Re: [CF-metadata] Surface temperatures

[Jonathan Wrotny]

Dear Philip,

Thanks for the message and your perspective on equivalence of standard 
names.  In reviewing the surface_temperature definition again and 
thinking about the observational data product motivating my original 
submission, I do agree with you that the intent of the two quantities is 
the same and, thus, the surface_temperature standard name will suffice 
for the datal product that I have submitted.  Namely, they both refer to 
the interface or skin temperature of the land surface. For a model 
surface temperature, this is the temperature of the surface exactly at 
the interface of the land and atmosphere, while the land surface 
temperature inferred from an infrared radiometer, for instance, is the 
effective temperature of the land interface/skin and other possible 
terrestrial features (e.g. snow, vegetation) at the surface.  While both 
the model and observed surface temperature will have there own 
associated uncertainties, these two quantities are compared directly in 
practice in the community.  In other words, I think they would pass the 
test proposed in your post...

If I want to validate one dataset against another (model-model, model-obs, or obs-obs), would I be 
comfortable doing a simple subtraction between the variables in the two datasets (ie, without any sort 
of correction)?"     If the answer is "yes", then I consider the quantities to be 
comparable, and they should have the same std_name.  If the answer is "no" then they should 
have different std_names.

Related to this is another point...for the case of a remotely sensed 
temperature over land, the temperature will be sensitive to the 
penetration depth of the radiation used for the measurement. The 
penetration depth of the infrared measurements, for example, varies with 
surface type (e.g., soil, inland water, snow, ice), therefore is is not 
practical to assign a particular depth to the remotely sensed surface 
skin temperature measurements.  This lack of preciseness argues for not 
creating a unique standard name for the observable land surface 
temperature.  On the other hand, it is possible to be more 
radiometrically precise with the sea surface skin temperature inferred 
from an observation and this bears it out in the 
sea_surface_skin_temperature definition.

So, I would like to retract my proposal for the 
land_surface_skin_temperature.  I believe that the current standard name 
of surface_temperature will work fine.

Sincerely,

Jonathan Wrotny

On 10/4/2013 11:36 PM, Cameron-smith, Philip wrote:
Hi John,

I apologize for not being clear :-(, so thank you for asking :-).

I think you are raising a couple of issues:

1) Is it appropriate for an observation and a model variable to have the same 
std_name?

This wasn't the point I was trying to discuss.  So hopefully I can clarify my 
thoughts a bit.

Consider the (hypothetical) case where there is a theoretical quantity (e.g. 
skin temperature), which is calculated as part of a PERFECT model, and is also 
measured by a PERFECT instrument.   In that case, both the model and 
observational data should have the same std_name.

Now consider the case where the model and/or instrument are NOT perfect (e.g., 
the model equations actually use the average temperature of the top 20m of 
soil, or the instrument actually measures the air temperature 10m above the 
surface).

The test I apply is: "If I want to validate one dataset against another (model-model, model-obs, or 
obs-obs), would I be comfortable doing a simple subtraction between the variables in the two datasets (ie, 
without any sort of correction)?"     If the answer is "yes", then I consider the quantities 
to be comparable, and they should have the same std_name.  If the answer is "no" then they should 
have different std_names.    Hence, the question of whether the data is from models or observations is not 
relevant.

For my imperfect example above I would say it fails the test.  If I was faced 
with that situation I would want to apply corrections to account for the 
difference between an air temperature and a soil temperature below the surface.

If we improve the quantities (modeled or measured), then at some point they 
become 'close enough' to pass the test.

Now for the specific case at hand, would I directly compare the skin temperature in my 
model against  a spectroscopic measurement of the skin temperature from space (which 
presumably includes assumptions about fitting a black body and accounting for atmospheric 
absorption)?   My answer would be "yes", and hence my objection to the proposed 
std_name.

As a side note, I believe the decision of which std_name is appropriate for a 
quantity is generally best made by the person who creates the data, because 
they understand their own model or instrument, and they can also make the 
decision at the time the CF file is being created.   But this is getting into a 
separate issue.

2) Why am I objecting to land_surface_skin_temperature when 
sea_surface_skin_temperature has already been accepted?

There are many quirks and inconsistencies in CF, as you know ;-), mostly 
because we discuss std_names one at a time (so we don't always see the big 
picture).  I don't recall discussing sea_surface_skin_temperature.  Perhaps it 
was before I joined the mailing list, or I wasn't paying attention.  It is also 
possible that someone at the time made the argument that the temperature 10-20 
microns below the surface of the ocean is sufficiently physically distinct from 
the theoretical skin temperature. [Do we want to reopen this discussion?]

If the experts on this mailing list speak up and say that the distinction 
between 10-20 microns depth and the infinitesimal skin is physically important 
in practice, then I shall drop my objections.  I just note that the proposal 
description for land_surface_skin_temperature states that the distinction is 
not very important.

Best wishes :-),

      Philip

-----------------------------------------------------------------------
Dr Philip Cameron-Smith, p...@llnl.gov, Lawrence Livermore National Lab.
-----------------------------------------------------------------------



-----Original Message-----
From: CF-metadata [mailto:cf-metadata-boun...@cgd.ucar.edu] On Behalf
Of John Graybeal
Sent: Friday, October 04, 2013 6:04 PM
To: CF Metadata List
Subject: Re: [CF-metadata] Surface temperatures

Philip, to be clear, are you saying one should, or should not, compare values
from a model to observational values? We don't distinguish between them in
CF names that I know of, so I'm assuming it's OK....  (And to Jonathon
Wrotny's point: Considering a fundamental concept like "temperature of an
observable feature" somehow different just by virtue of being in a model, is
just way too big a Pandora's box to open up. In my humble opinion.)

For me, 'equivalent names' means equivalent _names_.  If land_surface_skin
temperature is equivalent to surface_temperature:cell_methods="area:
mean where land", then sea_surface_skin_temperature must be equivalent
to surface_temperature:cell_methods="area: mean where sea".  And I bet I
could find quite a few other 'equivalences' by using a cell method like "area:
mean where land/sea".

Not only are such equivalences quite uncommon so far, to me it is not all that
equivalent. (Perhaps that is my ocean background, where 'surface' is still a bit
indeterminate -- rightly or wrongly!)

John

P.S. IIRC, sea_surface_temperature used to be the only sea surface
temperature; other definitions were added because our view of the top of
the ocean -- through satellites or models or thermometers -- was much more
refined. So we needed more refined terms to make things comparable again.


On Oct 4, 2013, at 17:28, "Cameron-smith, Philip" <cameronsmi...@llnl.gov>
wrote:

Hi Jonathan (Wrotny),

The general practice of CF is that quantities that are 'equivalent', ie close
enough that it is meaningful to take the difference between them, should
have the same std_name (ie, they are both trying to calculate or measure
the same physical quantity).
IMHO, this provides huge value to users, since it tells them when they can,
or shouldn't, compare two quantities (eg, compare the surface temperatures
from a model dataset with satellite observations of surface temperatures).  If
'equivalence' is treated too strictly, then no variable can ever be compared to
another.
Unfortunately, there is a grey zone between quantities are equivalent and
quantities that are not, and then long discussions usually occur.
 From the description of the quantity you describe, it seems to me that
land_surface_skin_temperature and
surface_temperature:cell_methods="area: mean where land", should be
deemed to be 'equivalent'.
If you agree, then one advantage for you is that you don't have to do any
more work on this email list ;-).
Best wishes,

       Philip



----------------------------------------------------------------------
- Dr Philip Cameron-Smith, p...@llnl.gov, Lawrence Livermore National
Lab.
----------------------------------------------------------------------
-


-----Original Message-----
From: Jonathan Wrotny [mailto:jwro...@aer.com]
Sent: Thursday, October 03, 2013 12:40 PM
To: Cameron-smith, Philip
Cc: Jonathan Gregory; cf-metadata@cgd.ucar.edu
Subject: Re: [CF-metadata] Surface temperatures

Dear Philip,

My take is that the land_surface_skin_temperature and the
surface_temperature are likely very close in value, since the
surface_temperature is an infinitesimally thin layer at the bottom
level of the atmosphere which interfaces with the land skin (soil)
below - hence, the definition stating that they can be taken to be
equivalent.
The land_surface_skin_temperature proposal is motivated by a new
observational data product which is the radiating temperature of a
very thin, top layer of the land surface.  This quantity does not
currently exist in the CF standard name set, but has an analogue in
sea_surface_skin_temperature.
The surface_temperature name was added to CF because it is a standard
model variable, I believe. Someone please correct me if I am wrong,
but the radiating temperature of the Earth in models is often simply
referrred to as the "surface temperature," so I wanted to draw a
connection between the model quantity and the observable
land_surface_skin_temperature in the definition such that they are
effectively the same thing.  This seems to be one of those situations
where there are two quantities, one created for an observed quantity
and the other for a model quantity, but the two quantities likely
have very similar values.  I guess the question is whether or not this is
permissible within CF.
Sincerely,

Jonathan Wrotny

On 10/3/2013 1:30 PM, Cameron-smith, Philip wrote:
Hi Jonathan (Wrotny), Jonathan (Gregory), et al.,

I am a little surprised.

It is explicitly stated in the proposed description that
land_surface_skin_temperature "can be taken to be equivalent to"
surface_temperature over land areas.
In the description for surface_temperature, it indicates that it can
apply to
just land using cell_methods.   Indeed, in the CF convention, example 7.6
explicitly states this:
Example 7.6.  Mean surface temperature over land and sensible heat
flux
averaged separately over land and sea.
     float surface_temperature(lat,lon);
     surface_temperature:cell_methods="area: mean where land";

I also note that surface_temperature is already an alias for
surface_temperature_where_land (which I think is deprecated)

Why is a new std_name needed? What am I missing?

It is true that there is a variable called
sea_surface_skin_temperature, but
it appears that this was introduced for different reasons.   Specifically, it
looks
like sea_surface_temperature was created to refer to the water _near_
the
surface to distinguish it from the 'skin'.   sea_surface_skin_temperature
then
differs from surface_temperature because it refers to the interface
under sea-ice rather than above sea-ice.
Best wishes, as always :-),

     Philip

--------------------------------------------------------------------
--
- Dr Philip Cameron-Smith, p...@llnl.gov, Lawrence Livermore National
Lab.
--------------------------------------------------------------------
--
-


-----Original Message-----
From: CF-metadata [mailto:cf-metadata-boun...@cgd.ucar.edu] On
Behalf
Of Jonathan Gregory
Sent: Thursday, October 03, 2013 9:35 AM
To: cf-metadata@cgd.ucar.edu
Subject: Re: [CF-metadata] Surface temperatures

Dear Jonathan

The new proposal looks fine to me. Thanks. I see that you don't
have to define the thickness of the layer; instead, you are
defining it implicitly through the method of diagnosis. Others may
have views, of
course.
Cheers

Jonathan

----- Forwarded message from Jonathan Wrotny <jwro...@aer.com>
-----

Date: Wed, 2 Oct 2013 11:26:27 -0400
From: Jonathan Wrotny <jwro...@aer.com>
User-Agent: Mozilla/5.0 (Windows NT 5.1; rv:17.0) Gecko/20130801
        Thunderbird/17.0.8
To: Jonathan Gregory <j.m.greg...@reading.ac.uk>, "cf-
metad...@cgd.ucar.edu"
        <cf-metadata@cgd.ucar.edu>
Subject: Re: [CF-metadata]  Surface temperatures

Dear Jonathan Gregory,

I am getting back to this reply after a long time - sorry, I was
pulled in a few different directions lately.  Hopefully, it is
possible to bring back to life a submission that I had made for
the land_surface_skin_temperature.

Revisiting my previous proposal and a few e-mails by Karl Taylor
and Evan Manning, I have made some modifications to the definition
of this standard name so that I can incorporate some suggestions
by Karl and Evan.  Here is my current proposal:

Standard Name:land_surface_skin_temperature

Definition:The land surface skin temperature is the temperature of
a land point or the land portion of a region as inferred from
infrared radiation emitted directly towards space through the
atmosphere. Not all of the emitted surface radiation originates at
the soil.Some may come from various terrestrial features (e.g.,
vegetation, rivers, lakes, ice, snow cover, man-made
objects).Thus, the land surface skin temperature is the aggregate
temperature of an effective layer which includes the soil and
terrestrial features at the surface (if they occur).In models, the
radiating temperature of the surface is usually the
"surface_temperature", which then can be taken to be equivalent to
land_surface_skin_temperature or sea_surface_skin temperature,
depending on the underlying medium.
Canonical Units:K

Thanks for still considering this proposal.  Sincerely,

Jonathan Wrotny

On 8/1/2013 12:56 PM, Jonathan Gregory wrote:
Dear all

I agree with Karl than in CF standard names "land" means
"non-sea", whereas sea-ice is part of sea. Hence I would support
adding land_surface_skin_ temperature, for use by applications
which classify
locations as land or sea.
However I also agree with Evan that one can approach this more
generally, and therefore I would also support the addition of
surface_skin_temperature, with which an area-type could be
specified, if anyone wants to follow that approach (we only add
names when they
are needed).
The quotations that Evan made show that we need to change the
definitions where they mention "skin". This is because in these
new names "skin" is being given a more precise and practical
meaning, motivated by observational methods, whereas the
surface_temperature
names were introduced for models, in which the skin can be a
notional
and infinitesimally thin layer.
Best wishes

Jonathan
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
----- End forwarded message -----
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
------------------------------------
John Graybeal
Senior Data Manager, Metadata and Semantics

T +1 (408) 675-5545
F +1 (408) 616-1626
skype: graybealski

Marinexplore
920 Stewart Drive
Sunnyvale, CA



_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata

_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata