Calendar and dates
The GCM needs to know the current date since the forcings, declination of the sun, surface conditions, ozone and aerosol fields, depend on the date.
You can run the GCM with a real calendar or with a fictive calendar where years have a constant number of days, either 360 or 365. With a fictive calendar, everything happens as if the Earth moved a little bit faster along its orbit around the sun.
Choose the calendar as soon as you prepare surface conditions with
the program ce0l
. In the file run.def
, use the parameter
calend
(character variable). For years with constant length, 360 or
365 days:
calend=earth_360d
or:
calend=earth_365d
For the real calendar (with leap-years), run ce0l
for each year you
want boundary conditions for:
calend=gregorian
anneeref=<the right year>
where anneeref
is an integer variable in run.def
.
When you create an initial state with program ce0l
, you also choose
the date of that state, using the variable anneeref
and another
integer variable in run.def
, dayref
, the day number in the
year. Note that the date of the created initial state is not read from
the date of the data in files ECDYN.nc
or ECPHY.nc
(see Initial
state and surface conditions).
When the GCM runs, it normally reads the intial date of the run from
the file start.nc
, and writes the final date in the file
restart.nc
. So you do not change calend
, nor anneeref
, nor
dayref
when you continue a simulation, starting from the end of a
previous simulation. anneeref
and dayref
serve as a reference
point in time for a time-step counter in start.nc
.
Restarting from an arbitrary date
You can set raz_date = 1
in run.def
. Then the GCM uses dayref
and anneeref
in run.def
for the initial date. This way, you can
start the GCM at any date from any restart.nc
and
restartphy.nc
. But watch out for the initial shock if there is a
large seasonal difference between the original date of the restart
files and the reset date. The shock will come from the discrepancy
between the state of the atmosphere and the forcing conditions.
Do not forget to set raz_date
back to 0 if you want to continue a
simulation.