• Created by Unknown User (gdcarver113@outlook.com) on Oct 31, 2018

Hi,

I want to use the older Euler dynamics in OpenIFS for some testing rather than the semi-Lagrangian scheme.

What namelist settings and other changes do I need to make to the model?

Thanks,

   RH.

1 Comment

  1. Unknown User (nagc)

    Hi Ryan

    To move from SL advection to Eulerian you need to do following :

    1/ Ensure your model is using quadratic (or cubic) truncation. The linear one is not appropriate to treat the aliasing of advection terms. The grib is set by the model initial conditions. The T21 test directory that comes with OpenIFS contains initial files on the quadratic grid. If you need more at different resolutions contact openifs-support@ecmwf.int and we can create some for you.

    2/ Namelist switches:

    LTWOTL=.F.
    LSLAG=.F.
    LSETTLST = false
    LSETTLS = false

    Reduce TSTEP by factor 5 to ensure stability and activate the Asselin time filter by setting REPS*=0.05 (multiple variables)

    There might be some other switches needed, basically you have to satisfy all the consistency checks hardcoded for the SL advection (for example the SETTLS scheme has to be off.)

    3/ Finally make sure the horizontal diffusion is appropriate. The current setting for the OpenIFS makes it relatively low, as the SL interpolation are also contribution to the model diffusivity. This additional source will be gone with the Eulerian scheme on. And the non-linear character of the Euler advection requires the additional diffusion term otherwise your equation would become hyperbolic which leads to a discontinuity developing in a finite time.

    Hope that helps.

    Cheers,  Glenn

    (with credit to Filip Vana)