Dear all,
I agree that standard name should describe the quantity itself, but not
how it's computed.
I note that for CMIP5 data we tried to make it clear how the quantity
identified by "sea_ice_thickness" was calculated, first by including the
cell_measures:
time: mean area: mean where sea
and then by including a "comment" attribute that reads:
"the mean thickness of sea ice in the ocean portion of the grid cell
(averaging over the entire ocean portion, including the ice-free
fraction). Reported as 0.0 in regions free of sea ice."
I think users of this data could easily avoid misinterpreting it simply
by reading the attributes describing it.
best regards,
Karl
On 7/21/16 9:12 AM, Jonathan Gregory wrote:
Dear Dirk
I understand your concern, but there are plenty of other instances where
important distinctions are made by coordinates and cell_methods rather than
by standard_name. For example, maximum, minimum, mean and instantaneous
temperatures are distinguished by cell_methods only. I can't think of another
instance where a given geophysical variable has more than one standard name.
I think that the users of the data have to be aware that the standard name
alone is not sufficient. In practice people often also identify variables by
the variable name, although CF does not standardise these, so that is not a
reliable method in general but could work in a particular project.
Best wishes
Jonathan
----- Forwarded message from Dirk Notz <[email protected]> -----
Date: Thu, 21 Jul 2016 15:32:37 +0200
From: Dirk Notz <[email protected]>
To: Jonathan Gregory <[email protected]>
Subject: Re: [CF-metadata] CMIP6 Sea Ice MIP: Ice thickness
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
Thunderbird/45.2.0
Dear Jonathan,
I appreciate that a CF variable is not solely defined by its variable
name, but also by its cell_methods.
I hence believe that the question of introducing
"sea_ice_equivalent_thickness" boils down to a question of the
overarching CF philosophy:
I agree that usually a new variable name should only be introduced if a
certain quantity cannot be described by existing variables.
However, I believe that in some cases, introduction of a new variable
name can also be warranted simply to prevent significant errors that can
occur because the actual data does not describe the quantity given by
the variable name because of applied cell_methods.
Even if we can create "equivalent_sea_ice_thickness" from the existing
variable "sea_ice_thickness", its physical usefulness does not primarily
relate to the actual sea-ice thickness implied by the variable name.
Instead, equivalent_sea_ice_thickness is primarily of interest for
oceanographers to easily appreciate the volume of sea water in the
vertical column that is on average frozen to sea ice. Given the
sometimes significant differences in magnitude between actual thickness
and equivalent thickness, I am worried that by keeping the same variable
name for such different quantities, significant confusion might arise.
Best,
Dirk
----
Dr. Dirk Notz
http://www.mpimet.mpg.de/~notz.dirk
Am 21.07.2016 um 15:05 schrieb Jonathan Gregory:
Dear Dirk
I think I get your point, but nonetheless I would say that the "equivalent
sea ice thickness" can be correctly described by existing conventions. If the
standard_name is sea_ice_thickness and the cell_methods says "area: mean",
the interpretation is that the sea_ice_thickness is integrated over the area
of the grid box and then divided by the grid box area. That is what you want,
isn't it? By contrast the local thickness of sea-ice, at a point, would have
cell_methods with "area: point", and the thickness averaged over the sea-ice
area is "area: mean where sea_ice". Thus the cases are distinguished by the
cell_methods. In the CF convention, the standard_name is only part of the
description of the data variable.
Best wishes
Jonathan
----- Forwarded message from Dirk Notz <[email protected]> -----
Date: Thu, 21 Jul 2016 13:33:55 +0200
From: Dirk Notz <[email protected]>
To: Jonathan Gregory <[email protected]>
Subject: Re: [CF-metadata] CMIP6 Sea Ice MIP: Ice thickness
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
Thunderbird/45.2.0
Dear Jonathan,
thanks a lot for your guidance. I understand that
equivalent_sea_ice_thickness can be constructed from existing mechanisms.
However, the definition of equivalent_sea_ice_thickness is not related
to actual sea-ice thickness, but simply defined as sea-ice volume per
grid area. This then only happens to have the same units as
sea_ice_thickness. We hence initially proposed to have
equivalent_sea_ice_thickness recorded as sea_ice_volume, but Alison
pointed out that units of "m" are not possible for any variable that is
called "volume".
However, given the substantial confusion that arose in some papers
published with CMIP5 data, where people assumed that "sea_ice_thickness"
is actual thickness, we believe that it would be worthwhile to add a
variable with a distinct name to avoid such confusion in the future.
Best,
Dirk
----
Dr. Dirk Notz
http://www.mpimet.mpg.de/~notz.dirk
Am 20.07.2016 um 15:21 schrieb Jonathan Gregory:
Dear Dirk
I believe that this distinction can be recorded by the existing mechanisms
of "where" and "over" in cell_methods, using the existing standard_name of
sea_ice_thickness. Please see Section 7.3.3 of the CF standard. The grid-box-
mean sea-ice thickness is "area: mean where all_area_types" and the thickness
meaned over sea-ice only is "area: mean where sea_ice", I think.
Best wishes
Jonathan
----- Forwarded message from Dirk Notz <[email protected]> -----
Date: Tue, 19 Jul 2016 09:53:02 +0200
From: Dirk Notz <[email protected]>
To: [email protected]
Subject: [CF-metadata] CMIP6 Sea Ice MIP: Ice thickness
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
Thunderbird/38.8.0
Dear CF community,
traditionally, the variable sea_ice_thickness from CMIP-type model
output was calculated by dividing the entire volume of sea ice in a grid
cell by the entire area of the grid cell, independent of the area
fraction of the grid cell that was actually covered by sea ice.
This gave rise to substantial confusion for users who expected that
sea_ice_thickness as stored within CMIP simulations refers to the actual
sea-ice thickness that is used in the sea-ice model code to calculate
heat fluxes, for example, rather than the average ice thickness that the
ice would have if it were to cover the entire area of the grid cell
while conserving its volume.
To prevent such confusion in the future, we would like to add the
following variable to the CF convention:
1. equivalent_sea_ice_thickness (new variable with units 'm3 m-2' or 'm')
to describe sea-ice volume per grid-cell area
The term "Equivalent sea-ice thickness" is known within the sea-ice
community to refer to "sea-ice volume per grid-cell area". Ideally, we
would have liked to suggest a variable name containing the term
"volume", but this seems difficult within the CF convention as then
units couldn't be 'm'.
Thank you very much for any feedback, help and guidance.
Best,
Dirk Notz
--
----
Dr. Dirk Notz
http://www.mpimet.mpg.de/~notz.dirk
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
----- End forwarded message -----
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
----- End forwarded message -----
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
----- End forwarded message -----
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
