Status of NCARs Data Assimilation Initiative (DAI) Report to NCAR - - PowerPoint PPT Presentation

▶
status of ncar s data assimilation initiative dai
SMART_READER_LITE
LIVE PREVIEW

Status of NCARs Data Assimilation Initiative (DAI) Report to NCAR - - PowerPoint PPT Presentation

Jan 13, 2024 35 likes •298 views

Status of NCARs Data Assimilation Initiative (DAI) Report to NCAR Directors Committee 17 April, 2003 DAI aims to create and lead a research community for data assimilation where individu- als benefit from sharing ideas, methodologies, and

slide-1
SLIDE 1

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Status of NCAR’s Data Assimilation Initiative (DAI)

Report to NCAR Directors Committee 17 April, 2003 “DAI aims to create and lead a research community for data assimilation where individu- als benefit from sharing ideas, methodologies, and software tools as well as access to a data assimilation testbed.” “With a carefully controlled but rapid expansion of DAI, it should be possible for NCAR to be recognized as a leader in research data assimilation within 4 years (FY07).” Summary: On track, so far...

slide-2
SLIDE 2

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

History of DAI: Organization

  • Sep. ‘01
  • Sep. ‘02
  • Sep. ‘03

Nychka, Tribbia and Snyder formu- late DA plans Anderson on IPA to coordinate DAI Initiative Proposal prepared FY03 budget request and three year plan

slide-3
SLIDE 3

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

History of DAI: Funding

  • Sep. ‘01
  • Sep. ‘02
  • Sep. ‘03

Anderson funded by NOAA FY01 Carryover, $67K FY02: $370K FY03: $700K

slide-4
SLIDE 4

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

History of DAI: Personnel

  • Sep. ‘01
  • Sep. ‘02
  • Sep. ‘03

Nychka, Snyder, Tribbia (1/6) Anderson Tim Hoar (50%) Steve Aulenbach Dale Barker, Bill Skamarock,... (~25%) Kevin Raeder (80%) Alain Caya (new hire) Hui Liu (new hire)

slide-5
SLIDE 5

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Goals: DAI will provide:

  • 1. A data assimilation community within NCAR to produce leading-edge research and to

provide focus to disparate efforts;

slide-6
SLIDE 6

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Goals: DAI will provide:

  • 1. A data assimilation community within NCAR to produce leading-edge research and to

provide focus to disparate efforts;

  • 2. A software environment for supporting data assimilation research and evaluation; the

Data Assimilation Research Testbed (DART);

slide-7
SLIDE 7

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Goals: DAI will provide:

  • 1. A data assimilation community within NCAR to produce leading-edge research and to

provide focus to disparate efforts;

  • 2. A software environment for supporting data assimilation research and evaluation; the

Data Assimilation Research Testbed (DART);

  • 3. A mechanism for data assimilation research collaboration with strategically selected

partners from universities and government research labs;

slide-8
SLIDE 8

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Goals: DAI will provide:

  • 1. A data assimilation community within NCAR to produce leading-edge research and to

provide focus to disparate efforts;

  • 2. A software environment for supporting data assimilation research and evaluation; the

Data Assimilation Research Testbed (DART);

  • 3. A mechanism for data assimilation research collaboration with strategically selected

partners from universities and government research labs;

  • 4. Software tools for use in undergraduate and graduate education;
slide-9
SLIDE 9

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Goals: DAI will provide:

  • 1. A data assimilation community within NCAR to produce leading-edge research and to

provide focus to disparate efforts;

  • 2. A software environment for supporting data assimilation research and evaluation; the

Data Assimilation Research Testbed (DART);

  • 3. A mechanism for data assimilation research collaboration with strategically selected

partners from universities and government research labs;

  • 4. Software tools for use in undergraduate and graduate education;
  • 5. Basic research and implementation support for ‘operational’ partners, both within

NCAR and outside.

slide-10
SLIDE 10

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Strategic Approach: Use DART (software infrastructure) to Focus Efforts

  • 1. Cross-cutting initiatives in danger of becoming too diffuse (everything to everyone)
  • 2. Strategy used successfully to coordinate modeling/dynamics groups (CCSM, MM5)
slide-11
SLIDE 11

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Strategic Approach: Use DART (software infrastructure) to Focus Efforts

  • 1. Cross-cutting initiatives in danger of becoming too diffuse (everything to everyone)
  • 2. Strategy used successfully to coordinate modeling/dynamics groups (CCSM, MM5)

What is DART?

  • A. Allows combinations of assimilation algorithms, models, and observation sets
  • B. Diagnostic tools for assimilation experimentation
  • C. Set of compliant models, observational system designs, and observation sets
slide-12
SLIDE 12

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Strategic Approach: Use DART (software infrastructure) to Focus Efforts

  • 1. Cross-cutting initiatives in danger of becoming too diffuse (everything to everyone)
  • 2. Strategy used successfully to coordinate modeling/dynamics groups (CCSM, MM5)

What is DART?

  • A. Allows combinations of assimilation algorithms, models, and observation sets
  • B. Diagnostic tools for assimilation experimentation
  • C. Set of compliant models, observational system designs, and observation sets

D A R T Educational Applications Basic Assimilation Research CAM, WRF, WACCM, .... Universities, NCEP, labs

slide-13
SLIDE 13

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI Core Activities

DAI Principal Scientists (Nychka, Snyder, Tribbia, Anderson) and Core Staff

Affiliated Affiliated External External Collaborator(s) Scientist(s) Scientist(s) Collaborator(s) MMM, WRF Development Team, Ensemble Filter for WRF CGD, CCSM Ensemble Filter for CAM MIT Atmos. Sciences DA for Rotating Annulus NCEP, Devel. Division Ensemble filter for MRF

Goal: Identify set

  • f meaningful

internal and exter- nal collaborations.

Structure of Data Assimilation Initiative

slide-14
SLIDE 14

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI Provides: Expertise on Assim. Methodologies and Applications Software Infrastructure (DART) Basic Research Support Support for Core Scientists and Staff Collaborator Provides: Expertise on particular application Model or observational data sets Affiliated scientists with expertise on application

slide-15
SLIDE 15

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 1. Data Assimilation Research Testbed (DART)

Basic framework implemented Currently using GFDL FMS infrastructure Switch to ESMF infrastructure when available Primarily implementing ensemble (Kalman) filters Variational for low-order models only Plans MAY include a variational (4D-Var) capability DART compliant models (largest collection ever available with assim system) CGD’s CAM 2.0 GFDL FMS B-grid GCM incorporated and in use Many low-order models available MMM’s WRF model in process of being incorporated NCEP MRF being tested quasi-operationally in partial implementation GFDL MOM ocean model partially incorporated in earlier version

slide-16
SLIDE 16

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 2. Supporting ASP Summer Colloquium on data assimilation

DART will be used for student exercises

  • 3. Hosting numerous data assimilation visitors

Luc Fillion, Canadian Meteorological Center Jim Hansen, MIT Xiaolei Zou, FSU Ron Errico, NASA/DAO Shree Khare, Princeton Ryan Torn, U. Washington

slide-17
SLIDE 17

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 4. An Ensemble Filter DA system for CGD’s CAM

Testing underway at low resolution Standard T42 configuration also working

5 10 15 1 2 3 4 5 6 7 8

Truth 220 230 240 250 260

5 10 15 1 2 3 4 5 6 7 8

Ensemble Mean Analysis 220 240 260

5 10 15 1 2 3 4 5 6 7 8

RMS ERROR = 0.61412 −1 1

Results from assimi- lating only 1800 surface pressure

  • bservations every

12 hours Temperature at 700mb shown Mean error reduced to about 0.6 C

slide-18
SLIDE 18

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 5. DART filter being used for parallel tests in NCEP operational global model
slide-19
SLIDE 19

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 6. Filter implemented for GFDL B-grid GCM

5 10 15 20 0.05 0.1 0.15 0.2 0.25 0.3

ASSIMILATION FREQUENCY, HOURS ENSEMBLE MEAN PRIOR RMS TEMP ERROR (K)

LEVEL 1 LEVEL 2 LEVEL 3 LEVEL 4 LEVEL 5

Perfect model experi- ment Assimilation of only surface pressure obser- vations RMS error for tempera- ture is plotted Impact of increasing

  • bservation frequency

1800 PS obs every 24 hours down to every 5 minutes

slide-20
SLIDE 20

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 7. Fundamental research on filtering data assimilation

Ability to use limited observations Observation system simulations Localizing impact of observations (GSP Postdocs helping out here)

  • 8. Incorporation of WRF into DART

Compiling but untested Initial version available by summer

  • 9. WRF / CAM integrated filter assimilation system

Fundamental problem for regional models is boundary Use CAM assimilations for WRF boundaries Prototype for embedded regional models? Encourage CGD/MMM assimilation collaboration

slide-21
SLIDE 21

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

DAI: Accomplishments and Plans

  • 10. Ability to assimilate real data (operational streams plus GPS, etc.)

Will allow DART models to be used as ‘operational’ prediction models Allows evaluation of data value New hire, Hui Liu, will do this as first task (initial version Sep. ‘03)

  • 11. Variational assimilation methods

Luc Fillion has spent 2 months on adjoint for Errico regional model (MAMS) Adjoint nearing completion Variational assimilation capability to follow Planned as first large model with variational assimilation capability in DART

slide-22
SLIDE 22

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Some Additional Long-term Science Goals

  • I. Ensemble filtering

Performance vs. ensemble size Theory for tuning parameters Use of difficult non-linear observations (radar, satellite radiances...) Assimilating ‘discrete’ structures (thunderstorms, ocean eddies...)

  • II. Using assimilation to address model deficiencies

Assimilating model parameters Evaluating relative error characteristics of different parameterizations

  • III. Design and evaluation of observing systems

Measuring information content of an observation Cost function constrained observation system design

  • IV. Variational data assimilation (requires enhanced funding)

Theory Synergy with ensemble methods;

slide-23
SLIDE 23

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Reaching out to Other NCAR Divisions

  • 1. HAO:

Co-sponsoring GSP Postdoc to look at ionospheric assimilation

  • 2. ACD:

Plans to incorporate WACCM and chemical observations into DART

  • 3. SCD:

Coordinating with NASA ESMF project Interest in other computational issues and new models

  • 4. ATD:

Observing system simulation experiments to estimate value of instruments???

slide-24
SLIDE 24

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Summary Initial progress has been very good, mostly due to building an excellent core staff Strategy for focusing the effort has been key to progress Initial growing pains (budget, hiring, identifying key collaborators) going away Several high impact assimilation applications are being developed by DAI Continued progress with NCEP assimilation should increase our credibility Primary fear: initial successes will lead to expanding too rapidly; Must avoid letting our success kill us.

slide-25
SLIDE 25

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

What’s worked well: Interactions with MMM on WRF, convective scale Interactions with CGD on statistics, theoretical DA Interactions with selected university researchers through visits (Jim Hansen, MIT) What’s been difficult: Understanding process (not a surprise for a developing program) Budget time lag Acquiring dedicated staff (we’ve been lucky to date)

slide-26
SLIDE 26

/home/jla/dart/ncar_division_directors_mtg/presentation.fm April 18, 2003

Key points (personal opinion):

  • 1. Obvious existing gap at NCAR; limited existing activity (in core, not applications)
  • 2. Congenial relations between principals in several divisions
  • 3. Mixture of basic science, applied science, and collaborative deliverables; some are very

high visibility

  • 4. Aggressive (but realistic?) proposal
  • 5. High visibility area due to advances in observing systems, software engineering, and

assimilation methodology