[CF-metadata] CMIP6 Sea Ice MIP: Integrated quantities

From: Dirk Notz <dirk.notz>
Date: Fri, 5 Aug 2016 08:08:55 +0200

Dear Alison,

thank you very much for your thoughtful comments, which we had
considered carefully before submitting our proposal to the CF list. As
you will have noticed, we followed your advice on almost all other
variables, but for the integrated properties area/extent/volume, our
suggestion to the list differed from your suggestion.

This is because in a sea-ice context, combining the two terms "area" and
"extent" within single variable names would give rise to substantial
confusion. The two terms simply describe two very different things,
which are used to describe sea-ice coverage in all climate literature:

"area" always describes the areal coverage of the sea ice itself.

"extent", in contrast, always describes the areal coverage of all grid
cells that have a certain amount of sea ice in them. This is the only
variable that relates to a threshold within individual grid cells in our
variable proposal, and there is no meaningful definition of
sea_ice_extent without such threshold. Despite it's unusual definition,
this variable has been the standard variable to evaluate model
simulations of sea-ice coverage, and is shown in virtually all satellite
time series that describe the sea-ice loss in the Arctic, for example.
This is because observational uncertainty of sea-ice extent is much
small than the uncertainty for sea-ice area, which is why sea-ice extent
has been the preferred way of evaluating models. There are substantial
issues with this approach, but that's a different story :-)

Any variable containing in its name the phrase "area_extent" would hence
make it very difficult to figure out what this variable actually
describes. We hence think that we would do more harm than good in
combining these two words within individual variable names. It was
therefore that we decided to not follow your naming suggestion on these
two variables.

Regarding volume the terminology "volume_extent" would be highly
confusing to the community, as "extent" describes an areal coverage
(units m?), whereas volume has units m?.

We therefore found Jonathan's suggestion to simply use sea_ice_area,
sea_ice_extent and sea_ice_volume very good, since they most clearly
describe the physical meaning of the variables we want to describe.
Given that sea_ice_extent already exists within the CF convention,
simply updating its definition should be sufficient, since in the
published literature sea_ice_extent only ever has the single meaning of
an integrated quantity over several grid cells.

Regarding your individual comments:


> Initially, Dirk sent me a list of variables to preview before proposing the standard names to the mailing list. In it there were several variables relating to sea ice extent, area and volume, some of which refer to a threshold and others which don't. I've reproduced my original comments below with some additional explanation.

sea-ice area and volume: These are defined in individual grid cells, and
we additionally suggest to use them to describe integrated measures over
full hemispheres.

sea-ice extent: This is only defined over several grid cells, and is the
only variable that refers to a threshold. There is no meaningful
application of the terminology sea-ice extent for individual grid cells.

>
> 1. Current CF proposal: sea_ice_extent_in_region
> CMIP6 short name: siextentn
> Long name: Sea ice extent North
> Units: 10^6 km?
> Description: Total area of all Northern-Hemisphere grid cells that are covered by at least 15 % areal fraction of sea ice
> My suggestion:
> Standard name: area_extent_of_grid_cells_with_sea_ice_above_threshold (canonical units: m2)
> + new standardized region name 'northern hemisphere' which could be supplied in a scalar coordinate variable.
> The standard name definition can be written so as to allow a threshold of sea_ice coverage to be specified, in this case 15%, using a scalar coordinate variable. If no threshold is specified it is assumed to be zero. This would follow the practice adopted for existing names such as time_when_flood_water_falls_below_threshold.
>
> I appreciate that the existing name sea_ice_extent doesn't have a definition, and maybe we could write something that would allow us to use it for the aforementioned variable, but I don't like that solution for the following reason. The existing name sea_ice_area has the definition ' "X_area" means the horizontal area occupied by X within the grid cell', i.e. it is the area of the sea_ice itself. By analogy, I would expect sea_ice_extent to be defined as something like ' "X_extent" means the horizontal area occupied by X summed across the horizontal domain of the data variable as described by the associated coordinate variables and coordinate bounds or a scalar coordinate variable with a standard name of "region" '. However, if I have understood Dirk's variable correctly, it is the sum of the area of the grid cells that contain at least 15 per cent sea ice, not the area of the sea ice itself, and I would favour introducing a new standard name for that concept. Plus it gives us the opportunity to mention t
he threshold.
>
> Dirk's original list contained an analogous variable for the southern hemisphere for which we could use the same standard name. Instead of using a "region" scalar coordinate variable, we could specify the latitude and longitude extents of the northern and southern hemispheres using coordinate variables as others have already suggested. I'd be happy with either approach, so on that point I'll go along with the majority decision.

We find the definition suggested by Jonathan very helpful and good:
"sea-ice_extent: Total area of all grid cells in which the sea ice area
fraction equals or exceeds a threshold. By default the threshold is 15%.
The threshold can be specified by supplying a coordinate variable or
scalar coordinate variable with standard_name of sea_ice_area_fraction."

Regarding the variable name, there is no other possibility to define the
existing variable sea_ice_extent in a meaningful way, we believe. Unless
we drop that variable from the CF convention, there doesn't seem to be
any real need to define a new variable from our point of view.

Regarding the decision on whether the hemispheres should be defined by
lat/lon, or simply by a new region definition, I don't have a strong
opinion, and would be happy to follow either common practice or the
majority.



>
> 2. Current CF proposal: sea_ice_area_in_region
> CMIP6 short name: siarean
> Long name: Sea ice area North
> Units: 10^6 km?
> Description: Total area of sea ice in the Northern hemisphere
> My suggestion:
> Standard name: sea_ice_area_extent (canonical units: m2)
> + new standardized region name 'northern hemisphere' or specify horizontal domain via coordinate variables.
> I think the existing name of sea_ice_extent (with added definition as above: "X_extent" means the horizontal area occupied by X summed across the horizontal domain of the data variable as described by the associated coordinate variables and coordinate bounds or a scalar coordinate variable with a standard name of "region") can definitely be used for this variable, and for added clarity I'm suggesting we also modify the name itself. Thus sea_ice_extent would become an alias of sea_ice_area_extent. This name would also work for the analogous southern hemisphere quantity in Dirk's original list.

If we were to use the existing term "sea-ice extent" to describe actual
"sea-ice area", the CF convention would get in conflict with the entire
scientific literature, as described above. We don't think this would be
a good idea, and rather suggest to simply follow Jonathan's suggestion
to use the existing sea_ice_area, which then should be summed over a
given region. This would also underpin how closely related sea-ice area
in individual grid cells is to the integrated sea-ice area that we ask
for here.

>
> 3. Current CF proposal: sea_ice_volume_in_region
> CMIP6 short name: sivoln
> Long name: Sea ice volume North
> Units: 10^3 km?
> Description: Total volume of sea ice in the Northern hemisphere
> My suggestion:
> Standard name: sea_ice_volume_extent (canonical units: m3)
> + new standardized region name 'northern hemisphere' or specify horizontal domain via coordinate variables.
> We have an existing name sea_ice_volume which does not have a definition. I think we should add a definition and by analogy with sea_ice_area I suggest it should be ' "X_volume" means the volume occupied by X within the grid cell'. To indicate the sum of the volumes across a horizontal domain we would then need a new standard name. For consistency with sea_ice_area_extent I suggest 'sea_ice_volume_extent' for this name. This would be defined as ' "X_volume_extent" means the volume occupied by X summed across the horizontal domain of the data variable as described by the associated coordinate variables and coordinate bounds or a scalar coordinate variable with a standard name of "region" '. Again, this name would work for both northern and southern hemispheres.

We don't think it's helpful to use extent (which has units m?) to
describe a quantity that has units m?. Hence, as for sea-ice area, we
feel it's best to use the existing standard name sea_ice_volume, and to
ask that this should be summed up over a certain region. We also here
find Jonathan's respective remark very intuitive.

I hope this makes sense - and sorry for the lengthy reply. It's length
was also born out of the fact that I'll be offline for the coming three
weeks, and tried to clarify our view in one go.

Alexandra Jahn, who is co-chair of our CMIP6 sea-ice MIP, might have
additional comments while I'm gone.

Thank you very much once again for this constructive exchange!

All the best,

  Dirk
Received on Fri Aug 05 2016 - 00:08:55 BST