Hi,
On 26-11-14 17:52, alison.pamment at stfc.ac.uk wrote:
> Dear Maarten,
>
> Thank you for your proposals. I have now created entries for them in
> the standard names editor at
> http://cfeditor.ceda.ac.uk/proposals/1?status=active&namefilter=&proposerfilter=Maarten&descfilter=&unitfilter=&yearfilter=&filter+and+display=Filter,
> so they certainly will not be lost. (At the moment they appear with
> blue headings to indicate that they are new proposals - I wanted to
> demonstrate this feature of the editor - but I will shortly change
> their status to "Under Discussion"). By proposing your names to the
> list they are already "in the process" so now it is a case of
> responding to any further comments as they arise. Names are formally
> accepted when consensus has been reached on the terms themselves, the
> definitions and units. Once accepted, they are automatically included
> in the next update of the standard name table.
Thank you.
> I have given my detailed comments on the individual proposals and
> your additional questions below.
>
> 1. toa_incoming_photon_solar_irradiance_per_unit_wavelength
>
> I agree with Jonathan's comment that "solar" is not really necessary
> and you have indicated that it is OK to remove it, so we would then
> have toa_incoming_photon_irradiance_per_unit_wavelength.
>
> Taking your own definition and adapting the wording slightly for
> consistency with existing "toa_incoming" and "irradiance" names I
> arrived at the following:
>
> "toa" means top of atmosphere. The TOA incoming photon irradiance is
> the photon flux from the sun on a surface perpendicular to the
> incoming radiation i.e. the "downwelling" flux. In accordance with
> common usage in geophysical disciplines, "flux" implies per unit
> area, called "flux density" in physics. A coordinate variable for
> radiation wavelength should be given the standard name
> radiation_wavelength. A photon flux is specified in terms of numbers
> of photons expressed in moles.
>
> Canonical unit: mol m-2 m-1 s-1
This sounds fine to me.
> As it stands, this looks fine and could probably be accepted for
> inclusion in the standard name table straight away. However, you
> included some notes about the coordinate variable and unit which need
> some further consideration. I'll address the units issues first, as
> these are the easiest to answer!
[...]
> It looks as though we would simply need to request an additional
> alias of 'photon' for the existing definition of mol/6.02214129e23.
> Do you agree?
Yes, this is exactly what I had in mind.
> On 12/11/14 Maarten wrote:
>> Note 4: Other instruments may use other units, depending on the
>> construction of the instrument. These can not be converted easily
>> into one another as they require additional knowledge, in
>> particular knowledge of the wavelength. I'm unsure if UDUnits is
>> equiped to handle these types of conversion. Additional names may
>> be needed for using "W m-2 m-1" or "W m-2 m-1 sr-1" respectively
>> (similar name, but leaving out the 'photon') Another quantity that
>> is commonly encountered expresses the radiation_wavelength as
>> wavenumber or frequency, and therefore also alters the
>> (frequency/energy) interval over which the quantity is integrated.
>> I'll leave the addition of these quantities to communities that
>> actually use them, IASI on MetOp comes to mind.
>
> It is accepted CF practice that different units for similar
> quantities, where conversion is not simply a matter of scaling,
> require different standard names. An existing example would be the
> ozone quantities
> equivalent_thickness_at_stp_of_atmosphere_ozone_content (canonical
> unit of metres) and atmosphere_mole_content_of_ozone (canonical unit
> of mol m-2). It is also accepted practice that we do not create new
> standard names unless someone actually needs to use them, so I think
> you are correct that other communities should propose the additional
> names as and when they are required.
Agreed.
> On 12/11/14 Maarten wrote:
>> Note 1: The coordinate variable for radiation wavelength may
>> require a mapping that is similar to the mapping of geolocation
>> coordinates for satellite observations, i.e. not a dimension, but a
>> multi dimensional mapping. This is the case for the OMI instrument,
>> where a 2D-dectector is used for wavelength (columns of CCD) and a
>> spatial dimension across the swath (rows of CCD). The mapping is
>> imperfect, and the wavelength depends on the row-index.
>
> I don't think we have encountered this type of data product before!
> The question of how to deal with the coordinate variables should not
> necessarily delay agreement of the standard name as coordinates are
> more properly in the realm of the main CF Conventions document. If
> writing your data will require coordinate variables to be treated in
> a new way then it would be necessary to propose a modification to the
> conventions as well as proposing the standard name.
I don't think so, but to be sure: adding a reference to a coordinate
variable in a standard name is probably causing a problem somewhere down
the line. It would be fine if instead of coordinate variable, it would
read: "coordinate variable or associated / ancillary / auxiliary
variable". In any case a link that is less strong than a dimension.
Note that I'm not suggesting to alter all standard_names where this may
occur, but it is worth noting that dimensions (and therefore coordinate
variables) are not always obvious.
> Certainly the conventions already allow 2D coordinate variables for
> use with geospatial grid mappings, but I don't think we have ever had
> a situation where one dimension is wavelength and the other a
> location. If I understand you correctly, the wavelength of a single
> data value also depends on where in the swath the observation was
> taken. Is that right? Is this because the instrument is scanning
> through a range of wavelengths as the satellite moves along its
> track?
Here is the measurement principle:
http://www.knmi.nl/omi/research/instrument/characteristics.php?tag=principle
A CCD is used to record the spectrum, with one dimension for the
wavelengths (CCD columns), and the other for the spatial dimension
across track (CCD rows). Due to the geometry the wavelength registration
differs slightly for each row. There is no scanning, either across track
or through the wavelengths.
> If so, I am wondering if it might be possible to get around
> this by using a spatial coordinate variable and also a
> radiation_wavelength auxiliary coordinate variable on the same
> dimension of the data. Would that work?
I think that would work, although we may want to skip the dimension for
the flight direction (along track dimension). Conceptually this sounds
very similar to the situation with a coordinates attribute for satellite
observation swaths.
> How many spatial dimensions
> are there in your data product?
time (T), ground_pixel (X), scan_line (Y), spectral_channel.
In our product time is a length-1 dimension, with the the offset to the
actual time recorded in an ancillary variable.
ground_pixel is across track, scan_line is the along track dimension.
The spectral_channel is the wavelength direction, but with a (slightly)
different wavelength scale for each ground_pixel. latitude & longitude
are attached via the normal coordinates attribute. Given the orbit (sun
synchronous polar), the axes are as indicated.
> I'd welcome input from anyone else
> who can help with this issue.
I agree that this is more a CF issue, and I will bring it up in the CF-2
discussions.
> 2. toa_outgoing_photon_radiance_per_unit_wavelength
>
> Putting together your own definition with existing definitions of
> photon_radiance names I arrived at the following:
>
> "toa" means top of atmosphere. Photon radiance is the photon flux in
> a particular direction, per unit of solid angle. The direction from
> which it is coming must be specified, for instance with a coordinate
> of zenith_angle. If the radiation does not depend on direction, a
> standard name of isotropic radiance should be chosen instead. A
> photon flux is specified in terms of numbers of photons expressed in
> moles. In accordance with common usage in geophysical disciplines,
> "flux" implies per unit area, called "flux density" in physics. A
> coordinate variable for radiation wavelength should be given the
> standard name radiation_wavelength.
>
> Canonical unit: mol m-2 m-1 s-1 sr-1
>
> Is this OK?
Yes, with a note that the CF conventions may need to say something about
attaching a viewing geometry (solar zenith angle, instrument zenith
angle, azimuth angles) to a variable that needs this.
> My earlier comments on units and coordinate variables apply to this
> name also. If necessary, we can start a new thread on the mailing
> list to discuss the coordinate issue more thoroughly.
Some discussion here never hurts, but I think adding this to CF-2 will
provide a little more room for a neat solution. Opinions are very
welcome to get it right in CF-2.
Best,
Maarten Sneep
--
KNMI
T: 030 2206747
E: maarten.sneep at knmi.nl
R: A2.14
Received on Wed Nov 26 2014 - 12:32:08 GMT
