1. Ideas on WoF NWP design - Relationship to HRRR(E) -Possible contributions from AMB/GSD Stan Benjamin Steve Weygandt Rapid Refresh domain RUC-13 domain HRRR 3-km future 1-km nests rapidrefresh.noaa.gov/hrrrconus Assimilation and Modeling Branch Global Systems Division WoF Kickoff Meeting 18 Feb 2010

2. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics

3. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics

4. HRRR web page – http://ruc.noaa.gov/hrrr

5. RUC / RR hourly updating Use latest observations to obtain the freshest analysis and forecasts for aviation, severe weather, energy, and general forecast applications 1-hr fcst 1-hr fcst 1-hr fcst 11 12 13 Time (UTC) Analysis Fields 3DVAR Obs 3DVAR Obs Back- ground Fields Data types used Rawinsonde (balloons) Wind Profilers (405 MHz, 915 MHz) RASS virtual temperatures VAD winds (WSR-88D radars) Aircraft (ACARS, TAMDAR) Surface (METAR, Buoy, Mesonet Precipitable water (GPS, GOES, SSM/I) GOES cloud-drift winds GOES cloud-top pressure/temp Radar reflectivity, lightning Ship reports/dropsondes Satellite radiances (Rapid Refresh)

6. RUC / RR and HRRR models Observations Data Assimilation cycle Hourly updating model Radar Assim HRRR RUC  RR

7. NSSL verification 06z 16 Aug 2007 00z+6h HRRR RUC radar 00z+6h HRRR No radar RUC radar assimilation on 13-km grid improves HRRR 3-km forecast

8. 3km HRRR verification -From NCAR report - 16 Dec 2009 “Model Performance and Sensitivity” (Mei Xu, David Dowell, Jenny Sun) RUC grids provided much improved initial condition for HRRR than NAM or GFS grids, especially in 1-6h RUC-based HRRR skill (even without radar assimilation) due to effective analysis of convective environment using other observations (ACARS, profiler, surface, cloud obs) and digital filter initialization (focuses vertical motion for convection)

9. 3km HRRR verification - From NCAR report - 16 Dec 2009 “Model Performance and Sensitivity” (Mei Xu, David Dowell, Jenny Sun) Addition of radar assimilation to RUC convective environment adds further improvement for first ~6h (representative example from individual case from late July 2009) RUC grids provided much improved initial condition for HRRR than NAM or GFS grids, especially in 1-6h

10. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics

11. 2010-2011 11 NAM NEMS based NMM Bgrid replaces Egrid Parent remains at 12 km Multiple Nests Run to 48hr –~4 km CONUS nest –~6 km Alaska nest ~1.5-2km DHS/FireWeather/IMET –~3 km HI & PR nests, and/or a ~1.5-2km DHS/FireWeather/IMET are possible Rapid Refresh WRF-based ARW NCEP’s GSI analysis (RR-version) Expanded 13 km Domain to include Alaska Experimental 3 km HRRR @ ESRL RUC-13 CONUS domain WRF-Rapid Refresh domain – 2010 Original CONUS domain Experimental 3 km HRRR Planned NCEP Operation Meso- and Storm-scale models

12. - Hourly cycling of land surface model fields - 6 hour spin-up cycle for hydrometeors, surface fields Rapid Refresh Partial Cycling RR Hourly cycling throughout the day RR spin-up cycle GFS model GFS model RR spin-up cycle 00z 03z 06z 09z 12z 15z 18z 21z 00z Observation assimilation Observation assimilation

13. Planned Operational Meso- and Storm-scale Ensembles 2012-2013 NAM/Rapid Refresh ENSEMBLE – NRRE Initially ~6 member ensemble made up of equal numbers of NMMB- & ARW-based configurations Hourly updated with forecasts to 24 hours High Res Rapid Refresh ENSEMBLE – HRRRE Nest HRRRE ensemble within NRRE Opportunities to: –Provide improved probability guidance for hazardous wx –Use for Nextgen, Warn-on-Forecast, sensible wx

14. Outline - ConfigurationDA #mem Current3km HRRRRUC-DFI )13km1 Spring 2010“RR-DFI +HRRR-DFI (3km) + Conv Init (GOES) ~Fall 20103km-HRRR+1km subnests+15min full output (hubs, RE site(s)?) Spring 2011RR – add chem 2011?3kmHRRR+1km-nest +chem Late 2011NARRE-6 member (~13km) (RR-EnKF in testing? -OU-Xuguang/AMB

15. Outline - ConfigurationDA #mem 2012-testing3km HRRREnKF(?) / NARRE 1 NextHRRRE(6x CPU) 6 Add 1km HRRR (27x 3km HRRR) (grow 1km nest inside 3km HRRR as CPU becomes available, similar) Our dream so far (but not yet WoF dream) 2015-18?1km HRRRE-chemHybrid EnKF/4dvar/DFI? + 300m fixed nests (for avx, RE) ~30 = “SuperHRRRE” (15-30min update frequency)

16. Data assimilation ideas Current – 3dvar, then 13kmDFI Next – add 3km DFI Next – incorporate DFI within 3dvar outer loop EnKF@13-20km, then 13km/3km DFI - may be significant improvement for mesoscale environment - initial work in FY10 under FAA (Ming Xue, Xuguang Wang w/ RR, RRgroup Hybrid EnKF/3dvar @13km, then 13km/3km DFI Hybrid @3km. (Does 3km DFI still add?)

17. HRRR Domain(s) RUC Domain HRRR 2010 September 2007 Initial HRRR domain over the northeastern United States “aviation corridor” 745 x 383 grid points, 200 processors March 2009 Domain expanded to cover approximately eastern 2/3 of the US 1000 x 700 grid points, 568 processors October 2009 Domain expanded to cover CONUS 1800 x 1060 grid points, 840 processors 17 HOURLY FREQUENCY MAINTAINED

18. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics

19. Crude radar refl QC in RUC (and RR) - Start with NSSL QC Comparison with GOES satellite (clear satellite results in radar clearing) Water vapor moistening (reduction of subsaturation) applied PW comparisons with GPS and RUC led to discovery of bird contamination Complaints from Seth Gutman – “Dr. GPS-PW” Reflectivity vs. temperature condition developed 3-d Reflectivity 4°C

20. Crude radar refl QC in RUC (and RR) - Start with NSSL QC Comparison with GOES satellite (clear satellite results in radar clearing) Water vapor moistening (reduction of subsaturation) applied PW comparisons with GPS and RUC led to discovery of bird contamination Complaints from Seth Gutman – “Dr. GPS-PW” QC working well in general, glitch in early Aug? Typical season variation, radar contribution to poorer fit but no regrets…

21. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics

22. Digital Filter-based reflectivity assimilation initializes ongoing precipitation regions Radar reflectivity assimilation Forward integration,full physics with radar-based latent heating -30 min -15 min Initial +15 min + 30 min RUC / RR HRRR model forecast Backwards integration, no physics Initial fields with improved balance, storm-scale circulation + RUC/RR Convection suppression – ask us about it…

23. Rapid Refresh (GSI + ARW) reflectivity assimilation example Low-level Convergence Upper-level Divergence K=4 U-comp. diff (radar - norad) K=17 U-comp. diff (radar - norad) NSSL radar reflectivity (dBZ) 14z 22 Oct 2008 Z = 3 km

24. RUC / RR and HRRR models Observations Data Assimilation cycle Hourly updating model Radar Assim HRRR RUC  RR

25. RUC / RR and HRRR models Observations Data Assimilation cycle Hourly updating model Radar Assim HRRR RUC  RR 2 nd pass 3-km radar DFI analysis

26. 18z Radar RUC NO HRRR NO RUC NO HRRR RAD 17 June 2009 RUC NO HRRR radar + 0h fcsts HRRR with radar DA only on 3-km domain hydrometeor

27. 19z Radar 17 June 2009 + 1h fcsts HRRR with radar DA only on 3-km domain RUC NO HRRR NO RUC NO HRRR RAD RUC NO HRRR radar hydrometeor

28. 20z Radar 17 June 2009 + 2h fcsts HRRR with radar DA only on 3-km domain RUC NO HRRR NO RUC NO HRRR RAD RUC NO HRRR radar hydrometeor

29. 20z Radar 21z Radar 17 June 2009 + 3h fcsts HRRR with radar DA only on 3-km domain RUC NO HRRR NO RUC NO HRRR RAD RUC NO HRRR radar hydrometeor

30. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics

31. WoF/HRRR nesting ideas - Possibilities WoF (HRRR dx/3) - 1-way nest in small domain Re-initialize WoF over multi-state subset of HRRR domain at HRRR-dx 2-way nests to HRRR-dx/3 (and beyond)

32. WoF/HRRR nesting ideas - Possibilities WoF (HRRR dx/3) - 1-way nest in small domain Re-initialize WoF over multi-state subset of HRRR domain at HRRR-dx 2-way nests to HRRR-dx/3 (and beyond) WoF domain du jour WoF dx -Outer = HRRRdx -Inner = HRRR/3 -……

33. WoF/HRRR/brain dump Outline - Current HRRR status HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid Nesting WRF (and other) model design issues physics microphysics, PBL, radiation, LSM, LES at ≤500m resolution, chemistry

34. NCAR-Thompson Microphysics RUC uses Dec 2003 version of scheme Version in WRF v3.1 (mp_physics = 8) has many changes - 2-moment (mixing ratio and number concentration) rain helps better simulate difference in drop-size distribution between rain resulting from melting snow and that from collision-coalescence of cloud drops - Greater ice supersaturation allowed (up to water saturation) - Snow particles assumed to be more 2-d than spherical (affects deposition, collision and fall speed) - Revised collection of snow and graupel by rain - Extensive use of lookup tables - Option for Gamma distribution for all precip hydrometeors Subjective impressions for RR: Less graupel, more cloud ice and snow than in RUC version 34

35. WRF-Chem and RR Primary WRF-Chem development and coordination occurring in GSD (Georg, Steven, Mariusz) Next few years: introduce simple version of WRF-Chem into the RR (or even HRRR) as a first step toward integrated operational weather--air quality forecasting - Aerosol direct effect on radiation (e.g. solar direct-beam irradiance, surface temp forecasts) - Improved warm-rain and ice nucleation in microphysics (aerosol indirect effects) for better cloud/precip forecasts (impact on ceiling, visibility, icing, surface temp) - First step: RR-Chem put together by Steven and Tanya * Once per day to 48h * Aerosol cycling only 35

36. (HRRR-Chem Vertically Integrated Small Aerosol Concentration (relative units) 1200 UTC 2 Sep 2009 Sources are primarily wildfires, biggest in San Gabriel Mtns, southern CA

37. Alternative PBL schemes available in WRF-ARW:  First-order bulk scheme.  Includes a countergradient term to parameterize nonlocal mixing.  Explicit entrainment which is proportional to surface buoyancy fluxes.  Stronger vertical mixing may alleviate the bias found in the MYJ.  2.5 and 3.0 level closure.  The master length scale is a function of 3 independent length scale (turbulent, surface layer, and stable layer).  Updated stability functions  Condensation Module.  Similar physics as MYJ, but tuned to LES simulations for more aggressive vertical mixing. MYNNYSUQNSE  2.5 level closure; similar to MYJ in neutral-unstable conditions, but in stable conditions, QNSE scheme is activated.  Turbulent eddies and waves are treated as one entity in the stable regime.  Similar physics as MYJ, but enhanced treatment of stable nocturnal boundary layer.

38. PBL Scheme Testing New candidate PBL schemes need to show skill across RR domain and reduce biases compared with MYJ. Given recent interest in the RR (and HRRR) for wind energy applications, low-level jets and coastal jet cases are good tests for the new PBL schemes. LLJ case(s) of 20070818-19 WRF-ARW Configuration (v3.1.1): 13.2 and 3.3 km grid spacing 51 vertical levels RUC LSM Grell-3 Cumulus Scheme Thompson Microphysics Scheme RRTM LW Radiation, Dudhia SW radiation MYJ/MYNN/QNSE/YSU PBL Initial Conditions: GFS 6-hourly analyses (Actual RR configuration covers all of North America)

39. Vertical cross-section @ 09Z 20070819 Temperature Wind Speed YSUMYNN QNSEMYJ QNSE produces the strongest and widest LLJ. YSU has the weakest and most vertically diffuse LLJ. Of the 3 TKE-based schemes, the MYNN has stronger vertical mixing, with the jet top ~100 m higher than MYJ or QNSE. Strength of daytime vertical mixing is similar in rank, but has more variation (not shown).

40. Convective probability forecasts from HRRR time-lagged ensemble (shown with deterministic fcst) 15z + 6h HRRR and HCPF Probability (%) Reflectivity (dBZ) 21z 16 July ‘09 Verification

41. 3km HRRR verification - From NCAR report - 16 Dec 2009 “Model Performance and Sensitivity” (Mei Xu, David Dowell, Jenny Sun) Addition of radar assimilation to RUC convective environment adds further improvement for first ~6h (representative example from individual case from late July 2009) RUC grids provided much improved initial condition for HRRR than NAM or GFS grids, especially in 1-6h

42. WoF/HRRR Areas of AMB contribution- HRRR / RR  HRRRE / NARRE  superHRRRE Radar QC hybcloud – satellite, bird, refl only Data assimilation DFI, EnKF, hybrid (mesoscale DA in addition to stormscale) Nesting WRF (and other) model design issues physics microphysics, PBL, radiation, LSM, LES at ≤500m resolution, chemistry