6 Standalone programs

6.1 Program assim

The standalone program assim can be used to test the assimilation. The program can be called from the command line:

The first argument is the initialisation file and the second argument is the time index of the observation to assimilate. All keys described in 4.2 have the same meaning for the program assim. But the forecast has to be specified by the following keys.

If the program is called with three arguments:

All assimilation cycles be between the two time indexes are performed in chronological order.

6.2 Program genobsoper

The standalone program genobsoper generate the observation matrix.

The first argument is the initialisation file and the second argument is the time index of the observation for witch the observation operator has to be created. All keys described in 4.2 have the same meaning for the program genobsoper. But the only diagnostic key used is DiagXXX.H.

If the program is called with three arguments:

The action is repeated for all time indexes between the start and the end time index.

6.3 Program applyobsoper

The standalone program applyobsoper extracts from a state vector the observations.

The first argument is the initialisation file and the second argument is the time index of the observation for witch the observation operator has to be created. All keys described in 4.2 have the same meaning for the program applyobsoper. But the only diagnostic key used are DiagXXX.Hxf and DiagXXX.invsqrtR. If a scalar observation point has been eliminated (out of the model grid for example) its weight in DiagXXX.invsqrtR is zero. The state vector is specified as it is described in 6.1.

If the program is called with three arguments:

The action is repeated for all time indexes between the start and end time index.

6.4 Program filteroper

The standalone program filteroper generates a sparse matrix witch acts as a spatial filter in the model space.

For each variable the filter is a Gaussian function:

N is a normalisation factor taking in to account the land-sea mask. The parameters {L}_{x}, {L}_{y} and {L}_{z} may be space dependent and have thus the same dimension as the state vector. The required keys are:

6.5 Program opermul

opermul is a general purpose program witch multiply two sparse operators. It can be used for example for multiplying a filter operator and a observation operator.

{O}_{1} is a operator mapping from space {S}_{1} to {S}_{2}, {O}_{2} from {S}_{2} to {S}_{3} and thus the product from {S}_{1} to {S}_{3}.

The required keys are:

6.6 Matlab utility GenObsFile

The utility ”GenObsFile” provides an easy way to save all the observations, coming from various sources, in a few files with the NetCDF format, and creates the .INIT file required by the assimilation routines.

Options for GenObsFile must be specified in the header of the Matlab routine, as described below:

• initheader: complete path & file name, of the file that must be copied on top of the .INIT file. This could be the ”model” part of the .INIT file.
• diags: complete path & file of a sample ”diagnostic” part of the .INIT file. The observation number should be replaced with <INDEX > and variable names with <EXT >. This part will be (adapted and) copied for each observation set.

  Example:

  Diag<INDEX>.Hxf =  [’xf.<EXT>’]

• Outdir : path where to store the new observations and .INIT file.
• Outfile : prefix of the new observation files
• maxX, minX, maxY, ..., minMJD: observations not within these ranges will be ignored when creating the new observation files
• rmse : vector containing errors on the observations, in the following order:
  [TEM SAL ETA other]=[...]

  It will only be used by the assimilation routine if no other observation error covariance \mathbf{R} matrix is specified. GenObsFile only uses values corresponding to variables present in your observations list.

• obstime : time of the day at which observations should be assimilated
• listfile : complete path+file name for the listfile, which contains the original observations. It is build using sections. There must be at least one section in the listfile. Each section contains a ”config” line followed by an arbitrary amount of data lines. The config line starts with the keyword ’config’, and has the following format: config VAR X Y Z MJD
  
  • VAR indicates how the observed data should be named in the .INIT file (TEM ...)
    
  • X might be (a) a complete path+file name with the longitude data, corresponding to the observations, (b) the keyword ’file’ if the longitude data is written in a file with the same file name as the actual data, with extension .X
    
  • Y (idem)
    
  • Z (idem)
    
  • MJD points to the file containing the MJD-time corresponding to the
  • observations, and might be (a) a complete path+file name, (b) the
  • keyword ’file’, (c) a datum in the format 1999-12-31, (d) a datum in
  • the MJD format ’51251’, (e) character limits to be found in the
  • actual observations file name.
    

  For example, if the actual file name is /home/johndoe/51657.TEM , MJD could be 15-19 as those are the indexes pointing to 51657 in the file name. After each config line, an arbitrary amount of observation files may be given. The filenames may contain matrix delimiters, as in (1:100,2:5,:)
  

Example listfile: