1. Weather Model Development for Aviation Stan Benjamin and Steve Weygandt: Assimilation and Modeling Branch, Chief/Deputy NOAA Earth System Research Laboratory, Global Systems Division, Boulder, CO Stan/Steve: Lead/Expert Model Development and Enhancement Product Development Team, AWRP/FAA 12h NOAA HRRR model forecast Valid 03z NOAA/ESRL/GSD30 Oct 2013 1 Aviation Model Development Observed radar 03z June 30 2012

2. An Important Pinpoint Prediction Challenge: The 29 June 2012 Mid-Atlantic Derecho A fast-moving damaging wind event… 700 mile long swath of damage, 5 million without power, 22 fatalities 2 PM 4 PM 6 PM 8 PM 10 PM MID 11 AM Start HRRR run

3. Computer weather modeling: What is the potential? HRRR 2012 derecho loop Observed radar HRRR forecast initialized 15z (11am Eastern Time) 29 June 2012 – Mid-Atlantic/DC thunderstorm/derecho event

4. Computer weather modeling: How is it done? 4 Weather computer model: Solve physics equations at many points repeatedly to mimic time- evolution of 3-D of temperature, wind, moisture, clouds, etc. 1800 points Model Terrain 1800 x 1060 points x 50 levels = 95,000,000 3-d points every 20 seconds 50- levels

5. NOAA Next-Generation Model Development RAP HRRR RAP - Rapid Refresh –NOAA “situational awareness” model for high impact weather –New 18-hour forecast each hour –NOAA operational – 1 May 2012 –Hourly use by National Weather Service, Storm Prediction Center, FAA, private sector HRRR – High-Resolution Rapid Refresh -Next-generation storm/energy/aviation guidance -New 15-h forecast each hour -Real-time experimental on ESRL supercomputer -Open ftp access

6. RAP and HRRR data assimilation 6 RAP Data Assimilation cycle Observations Hourly cycling model HRRR EnKF- Hybrid + Radar and Cloud anx Radar and Cloud anx + 3DVAR

7. Operational Prediction Process Observations Objective Analysis ( adjust background) Model Prediction Analysis Update Cycle Human Forecaster Statistical post- processing (downscaling, probability) Data Assimilation

8. HRRR (and RAP) Future MilestonesHRRR Milestones Rapid updating – Why do it? Better forecasts 6 AM time 9 AM noon 3 PM6 PM 12-h fcst Truth 12-h update to previous forecast More frequent model updates with newer obs Smaller adjustments 9-h fcst 6-h fcst  3-h fcst  3-h update to previous forecast Next forecast

9. Benefits of Rapid Cycling NWP Rapid update cycling with latest observations improves short-range forecasts ( including upper-level winds) RUC jet-level (35 kft) wind forecast errors 3-h fcst wind errors 6-h fcst wind errors 12-h fcst wind errors LAX ORD LAX ORD LAX ORD NOAA/ESRL/GSD12 July 2012 9 Aviation Model Development NOAA/ESRL/GSD30 Oct 2013 9 Aviation Model Development

10. RAP error reduction to 1-h forecast 1h 3 6 12 18h Rapid Refresh Wind forecast accuracy vs. forecast length The Rapid Refresh is able to use recent obs to improve forecast skill down to 1-h projection 1 Jan - 7 Mar 2012 - Verification against weather balloon data NCEP Production Suite Review4-5 December 2012Rapid Refresh / HRRR 10 NOAA/ESRL/GSD30 Oct 2013 10 Aviation Model Development

11. Rapid Refresh Hourly Update Cycle 1-hr fcst 1-hr fcst 1-hr fcst 11 12 13 Time (UTC) Analysis Fields 3DVAR Obs 3DVAR Obs Back- ground Fields Partial cycle atmospheric fields – introduce GFS information 2x/day Cycle hydrometeors Fully cycle all land-sfc fields (soil temp, moisture, snow) Hourly Observations RAP 2013 N. Amer Rawinsonde (T,V,RH)120 Profiler – NOAA Network (V)21 Profiler – 915 MHz (V, Tv)25 Radar – VAD (V)125 Radar reflectivity - CONUS1km Lightning (proxy reflectivity)NLDN, GLD360 Aircraft (V,T)2-15K Aircraft - WVSS (RH)0-800 Surface/METAR (T,Td,V,ps,cloud, vis, wx) 2200- 2500 Buoys/ships (V, ps)200-400 Mesonet (T, Td, V, ps)flagged GOES AMVs (V)2000- 4000 AMSU/HIRS/MHS radiancesUsed GOES cloud-top press/temp13km GPS – Precipitable water260 WindSat scatterometer2-10K Nacelle/Tower/Sodar20/100/10

12. Observations assimilated in hourly updated models (Rapid Refresh) - All used to initialize 3km HRRR Radar reflectivity 12

13. HRRR (and RAP) Future MilestonesHRRR Milestones High Resolution – Why do we need it? RAP HRRR Thunderstorm ~3km horizontal resolution needed to “resolve” thunderstorms

14. HRRR (and RAP) Future MilestonesHRRR Milestones High Resolution – Why do we need it? RAP HRRR Thunderstorm ~3km horizontal resolution needed to “resolve” thunderstorms …but 4x resolution costs 64x computer power

15. 13-km 6hr forecastHRRR 6hr forecast 13-km Resolution Parameterized Convection 3-km Resolution Explicit Convection 5 PM EDT observed 07 June 2012 NO STORM STRUCTURE NO ESTIMATE OF STORM PERMEABILITY ACCURATE STORM STRUCTURE ACCURATE ESTIMATE OF STORM PERMABILITY HRRR (and RAP) Future MilestonesHRRR Milestones 3-km HRRR – what it gets you...

16. Radar Obs 06:00z 18 May 2013 05z + 1 hour Radar data assimilation: Getting storms in the right places 1-hr fcst radar DA (13-km and 3-km) 1-hr fcst NO radar DA

17. Run model backwards in time (reversible processes only) Run model forward in time (heating from radar observations) Digital filter after backward and forward step Forward integration,full physics with obs-based latent heating -20 min -10 min Initial +10 min + 20 min RAP / HRRR model forecast Backwards integration, no physics Initial fields with improved balance, storm-scale circulation 17 Radar data assimilation: How it works for RAP and HRRR

18. 00z init 00z 12 Aug 2011 Convergence Cross-Section RAP HRRR RADAR RAP HRRR no radar Rapid convective spin-up with radar data Radar data assimilation: How it works for RAP Reflectivity

19. +1 hr fcst 01z 12 Aug 2011 Convergence Cross-Section RAP HRRR RADAR RAP HRRR no radar Rapid convective spin-up with radar data Radar data assimilation: How it works for RAP Reflectivity

20. Cloud and Hydrometeor Analysis Hydrometeor designation from radar Adjust cycled explicit cloud fields using METAR and satellite data YES HM 29 th Conf on EIPT (IIPS)08 January 2013High-Resolution Rapid Refresh 20 NOAA/ESRL/GSD30 Oct 2013 20 Aviation Model Development

21. Observations Data Assimilation Cycle Rapid cycling NWP Data Assimilation and Rapid Cycling Numerical Weather Prediction (model) Air transportation (NextGen) Detailed, precise short-range weather guidance needed for: Required for improved weather guidance for: Turbulence Ceiling/visibility Convective weather Icing Terminal/enroute weather Safety and efficiency

22. Aviation hazard forecasts – all based on RAP and HRRR models (out to 15-18h) Hourly updated 13km Rapid Refresh model forecasts (development supported by FAA/MDE, NOAA) Refreshing from latest observations every hour gives better accuracy

23. 23 Subset of full domain An example of computations needed 1800x1059x50 grid points = 95 E6 grid points x 50,000 floating pt ops per grid point = 4.75 E12 FPA / time step x 2160 time steps / 12h forecast = 10 E15 FPA / 12h forecast 10,000,000,000,000,000 calculations for one 12h HRRR CONUS forecast Weather computer model: Solving physics equations on many points repeatedly to provide 3-D forecast forecast of temperature, wind, moisture, clouds, etc. 1800 points 1060 points

24. ModelVersionInitialized Forecast Length Run Time# CPUsDisk Space RAPWRFv3.3.1+Hourly18 hrs~30 min200230 GB (per run) HRRRWRFv3.3.1+Hourly15 hrs~50 min1128800 GB (per run) ModelRun at:Domain Grid Points Grid Spacing Vertical Levels Height Lowest Level Pressure Top Initialized RAP GSD, NCO North America 758 x 567 13 km508 m10 mb Hourly (cycled) HRRRGSDCONUS 1799 x 1059 3 km508 m20 mb Hourly (no-cycle) RAP and HRRR Resources CW Overview Meeting12 June 2012High-Resolution Rapid Refresh 24 NOAA High-Performance Computer System Number of Filesystems Total Reserved Disk Space CPU Type Total Reserved CPUs Performance Increase Jet (current)4150 TBIntel Nehalem1736- Zeus (new)2230 TBIntel Westmere2000-400030% NOAA/ESRL/GSD12 July 2012 24 Aviation Model Development NOAA/ESRL/GSD30 Oct 2013 24 Aviation Model Development

25. HRRR Primary FAA-CoSPA NCEPESRL/GSD/AMB RAP Dev1 RAPv2 Primary HRRR Dev1 RAPv1 NCO RAP Dev2 RAP Retro HRRR Retro Retrospective Real-Time HRRR (and RAP) Future MilestonesHRRR Milestones RAP/HRRR internal verification NWS-AWIPS WEB RAP/HRRR Model Development and Eval FPAW - 201324 October 2013RAP/HRRR Model Status 25 NOAA/ESRL/GSD30 Oct 2013 25 Aviation Model Development

26. HRRR 12 hr fcst availability Includes all missed/incomplete runs HRRR (and RAP) Future MilestonesHRRR Milestones HRRR Availability – getting it out CoSPA Evaluation Periods Development and Upgrade Periods 3 month running average FPAW - 201324 October 2013RAP/HRRR Model Status 26 NOAA/ESRL/GSD30 Oct 2013 26 Aviation Model Development

27. Jet or Zeus (union) HRRR 12 hr fcst availability Includes all missed/incomplete runs Jet (HRRR primary)Zeus (HRRR backup) HRRR (and RAP) Future MilestonesHRRR Milestones HRRR Dual-Computer Availability FPAW - 201324 October 2013RAP/HRRR Model Status 27 NOAA/ESRL/GSD30 Oct 2013 27 Aviation Model Development

28. Jet or Zeus (union)Jet (HRRR primary)Zeus (HRRR backup) HRRR 12 hr fcst availability Excludes two (or fewer) consecutive missed/incomplete runs HRRR (and RAP) Future MilestonesHRRR Milestones HRRR Dual-Computer Availability FPAW - 201324 October 2013RAP/HRRR Model Status 28 NOAA/ESRL/GSD30 Oct 2013 28 Aviation Model Development

29. http://rapidrefresh.noaa.gov/hrrraviation/http://rapidrefresh.noaa.gov/hrrraviation/ (hourly output) Feedback from AWC Aviation specific forecast fields: Convection (intensity and depth) Ceiling and Visibility Flight Rules Surface Wind Precip Type Snowfall http://rapidrefresh.noaa.gov/hrrraviation15min/http://rapidrefresh.noaa.gov/hrrraviation15min/ (sub-hourly output) HRRR (and RAP) Future MilestonesHRRR Milestones HRRR Aviation Websites FPAW - 201324 October 2013RAP/HRRR Model Status 29 NOAA/ESRL/GSD30 Oct 2013 29 Aviation Model Development

30. HRRR (and RAP) Future MilestonesHRRR Milestones CoSPA Evaluation Periods Development and Upgrade Periods 3 month running average HRRR Latency – getting it out fast FPAW - 201324 October 2013RAP/HRRR Model Status 30 2013 HRRR changes permitted ~45 min reduction in latency 2013 CoSPA period 2012 CoSPA period +2:30 +1:45 NOAA/ESRL/GSD30 Oct 2013 30 Aviation Model Development

31. 3-km Interp 2012 HRRR Initialization from RAP GSI 3D-VAR GSI HM Anx Digital Filter 1 hr fcst 18 hr fcst 15 hr fcst 3-km Interp GSI 3D-VAR GSI HM Anx Digital Filter 1 hr fcst 18 hr fcst 15 hr fcst 3-km Interp GSI 3D-VAR GSI HM Anx Digital Filter 18 hr fcst 15 hr fcst 13 km RAP 3 km HRRR 13z 14z 15z

32. 3-km Interp 2013 HRRR Initialization from RAP GSI Hybrid GSI HM Anx Digital Filter 1 hr fcst 18 hr fcst GSI Hybrid GSI HM Anx Digital Filter 1 hr fcst 18 hr fcst GSI Hybrid GSI HM Anx Digital Filter 18 hr fcst 3 km HRRR 13z 14z 15z 13 km RAP Refl Obs 1 hr pre-fcst Obs GSI HM Anx GSI 3D-VAR HM Obs 15 hr fcst GFS Ens GFS Ens GFS Ens 45 min of reduced latency

33. Radar Obs 05:00z 18 May 2013 0-hr fcst 3-km radar DA 0-hr fcst NO 3-km radar DA 05z + 0 min 3-km radar data assimilation: Eliminating the spin-up problem

34. Radar Obs 05:15z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 15 min 15-min fcst 3-km radar DA 15-min fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 34 NOAA/ESRL/GSD30 Oct 2013 34 Aviation Model Development

35. Radar Obs 05:30z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 30 min 30-min fcst 3-km radar DA 30-min fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 35 NOAA/ESRL/GSD30 Oct 2013 35 Aviation Model Development

36. Radar Obs 05:45z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 45 min 45-min fcst 3-km radar DA 45-min fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 36 NOAA/ESRL/GSD30 Oct 2013 36 Aviation Model Development

37. Radar Obs 06:00z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 1 hour 1-hr fcst 3-km radar DA 1-hr fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 37 NOAA/ESRL/GSD30 Oct 2013 37 Aviation Model Development

38. Radar Obs 06:30z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 1:30 min 1-hr 30m fcst 3-km radar DA 1-hr 30m fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 38 NOAA/ESRL/GSD30 Oct 2013 38 Aviation Model Development

39. Radar Obs 07:00z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 2hr min 2-hr fcst 3-km radar DA 2-hr fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 39

40. Radar Obs 08:00z 18 May 2013 Improved 0-2 hr Convective Fcsts 05z + 3hr 3-hr fcst 3-km radar DA 3-hr fcst NO 3-km radar DA FPAW - 201324 October 2013RAP/HRRR Model Status 40

41. HRRR (and RAP) Future MilestonesHRRR Milestones Continued HRRR Improvement in 2013 2013 HRRR 2012 HRRR Critical Success Index (CSI) skill score Verification June 1- Sept. 30 (respective seasons)

42. CoSPA: Collaborative effort: ESRL/GSD, NCAR/RAL, MIT/LL Provide 0-8 hr thunderstorm intensity and echo top guidance to aviation community HRRR 15 UTC 08 July 2011 6 hr forecast valid 21 UTC CoSPA 17 UTC 08 July 2011 4 hr forecast valid 21 UTC Observation Valid 21 UTC 08 July 2011 Blend with CIWS HRRR (and RAP) Future MilestonesHRRR Milestones HRRR use in CoSPA FPAW - 201324 October 2013RAP/HRRR Model Status 42

43. 10-11 hr fcst 09-10 hr fcst 08-09 hr fcst 11-12 hr fcst 10-11 hr fcst 09-10 hr fcst Forecasts valid 21-22z 27 April 2011Forecasts valid 22-23z 27 April 2011 All six forecasts combined to form probabilities valid 22z 27 April 2011 HRRR 11z Init HRRR 12z Init HRRR 13z Init Time-lagged Ensemble Spatial radius 45 km Time radius 1 hr UH threshold 25 m 2 /s 2

44. FPAW - 201324 October 2013RAP/HRRR Model Status 44 Example: 28 August 2013 Convection develops in NYC terminal areas around 16 UTC

45. FPAW - 201324 October 2013RAP/HRRR Model Status 45 Obs valid 16z 28 Aug 2013 HRRR Conv Prob Fcst (HCPF) 02 hr fcst valid 16z 28 Aug 2013 Convective coverage now over much of extreme E PA and NJ Probability exceeds 40%

46. Aviation hazard forecasts – all based on RAP and HRRR models (out to 15-18h) Hourly updated 13km Rapid Refresh model forecasts (development supported by FAA/MDE, NOAA) Refreshing from latest observations every hour gives better accuracy

47. 2013 - early 2014 – 2 computers running identical HRRR (and RAPv2) – interim solution – Boulder – computer 1 – Fairmont, WV – computer 2 – Expected reliability at 98.5-99% via coordination of downtimes for Boulder vs. Fairmont computers 2014 – NCEP running RAPv2 (FYQ1) and HRRR (FYQ2) – RAPv2 running in NCEP/NCO testing now – Effort underway to begin testing HRRR at NCEP HRRR Transition to NCEP FPAW - 201324 October 2013RAP/HRRR Model Status 47

48. 48 NCEP ESRL

49. Rapid Refresh for Aviation, Severe Weather, Energy Air Quality WRF NMM WRF ARW WRF: NMM+ARW ETA, RSM GFS, Canadian Global Model Mostly Satellite +Radar North American Mesoscale WRF NMM North American Ensemble Forecast System Hurricane GFDL HWRF Global Forecast System Dispersion ARL’s HYSPLIT For eca st Severe Weather Climate CFS ~2B Obs/Day Short-Range Ensemble Forecast NOAA’s Current NWP Model Production Suite MOM3 Coupled Global Data Assimilation Oceans HYCOM WaveWatch III NAM+CMAQ 49

50. 50 History – NCEP (NMC) Implementations of FAA- funded high-frequency models with aviation focus 1994First operational implementation of RUC 60km resolution, 3-h cycle 199840km resolution, 1-h cycle, Cloud physics, land-surface model 200220km resolution GOES cloud data assimilation 2003Change to 3dVAR analysis for improved assimilation 200513km resolution New observation types (METAR cloud, GPS-PW, new cloud physics) 2008 Assim of radar reflectivity, mesonet winds, 2012 WRF/GSI-based Rapid Refresh (RAP) replaced RUC 2013 (Dec) RAP version 2 – improved storm environment 2014 (Mar) HRRR version 1 – 3km

51. Weather Model Development for Aviation – development of the Rapid Refresh (RAP) and HRRR models – application of HRRR and RAP to aviation applications – hourly data assimilation of latest observations to initialize the RAP and HRRR – Implementation of RAP and HRRR to NOAA operational models at NCEP – inter-lab/center interactions Development - NCAR, MIT/LL, U. Oklahoma, NCEP Operational evaluation - NWS Regions/Offices, AWC, SPC, HPC, FAA Implementation – NCEP (EMC, NCO) Obs radar 21z Sept 16 2010 08z+13h HRRR Valid 21z

52. Weather Model Development for Aviation – inter-lab/center interactions Development – – NCAR, – MIT/LL – U. Oklahoma – NCEP (National Centers for Environmental Prediction – within NWS/NOAA) Operational evaluation – NWS Regions (Eastern, Central, Southern, Western, Alaska) – NWS Forecast Offices (~125 of them) – Aviation Weather Center, Kansas City, MO – Storm Prediction Center, Norman, OK – FAA Implementation – NCEP (EMC, NCO) at NOAA/National Weather Service Obs radar 21z Sept 16 2010 08z+13h HRRR Valid 21z

53. Purpose of FAA Model Development and Enhancement (MDE) Team Improve accuracy of hourly updated weather models, needed especially for increasing aviation requirements. NOAA Research (GSD) leads and coordinates MDE project with other labs -- NCAR, NOAA operations (NWS, NCEP), universities, MIT/Lincoln Labs Steady improvement in weather model guidance is critical to meet future requirements for National Air Space in NextGen era. MDE Product Development Team is the focal point for this effort (along with NOAA operations). NOAA/ESRL/GSD12 July 2012 53 Aviation Model Development

54. 13-km 6hr forecastHRRR 6hr forecast 13-km Resolution Parameterized Convection 3-km Resolution Explicit Convection HRRR (and RAP) Future MilestonesHRRR Milestones Key Advantage of 3-km HRRR ACCURATE STORM STRUCTURE ACCURATE ESTIMATE OF PERMABILITY NO STORM STRUCTURE NO ESTIMATE OF PERMEABILITY 07 June 2012 5 PM EDT observed FPAW - 201324 October 2013RAP/HRRR Model Status 54

55. Take-home Messages NOAA HRRR/RAP weather model grids are the backbone for most aviation hazard guidance products for 1-18h duration, including – G2G-turbulence, – FIP-icing – Upper-level winds – Ceiling/visibility – Terminal forecasts Improvements dependent on transforming increases in atmospheric scientific understanding and computer resources Sustained FAA support has given substantial improvement to US aviation weather guidance 55