Previous Section Headers
4.5 Adding Parameterizations
New parameterizations should be independent of the order in which
model arrays are stored in memory. This implies that if a
parameterization requires two time levels of a particular field, these
two levels should be passed as input by the interface
routine. Parameterizations should also provide separate output
arguments for return to the calling routine. Furthermore, they should
not modify input arguments which point into the main model buffers or
other data structures. Time tendencies may be calculated from the
input and output arguments at the level of the parameterization
interface routine.
New parameterization variables that must be present at restart must
appear on the restart dataset. If a new field is stored in one of the
model buffers, it will automatically be written to the restart
dataset. Otherwise, code must be added to routine wrtrs1 to
write the new field to a the restart dataset. Code must also be added
to routine splitf to read the new field upon restart. Since
these routines have blocks of code for both in-core and out-of-core
cases, changes must be made to both blocks to ensure proper restart
capability.
Search page
Questions on these pages can be sent to...
erik@ucar.edu .
$Name: ccm3_6_6_latest3 $ $Revision: 1.35.2.1 $ $Date: 1999/03/25 21:37:33 $ $Author: erik $