At the beginning of May, approximately 30 researchers from across the globe attended the NCEO/DARC Data Assimilation Training Course held at the University of Reading.

During an exceptionally sunny week at the beginning of May, we held the NCEO training course on data assimilation at the University of Reading. This annual event brought together approximately 30 researchers from across the globe to learn how data assimilation is used to combine numerical models and observations to provide something far more useful than either alone. Data assimilation has proven itself to be essential to modern numerical weather prediction but has practical applications to a much larger variety of geophysical problems. This was reflected in the breadth of the participants’ research areas.

Data assimilation is a fundamentally mathematical concept largely based on ideas from probability theory, numerical analysis, optimisation and more. However, the successful application of data assimilation also requires knowledge of the physics of the numerical models and the available observations. This includes an understanding of the precision and accuracy of each so that we know how much to trust each to provide a physically plausible output.

A challenge to understanding data assimilation as a newcomer is the number of different algorithms available to solve the data assimilation problem.  The course aimed to explain the fundamental differences and similarities between the competing algorithms, such as variational verses ensemble methods. The benefits and limitations of each were further illustrated with hands-on computer practicals.

As models and observations evolve to serve an increasing number of geophysical problems, so too must data assimilation. Therefore, next year the course will also evolve to help give researchers the knowledge to tackle these new challenges.

If you are interested in attending next year’s training course, please keep an eye on the DARC website.