RepeatYear forcing for JRA55-do reanalysis is 1st May 1990 to 30th April 1991

[navidcy]

Based on the analysis by Stewart et al. 2021, the Repeat-Year forcing is 1st May 1990 to 30th April 1991. Is there a reason why this is not the default, e.g., here:

ClimaOcean.jl/src/DataWrangling/JRA55/JRA55_metadata.jl

Lines 30 to 36 in c7ce952

	function all_dates(::JRA55RepeatYear, name)
	if name == :river_freshwater_flux || name == :iceberg_freshwater_flux
	return DateTime(1990, 1, 1) : Day(1) : DateTime(1990, 12, 31)
	else
	return DateTime(1990, 1, 1) : Hour(3) : DateTime(1990, 12, 31, 23, 59, 59)
	end
	end

I propose we change that.

[navidcy]

Would such a change make other things cumbersome?

[navidcy]

cc @taimoorsohail

[glwagner]

I think the data is shipped to go fro Jan 1 to Dec 31 ?

[glwagner]

Check the NetCDF files

[navidcy]

OK, will do

[taimoorsohail]

JRA55 seems to go from May 1990 to April 1991, but the underlying ECCO data used for initialisation and relaxation (for sea ice) does not have data before January 1992. So, the simulation is being initialised/restored to a different year from the surface forcing field. We need to develop an alternative source of data for this purpose if we stick to RYF 1990-1991.

[simone-silvestri]

if you want there is this PR open that should allow using all the JRA55 data. #437

[taimoorsohail]

Thanks @simone-silvestri! I think 1990-1991 is an approximately "neutral" year so a good choice for JRA55. We just need to find an alternative initialisation source to ECCO that does go back to 1990.

[simone-silvestri]

There is World ocean atlas for example, you might want to look at https://github.com/briochemc/WorldOceanAtlasTools.jl. Otherwise it would be nice to have support for Copernicus or ORAS5

[glwagner]

Copernicus 1/12th degree would be very nice to support

[taimoorsohail]

Yes, I thought about World Ocean Atlas, but the quality of observations around the poles is pretty bad for the purposes of a full-depth relaxation At least ECCO evolves the ocean with a sea ice model, so any spurious observations are hopefully sorted out by the sea ice model. The Copernicus 1/12th-degree reanalysis also only begins from 1993..

IMO, ORAS5 is the best option as it goes back to 1958 without gaps in the data.

[glwagner]

Which is more important: the atmospheric forcing, or the ocean initial condition? And what is the use case?

[glwagner]

or use 2003

[glwagner]

I think it would be interesting to look at a few repeat years and focus on the ones we can run today for that study

[taimoorsohail]

The Stewart et al study (https://doi.org/10.1016/j.ocemod.2019.101557) explores 1984-85, 1990-91, and 2003-2004 years as options. I think they say all three are good candidates, though 03-04 has more of the anthropogenic warming signal in it.

I was actually thinking about this and I think WOCE/WOA implementation might be the best, because it aligns with OMIP2 protocols, and following sea ice implementation will probably be the way to go for any surface salinity restoring (if necessary).

[glwagner]

Notably that study does not include ocean model solutions forced by years that they would judge to be "bad" to illustrate why there could be a "bad year" to choose. I'm not sure all questions are settled here.

[glwagner]

My suggestion would be to move away from the notion that there needs to be a "commitment" to products of various kinds, based on a theoretical analysis or assumption. Better to just run cases! If a particular solution needs to be improved somehow that isn't too hard. In any case, any decision can be justified by running both possibilities.