1 / 44

HIWPP Hydrostatic Model Sub–Project (3.1) Progress Report

HIWPP Science Team Meeting 13-14 May 2014. HIWPP Hydrostatic Model Sub–Project (3.1) Progress Report. Stan Benjamin NOAA Earth System Research Laboratory. HIWPP Science Team Meeting 13-14 May 2014. HIWPP Hydrostatic Model Sub–Project (3.1) Progress Report. Stan Benjamin

skylar
Télécharger la présentation

HIWPP Hydrostatic Model Sub–Project (3.1) Progress Report

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. HIWPPScience Team Meeting 13-14 May 2014 HIWPPHydrostatic Model Sub–Project (3.1) Progress Report Stan Benjamin NOAA Earth System Research Laboratory

  2. HIWPPScience Team Meeting 13-14 May 2014 HIWPPHydrostatic Model Sub–Project (3.1) Progress Report Stan Benjamin NOAA Earth System Research Laboratory 3.1.1 - assimilation/ensembles/stochastic physics - Jeff Whitaker and Tom Hamill 3.1.2 - parameterization development Georg Grell and Tom Hamill 3.1.3 - GFS and global ensemble/NAEFS – Yuejian Zhu 3.1.4 – FIM Stan Benjamin 3.1.5 - Navy – Melinda Peng and Tim Whitcomb

  3. HIWPP OAR Goals • Accelerate NOAA/US global model development • Non-hydrostatic model (recap from plan) • Determine non-hydrostatic global model that will be the focus of concerted development within OAR as a risk-reduction alternative to non-hydrostatic GSM-GFS and NMMB model development by NCEP • Hydrostatic model 3.1 sub-task goals: • Establish “advanced hydrostatic model” benchmark by which to measure performance of upcoming global non-hydrostatic models • Improve hydrostatic-scale medium-range forecast capability via advanced models and ensembles • Improve components (physics, DA) applicable also to non-hydrostatic models HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  4. Components of HIWPP Hydrostatic Model Methodology • Development • Data assimilation • Ensemble forecasts • Physical parameterizations and hydrostatic models • Verification and evaluation • Retrospective testing • Provides baseline skill for GFS, NAVGEM, FIM models • Modifications to parameterizations, DA, numerics • Multi-model ensembles (e.g., GFS and FIM) for possible NAEFS extensions • Quasi-real-time testing • Advanced higher-resolution deterministic runs • Experimental NAEFS and GEFS extensions (addition of ~10 members each from FIM and NAVGEM) HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  5. Ensemble forecasts for HIWPP hydrostatic models/ensembles • Experimental extension to NAEFS • Add ~10 members each of FIM and NAVGEM at highest resolution possible to GEFS/NCEP (and CMC) • Can value be added to current NAEFS (GEFS + CMC)? • Can value be added to GEFS with multi-model approach (with FIM replacing some GEFS members at same CPU cost)? • High-resolution mini-ensemble from HIWPP deterministic models • ~3-member ensemble at up to ~15km resolution – GFS, FIM, NAVGEM • Is value added with this mini-ensemble?

  6. HIWPP 3.1 – hydrostatic modeling - components • 3.1.1 - assimilation/ensembles/stochastic physics • - Jeff Whitaker and Tom Hamill • 3.1.2 - parameterization development • Georg Grelland Tom Hamill • 3.1.3 - GFS and global ensemble/NAEFS – • Yuejian Zhu • 3.1.4 – FIM • Stan Benjamin • 3.1.5 - Navy – • Melinda Pengand Tim Whitcomb HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  7. 3.1 Milestones – Status – 5/13/14 • First Quarter of FY14: • Identify participants, computer resources, and determine model configurations for hydrostatic tests. [ESRL, EMC, Navy] • Completed in 2nd quarter • Second Quarter of FY14: • Begin retrospective testing and tuning advanced 4D ensemble-variational assimilation system (4D-En-Var) with high-resolution hydrostatic GFS, in collaboration with NCEP [EMC] • Started in 3rd quarter • Begin retrospective runs of hydrostatic models [ESRL-FIM, EMC-GFS, Navy-NAVGEM] • Starting in late 3rd quarter HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  8. 3rd to 4th Quarter of FY14: • Expected implementation of improved, higher-resolution GFS, GEFS, and GSI [EMC] • Planned for 4th Quarter (August) for GFS upgrade, 1st Quarter FY15 for GEFS upgrade • Initial verification results produced for retrospective runs of hydrostatic models (retrospective runs continue) [EMC, Navy, ESRL] • Still can be completed by end of 3rd quarter • Final configuration of a 4D-En-Var system for initializing high-resolution hydrostatic real-time runs. Begin cycling forecasts [ESRL] • Estimated now as 30 September • By August, some preliminary retrospective runs over 1-year period of hydrostatic models for deterministic forecasts may be made available to foster development and evaluation • Estimated now as completed by August • Coordinate with HIWPP Verification team (section 3.5.3) on evaluation of retrospective deterministic and ensemble forecasts from each participating laboratory [ESRL, Navy, EMC] • Will start in 3rd quarter (even on Wed 14 May in Science Meeting) HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  9. Fourth Quarter of FY14: • By September, retrospective runs over 1-year period of hydrostatic models for ensemble forecasts are complete using whatever configuration deemed optimal for each laboratory [ESRL, NRL, EMC] • Complete by 1st Quarter FY15. • Coverage will be for the majority of a 1-year period. • NCEP will complete test runs with the T1534 GFS for selected periods: • 2013 – 15 May – 25 October (retrospective), • 2014 – real-time from Feb 2014ongoing • NCEP will provide T1534 GFS initial files to ESRL, who will, in turn, provide to NRL • Both FIM and NAVGEM for these test period will be initialized from the GFS T1534 initial conditions

  10. Configurations for HIWPP experimentalhydrostatic retrospective and real-time runs HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  11. Recent progress toward HIWPP retrospective testing • FIM code (variation of ss2icos routines) can now efficiently read T1534 GFS sanl/sfcanl (native/native spectral/gaussian data) • Completed in April-May 2014 – thanks to NCEP for access via HPSS • Code will be provided to NRL • FIM providing efficient 1/8 deg output • Now used routinely in FIM experimental runs. • Also available – 13km CONUS (AWIPS 130 grid) and 13km Rapid Refresh (AWIPS 83 grid) HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  12. First Quarter of FY15: • By October, decision made based on initial retrospective results on hydrostatic model for quasi-real-time ensembles: quasi-real-time forecasts from all HIWPP hydrostatic models will be made available for deterministic and ensemble forecasts through HIWPP. A decision on mini-ensemble from high-resolution models will be made by a working group [ESRL, EMC, Navy members] • OK • Configuration is finalized for quasi-real-time experimental forecasts from hydrostatic models at up to 10-15-km resolution [ESRL, EMC, Navy] • OK for completing this also by 31 October • Collect and synthesize feedback from beta testers on their impressions of real-time demonstration of hydrostatic models [ESRL] • This is expected to start in November 2014 to January 2015. • *NB: EMC inputs for these tasks, here and subsequently, are derived from operational forecasts (GFS 13-km or GEFS as appropriate) HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  13. Second Quarter of FY15: • January: Start of quasi-real-time experimental forecasts from hydrostatic models at up to 10-15-km resolution [ESRL, EMC*, Navy] • Current best estimate – During November 2014 • January: Start of quasi-real-time ensemble forecasts (0-16 days) from hydrostatic models [ESRL, EMC*, Navy] • Same as above • Third Quarter of FY15: • Continue quasi-real-time experimental runs from hydrostatic models at up to 10-15-km resolution [ESRL, EMC*, Navy] • Continue quasi-real-time experimental ensemble forecasts (0-16 days) from hydrostatic models [ESRL, EMC*, Navy] • Produce draft report on hydrostatic real-time tests, and submit these results to a peer-reviewed journal [ESRL, EMC, Navy] HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  14. HIWPP 3.1 – hydrostatic modeling - components • 3.1.1 - assimilation/ensembles/stochastic physics • - Jeff Whitaker and Tom Hamill • 3.1.2 - parameterization development • Georg Grelland Tom Hamill • 3.1.3 - GFS and global ensemble/NAEFS – • Yuejian Zhu • 3.1.4 – FIM • Stan Benjamin • 3.1.5 - Navy – • Melinda Pengand Tim Whitcomb HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  15. 3.1.1 Assimilation/Ensembles/Stochastic Physics: progress so far • Assimilation: • New hire in PSD (Lili Lei) started April 1. • She will be testing aspects of the new 4D-EnVar system with the GFS • Incremental Analysis Update (IAU) capability introduced in GSI/GFS to reduce imbalances in the assimilation cycle. Now testing (anticipated to be part of FY15 4D-EnVar upgrade). • Stochastic Physics (Philip Pegion): • New stochastic physics schemes implemented and tested in the GFS, will be implemented within the EnKF cycle in FY14 upgrade. Code now being transferred to NEMS framework for FY15 GFS upgrade.

  16. HIWPP 3.1 – hydrostatic modeling - components • 3.1.1 - assimilation/ensembles/stochastic physics • - Jeff Whitaker and Tom Hamill • 3.1.2 - parameterization development • Georg Grelland Tom Hamill • 3.1.3 - GFS and global ensemble/NAEFS – • Yuejian Zhu • 3.1.4 – FIM • Stan Benjamin • 3.1.5 - Navy – • Melinda Pengand Tim Whitcomb HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  17. 3.1.2 Parameterizations: a) unified representation of turbulence and cloudsb) scale- and aerosol-aware stochastic convective parameterization • Two hires involved • Position 1 • Position advertised when funds became available at NCEP, early calendar 2014. • Interviewed and selected Alex Belochitski, who will start at NCEP on 15 July after finishing pending work at GFDL. • Project delayed by roughly 3-6 months given late arrival of funds and partly because our postdoc needed two more months to fulfill existing commitments. • Alex has experience at NCEP; expect him to be efficient once he starts. • Position 2 • Hiring a post-doc (to free up Georg Grell for HIWPP physics development) took 3 attempts, the third attempt appears to be successful. Hopeful starting date June 2nd.

  18. 3.1.2 Parameterizations: a) unified representation of turbulence and cloudsb) scale- and aerosol-aware stochastic convective parameterization • Initial testing of using scale-aware Grell-Freitas (GF) scheme (deep and shallow) within GFS physics within FIM, also direct coupling of convective parameterization to radiation • Assessing option to fuse GF scheme directly into the GFS physics package (ultimately faster than going through WRF cumulus driver) • Version exists for Grell-Devenyischeme within GFS • Initial testing of aerosol awareness with GF with WGNE test case has begun. Currently limited to WRF experiments to compare simple approach to sophisticated approach

  19. COMPLETED (1) Unified representation of turbulence and clouds: First Quarter of FY14 (unified physics): • Identify and hire postdoctoral associate at NCEP. • Begin including new physics modules in the single-column model (SCM) based on the Global Forecast System (GFS) model. Second Quarter of FY14: • Complete implementation of new physics modules into GFS SCM. • Perform SCM experiments with both the current operational GFS physics and new physics; compare results. Third Quarter of FY14: • Use DOE ARM program observations/analyses to initialize the SCM and to evaluate/validate the SCM performance. • Tune physics if necessary. • Begin including advanced physics into the global forecast model. (2) Scale- and aerosol-aware stochastic convective parameterization Second Quarter of FY14: • Hire of CIRES scientist Fourth Quarter of FY14: • Tuning of stochastic convective parameterization using EMC evaluation metrics finished

  20. HIWPP 3.1 – hydrostatic modeling - components • 3.1.1 - assimilation/ensembles/stochastic physics • - Jeff Whitaker and Tom Hamill • 3.1.2 - parameterization development • Georg Grelland Tom Hamill • 3.1.3 - GFS and global ensemble/NAEFS – • Yuejian Zhu • 3.1.4 – FIM • Stan Benjamin • 3.1.5 - Navy – • Melinda Pengand Tim Whitcomb HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  21. Update of HIWPP Hydrostatic Model- Section 3.1.3 Yuejian Zhu HIWPP quarterly meeting 13-14 May 2014 Boulder, CO Acknowledgment: John Derber

  22. Components of HIWPP Hydrostatic Model Methodology (Stan’s earlier slide) • Development • Data assimilation • Ensemble forecasts • Physical parameterizations • Verification and evaluation • Retrospective testing • Baseline for GFS, NAVGEM, FIM • Modifications to parameterizations, DA, numerics • Multi-model ensembles (e.g., GFS and FIM) for possible NAEFS extensions • Quasi-real-time testing • Advanced higher-resolution deterministic runs • Experimental possible NAEFS extensions (addition of ~10 members each from FIM and NAVGEM) and multi-model GEFS configuration (10 FIM members replacing 10 GSM/GFS members)

  23. Next NCEP Analysis Implementation Highlights • Structure • T574 analysis for T1534 deterministic • Single scale background error (for efficiency) • Code optimization • Observations • GPSRO enhancements • Updates to radiance assimilation • Assimilate SSM/IS UPP LAS data • CRTM v2.1.3 • New bias correction • Additional satwind data – hourly GOES, EUMETSAT • EnKFmodifications • Stochastic physics in EnKF forecasts • T574L64 EnKFensembles

  24. Next NCEP GFS Model Highlights • T1534 Semi-Lagrangian (~13km) version of Global Spectral Model • Use of high resolution daily SST and sea ice analysis • High resolution until 10 days • Physics • Radiation modifications • Reduced drag coefficient at high wind speeds • Stationary convective gravity wave drag • Consistent GFS diagnosis snow accumulation in post and model • Compute and output frozen precipitation fraction • Hybrid EDMF PBL scheme and TKE dissipative heating • Land Surface • Soil Moisture climatology from CFSv2 • Changes to roughness length calculations

  25. Next GFSimplementation Highlights • Retrospective runs (GFS/GDAS) • 2014: 01/01/2014 – current • 2013: 05/15/2013 – 10/25/2013 (plan to 10/31/2013) • 2011: 05/20/2011 – 09/01/2011– on R&D machine (zeus) • Same files to be used for FIM and NAVGEM HIWPP retro tests • Post-processed Forecast data • Resolution – 0.25 lat/lon • Frequency – hrly out to 12 hrs, 3hrly out to 10 days • Format – GRIB II • Tentative schedule for implementation • August 2014

  26. Next NCEP GEFS Highlights • Model • GFS Semi-Lagrangian model • Resolution • T574 (T382 physics - 34km for 0-168 hours???), T382(T254 physics – 55km for 168-384 hours) for 64 vertical hybrid levels • Output: • Every 6-hr for 0.5*0.5 degree pgrb files • Initial perturbations • Base: EnKF 6hr forecast perturbations • Possible additions for initialization: • Tropical Storm relocation • Centralization • Ensemble transform (under investigation, un-necessary if there is no significant difference of the performance) • Rescaling (un-necessary if we confirm our parallel EnKF’s changes) • Stochastic perturbations • Tune STTP for model change and initial perturbation changes • Turn off stochastic perturbations for surface pressure in STTP • Tentative schedule for next implementation – Q1FY2015

  27. Ensemble forecasts for HIWPP hydrostatic models/ensembles (Stan’s slide) • Experimental extension to NAEFS • Add 10-20 members each of FIM and NAVGEM at highest resolution possible to GEFS/NCEP (and CMC) • Can value be added to current NAEFS (GEFS + CMC)? • Can value be added to GEFS with multi-model approach (with FIM replacing some GEFS members at same CPU cost)? • High-resolution mini-ensemble from HIWPP deterministic models • ~3-member ensemble at up to ~15km resolution – GFS, FIM, NAVGEM

  28. HIWPP 3.1 – hydrostatic modeling - components • 3.1.1 - assimilation/ensembles/stochastic physics • - Jeff Whitaker and Tom Hamill • 3.1.2 - parameterization development • Georg Grelland Tom Hamill • 3.1.3 - GFS and global ensemble/NAEFS – • Yuejian Zhu • 3.1.4 – FIM • Stan Benjamin • 3.1.5 - Navy – • Melinda Pengand Tim Whitcomb HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  29. FIM numerical atmospheric model • Horizontal grid • Icosahedral, Arak A grid – testing at 60km/30km/15km/10km • Vertical grid • Staggered Lorenz grid, ptop = 0.5 hPa, θtop ~2200K • Generalized vertical coordinate • Hybrid θ-σ option (64L, options for 100L, 38L, 21L) • GFS σ-p option (64 levels) • Numerics • Adams-Bashforth3rd-order time differencing • Flux-corrected transport • Physics • GFS physics suite (May 2011 version), options for Grell-Freitas cumulus, MYNN PBL, GFS-2012 suite, McICA cl/rad • Coupled model extension options • Chem – WRF-chem/GOCART • Ocean – icosahedral HYCOM (no flux coupler), tri-polar HYCOM (with coupler)

  30. FIM changes for early 2014 • Numerics/grid • More accurate interpolation for edge momentum values • More regular icosahedral grid (following slide) • Hybrid vertical coordinate – limit regridding to 1dz • Alt_landuse option (with 1km USGS) added and in use • Finer coastline land/water resolution with FIM 15km/10km versions • Post-processing • Icos-grid smoothing by variable • Output of 1/8 grid data including for track/intensity TC forecasts HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  31. Standard recursive icosahedral grid Used in FIM until 2013 Icosahedral grids at refinement level 5 (rhombus sides with 2**5 grid points) Modified recursive icosahedral grid (less distortion) Used in FIM starting early 2014

  32. 3-year (2011-2013) Atlantic hurricane intensity error results • Smaller (negative = underforecast) is better • 2014 FIM model has lowest intensity errors vs. GFS and 2013 FIM at almost all forecast times

  33. RMS errors (smaller better) - verification with rawinsonde observations Recent 3-month period – Feb-May2014 Wind/RH RMS error vs. raobs 72h forecasts N. America GFS// FIM 72h wind forecast GFS// FIM 72h RH forecast GFS// FIM FIM GFS BetterBetter FIM GFS BetterBetter

  34. FIM changes for later 2014 Areas of ongoing work for consideration by late 2014 • Numerics • Revised vertical hybrid isentropic-sigma coordinate • Physics • Use of physics tendency at each dynamics time step • Significantly reduced vertical momentum diffusion (in GFS physics) • May 2012 GFS physics implementation • Grell-Freitas (2013, ACPD) physics - scale-aware, aerosol-aware cumulus (deep/shallow) • Stochastic versions for GFS physics (with HIWPP) • Data assimilation • Complete development of ensemble DA with FIM/GSI following GFS EnKF/hybrid HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  35. HIWPP 3.1 – hydrostatic modeling - components • 3.1.1 - assimilation/ensembles/stochastic physics • - Jeff Whitaker and Tom Hamill • 3.1.2 - parameterization development • Georg Grelland Tom Hamill • 3.1.3 - GFS and global ensemble/NAEFS – • Yuejian Zhu • 3.1.4 – FIM • Stan Benjamin • 3.1.5 - Navy – • Melinda Pengand Tim Whitcomb HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

  36. Navy HIWPP Status Tim Whitcomb Melinda Peng Marine Meteorology Division Naval Research Laboratory

  37. NAVGEM HIWPP Configuration • Horizontal resolution increased to T639 (~21km at equator) • Vertical resolution increased to 64 levels • accommodate higher model top at 0.00566 hPa • places top level above stratospheric jet • 450 second timestep, perturbation theta dynamics, non-orographic GWD, water vapor photochemistry • Added new capability for ¼ degree output

  38. 500.0 hPa Dewpoint Depression: 26 June 2013 T+0h

  39. 500.0 hPa Dewpoint Depression: 26 June 2013 T+120h Sharper Gradients Greater fidelity representing local variability

  40. Done: Read in IC files Current: Repackage fields for use by NAVGEM Shortly Thereafter: Validation tests HIWPP Initial Conditions NAVGEM Forecast Model HIWPP Forecast Output

  41. Done: Set up test model configuration and modifications for output Current: Scaling analysis and I/O requirements Shortly Thereafter: Parameterization adjustment and tests with HIWPP initial conditions HIWPP Initial Conditions NAVGEM Forecast Model HIWPP Forecast Output

  42. HIWPP Initial Conditions NAVGEM Forecast Model Done: Adapt COAMPS GRIB2 Conversion package for NAVGEM Current: Iterate with NOAA team to check packaging Shortly Thereafter: Test with higher-resolution output and add required variables HIWPP Forecast Output

  43. Remaining Issues • Computational • working on faster I/O methods • scaling analysis to determine optimal compute configuration • Model Configuration • Subsample 1-year range (depends on computer) • Parameterization adjustment • What should we use as a lower boundary condition (e.g. SST, sea ice)?

  44. HIWPP Science Team Meeting 13-14 May 2014 - Hydrostatic (3.1) Sub-Project

More Related