Apart from adding the LSM module into V3 MM5, there are other
changes made in the model. These are:
6.1 Input and Output Files
- Use Fortran OPEN to access input files and write output files.
This has resulted in a great simplification in the deck. There
are no script variables to set in a mm5.deck (Cray deck may be
an exception). Now to run a model interactively, the model expects
to see the following files in the Run directory:
MMINPUT_DOMAIN1
BDYOUT_DOMAIN1
LOWBDY_DOMAIN1
TERRAIN_DOMAIN2 or MMINPUT_DOMAIN2 (if running a nest)
RESTART_DOMAIN1 (if it is a restart run)
RESTART_DOMAIN2 (if it is a restart run, and if running
a nest)
MM5OBS_DOMAIN1 (if doing observational nudging on domain
1)
....
The file named LOWBDY_DOMAIN1 contains the substrate
temperature array that used to be in BDYOUT file in V2.
The output from the model, also appearing in the Run directory,
has a name, rather fort.41, and fort.42:
MMOUT_DOMAIN1
MMOUT_DOMAIN2 (if you run two domains)
SAVE_DOMAIN1 (if IFSAVE = .T.)
SAVE_DOMAIN2 (if IFSAVE = .T., and if running a nest)
Most files use the same Fortran unit number as they did in V2.
The new lower boundary file makes use of Fortran unit number 8,
which was no longer needed because the removal of ehtran file
among the input files.
All variables in MMINPUT_DOMAINx and MMOUT_DOMAINx files are
not coupled any more.
Another input file expected from MM5 is LANDUSE.TBL.
This file is provided in the MM5 tar file. It lists landuse characteristics
and may be modified by users.
- The ehtran is removed from the tar file, and the array is
now initialized in the model.
- A new namelist variable, BUFFRQ, is added. This defines
the time interval (in minutes) the output files are split. For
example, if BUFFRQ=720., the model will create an output file
every 12 hours (applies to history file only). The splitted output
file will have file extension _00 for hour 0, and _01 for 0 -
12 hours, and _02 for 13 - 24 hours, etc.. To turn the option
off, set BUFFRQ < TAPFRQ. In addition, the output files from a
restart run will also have an file extension (whether you choose
to split files or not): _01 for first restart run, and _02 for
the second, etc..
6.2 New solve.F
We have removed the hydrostatic option in V3, and made significant
restructuring changes to the solver routine (now called solve.F
The tendencies and decoupled variables are now stored as three-dimensional
arrays instead of as J-slices. The advection and diffusion
routines are now three-dimensional in that the J-loops
are inside them, and the solver now has a J-loop for each
physics process. This leads to many more and much shorter multi-tasked
sections than previously. These changes will allow for easier
coding of an efficient multi-tasked adjoint, and have eliminated
many repeat calculations.
The new solve routine has improved the model performance in
V3. Timing of benchmark case SESAME shows a 4.8 % speedup on DEC
(or COMPAQ) AlphaStation. The shorter multitask loop apparently
also helps to speedup compile time on Cray and for MPP processing.
The V3 MM5 may require more memory to run for large domain cases.
6.3 New and Improved Physics Options
There is one new physics option in MM5 V3-1: the Gayno-Seaman
PBL (IBLTYP=6). This is based on Mellor-Yamada TKE prediction.
It is distinguished from other TKE schemes by the use of liquid-water
potential temperature as a conserved variable, allowing the PBL
to operate more accurately in saturated conditions. This option
will also be available in V2 release 2-13. (This scheme is developed
and tested at Penn State.)
The Reisner II microphysical scheme is updated from that in
V2 release-2-12. Ice number concentration is again prognosed.
Limitted test shows improvement over previous versions. (Thanks
again to Greg Thompson and Roy Rasmussen of NCAR/RAP.)
6.4 Namelist Change
There are a few changes in the namelist (mmlif):
- A new namelist, NPARAM, is added to include those variables
related to nest runs. These variables were scattered in other
namelists in V2.
- Some namelist variables are moved to new locations. For example,
the forecast length TIMAX is now at the top of the namelist for
easy editing.
6.5 Reference State with Isothermal Layer
MM5 Version 3.1 will allow for an isothermal layer to be part
of the reference state. This will better approximate the stratospheric
lapse rate leading to smaller values of p' there, and allowing
the model top (PTOP) to be placed at lower pressures (e.g. 10
mb).
This is implemented by setting a parameter called TISO in the
INTERPF namelist with the other reference-state parameters (TS0,
TLP, P0). If TISO is zero Kelvin or less than the reference-state
temperature at PTOP, the reference state will be, as before, just
a log-p lapse rate. If TISO is set to a typical stratospheric
value (e.g. 220 K), that value will be used as a lower limit for
the reference temperature. At temperatures above TISO the log-p
lapse rate will still be used. There is a discontinuity in the
lapse rate similar to that at the tropopause.
The isothermal layer is not designed for use at heights below
topography because the reference surface pressure calculation
is based on the given log-p lapse rate. For realistic topography
and stratospheric temperatures this is not a major restriction.