search

cmip6dr / CMIP6_DataRequest_VariableDefinitions

Definitions of variables in the CMIP6 Data Request
7 stars 0 forks source link

area_fraction and cell_methods #411

Open taylor13 opened 2 years ago

taylor13 commented 2 years ago

In the CF standard name descriptions of names that include the string "area_fraction", the guidance provided includes the following:

"Area fraction" is the fraction of a grid cell's horizontal area that has some characteristic of interest. 
It is evaluated as the area of interest divided by the grid cell area.

This seems to rule out defining such fractions as the area of interest divided by the area of some portion of the grid cell (e.g., the area of snow on sea ice divided by the area of sea ice, which in general differ from the area of snow in a grid cell divided by the area of a grid cell). I think the current CF guidance is too restrictive and should be revised. If it is not, then some fractions defined in the CMIP6 data request should not have standard name area_fraction.

Suppose that the current CF guidance is revised to allow for more general area fractions. Then I still have some questions about some of the fractions requested in CMIP6:

  1. Why does variable cnc ("vegetation_area_fraction") have cell methods "area: mean where land time: mean ", whereas all the individual types of vegetation (e.g., cropFrac) have cell methods "area: mean where land over all_area_types time: mean"? The first method calculates the area of vegetation in a grid cell divided by the area of land in the grid cell. The second method calculates the area of crop in a grid cell divided by the area of the grid cell (including any non-land areas).
  2. It's not clear to me why both siconc with cell methods "area: mean where sea time: mean" and siconca with cell methods "area: time: mean" are both needed? Doesn't siconca=siconc*sftof, where sftof is the fraction of the grid cell occupied by sea.
  3. In the cell_methods for some variables you include as a comment the name of the "mask" that is used to determine what fraction of the cell is of a certain type, but in other cases you don't include the name of the "mask". Is there a reason for this, or perhaps is this something you never got around to cleaning up. (I wouldn't say it would be high priority on my list.)
martinjuckes commented 2 years ago

Hi @taylor13 , thanks for raising this.

I agree with your point that the definition of area_fraction. This looks like an issue for CF. There might be a case for a new standard name, e.g. partial_area_fraction, if anyone wants to keep the restricted for area_fraction (which could perhaps become cell_area_fraction). This means that the usage in CMIP6 variable SImon.simpconc is currently invalid.

  1. I'm afraid I don't recall the discussion, but as it stands cropFrac is a cell area fraction as you say and cnc is a partial area fraction. To me, this example suggests that it would be good to have separate standard names in order to make the difference really clear, rather than relying on cell methods.
  2. The two siconc variables were needed if a modeling centre provides sea-ice concentration on both atmosphere and ocean grids. This might be redundant with a new file naming scheme.
  3. The inclusion of the name of the mask was a new feature which was not fully implemented. I'd like to complete it at some point, but feedback on the utility of the current formulation would be useful. The absence of the name of the mask from the file metadata becomes an increasing problem as the number of mask variables expands. At present, the mask name does not appear in the external_variables attribute because it is only a comment, but it might be useful to have it there -- it is essential for many analysis calculations.