Horizontal advection options in MPAS

microted

New member
I'm wondering if it is possible to tune the monotonic filter for scalar advection? I just added the NSSL 2-moment microphysics to MPAS 7.3, and it shows a familiar issue of noisy edges (oscillations -- especially on strong gradients ), which is pretty obvious in the reflectivity in the ideal supercell test (attached figure). This comes from the number and mass having different errors at the edges, resulting in small numbers of large particles that then fall along the reflectivity boundary. The positive definite filter in addition to the monotonic doesn't seem to affect the result much.

Cranking up the value of config_coef_3rd_order definitely makes the result smoother, and at 1 it is not too bad, but of course there are trade-offs there. Also, is it correct that the namelist options for advection order have no effect (e.g., config_scalar_adv_order)?

Any suggestions?

- Ted
mpas_dbz_vert.png

GitHub - MicroTed/MPAS-Model: Repository for MPAS models and shared framework releases. -- branch 'nssl2m'

mpas_dbz.png
 

skamarock

New member
Hi Ted

Yes, the advection order inputs in the namelist have no effect. These were dropped in the solver during the optimization that occurred before the V5 release in 2017; we need to remove them and fix the documentation. The only options available are either 3rd or 4th order flux operators. The config_coef_3rd_order controls this; equal to 1 and you have the same scheme as the 3rd-order scheme in WRF, equal to zero you get the 4th-order scheme. Our default value of 0.25 comes from tests reported in Skamarock and Gassmann (2011; doi:10.1175/MWR-D-10-05056.1).

The montonic limiter is FCT, and it is the same as that in WRF so I'm not surprised you are seeing the same behavior. I believe you are using the WENO scheme you implemented in WRF to address the bad behavior you are seeing? We do not have a WENO scheme available in MPAS, so that is not a near-term option.

The problem seems to be that the flux limiters is nonlinear and can act differently on the mass and number concentration fields, and perhaps more specifically it does not preserve nonlinear correlations.
 
Top