Lateral dissipation in LMDZ

Some information about lateral dissipation in LMDZ (and related input parameters)

Overview

In a nutshell: It is necessary to add some lateral dissipation in order to fulfill the observed energy cascade from resolved scales (by the GCM) to unresolved scales (sub-grid scales, from the point of view of the GCM). In practice this is done by adding a dissipation term in the form of an iterated laplacian acting on winds and temperature.

Dissipation parameters

These are set in gcm.def. The main ones are:

• dissip_period : Apply dissipation every dissip_period dynamical steps (default is 0, which implies let the model pick an appropriate value)
• nitergdiv : number of iterations on velocity dissipation operator grad.div (typically 1)
• nitergrot: number of iterations on velocity dissipation operator grad.rot (typically 2)
• niterh: number of iterations on temperature dissipation operator div.grad (typically 2)
• tetagdiv: dissipation time scale (s) for smallest wavelength for u,v (grad.div component)
• tetagrot: dissipation time scale (s) for smallest wavelength for u,v (grad.rot component)
• tetatemp: dissipation time scale (s) for smallest wavelength for potential temperature (div.grad)

In addition there is a multiplicative factor for the dissipation coefficient, which increases with model levels (see dyn3d_common/inidissip.F90), which can be controlled by flag “vert_prof_dissip” (tampering with this flag is for experts only)

Good to know and rules of thumb

• Dissipation parameters should preferably be those stated in the example gcm.def files provided in the DefLists subdirectory.
• If there are some numerical instabilities then one should try to increase dissipation (i.e. reduce teta* time scales) as this tends to stabilize the model.
• Optimal values for tetadiv/tetarot/tetatemp depend on the horizontal resolution of the GCM. In practice, the higher the horizontal resolution, the more dissipation is needed.

29/11/2021