Cherry pick ice_in and year_align changes from MOM6-CICE6 repo

[anton-seaice]

I cherry picked recent config changes from MOM6-CICE6 repo:

I noted two things when running this:

• Do we need to set the dates in 'ww3_shel.nml' now we have changed them in datm/drof streams xml files?
• The dates on the history output are wrong:

ncdump archive/output001/GMOM_JRA_WD.ww3.hi.1900-01-03-00000.nc | grep time
        time = UNLIMITED ; // (1 currently)
        double time(time) ;
                time:units = "seconds since 1900-01-02 00:00:00" ;
                time:calendar = "noleap" ;
        int mapsta(time, ny, nx) ;
        float UAX(time, ny, nx) ;
        float UAY(time, ny, nx) ;
        float ICE(time, ny, nx) ;
        float HS(time, ny, nx) ;
        float T02(time, ny, nx) ;
        float T0M1(time, ny, nx) ;
        float T01(time, ny, nx) ;
        float FP0(time, ny, nx) ;
        float THM(time, ny, nx) ;
        float EF(time, freq, ny, nx) ;
        float USSX(time, ny, nx) ;
        float USSY(time, ny, nx) ;
 time = 172800 ;

Note the time:units are seconds since 1900-01-02, but the the time is 172800 (= 2 days). This gives a date of 1900-01-04 00:00:00, however the expected result is the output represents the average from 1900-01-02 00:00:00 to 1900-01-02 24:00:00. Is this a known issue @ezhilsabareesh8 ?

(n.b., I set nuopc.runconfig stop and restart to 1 day, and wav history_n to 1 day, and ran twice in Payu)

[anton-seaice]

xarray also interprets this with the 1900-01-04 date

array([cftime.DatetimeNoLeap(1900, 1, 4, 0, 0, 0, 0, has_year_zero=True)],
      dtype=object)

[ezhilsabareesh8]

• Do we need to set the dates in 'ww3_shel.nml' now we have changed them in datm/drof streams xml files?

For gridded outputs the dates in ww3_shel.nml need not to be changed, since the dates from the nuopc.runconfig file will be used for the output.

• The dates on the history output are wrong:

Upon checking your configuration I also found that the history_options = nmonths is not changed to ndays when you run the simulation for 1 day. The history options in these lines to be changed.

[anton-seaice]

For gridded outputs the dates in ww3_shel.nml need not to be changed, since the dates from the nuopc.runconfig file will be used for the output.

And non-gridded output? (Can we delete those lines from ww3_shel.nml? )

• The dates on the history output are wrong:

Upon checking your configuration I also found that the history_options = nmonths is not changed to ndays when you run the simulation for 1 day. The history options in these lines to be changed.

Yeah I did that, which gave me the daily output above, but the dates in that output are set wrong.

[ezhilsabareesh8]

And non-gridded output? (Can we delete those lines from ww3_shel.nml?)

The point output feature is not fully implemented in the current RYF configuration since the point outputs are binary files and I am currently using it only for testing purposes. The stride is currently set to zero (refer here), preventing the generation of any point outputs. Nevertheless, it's essential to include these lines as we plan to incorporate this feature in the future.

[aekiss]

why is it

                time:units = "seconds since 1900-01-02 00:00:00" ;

rather than

                time:units = "seconds since 1900-01-01 00:00:00" ;

?

[anton-seaice]

why is it

                time:units = "seconds since 1900-01-02 00:00:00" ;

rather than

                time:units = "seconds since 1900-01-01 00:00:00" ;

?

It looks like the time:units get set based on current time before the model is run:

https://github.com/ESCOMP/WW3/blob/97740a1abb36b7ba6e52a7d6f7a37dd06f915633/model/src/wav_comp_nuopc.F90#L607

but then the elapsed time used for the time value is based on the model start time:

https://github.com/ESCOMP/WW3/blob/97740a1abb36b7ba6e52a7d6f7a37dd06f915633/model/src/wav_comp_nuopc.F90#L1063

Ill investigate more!

[aekiss]

So does that mean the units will be different for each successive run of an experiment? I'm wondering whether xarray would handle concatenation in that case.

[anton-seaice]

So does that mean the units will be different for each successive run of an experiment?

That's what it looks like. e.g., if a run is 1 day long, the first run would be correct (day 1 + 1 day), the second would be day 2 + 2 days (i.e. day 4) , the third would be day 3 + 3 days (i.e. day 6) etc.

I'm wondering whether xarray would handle concatenation in that case.

I think it handles different time-steps ok, as long as they don't overlap. i.e. I have concatenated monthly and daily data before.

[anton-seaice]

I raised ESCOMP/WW3#17

As its specific to the nuopc cap, we might need a patch. I guess we should also raise an issue in the access-om3 repo.

I think I am happy enough this all works otherwise. and the history output problem is unrelated to this change. I suggest we turn on floe size distribution in a subsequent PR (rather than this one).

[anton-seaice]

@aekiss @ezhilsabareesh8 :- Can you review please? :)

[ezhilsabareesh8]

It looks good to me.

[ezhilsabareesh8]

tr_fsd is required for floe size distribution, we can include it in the next PR

[aekiss]

Looks good to me, thanks @anton-seaice

[aekiss]

Happy for you to merge @anton-seaice

[anton-seaice]

I dont have write permission :)