Currently employed at Para//ab, BCCS, where I mostly work on projects related to ocean modeling, but also with parallel computing and optimization of applications for supercomputing, e.g. climate models or other projects where Para//ab participate.

The equations we try to solve, also called "the model", is a system of partial differential equations that describes the dynamics of the ocean. This system is very complex. To get anywhere we do as most scientists; divide the problem into smaller subproblems that we know how to deal with efficiently. A "numerical model" is a result of such a process, it is a reformulation of the problem into a form suitable for solution by computers. One example is the Bergen Ocean Model that is developed and frequently used here in Bergen - look at this page or this page for graphical presentations of some of our results. Doing lots of simplifications and assumptions comes at a cost; we introduce error. Our primary concern is called numerical error, which we try to identify and remove/minimise. The underlying physical model has of course also weaknesses, and it is not always obvious if errors are numerical or due to the model itself. On our way we utilize tools from many fields of science and also try to pick up the essential ideas. Examples of such fields are e.g. oceanography, linear algebra, programming, visualisation and parallel computing.

Some of my recent activities are reflected in:

I finished my studies for the Dr degree modeling the Skagerak area, some animations can be found here.

Abstracts for work on the TODO list...

"On the applicability of implicit methods in regional scale baroclinic ocean models"

In present ocean models both implicit and explicit methods are commonly used to find the surface elevation. When applying implicit methods in ocean modelling with Courant numbers greater than unity, one will fail to represent short modes of the solution. If these modes are of minor importance to the overall problem, implicit methods may still be a cost effective way of propagating the solution in time.

In the present paper the domain of validity of implicit methods for solving the shallow water equations is investigated for a baroclinic benchmark, using both implicit and explicit versions of a sigma coordinate ocean model.

It is also demonstrated that the commonly used boundary conditions for the water elevation in ocean models, using semi-implicit time stepping methods, may not be in agreement with the thermal wind equations for geostrophic flow.

We find that the implicit method performs well with high spatial resolution, but on coarse grids the explicit method give much more accurate results. We believe that the enhanced resolution improve the results for the implicit method due to a better representation of the surface gravity waves, and that it also reduce the above mentioned problem with the boundary condition.