANALYSIS

  33. Real-time WRF EnKF 36km Mean 4km Mean Near-surface Windspeed, MSLP 01-HR

  34. Real-time WRF EnKF 36km Mean 4km Mean Near-surface Windspeed, MSLP 02-HR

  35. Real-time WRF EnKF 36km Mean 4km Mean Near-surface Windspeed, MSLP 03-HR

More Related