Open aidanheerdegen opened 1 week ago
Linking https://github.com/ACCESS-NRI/access-esm1.5-configs/pull/18 (specifically https://github.com/ACCESS-NRI/access-esm1.5-configs/pull/18/files#diff-b335630551682c19a781afebcf4d07bf978fb1f8ac04c6bf87428ed5106870f5R10) as this is relevant to the documentation in that PR.
as this is relevant to the documentation in that PR.
Can you guess what I was reviewing when I decided to post this?
I don't necessarily have any expertise in this bit, but for whatever is decided we should still prepend [dev|release]-<config>
for whatever <config>
.
Xdeg
would be good to keep if it's common between all/most models.
What is meant by n96
?
If we go with long names, can we make -
reserved, as well as _
? We currently use -
as a separator for [dev|release]-<config>-<tag>
. pre-industrial
would break that convention.
n96
is the atmosphere "resolution" (it's a spectral model so this doesn't really have a well defined spatial resolution I believe but @blimlim is better placed to comment).
I agree -
should be reserved, so pre_historical
if we want to "go long or go home".
Even with pre_historical
, we've been using _
as a separator between 'fields', for lack of a better word. Like 1deg_jra55_iaf_bgc
, _
separates resolution, forcing product and forcing mode (and the BGC bit). I think we're running out of freely usable separators :D
Even with
pre_historical
, we've been using_
as a separator between 'fields'
So prehistorical
it is! :)
n96
is the atmosphere "resolution" (it's a spectral model so this doesn't really have a well defined spatial resolution I believe
From what I understand, the UM uses a finite difference scheme with regular lat/lon grid. The n96
grid has a lat x lon grid cell size of 1.25 x 1.875 degrees, which results in there being 145 latitudes and 192 longitudes. I'm not sure how it originated, but the nXY
naming scheme seems to just halve the number of longitudes, e.g. n48
has 96 longitudes, and I think n216
has 432.
Not sure if I have any ideas that are too useful here! We could go with the CMIP6 experiment names, e.g. piControl
and historical
from the experiment list here (scroll down to the CMIP activity ID
to see the relevant ones), though that might cause problems down the line. E.g. many do use dashes in their name, and the emissions driven experiments are named things like esm-piControl
(I'd probably get confused thinking this was referring to the model rather than the experiment). If a configuration is from a CMIP experiment though, it would probably be good state which one it is somewhere in the configuration (maybe in the readme?).
This may also be informed by/be consistent with the input file naming scheme.
If we copy the naming schemes we've been thinking about for the input files , do you think it reads ok or is it a bit unwieldy? E.g. release-global.oi_1deg.a_N96_historical_conc
From what I understand, the UM uses a finite difference scheme with regular lat/lon grid. The
n96
grid has a lat x lon grid cell size of 1.25 x 1.875 degrees, which results in there being 145 latitudes and 192 longitudes. I'm not sure how it originated, but thenXY
naming scheme seems to just halve the number of longitudes, e.g.n48
has 96 longitudes, and I thinkn216
has 432.
Sorry for the wrong information and thanks for the correction @blimlim!
Not sure if I have any ideas that are too useful here! We could go with the CMIP6 experiment names, e.g.
piControl
andhistorical
from the experiment list here (scroll down to the CMIPactivity ID
to see the relevant ones), though that might cause problems down the line. E.g. many do use dashes in their name, and the emissions driven experiments are named things likeesm-piControl
(I'd probably get confused thinking this was referring to the model rather than the experiment). If a configuration is from a CMIP experiment though, it would probably be good state which one it is somewhere in the configuration (maybe in the readme?).
We've chosen to define an experiment as a "realisation of a configuration". What we call an experiment CMIP calls a simulation. A CMIP experiment is a description of a protocol for a simulation.
That is a long-winded way of saying in our way of working piControl
should be the name of the branch (experiment) that is a specific realisation (series of runs) of a pre-industrial configuration.
This may also be informed by/be consistent with the input file naming scheme.
If we copy the naming schemes we've been thinking about for the input files , do you think it reads ok or is it a bit unwieldy? E.g.
release-global.oi_1deg.a_N96_historical_conc
Yes I think that is probably unwieldy. Because we get to define what a name means I think we can say that it is, for example:
<spatial_extent>_<ocean_resolution>_<atmosphere_resolution>_<scenario>+<modifier>
and assume the oi
and a
bits. I think they tend to make it more difficult to comprehend.
If we break it up into those fields it might help
Field | Possible values |
---|---|
Spatial extent | global ,panant ,regional ,aust |
Ocean resolution | 1deg , 2deg , 025deg |
Atmosphere resolution | n48 , n96 , n512 |
Scenario | historical , preindustrial , pliocene , lgm , ssp585 |
Modifiers | emiss , conc |
Is "scenario" a good description? It could be more specific, e.g. "chronological period", if that was all that it is ever required for, but as I included a CMIP future scenario if we agree that is the appropriate then scenario is probably a better term.
I added "Modifiers" as a general bucker for some extra specification, and there could be multiple but I can't think of any.
Any opinions @MartinDix?
scenario is better than chronological period because some cases don't have a real date range associated with them, e.g. 4xCO2. There are more possible modifiers, e.g. land use only, GHG only. Not sure that conc/emiss is the best choice. We should check with Tilo and Tammas.
Probably not useful input at this point, but having experiment names that start with a number can be a little annoying sometimes because many languages do not allow variable names that start with a number. I guess we're already pretty much already locked in to these so... just ignore me.
For ESM1.5/6 we don't need the spatial extent. It might be useful sometime further into the future. To avoid having a number at the start could put the atmosphere first, n96_1deg ...
There are more possible modifiers, e.g. land use only, GHG only. Not sure that conc/emiss is the best choice. We should check with Tilo and Tammas.
Who/How do we do this?
I only suggested a +<modifier>
in case that was a useful idea. Particularly if modifiers are optional, if there was the possibility of multiple modifiers for a given configuration, or if there was a "base" configuration and an equivalent with some modification. If there is always a modifier field then it probably makes sense to use a _
separator.
having experiment names that start with a number can be a little annoying sometimes because many languages do not allow variable names that start with a number.
For ESM1.5/6 we don't need the spatial extent. It might be useful sometime further into the future. To avoid having a number at the start could put the atmosphere first, n96_1deg ...
So proposal for configuration name is:
<atmosphere_resolution>_<ocean_resolution>_<scenario>+<modifier>
with corresponding branches with release-
and dev-
prepended.
This issue has been mentioned on ACCESS Hive Community Forum. There might be relevant details there:
https://forum.access-hive.org.au/t/access-esm1-5-access-nri-flagship-release-plans/1772/16
Should probably be named something like concentration-driven vs. interactive carbon cycle, because the pre-industrial simulation does not have emissions but can have an interactive carbon cycle enabled. The emissions, whether 0, constant of variable are indicated by the "scenario" descriptor.
Anyone with opinions about what the names of the initial configurations should be please chime in here. The CMS repo just called them
historical
,pre-industrial
,miocene
etc.The OM2 configurations, by necessity, have resolution, forcing product and forcing mode in the name, e.g.
release-1deg_jra55_iaf
.Do we want to adopt any naming convention like this? Clearly a forcing product and mode is not required, but perhaps it useful to include a resolution as a standard part of our configuration naming for consistency and transparency, to make it obvious to users what this refers to. This may also be informed by/be consistent with the input file naming scheme.
So we could start with:
1deg_n96_hist_conc
1deg_n96_pi_conc
given that emissions driven will be near-term targets they would be something like
1deg_n96_hist_emiss
1deg_n96_pi_emiss
Those are fairly terse, could go for longer and more descriptive
historical
andpre-industrial
.Thoughts?