About the Scripts and General Flow

MLEF Algorithm

Options

• Models:                 KdVB, GEOS, SWM, RAMS, GFS, WRF, LPDM, SiB-CASA-RAMS, PCTM etc…
• Observations:        Synthetic or include various real observations
• Obs. operators:      Include various forward operators
• Solution Type:       Mode (max likelihood-MLEF) or Mean (ensemble mean ETKF)
• Estimator:              Filter or Smoother
  Control variable:   Initial conditions, Model bias, Model parameters
• Covariances:         Localized, or Non-localized forecast error covariance
• Minimization:        Minimization algorithm (C-G, L-BFGS)
• MPI:                     Parallel MPI run or a Single processor run
• Verifications:         Innovation statistics (chi-square test, K-S test), RMS-errors

Tasks

Decide which MODEL will be used

• On IBM use MASS Math library
• On other machines use LAPACK

1. Install LAPACK on your computer, unless it already exists. Compile with 32 bit object code option (LAPACK can be found at http://www.cs.colorado.edu/~lapack)
2. Install MLEF algorithm on your computer
3. Include MODEL into MLEF
   1. Develop an interface between MODEL and MLEF (a template from SWM, RAMS is available)
   2. Prepare all input files for the mode
4. Prepare observations
   1. Synthetic observations of atmospheric and carbon variables
   2. Separate observations into groups (if needed)
   3. Define data assimilation interval
5. Perform initial data assimilation experiments with synthetic observations
   1. Perfect model assumption experiments
   2. Parameter estimation experiments
   3. Model error estimation experiments
6. Develop observation operators (for example):
   1. Atmospheric observations (u,v,T,p,q, etc.)
   2. Carbon observations

Go back to the main page | Go to Ensemble Site