-
Notifications
You must be signed in to change notification settings - Fork 9
Description
Describe the bug
In the netcdf output, when using a restart file, then time:units attribute is set in a way that is inconsistent with the time variable.
For an initial run of one day long, in the netcdf output:
time:units = "seconds since 1900-01-01 00:00:00" ;
time = 86400 ;
Which is correct
If another day is run, the time variable in netcdf output is:
time:units = "seconds since 1900-01-02 00:00:00" ;
time = 172800 ;
This is inconsistent, as both the time and the time origin have stepped forward one day.
Interestingly the file name has the correct date (i.e. the end of the time period):
GMOM_JRA_WD.ww3.hi.1900-01-03-00000.nc
To Reproduce
Start the model from a restart file, with netcdf output turned on.
Expected behavior
Time:units should be set from the model start time, not the restart time.
Additional context
The time_origin is set based on the 'Current Time' before the model is time stepped:
WW3/model/src/wav_comp_nuopc.F90
Line 613 in 97740a1
| time_origin = 'seconds since '//time_origin(1:10)//' '//time_origin(12:19) |
call ESMF_ClockGet( clock, currTime=esmfTime, rc=rc )
if (ChkErr(rc,__LINE__,u_FILE_u)) return
endif
! Determine time attributes for history output
call ESMF_TimeGet( esmfTime, timeString=time_origin, calendar=calendar, rc=rc )
if (ChkErr(rc,__LINE__,u_FILE_u)) return
time_origin = 'seconds since '//time_origin(1:10)//' '//time_origin(12:19)
But then the 'elapsed_secs' is set from the 'Start Time'
WW3/model/src/wav_comp_nuopc.F90
Line 1062 in 97740a1
| elapsedTime = nextTime - startTime |
elapsedTime = nextTime - startTime
call ESMF_TimeIntervalGet(elapsedTime, s_i8=elapsed_secs,rc=rc)
I think setting time_origin based on Start Time would make sense.