Hi Simon,
When I started writing code to handle NetCDF data, the files I was
dealing with used the axis = X and Y attribute for rotated pole
coordinates. So I took that interpretation to be standard and, like you,
coded to use that attribute as one way to identify the spatial
coordinate variables and dimensions. This seemed to make sense, I guess
since I'm used to seeing (x,y,z) most often as cartesian coordinates,
but sometimes a more generalised coordinate system, whereas I've more
often seen (phi,lambda) as first choice symbols for latitude and longitude.
The example I'm thinking of had variables like
A(dim1,dim2,dim3,dim4)
latitude(dim2,dim3)
longitude(dim2,dim3)
In locating A in (lat,lon) space, it seemed useful to look for an axis
attribute of "dim2" and "dim3", for the most general case in which their
names weren't recognizable, or we didn't want to rely on the dimension
ordering. Of course we should be able to rely on recognising the
"latitude" and "longitude" variables, so could work back from there.
Not that it would create a real problem whichever way the inconsistency
in CF is resolved: just pointing out how and why I found this use of the
axis attribute useful.
Regards,
Richard
On 20/11/2006 15:05 p.m., Simon Wood wrote:
> Karl, Jonathan and others,
>
>> I'll reiterate what I said earlier (but I think somehow the email
>> didn't get distributed). I'm in favor of eliminating the apparently
>> contradictory statements. Of the two choices, perhaps least confusion
>> would result from adopting Jonathan's suggestion: restrict X and Y to
>> genuine longitude and latitude coordinates.
>
> By 'genuine longitude and latitude coordinates' would you intend to
> exclude auxiliary (lat/lon) coordinates (eg lat(y,x) and lon(y,x)), and
> allow only for datasets with lat lon dimensions (and hence lat(lat) and
> lon(lon) coordinate variables)?
>
> I'm a bit confused since I originally interpreted CF 4 to mean that the
> axis attribute as an extra way of identifying a coordinate variable. I
> assumed it would be good to use it with auxiliary coordinate variables
> for datasets using x,y dimensions and, say, lambert conformal
> projection. But then by analogy to the discussion about rotated pole
> projections on the earlier part of this thread I got to thinking that
> maybe the axis attribute should be associated with the x(x) and y(y)
> 'projection_x/y_coordinate' variables instead (ie so it can be used as a
> hint to the plot orientation).
>
> However, it seems to me that restricting the attribute to only be used
> for 'proper' coordinate variables (ie lat(lat), lon(lon)) is a bit
> drastic (and it doesn't really add much value to these variables does
> it?). All thats really needed is to resolve the ambiguity in intended
> usage (ie is it to hint at plot orientation or to help identify
> coordinate variables including auxiliary coordinate variables). Of
> course if both interpretations are in common use then it gets a bit
> trickier...
>
> The reason I was intending to use the axis attribute was just in an
> attempt to include as much metadata as possible so that (hopefully) a
> wider range of data readers would be able to understand our data; I
> guess if its just going to add confusion I won't use it. Does anybody
> know of data readers that currently rely on the axis attribute in any way?
>
> regards
>
> Simon
>
--
=========================================================
Richard Gorman
National Institute of Water and Atmospheric Research
PO Box 11-115, Hamilton, 3251, New Zealand
Tel: +64 7 856 1736 Mob: 021 074 7490 Fax: +64 7 856 0151
Email: r.gorman at niwa.co.nz Web: http://www.niwa.co.nz
=========================================================
Received on Tue Nov 21 2006 - 15:16:38 GMT
