New book on data assimilation available:
Geir Evensen: Data assimilation, The Ensemble Kalman Filter, Springer, 2006
Springer link and
Download EnKF ECLIPSE V3.1
Version 3.2 of the EnKF distribution,(Sep 12, 2006).
Mostly minor updates related to the compilation with the linux intel compiler.
The ensemble io is now given by the memlist.inc to make the states_io identical on
all platforms (F90 does not have a specification of how to write a type variable as
a unformatted direct file).
Version 3.1 of the EnKF distribution,
specially tailored for setup with new Eclipse models, is now available (May 24, 2006).
The user interface appears similar to the version 2.x but has been completely recoded in
Fortran 90. Thus, the ksh scripts previously used to run all the sub-programs have been
replaced with a single F90 program, and this allows for simpler debugging and a more
Other major changes include:
- Automatic generation of include files representing the interface to ECLIPSE, thus
no manual programming is needed when setting up a new case.
- All the statistical priors are now specified in an xls sheet providing an interface which is rather simple to use.
- Additional parameters such as the OWC, GOC, MULTZ, and MULTFLT are now included as variables to be estimated.
- There are now extensive help menues included.
- Since the previous documentation and installation menues are obselete we now refer to the enclosed
- A known bug: the Diag/summary files contains an empty line that needs to be removed. This can be done using
the script fix.sh.
This page is established as a reference page for users of
EnKF with the ECLIPSE reservoir simulation model. It contains a specific Fortran 90 implementation
which links the EnKF to the ECLIPSE reservoir simulator and documentation on how to use it. Much of the
code is the standard open EnKF code which is also available from the
EnKF home page.
The material on this page will provide new users of the EnKF and ECLIPSE with a quick start and spinup,
and experienced users with optimized code which may increase the performence of their implementations.
The EnKF was originally proposed by
and has later been further developed
and examined in a large number of published papers. A recent review and overview of the EnKF
is given by Evensen (2003),
which provides detailed information on the formulation, interpretation
and implementation of the EnKF, and now serves as a reference document for the basic methodology.
A later publication
provides a discussion of alternative square root schemes for the analysis step where the perturbation
of measurementes is avoided. Further, it provides a discussion of ensemble sampling strategies which
ensures a better conditioned ensemble, and it gives a theoretical discussion of issues related
to the use of a low rank representation for the measurement error covariance matrix.
Users of the EnKF are encouraged to e-mail useful example codes and upgrades, which can be
installed on this web site, to Geir Evensen.