High Level Overview of GSI System by bloved

VIEWS: 43 PAGES: 14

									High Level Overview of GSI
         System

             John C. Derber
          NOAA/NWS/NCEP/EMC




 4-6 Jan 2005       GSI User orientation meeting
    Orientation Meeting general
             comments
• This orientation meeting was set up because there
  were several external regional GSI users (or
  potential users) visiting EMC in Jan. and wanted
  instruction.
• Meeting is intended to be a description of current
  system not a scientific discussion.
• We acknowledge that not everything is done in
  “best” way –
   – suggestions welcomed
   – but not all suggestions will be adopted
      4-6 Jan 2005               GSI User orientation meeting
            Orientation Meeting
             general comments
• GSI analysis code is an evolving system.
   – Scientific advances – e.g.,
      • situation dependent background errors
      • new satellite data
      • new analysis variables
   – Improved coding – e.g.,
      • Bug fixes
      • Removal of unnecessary computations, arrays, etc.
      • More efficient algorithms
      • Generalizations of code to allow use on different compute
        platforms
      • Improved documentation
      4-6 Jan 2005                   GSI User orientation meeting
            Orientation Meeting
             general comments
• New versions released approximately every month
   – Old versions are not supported
• Developers are expected to keep up with current
  version
• Code is intended to become Operational
   – must satisfy NCEP coding requirements
   – must fit into NCEP infrastructure
• EMC does not have resources to provide high
  level of support.
      4-6 Jan 2005              GSI User orientation meeting
                        History
• The GSI system was initially developed as the
  next generation global analysis system
   – Wan-Shu Wu, R. James Purser, David Parrish
      • Three-Dimensional Variational Analysis with spatially
        Inhomogeneous Covariances. Mon. Wea. Rev., 130, 2905-
        2916.
   – Based on SSI analysis system
   – Replace spectral definition for background errors with
     grid point version based on recursive filters

      4-6 Jan 2005                 GSI User orientation meeting
                       History
• After development of initial GSI analysis system,
  desire for single global/regional analysis system
  expressed by EMC management
   – realization that GSI fairly straightforward to modify for
     both applications
• GMAO collaboration
• Modification to fit into WRF and NCEP
  infrastructure
• Evolution to ESMF

      4-6 Jan 2005                GSI User orientation meeting
          Basic analysis problem
J = Jb + Jo + Jc

J = (x-xb)TB-1(x-xb) + (H(x)-y0)T(E+F)-1(H(x)-y0) + JC

J = Fit to background + Fit to observations + constraints

x = Analysis
xb = Background
B = Background error covariance
H = Forward model
y0 = Observations
E+F = R = Instrument error + Representativeness error
JC = Constraint term
       4-6 Jan 2005              GSI User orientation meeting
      Overview of presentations
• Data usage
  – Input data (Keyser)
  – Observation error and basic QC (Wu)
  – Satellite radiances (Derber)
  – Radar and surface data (under development -
    Marshall/Pondeca)
  – Nonlinear QC (under development – Parrish)

      4-6 Jan 2005         GSI User orientation meeting
    Overview of presentations
• Background error
  – Recursive filters (Purser)
  – Background errors and their estimation (Wu)
  – Anisotropic capabilities (under development –
    Parrish/Pondeca)
• Solution of minimization problem
  – Overview of code (Treadon)
  – Minimization technique (Derber)
     4-6 Jan 2005          GSI User orientation meeting
    Overview of presentations
• Running the system
  – Connections to WRF (Parrish)
  – Analysis Input/Output files (Treadon)
  – Cycling issues (Wu)
• Coding issues (Treadon)



     4-6 Jan 2005          GSI User orientation meeting
      GSI Code - General Flow
                              Read in &
User input &                                        Additional
                              distribute
initializations                                   initializations
                             observations

     Outer loop
     a) (Read/distribute) guess & derived fields
     b) (Read/distribute) background error
     c) Set up right hand side of analysis equation
     d) Call inner loop
         i.     Compute gradient information
         ii.    Apply background error
         iii.   Compute search direction
         iv.    Compute step size
         v.     Update analysis increment


        4-6 Jan 2005                 GSI User orientation
                  Write analysis & related output meeting
                Expected Scientific
                  advancements
• Inclusion of new types of data
   – COSMIC GPS – radio-occultation
   – Radiance data
        •   SSM/I and SSM/IS radiances
        •   NASA instruments from AQUA, TERRA, CHEM
        •   NPOESS and METOP
        •   Next Generation Geostationary satellite
   –   Doppler radial velocities and reflectivies
   –   Precipitation observations
   –   Cloud observations
   –   Ozone observations
        4-6 Jan 2005                GSI User orientation meeting
              Expected Scientific
                advancements
• Improved specification of observational errors
   – Situation dependent representativeness errors
   – Increased granularity in specification of obs errors.
• Advanced DA techniques
   – Situation dependent background errors
   – 4-D var.
   – Improved balance constraints



      4-6 Jan 2005                GSI User orientation meeting
               Expected Scientific
                 advancements
• New analysis variables
  –   SST
  –   Clouds
  –   Aerosols
  –   Precipitation
  –   Constituents
• Others
       4-6 Jan 2005        GSI User orientation meeting

								
To top