[CF-metadata] Fwd: Re: Request for a new standard_name : surface_snow_extent

From: Jonathan Gregory <j.m.gregory>
Date: Wed, 14 Aug 2013 15:55:05 +0100

Dear Thomas

> Would that mean something like:
>
> float snow_extent(time) ;
> snow_extent:standard_name = "surface_snow_extent" ;
> snow_extent:cell_methods = "area: sum where land" ;

Yes, that's right. Also
     snow_extent:units="m2"; // or equivalent

> To define such a threshold might be tricky :) Intuitively it should be a threshold on area_fraction, but maybe not always. In our processing, we compute a "probability of snow", and apply a threshold on this probability for deciding on snow/no-snow. This probability and how we compute it in the first place has nothing standardized.
>
> But let's take the sea_ice_extent as an example. Sea ice extent is (usually) the area of sea ice where sea_ice_area_fraction is more than 0.15. How should we properly define a ice_extent variable that specifies the threshold?

We already have a standard_name of sea_ice_extent, as you pointed out, and
I think its definition should include the threshold to define what defines
the ice edge, as a default. However if you wanted to state the threshold
explicitly, I'd do it like this:

float ice_extent(time) ;
  ice_extent:standard_name = "sea_ice_extent"; // the existing stdname
  ice_extent:cell_methods = "area: sum";
  ice_extent:coordinates = "threshold"; // time is a Unidata coord var
  ice_extent:units = "m^2"

float threshold;
  threshold:standard_name = "sea_ice_area_fraction"
  threshold:long_name = "scalar value to separate ice-covered from ice-free area"

int time(time);
  time:units = "seconds since xxxx"

That's essentially what you said, but we don't need "where" in this case,
I think, because the sea ice extent can only apply where there is sea ice,
unlike the snow extent.

Best wishes

Jonathan
Received on Wed Aug 14 2013 - 08:55:05 BST