WGOMD Past_ Present and Future S. Griffies - CLIVAR

Document Sample
WGOMD Past_ Present and Future S. Griffies - CLIVAR Powered By Docstoc
					  WGOMD: Past, Present, and Future
                Stephen Griffies
         NOAA/GFDL/Princeton USA
Working Group for Ocean Model Development (WGOMD)
               Presentation to WGOMD
                   26 August 2007
                   Bergen, Norway
         WGOMD Terms of Reference (as of 2000 with
                 slight revision 2005)
•   To stimulate the development of ocean models for research in climate and
    related fields. (recently deleted qualifier: “with a focus on decadal and longer
    timescales at mid-and high-latitudes.”)
•   To encourage investigations of the effects of model formulation on the results
    of ocean models, making use of sensitivity studies and intercomparisons.
•   To promote interaction amongst the ocean modelling community and between
    this and other communities through workshops and other activities.
•   To stimulate the validation of ocean models when used in stand alone mode and
    as part of a coupled ocean-atmosphere model, using oceanographic data and
    other methods, and to advise on the observational requirements of such studies.
•   To publicise developments in ocean models amongst the climate modelling
•   To collaborate with other activities in areas of overlapping responsibility.
•   To advise on ocean modelling and related issues and to report on its activities to
    the JSC/CLIVAR WGCM and CLIVAR Scientific Steering Group. *
•   *Two sets of parent organizations are often cumbersome for WGOMD (e.g.,
    two different international meetings to report to each year).

    ToR are largely applicable today, with perhaps some
    modifications based on new developments in the community.
                        WGOMD Assumptions
•   Ocean models are relevant to understanding climate and predicting potential
    future changes.
•   The space-time scales relevant for WGOMD considerations extends from the
    global climate scale to the regional and increasingly the coastal scales.
•   Scientifically based model fundamentals produce robust model tools for use in
    climate science. This is the science of ocean models:
     –   Dynamical assumptions
     –   Numerical methods
     –   Physical parameterizations
     –   Rational, complete, and pedagogical model documentation
•   Well defined and fully documented experimental designs for model
    simulations are essential to realize robust results which can be reproduced by
    other groups. Absent the full documentation of model designs, one is doing
    irreproducible model integration (not science). This is the science of ocean
     –   Forcing datasets
     –   Bulk formulae
     –   Restoring terms
     –   Coupling methods
     –   Integration times
     –   Analysis methods
     State-of-science for models and modelling
• Ocean models are not mature.
• Ocean modelling practice is not mature (i.e., global ocean-ice
  simulations are not generally comparable between groups).
• Methods used are often not robust, with ad hoc and
  undocumented steps employed to “get the models running.”
  This situation leads to modelling being as much an “art” as a
  science. That is, the simulations are often not reproducible
  (even by the group performing the original simulation!).
• Given the growing importance of ocean models for
  understanding and predicting global and regional climate, the
  models and the experimental design must be given a well
  documented scientific foundation.
                  WGOMD mission
• A central mission of WGOMD is to facilitate the maturation of
  ocean models, and the use of ocean models in well defined and
  reproducible ocean modelling simulations.
• WGOMD aims to realize this mission by providing pedogogical
  peer-review survey papers that document models and the
  experimental design of simulations.
• It also does so by organizing topical workshops that bring
  elements of the oceanography community together to discuss
  research and development areas relevant to increasing the
  scientific integrity of models and their simulations.
• Realizing this mission (or some aspect of it) allows WGOMD to
  provide scientifically based advice to other Clivar panels and to
• This mission remains ongoing, with some success. However,
  further efforts are required to make routine use of ocean models
  by a scientifically literate researcher a process that produces
  useful scientific results.
     Key contributions of WGOMD
• Review paper: Pedagogically documents state-of-art in
  ocean climate models (Griffies etal (2000))
• Workshops: Topical workshops that facilitate
  collaboration, communication, and education
   – Princeton/GFDL 2004: State-of-art in Ocean Climate
   – Hobart/CSIRO 2005: Southern Ocean Modelling
   – Bergen 2007: Numerical Methods for Ocean Models
• CORE: Benchmark experiments for global ocean-ice
  models. Peer-review paper illustrates CORE-I with
  seven ocean-ice models each run for 500 years
  (Griffies etal in prep).
           WGOMD review article (2000)
• Pedogogical survey of
  ocean climate model
  methods and
• Highlighted vertical
  coordinates as key for
  model algorithms, with
  many complementary
  attributes between
• Influenced AR4 ocean
  climate model
• A basis for ongoing
  research efforts at
  improving model
             Princeton/GFDL Workshop 2004:
         State of the art in ocean climate modelling
•   Key developers of AR4 ocean
    climate models discussed their
    methods, parameterizations,
    and results.
•   Experts in ocean physics and
    numerics scrutinized the AR4
    models and made
    recommendations for next
    round of IPCC ocean models.
•   Community input to the
    WGOMD’s efforts at
    establishing an OMIP. A key
    outcome was to propose CORE
    as a science-based
    collaborative project, rather
    than push forward with a
    mandatory OMIP, as such was
    was considered premature.
                 Hobart Workshop 2005:
      State of the art in Southern Ocean Modelling
•   Southern Ocean is key to
    represent with high fidelity
    in climate models, as it
    represents a huge sink for
    heat and carbon, and the
    processes active have
    importance to all ocean
•   Science workshop discussed
    and debated methods of
    simulating and analyzing
    Southern Ocean physical
    and biogeochemical
•   Ten lectures with
    discussions provided
    pedagogical surveys of key
    aspects of the Southern
               Bergen Workshop 2007:
          Numerical Methods for Ocean Models
• Discuss and debate novel
  methods for developing the
  next generation of ocean
  models for global, regional
  and coastal applications.
• Bring together key
  practitioners and algorithm
  developers for eight
  provocative and pedagogical
• Enhance communication
  amongst a community of
  algorithm developers who
  typically do not have the
  opportunity to gather in such
  focused workshop settings.
  Coordinated Ocean-ice Reference Experiments
• Benchmark simulations for global ocean-ice
• CORE-I proof of concept project includes seven
  model groups with three ocean model classes
  (geopotential, isopycnal, hybrid).
• A step toward an Ocean Model Intercomparison
  Project (OMIP)
End of Saturday’s material
            WGOMD panel meetings: the road to CORE
•   Miami, Sep2000
     –   Led to the review paper “Developments in Ocean Climate Models” (Griffies etal, Ocean Modelling 2000).
•   Santa Fe, Mar2001
     –   Meeting was the first of many to discuss metrics for ocean climate model evaluation. Also launched the Pilot
•   MPI Hamburg, May2002
     –   Start of Pilot-OMIP using German (MPI) repeating annual mean forcing (based on ECMWF)
     –   Recognition of the nontrivial nature of developing a suitable dataset to force global ocean+ice models on
         centennial and longer time scales
•   Villefranche Apr2003
     –   NCAR (Bill Large) announced development of alternative ocean+ice model forcing with repeating annual
         year and interannual forcing; based on NCEP + satellites. GFDL to provide support.
•   GFDL Princeton Jun2004
     –   Science workshop (~100 participants): “State of the art in ocean climate models”
     –   WGOMD recommends the use of Large and Yeager (2004) dataset for ocean+sea ice model forcing; NCAR
         and GFDL to support future upgrades to the dataset.
•   CSIRO Hobart Nov2005
     –   Science workshop (~100 participants): “Modelling the Southern Ocean”
     –   Initial reports from groups using Large and Yeager (2004); aim to have peer-reviewed manuscript by end of
         2006 with 4-5 modelling groups participating.
•   Bergen Aug2007
     –   Science workshop (~100 participants): “State of the art in numerical methods for ocean climate models”; co-
         organized with Layered Ocean Model leadership.
     –   Report on the CORE-I results (seven ocean-ice models run 500 years).
                Miami/RSMAS: Sept 2000
•   Panel members: Boening (chair), Bryan,
    Chassignet, Griffies, Gerdes, Hasumi, Hirst,
    Treguier, Webb
•   Discussion of model development, with
    focus on global models planned for IPCC-
•   Highlighted need to develop suite of model
    metrics for use in evaluating simulations. An
    ongoing theme!
•   Panel members’ homework to summarize
    model development in their local sphere led
    to a review paper “Developments in Ocean
    Climate Models” (Griffies etal, Ocean
    Modelling 2000) which documented the
    state-of-the-art in ocean climate models.
•   Gerdes presented results from German
    OMIP (MPI-HOPE and AWI-MOM2) which
    initiated the idea of a broad based ocean-ice
    model comparison project.                       WGOMD
•   German OMIP used the MPI-dataset from
    Frank Roeske, based on ECMWF. This
    dataset is proposed for a pilot comparison
                           Santa Fe: May 2001
• Panel members: Boening (chair), Bryan,
  Chassignet, Griffies, Gerdes (missing),
  Hasumi, Hirst, Treguier, Webb
• Formal proposal for WGOMD sponsored
  pilot-OMIP managed by Frank Bryan. Seven
  groups volunteered to participate. Use the
  MPI-Roeske dataset to force the models.
  Allow modellers to use their own bulk
  formulae and methods for salinity restoring.
  100 year runtime.
• Discussed importance of standard metrics for
  use in evaluating ocean simulations.
• In general, the difficulty of designing an
  OMIP was under-estimated (forcing data,
  experimental design, bulk formulae,              WGOMD with guests
  integration length). Additionally, many of the
  main players were committed to developing
  coupled climate models for AR4, with that
  work complementary to the ocean-ice OMIP
  simulations, but often at the cost of fully
  developing the ocean-ice experimental design
  and code framework. These issues would
  plague the project for the next few years.
                       Hamburg/MPI: May 2002
•   Panel members: Boening (chair), Bryan, Chassignet, Griffies,
    Gerdes, Hasumi, Hirst, Treguier, Webb
•   Focused discussion on PILOT-OMIP.
•   Larry Gates (founder of AMIP) encouraged us to move
    forward with OMIP as coordinated through PCMDI.
•   Mojib Latif noted that modellers would not fully agree on a
    forcing dataset, so we should just agree on one set and run the
    models using the same protocol.
•   However, we were still missing some points:
     – Modellers agreed only on a rough experimental design. Some
       were even doing ocean-only runs, which are distinctly not the
       focus. Such is fine for testing purposes, but precluded a focused   WGOMD with guests
     – Not all centres had been routinely running ocean-ice simulations
       using prescribed atmospheric data; most either ran ocean-alone
       or fully coupled climate models. Hence, code needed to be
       developed, and that added much time to the project.
     – The nontrivial sensitivity of simulations to different bulk
       formulae was not yet appreciated.
     – 100 year simulations still considered adequate, but later were
       seen to be insufficient for overturning circulation (could have
       been guessed!).
     – The inability of the modelling centres to agree on the forcing
       dataset precluded the buy-in from some key players (e.g.,
                    Villefranche-sur-Mer 2003
•   Panel members: Boening (chair), Bryan,
    Chassignet, Griffies, Gerdes, Hasumi, Hirst,
    Treguier, Webb
•   Broad discussion and debate of datasets used
    to force ocean-ice simulations.
•   Bill Large from NCAR highlighted some
    egregious problems with the ECMWF and
    NCEP reanalysis data, which is the basis for
    the Roeske forcing fields from MPI
    (ECMWF). He proposed a new product
    based on NCEP with satellite products
    replacing certain reanalysis fields.
•   WGOMD chose to pursue the NCAR data in
    order to entrain the bulk of the modeling
    community. GFDL in particular agreed to        WGOMD and guests with Atlantic Panel
    work closely with NCAR to prototype their
    dataset with a GFDL-MOM configuration.
•   Another reason to eschew the MPI dataset
    was that is became clear it was poorly
    supported (Roeske leaving the field), and
    thus became a risky choice for large
    comparison projects.
                       Princeton/GFDL 2004
•   Panel members: Boening (chair), Bryan,
    Chassignet, Drange, Griffies, Gerdes, Hasumi,
    England (missing; replaced Hirst), Treguier,
    Banks (replaced Webb)
•   Community feedback: comparison of global ocean
    -ice models is of scientific interest, so the
    development of a WGOMD sanctioned common
    benchmark simulation is highly useful.
•   The interest in establishing a formal OMIP was
    less clear at this point, given the many questions
    of how to run an ocean-ice model with large
    uncertainties in forcing data
•   CORE proposed as a scientifically interesting
    exercise whereby a number of groups will run the
    Large and Yeager (2004) dataset forcing their
    global ocean-ice models for 500 years. No          WGOMD + workshop participants
    pretense (yet) that this is an OMIP, so no efforts
    (yet) made to involve PCMDI.
                          Hobart/CSIRO 2005
•   Panel members: Boening (co-chair), Banks,
    Chassignet, Drange, England, Gerdes (ex-
    officio), Greatbatch (missing), Griffies (co-
    chair), Hasumi, Holland (replaced Bryan),
•   Results from 500 year CORE-I simulations
    starting to be submitted (CCSM-POP, GFDL-
    MOM, ORCA-OPA). However, other interested
    groups remain unable to get the code together,
    develop an appropriate experimental design, nor
    garner sufficient computer time.
•   Early results point to the sensitivity of         WGOMD and guests
    simulations to fresh water forcing. How strong
    should salinity be restored? This question was
    left for each model group to decide. No formal
    recommendation made part of the CORE design.
•   With only three groups, and with no isopycnal
    nor hybrid models, the project remained sub-
                                  Bergen 2007
• Panel members: Boening (emiritus), Banks,
  Chassignet, Drange, England (absent), Gerdes
  (emiritus), Greatbatch, Griffies (chair), Tsujino,
  Danagasoglu (representing Holland), Madec,
  Treguier (emiritus)
• Seven ocean-ice models run for 500 years with
  Large and Yeager (2004).
    – four geopotential
    – two isopycnal
    – one hybrid
• CORE-I has reached a critical mass
• Wide variety of results, with more questions
  raised than answered.
                                                       WGOMD and guests in Bergen
    – Broad comparison projects such as this achieve
      much by raising questions, which then motivate
      further research.
    – Without the comparison, questions remain
      unasked, and thus unanswered.
• Peer-review document with ~20 authors to be
  submitted to Ocean Science (EGU online
  journal founded by David Webb—former
  WGOMD member).
      What is next for WGOMD?
• CORE-II (interannually varying)
• CORE-III (Greenland ice-melt perturbation)
• CORE-IV (heat perturbation)
• Intermediate coupled models
• “Data over-ride” coupled models
• WGOMD Repository for Evaluating Ocean
  Simulations (WGOMD REOS)
• Another workshop for 2009?
• Co-chair
• General membership issues
                CORE-I à OMIP?
• Are we ready to entrain PCMDI?
  – Maintain repository of model results; ancilliary code, such as
    bulk formulae and river mapping scheme
  – Advantage: expose datasets to larger community analysis,
    and maintain better version control on data and code
  – Disadvantage: PCMDI requires datasets to fit into their
    constraints (e.g., data must be mapped to spherical grid)
• WGCM: should CORE-I submission to PCMDI be
  part of the AR5 model database?
• Propose: make CORE-I a recommendation rather than
  a requirement.
• Propose: PCMDI involved through WGCM oversight,
  not WGOMD oversight.
   CORE-II: interannually varying forcing
• Only a few cases with interannually varying Large and
  Yeager (2004).
• How to initialize (ocean and ice)?
• What about salinity restoring? Same as CORE-I?
• How many cycles of the forcing dataset?
• What to analyze (metrics)?
• Action item: subgroup of interested WGOMD
  participants should develop an experimental design
  and run a mini-comparison.
   – Who is interested? Who will take charge?
          CORE-III: Greenland ice melt
• Have only run with 2-degree version of the GFDL-
  MOM, using MPI-OMIP dataset, and various other
  boundary conditions.
• Interactive atmosphere model desirable due to interest
  in perturbations, and dis-interest in mixed boundary
  condition instabilities.
• Survey questions:
   – Any group interested in pursuing this experiment with ocean
     -ice model?
   – What about coupled climate model? GFDL interested, but
     focus for next ~6months (at least) on model development…
   – Can/should WGOMD investigate a simple atmospheric
     model for use in comparisons? (e.g., SPEEDY? UVic
• Interested parties? Who will coordinate?
  CORE-IV: global warming perturbation
• Within the present CORE-I design, consider one
  or both of the following perturbations
  – Add xx W/m2 uniformly over globe
  – Increase the atmospheric air temperature uniformly
    over the globe by yy degC
  – Force ocean-ice models with multi-model ensemble
    mean (is this at PCMDI?). Could compare to CMIP
• Purpose: determine centenial-scale response of
  ocean-ice system to global warming sorts of
• Action item?
      WGOMD Repository for Evaluating Ocean
                  Simulations (REOS)
• Purpose: Web-based tool for evaluating ocean
• Proposal: WGOMD maintained web page containing
  input from the modelling, observational, and analysis
  communities which provide guidance and oversight for
  methods of evaluating ocean simulations.
   – Datasets
   – Analyses/syntheses:
       • already a Clivar page http://www.clivar.org/data/synthesis/directory.php
   – Tools (e.g., methods for sampling model output as in the
   – Annotated bibliography of relevant papers
   – Forum to discuss methods and best practices
       • e.g., how much confidence should one place in certain metrics or

• Action item: feedback and input to Anna and Stephen

• Should WGOMD expand its focus to include
  elements of the following:
  –   Prediction (especially decadal)
  –   Assimilation
  –   Regional modelling
  –   Operational modelling/forecasting
• Arguably we have the expertise on
  WGOMD to do so, at least to some level.
• Action items?
                2009 panel meeting
• Continuing our new tradition of meeting each ~18
  months, next panel meeting will be first half of 2009.
• Location: Given the tight budget constraints felt by
  many groups, it will need to be either North America or
• Past locations:
   –   Miami(2000)
   –   Santa Fe(2001)
   –   Hamburg(2002)
   –   Villefranche(2003)
   –   Princeton(2004)
   –   Hobart(2005)
   –   Bergen(2007)
• Proposal: April 2009 at Hadley Centre (check conflicts)
             2009 Workshop?
• Science workshop to accompany the WGOMD
  panel meeting.
• 2009 is roughly 20 years after the famous Gent
  & McWilliams (1990) paper which has been
  influential in the ocean parameterization world.
• Proposal: “Representing and parameterizing the
  ocean mesoscale: 20 years after Gent and
  McWilliams (1990)”
• Anne Marie Treguier now emiritus, with
  Gurvan Madec replacing.
• Claus Boening stepping down. Richard
  Greatbatch replacing.
• Marika Holland stepping down. Proposed
  Gokhan Danabasoglu to replace.
• Propose: one new member, with focus of this
  person’s expertise on topic of interest for future
  WGOMD; e.g., prediction? assimilation?
  Unconstrained by geographic region.
    Griffies à Griffies + co-chair
• I have been member of panel since 2000.
• Co-chair and chair since 2004 (three panel
  meetings and three workshops)
• Time for new ideas and energy.
• Proposal: Helene Banks to co-chair starting this
  year, to help balance the panel’s expertise on
  model applications, especially towards climate
  prediction, metrics, evaluation.
• Proposal: 2009 workshop/meeting will be my
  last as chair, at which time may wish to
  consider another co-chair…future discussions.
• Gerdes, Hurlin, and Griffies (2006): Sensitivity of a
  global ocean model to increased runoff from Greenland.
  Ocean Modelling, vol. 12, pages 416-435.
• Griffies, Boening, Bryan, Chassignet, Gerdes, Hasumi,
  Hirst, Treguier, and Webb (2000): “Developments in
  Ocean Climate Modelling”, Ocean Modelling, vol. 2,
  pages 123-192.
• Griffies and coauthors (2008): A proposal for
  Coordinated Ocean-ice Reference Experiments (CORE).
  In preparation.
• Large and Yeager (2004): Diurnal to decadal global
  forcing for ocean and sea-ice models: the data sets and
  flux climatologies, NCAR Technical Report.
Blank slide
“Ocean WGNE” - draft terms of reference
                       (from a WCRP perspective)

     In collaboration and consultation with other groups of sponsoring
(i) Review the development of ocean models for use in coastal to global
     scale operational and climate prediction taking into account the needs
     for representation of physical, biogeochemical and ecological
     components, and the coupling between these and other system
     components (in particular the atmosphere, land and cryosphere)
     .Propose numerical experiments aiming to refine numerical
     techniques and the formulation of physical, biogeochemical and
     ecological processes in ocean models, taking account of the optimal
     techniques for provision of ocean model forcing.
(ii) Promote the development of new methods for ocean prediction.
“Ocean WGNE” - draft terms of reference
                        (from a WCRP perspective)
(iii) Design and promote co-ordinated experiments for:
    – validating model results against observed oceanic properties and
      variations and diagnosis of shortcomings;
    – exploring the intrinsic and forced variability and predictability of the
      circulation of the ocean on all scales and its impacts on ocean
      biogeochemistry and ecology on short to extended ranges;
    – assessing the intrinsic and forced variability of the ocean on a range of
      time‑scales, including the interactions with biogeochemical and ecological
(iv) Promote the development of data assimilation methods for application
    to ocean predictions, and for the estimation of derived climatological
    quantities (ocean synthesis/reanalysis).
(v) Promote the timely exchange of information, data and new knowledge
    on ocean modelling through publications, workshops and meetings.
(vi) Advise the sponsoring organizations on progress in ocean modelling.

Shared By:
linzhengnd linzhengnd http://