Title: File-based model connections for data-assimilation with OpenDa

Authors: Julius Sumihar (Deltares, Delft, The Netherlands)
Martin Verlaan (Deltares and TU Delft)
Stef Hummel (Deltares)
Nils van Velzen (Vortech and TU Delft)

When starting a new data-assimilation application, the starting point is typically an existing model, some observations and a limited amount of time and funding to make a first assessment of the potential benefits of data-assimilation. In some applications the source code is available, but for many commercial models only the executable is available. In these cases, it is often a good solution to connect the data-assimilation methods to the model through the input files and output files of the model.

In the opensource data-assimilation toolbox OpenDA (www.openda.org), a generic approach, called black-box wrapper was developed for this purpose. The aim was to reduce the implementation time for new applications as much as possible. However, some parts of the code remain model dependent, such as the routines to read and write the input files, output files and restart files of the model. For some models these files are similar to earlier applications, so the adapter source code can be reused at least partially. The developments focussed on ensemble based techniques and model calibration, since the availability of an adjoint model is less likely in this scenario.

The use of the file-based black-box wrapper avoids several potential problems that may delay the early stages of development of a data-assimilation application. An important advantage is that the files that link the data-assimilation method to the model can be inspected to detect errors. It is also easier to use existing parallel computing capabilities of the model.

Ongoing developments try to extend the number of options for the description of uncertainty for model and observations. Although the number of options is limited at the moment, the noise properties can be specified through the OpenDA configuration files without writing additional code or compilation.

GMAO Head: Michele Rienecker Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: May 27 2011