Weather Research and Forecast (WRF) Modeling System

1. Weather Research and Forecast (WRF) Modeling System Ü Develop an advanced mesoscale forecast and assimilation system Ü Promote closer ties between research and operations Context: Design for 1-10 km horizontal grids Advanced data assimilation and model physics Accurate and efficient across a broad range of scales Well-suited for both research and operations Community model support http://wrf-model.org http://wrf-model.org

2. WRF Events for 2000 12 January First WRF Oversight Board Meeting 14 February WRF Planning Meeting 29-30 March WRF Planning Workshop 23 June First Annual WRF Users Workshop First Meeting of WRF Science Board 30 October Release of “bare-bones” WRF Model WRF Status,updates and codes available from: wrf-model.org

3. WRF Project Collaborators • Original Partners: • NCAR Mesoscale and Microscale Meteorology Division • NOAA National Centers for Environmental Prediction • NOAA Forecast Systems Laboratory • OU Center for the Analysis and Prediction of Storms • Additional Collaborators: • Air Force Weather Agency • NOAA Geophysical Fluid Dynamics Laboratory • NASA GSFC Atmospheric Sciences Division • NOAA National Severe Storms Laboratory • NRL Marine Meteorology Division • EPA Atmospheric Modeling Division • University Community

4. WRFProject Management S. Lord, Chair NOAA/NCEP S. MacDonald FSL & GFDL R. Gall NCAR/MMM S. Nelson NSF/ATM Col. C. Benson USAF/AFWA Capt. C. Gunderson NAVY G. Kulesa FAA WRF Oversight Board Joe Klemp NCAR/MMM WRF Coordinator WRF Science Board WRF Development Teams (5)

5. Morris Bender NOAA/OAR Stanley Benjamin NOAA/OAR Daewon Byun NOAA/ARL Mark DeMaria NOAA/NESDIS Jim Doyle NRL Jimy Dudhia NCAR Michael Farrar USAF/AFWA John Manobianco NASA/ENSCO Jeffrey McQueen NOAA/OAR Russell Schneider NOAA/NWS Nelson Seaman Penn State U. Danny Sims FAA/ACT-320 David Stensrud NOAA/OAR Wei-Kuo Tao NASA/GSFC Eric Thaler NOAA/NWS Greg Tripoli U. Wisconsin Robert Wilhelmson U. Illinois Ming Xue Oklahoma U./CAPS WRF Science Board

6. WRF Development Teams

7. WRF Software Objectives • Performance-Portable • Performance: scaling and time to solution • Architecture independence • No specification of external packages • Run-Time Configurable • Scenarios, domain sizes, nest configurations • Dynamical-core and physics • Maintainability & Extensibility • Single source code • Modular, hierarchical design, coding standards • Plug compatible physics, dynamical cores http://www.mmm.ucar.edu/wrf/WG2/WRF_conventions.html

8. Single version of code enabled for efficient execution on: Distributed-memory multiprocessors Shared-memory multiprocessors Distributed memory clusters of SMPs Logical domain 1 Patch, divided into multiple tiles Inter-processor communication WRF Multi-Layer Domain Decomposition • Model domains are decomposed for parallelism on two-levels • Patch: section of model domain allocated to a distributed memory node • Tile: section of a patch allocated to a shared-memory processor within a node • Distributed memory parallelism is over patches; shared memory parallelism is over tiles within patches

9. Top-level “Driver” layer Isolates computer architecture concerns Manages execution over multiple nested domains Provides top level control over parallelism patch-decomposition inter-processor communication shared-memory parallelism Controls Input/Output “Mediation” Layer Specific calls to parallel mechanisms Low-Level “Model” layer Performs actual model computations Tile-callable Scientists insulated from parallelism General, fully reusable Mediation Layer uv prep filter scalars physics big_step decouple recouple advance w advance Model Layer WRF Hierarchical Software Architecture Driver Layer wrf initial_ config alloc _and_configure init _domain integrate solve_interface solve

10. Alpha workstation (EV56) 30 25 20 15 10 5 81 41 Y tile dimension 0 21 21 41 81 X tile dimension VPP 5000 100 80 60 40 20 0 -20 -40 81 -60 41 Y tile -80 dimension 21 21 41 81 X tile dimension Penalty for IJK Loop & Storage Ordering • IJK versus KIJ for all patch dimensions X,Y=(21,41,81); 41 levels throughout • Penalty for IJK decreases with increased length of minor dimension, X • Penalty is most severe for sizes typical of a DM patch • IJK is strongly favored by vector for adequate length of X • Surprise: vector prefers KIJ for short X; but an unlikely result once full physics • IKJ has been chosen for loop and storage ordering

11. Numerics for Dynamical Solver • Numerical Modeling Issues: • Equations / variables • Vertical coordinate • Terrain representation • Grid staggering • Time Integration scheme • Advection scheme • Strategy • Identify and analyze alternative procedures • Evaluate alternates in idealized simulations • Evaluate in NWP applications as model complexity increases

12. Treatment of Terrain by Vertical Coordinate Terrain Following • Smooth topography well represented • Selective resolution enhancement near ground • Potential for spurious circulations above steep terrain • Can represent blocking due to step terrain • Reduced errors in computing horizontal gradients • Degraded representation of sloped topography • Maintains horizontal coordinate surfaces • Represents terrain slope accurately • Potential complications in numerics for shaved cells Step Mountain Shaved Cell / Partial Step

13. Prototype Nonhydrostatic Model Solvers • Split-Explicit Eulerian Model: • Pressure and temperature diagnosed from thermodynamics • Two time level split-explicit time integration • Flux-form prognostic equations in terms of conserved variables • Accurate shape preserving advection • Both terrain-following height and mass coordinates being tested • Semi-Implicit Semi-Lagrangian Model: • Unstaggered (A) grid • Forward trajectories with cascade interpolation back to grid • High order compact differencing • Terrain following hybrid coordinate • Runge-Kutta (3rd & 4th order) time integration

14. 5 min 10 min 15 min Comparison of Gravity Current Simulations Height Coordinate Mass Coordinate

15. Time-Split Leapfrog and Runge-Kutta Integration Schemes

16. Two Examples of Possible Vertical Coordinate Structures With The General Hybrid Coordinate

17. Strategy for WRF Model Physics • Define “plug-compatible” interface for physics modules • Implement and test basic physics in WRF: • Kessler-type (no-ice) microphysics • Lin et al. (graupel included) microphysics • Kain-Fritsch & Betts-Miller-Janjic cumulus parameterizations • Shortwave radiation (cloud-interactive) from MM5 • Longwave radiation (RRTM) • MRF (Hong and Pan) PBL • Blackadar surface slab ground temperature prediction • NCEP working on the NOAH LSM for WRF • Implement a complete suite of research physics packages • Encourage and facilitate community involvement in advanced model physics development and evaluation

18. WRF 3D-Variational Data-Assimilation System • Essential features of initial 3D-Var system: • Basic quality control • Assimilation of conventional observations (surface, radiosonde, aircraft) • Multivariate analysis • Adherence to WRF coding standards • Additional features to be added: • 3-D anisotropic background errors using recursive filters • Additional observation operators (radar, satellite, wind profiler, etc.) • Flexible choice of first guess • Further enhancements

19. WRF Model Testing and Verification Strategy • Analytic and converged numerical solutions • Inviscid dynamics (baroclinic instability, frontogenesis) • Buoyancy driven flow (gravity currents, warm thermals) • Topographic flow (nonhydrostatic, hydrostatic, inertial-gravity mountain waves) • Moist convection (idealized convection with constant eddy mixing) • Regime dependence of nonlinear flows • Topographic flow (finite amplitude waves, wave overturning, lee vortices) • Moist convection (convective behavior as a function of CAPE and shear) • Observational case studies • Extratropical cyclones (STORM-FEST case) • Topographic flow (downslope windstorm, orographic precip., cold-air damming) • Moist convection (supercell case, squall-line case, multi-parameter radar case) • PBL-surface physics (1-D diurnal cycle, sea-breeze case, marine inversion&CTD) • Tropical cyclone (COMPARE case)

20. Pre-implementation Strategy for WRF Model Testing & Validation GOAL: perform clean operational vs WRF comparisons • Convert existing Meso Eta Model into WRF modeling infrastructure • use selectable dynamics WRF option • use tested nonhydrostatic component of Meso • Compare computer performance of WRF vs operations • measure performance benefit or penalty of WRF design • if significant penalty is measured, then redesign is called for • if no penalty, then could immediately implement WRF modeling infrastructure into NCEP operations for both nested & continental Meso • Compare forecast performance of WRF vs operations • Emphasis on REAL-DATA retrospective case studies • Small and large-domain capabilities examined for nested and continental requirements of NCEP operations

21. Timeline for WRF Project Development Task 2000 2001 2002 2003 2004 2005-8 Basic WRF model (single dynamic core, limited physics, standard initialization) WRF model with selectable dynamic cores WRF model with hybrid vertical coordinate Research quality NWP version of WRF Basic Research suite Advanced suite Simple WRF model physics Advanced Basic Research 3-D VAR assimilation system Basic Advanced 4-D VAR assimilation system Testing for initial operational use at NCEP, AFWA and FSL Routine diagnosis of operational performance & of future refinements Release and support to community Implement into operations