SESAM

"The purpose of the SESAM code (developed in the MEOM group) is to perform the various basic operations that are required in sequential data assimilation systems.

These operations include square root or ensemble observational updates (with global or local parameterization of the forecast error statistics), adaptive statistical parameterizations, anamorphosis transformations, or the computation of truncated Gaussian estimators.

SESAM also provides diagnostic tools, to compute observation representers, EOF decompositions or regional RMS misfits, and various utilities for extracting observations, converting between file formats or performing simple algebraic operations.

SESAM data types: SESAM is a program that can perform a series of well defined operations on a limited set of data types, belonging to the 3 following categories:

1. vectors of system variables:
   • the state vector or control vector (dimemsion:x) contains the system variables that must be estimated from the observations;
   • the observed section (dimension:y) represents the part of the state vector that can be observed;
   • the observation vector (dimension:o) is a composite data type, containing both the vector of observed values at original location and the corresponding observation operator;
   • the local observed section represents the part of the observed section associated to every subsystem (for local applications).
2. ensemble of vectors or error covariance matrices:
   Any error covariance matrix in SESAM (other than diagonal matrices, which are stored as vector data types) must be described as an ensemble of vectors, representing the columns of a square root of the covariance matrix. This data type is consequently a set of r vectors that can be used as a single data type by SESAM (with dimension xr, yr, or or according to the vector data type).
3. information data sets:
   • the observation data base contains the original observed information, from which SESAM can extract observations;
   • the model grid defines the location of the state vector variables;
   • the model mask defines the model grid points that must be included in the state vector and in the observed section (it is part of the SESAM configuration).
   • the parameter file is a text file with the parameters that are required for some SESAM operations.

SESAM modules: The operations that can be performed for producing or transforming these data types are organised in modules, which are grouped into 3 categories:

1. the analysis modules (SEEK filter analysis, ...).
2. the utility modules (extract observations from a database, interface data from one file format to another, estimate a reduced rank error covariance matrix, ...).
3. the diagnostic modules (computation of RMS differences, projection, algebraic operations, ...). 

http://lgge.osug.fr/meom/Outils/SESAM/sesam.html

http://meom-group.github.io/code/