Facilitating Research into Operation: NCAR perspective
Roger WakimotoNCAR
Member Institutions
Board of Trustees
Finance & Administration
Katy Schmoll, VP
Corporate AffairsJack Fellows, VP
NCARRoger Wakimoto, Director
Maura Hagan, Deputy Director
UCAR Community Programs (UCP)
Jack Fellows, Director Hanne Mauriello, B&A
Computational & Information Systems Laboratory (CISL)
Al Kellie, Associate Director
NCAR Earth System Laboratory (NESL)
Greg Holland, Interim Associate Director
Research Applications Laboratory (RAL)
Brant Foote, Associate Director
Earth Observing Laboratory (EOL)
Sue Schauffler, Acting Associate Director
Cooperative Program for Operational Meteorology
Education & Training (COMET)
Tim Spangler
Constellation Observing System for Meteorology Ionosphere
Climate (COSMIC)
Bill Kuo
Unidata
Mohan Ramamurthy
Joint Office for Science Support (JOSS)
Gene Martin
Global Learning & Observations to Benefit the
Environment (GLOBE)
Ed Geary
Digital LearningSciences (DLS)
Karon Kelly
Visiting Scientists Program(VSP)
Meg Austin
National Science Digital Library (NSDL)
Kaye Howe
Advanced Study Program
(ASP)Chris Davis
UCARRichard Anthes,
President
Integrated Science Program (ISP) Peter Backlund
Budget & Planning(B&P)
Rena Brasher-Alleva
Research Relations
Peter Backlund
NCAR LibraryMary Marlino
High Altitude Observatory (HAO)
Michael Thompson, Associate Director
Computational & Information Systems Laboratory (CISL)
Al Kellie, Associate Director
Institute for Mathematics Applied to Geosciences (IMAGe)
Doug Nychka
Operations & Services DivisionAnke Kamrath
Technology Development DivisionRich Loft
NCAR Earth System Laboratory (NESL)Greg Holland, Interim Associate Director
Climate & Global Dynamics Division (CGD)Bill Large
High Altitude Observatory (HAO)Michael Thompson, Associate Director
Mesoscale & Microscale Meteorology Division (MMM)Rich Rotunno (Interim)
Research Applications Laboratory (RAL)Brant Foote, Associate Director
Aviation Applications Program (AAP)Bruce Carmichael
Hydrometeorological Applications Program (HAP)Roy Rasmussen
Joint Numerical Testbed (JNT)Barb Brown
National Security Applications Program (NSAP)Scott Swerdlin
Weather Systems & Assessment Program (WSAP)Bill Mahoney
Earth Observing Laboratory (EOL)Roger Wakimoto, Associate Director
Cyberinfrastructure & Data Services (CDS)Michael Daniels
Design & Fabrication Services (DFS)Jim Moore, Interim
Field Project Services (FPS)Brigitte Baeuerle
In-Situ Sensing Facility (ISF)Stephen Cohn
Research Aviation Facility (RAF)Jeff Stith
Remote Sensing Facility (RSF)Wen-Chau Lee (Acting)
Technology Development Facility (TDF)Alan Fried
Atmospheric Chemistry Division (ACD)Bill Randel
Governing Entities
UCAR (University Corporation for Atmospheric Research)
NCAR (National Center for Atmospheric Research) and Labs
Facility & Service Providers
Scientific Research Divisions
Education & Application Programs
UOP (UCAR Office of Programs)
Institute
Climate Science and Applications Program (CSAP)Lawrence Buja
NCAR involvement with DTC
• Approximately 2/3 of DTC staff reside in RAL’s Joint Numerical Testbed
• NCAR hosts the DTC Director’s Office• NCAR/NESL/MMM provides support for
the WRF modeling community• Modeling research & development
conducted at NCAR have significant potentials for operational applications
Worldwide WRF User Participation 130 Foreign Countries
Registered Users 8/08/10 U.S. Universities, govt. labs, and private sector 4759
Foreign users 9272 -------- 14031 5070 active subscribers to [email protected]
425 email inquiries per month
WRF Registered Users
WRF User Community• WRF is a modeling system jointly developed by the
research and operational community over the past decade• The number of registered user exceeds 14,000, and it
keeps growing.• This is a big community of which operational centers can
leverage its research results. • NCAR/MMM, NCEP/EMC, and DTC provide community
user support for WRF modeling system• NCEP is transitioning to the NEMS framework for its
modeling systems. • We need to ensure there is a path for research-to-operation
transition. Model software framework should not become an obstacle for collaboration.
NCAR’s Modeling R&D: Some examples
• NESL: – Development of Model for Prediction Across Scales
(MPAS)– WRF 4D-Var system– WRF EnKF data assimilation– Radar data assimilation– Cloud-scale ensemble prediction
• CISL:– DART (Data Assimilation Research Testbed)
• RAL:– Model verification tools development– Aviation weather forecasting
Based on Voronoi Tesselations (hexagons)
Jointly developed, primarily by NCAR and LANL, for weather, regional climate and climate applications.
MPAS infrastructure - NCAR, LANL, others.MPAS - Atmosphere (NCAR)MPAS - Ocean (LANL)MPAS - Ice, etc.
Bill Skamarock, Joe Klemp, Michael Duda, Sang-Hun Park, Laura Fowler NCAR/NSFTodd Ringler LANL/DOEJohn Thuburn Exeter UniversityMax Gunzburger Florida State UniversityLili Ju University of South Carolina
MPAS: Model for Prediction Across Scales
Prediction Across Scales
Vertical velocity contours at 1, 5, and 10 km (c.i. = 3 m/s)
30 m/s vertical velocity surface shaded in redRainwater surfaces shaded as transparent shellsPerturbation surface temperature shaded on baseplane
500 m cell spacingSupercell at 2 hours
Jablonowski and Williamson baroclinic wave test case.Relative vorticity (s-1), day 16 (jet level)
MPAS formulation is demonstrably accurate and efficient at both large and small scales, and should scale well on next-generation supercomputers.
MPAS-Atmosphere, global configuration MPAS-Atmosphere, cloud-model configuration
MPAS: Variable Resolution Capability
• Initial test results using variable resolution grids are encouraging (baroclinic waves and squall lines).• Atmospheric (hydrostatic and nonhydrostatic) and ocean solvers are
robust on these grids.
Local refinement capabilities are critical for NWP and regional climate applications.
MPAS - Current Status - August 20103D Solvers• Hydrostatic 3D SVCT solver (pressure coordinate).• Nonhydrostatic 3D SVCT solver (height coordinate).• Both solvers work on the sphere and on 2D and 3D Cartesian
domains.• Tests results confirm viability of Voronoi C-grid discretization at
large scales (global) and cloud-permitting scales for both solvers.
• Variable-resolution grid results are encouraging.
Present and Future Development• Weather, regional climate and climate physics suites.• Further testing of variable resolution meshes, physics
development.• Further development and testing of higher-order transport
schemes.
Expectations• NWP testing by early next year.• Friendly-user release summer/fall 2011.
Software Engineering
MPAS software:• Developed based on MPAS applications requirements.• Lightweight (for rapid prototyping, ease of maintenance).• We have developed and use a Registry similar to that used
in the WRF infrastructure.• We are exploring using other model physics (e.g. from
CCSM, WRF, GFS) directly from the other models’ repositories (in the spirit of Kalnay et al’s Interchange of Physical Parameterizations proposal in BAMS 1989 620-622).
• We are considering possible model couplers (e.g. CPL7).• We have not adopted any formal software framework.• We are considering lightweight, industry-standard
approaches for the MPAS software infrastructure (e.g. OOPS - Object-Oriented Prediction System being developed at ECMWF).
MPAS - Current Status - August 2010
The Ensemble Kalman Filter (EnKF)
• EnKF combines data assimilation and ensemble forecasting– Analysis step produces ensemble of analyses, given new
observations– Analysis step employs covariance, estimated from short-range
ensemble– In forecast step, make ensemble of short-range forecasts from
ensemble of analyses
• Attractions for mesoscale applications– Few assumptions about covariances, so applicable to range of
scales/phenomena– Flexible to details of model, such as complex microphysical
schemes– Ease of implementation and parallelization; no adjoints
• For applications here, use 50-100 members
Data Assimilation Research Testbed (DART)
• Provides general, model-independent algorithms for ensemble filtering
• Numerous DART-compliant models– ARW, CAM, NOGAPS, …– Applications to atmospheric, ocean, and space weather data
assimilation
• Parallel analysis scheme that scales well to 100’s of processors
• See http://www.image.ucar.edu/DAReS/DART/
Real-Time Analyses for Tropical Cyclones
• Analyses from WRF/DART provided ICs for NCAR’s high-res TC forecasts during 2009 season
• Produced 36-km analyses every 6 h – Assimilate conventional obs + satellite winds + vortex position, intensity– NO bogussing of any kind; no satellite radiances
• WRF configuration– “hurricane” physics + KF convection– 36 km, with stationary 12-km nest centered on each TC/TS/TD
• System cycled continuously for ~ 4 months
Real-Time Analyses for Tropical Cyclones (cont.)
Real-Time Analyses for Tropical Cyclones (cont.)
• Analyses captured all 2009 storms, from depressions to hurricanes.– No need to bogus– No spurious storms, despite not assimilating radiances
RMS fits of analysis and 6-h forecast to best-track estimates
Courtesy R. Torn
Real-Time Analyses for Tropical Cyclones (cont.)
• Analysis increment from position observation– Reflects covariance (wind speed, vortex position),
which in turn reflects vortex structure– Shifts vortex coherently and consistently in all
model fieldsHurricane Bill, 00Z 19 Aug 2009
Wind speed @ 1st level
contours: ens.-mean, 6-h forecast
colors: increment given obs of vortex position (analysis - forecast)
Courtesy S. Cavallo
Radar data assimilation slides
b) WRF 1h FCST
WRF-VAR radar data assimilation testing in central U.S.
• Radar data assimilation systems are being tested and improved over selected U.S. regions and foreign countries
• These systems include GSI, WRF-VAR, and WRF/DART.
• 3DVAR, 4DVAR, and EnKF are all involved in the testing/development.
• These systems are tested with WRF as the forecast model and with a goal to support operational applications.
WRF 3h FCSTRadar Mosiac
Beijing operational pre-testing
25 radars assimilated
Different requirements in Operation and Research NWP
• Operation:– Robustness– Efficiency– Easy maintenance– Prefers incremental
changes– Thorough testing
• Research:– Flexibility– Multiple-choices– User friendliness– Community support– Innovation
Effective Research to Operation Transition
• Need to challenge (attract) research community to work on problems of interest to operation
• Need a clear path for R2O• Need to minimize obstacles preventing
R2O (e.g., model software framework)• Need to provide sufficient support for
O2R and R2O activities
Concluding Remarks
• NCAR is committed to:– Support the WRF community for many
more years to come.– Support the development of next
generation modeling and data assimilation systems (e.g., MPAS, DART/EnKF)
– Support the research to operation transition through the DTC