[CF-metadata] standard names linked to melt ponds on sea ice

From: Jonathan Gregory <j.m.gregory>
Date: Thu, 6 Feb 2014 17:02:48 +0000

Dear Thomas

> > Yes, I think so too. Sea ice with melt ponds is still sea ice. In
> > terms of area types, sea = ice_free_sea + sea_ice.
>
> Agreed. But where is this written? Should it enter the "help string" in the table of Area Types?

I agree that would be helpful.

> > Given a new area_type, it is not necessary to
> > have a new standard_name as well, since you could use the generic area_fraction.
>
> This is interesting, I was not fully aware of that construct. I fear however that many application still expect a standard name as a string with underscores, maybe with a standard name modifier. I would not like to be the first one out with this construct...

I don't think you would be the first! The generic area_fraction was added so
that a data variable could supply area fractions for several area types at the
same time i.e. with a dimension for area type, as well as because this is a
way to limit the proliferation of standard names.

> I looked in the latest CF 1.7 draft1 and did not find a description of using area_fraction. Could someone propose an example, for example a (time, x, y) variable holding ice concentration (sea_ice_area_fraction) but using standard_name = "area_fraction"? Just as a thought experiment?

  float sea_ice_area_fraction(time,x,y);
    area_fraction:coordinates="area_type";
    area_fraction:standard_name="area_fraction";
  char area_type(maxlength);

with area_type="sea_ice" i.e. it's a string-valued aux coordinate variable
(CF section 6.1).

> Independent of this newer construct, and if I would go for proposing a new standard name: "sea_ice_melt_pond_area_fraction", how would it be defined? It is supposedly the relative areal extent of sea_ice_melt_pond, but I always have difficulties figuring out what the normalizing area is: is it the whole cell area, or the area covered by sea (as in ice_free_sea + sea_ice), or by sea_ice? sea_ice_melt_pond_area/sea_ice_area or sea_ice_melt_pond_area/sea_area? Depending on the answer, this could agree with what Siri is using for the dataset at NSIDC.
>
> For example, in my understanding, sea_ice_area_fraction is the area of sea_ice divided by the area of sea (again as ice_free_sea + sea_ice). So that a coastal cell with partial land cover of 0.6 could still have a sea_ice_area_fraction of 1.0 if the 0.4 sea area was fully ice covered. Correct? How can one tell from just "sea_ice_area_fraction"?

The default normalising area is always the grid-box area. Sect 7.3.3 describes
how to specify means normalised over portions of the grid-box area.

> Finally, I also need the area fraction of "melt pond free sea ice", so that calls for another surface type: "sea_ice_melt_pond_free_sea_ice".

That seems rather hard to read in English to me. What about pond_free_sea_ice?
Would that be clear enough?

Best wishes

Jonathan
Received on Thu Feb 06 2014 - 10:02:48 GMT