next up previous contents
Next: Atmospheric Boundary Layer Up: Model Physics Previous: Convective Closure Assumptions

Horizontal Diffusion Schemes

 

Two types of diffusions are used to control nonlinear instability and aliasing. The second-order diffusion is used for temperature and moisture equations with diffusion coefficients viscT and viscQ, respectively. A more scale-selective fourth-oder diffusion is used in the momentum equations with a tunable diffusion parameter visc4U. The second-order diffusion scheme takes the form

equation872

and center differencing is used for tex2html_wrap_inline2992 . The fourth-order diffusion is of the form

equation877

and a nine-point differencing is used following the scheme used in the PSU/NCAR MM5. With an isotropic x and y coefficients, and absorbing a factor of (grid space) tex2html_wrap_inline2994 , this is

eqnarray883

The nine points used (labeled by their weights) for this scheme are as follows (for the diffusion on the center point).

         0  0  1  0  0 

         0  0 -4  0  0 

         1 -4  6 -4  1 

         0  0 -4  0  0 

         0  0  1  0  0

In version 2.3, spherical geometry is considered for the differential equations. In horizontal diffusion schemes, the weights for the points in the difference stencil satisfy tex2html_wrap_inline2996 for any field A, and when cos(j)=1, the weights are the same as in Cartesian Coordinate shown above.

The use of the fourth-order diffusion scheme was originally motivated by the desire to have diffusion terms be negligible in the zonal and vertical average momentum budget (relative to the balance of surface stress and nonlinear momentum transport). Downgradient diffusion is a poorer approximation for momentum than for a conserved tracers, hence the decision to try and keep the impacts of horizontal diffusion at large scales small via fourth-order representation in the momentum equations, while retaining second order in temperature and moisture equations. The values of the diffusion parameters were tuned largely on the trade-off of lowering diffusion but keeping numerical stability in long runs. The lower diffusion also tends to increase the transient eddy variance which tends to improve surface stress climatology.



Climate Systems Interaction Group
Sun Aug 25 00:58:46 PDT 2002