Setup:
------

Environment:
------------
Scratch/data directories are set in EXPT.src. Scratch directories are where
models run, the preprocess.sh script will copy necessary files to that location
before running the model. Data directories are where model output is stored. The
postprocess.sh script will move data files from the scratch directory to the
data directory.

Note that by sourcing the REGION.src file (in the top-level region directory)
and the EXPT.src file (in the experiment directory), the shell wil set the
variables S and D which point to the scratch and data directories respectively.
type 
  cd $S  
or
  cd $D
to go to the respective directories.

initialize flag:
-------------
Create an empty file called "INITIALIZE" and start the model. 

If yrflag<2 in blkdat.input and this file is present you will start from 
climatology (if clmflg=2 in blkdat.input). The model will be initialized 
from the time specified in infile.in.

If yrflag==3 in blkdat.input, this allows you to start from a restart 
file without considering the value of "dtime" which is kept in the restart 
file header. This is useful when running a spinup, and then using the resulting
file to start an experiment with "realistic" forcing (ex era40).

Running jobs :
--------------
Before jobs are run the scripts preprocess.sh must be run, this can be set in
the job queue script. You can also run this script from the experiment
directory, to make sure you have all data files you need. It should cover most
of the data files we use. 

If you add new data files for the model, you will have to modify preprocess.sh
so that the script copies them to the scratch directory. Otherwise you must copy
them by hand to the scratch directory.

postprocess.sh copies files from the scratch directory to your data directory.
If you find some files are not copied from the  scratch directory you will have
to modify postprocess.sh. Also, if a job runs out of time, postprocess.sh will
not be run. In that case you can run the postprocess.sh script interactively to
retrieve the files to the data directory.

Job scripts will have to be modified from machine to machine, but the
preprocess.sh / postprocess.sh scripts should be able to run on all machines.