-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi,
At the danish and swedish meteorolical institutes, we have implemented the
prototype suggestion.
We have two points we would like to discuss here:
- - Wavelength: we find it necessary to add the minimum and maximum wavelength
(spectral range) to the metadata. We think it is not enough just to specify one
wavelength for a band.
- - Geolocalization: we have quite a lot of data that is easily represented on a
grid, but the CF-convention requires the grid of lons and lats to be calculated.
This is of course highly inconvenient for fixed grids (from a data storage and
transfer perspective). We suggest adding a proj-string (proj.4 string
definition) and an area-extent option in the convention for grids. To us, it
makes a lot more sense than maintaining a list of accepted pre-defined grids in
the convention.
Best regards,
Martin Raspaud, Adam Dybbroe (SMHI)
Esben Stigård Nielsen, Kristian Rune Larsen, Lars Ørum Rasmussen (DMI)
Tom Whittaker skrev:
> We would like to suggest the following as a "starting point" that we
> can refine toward the goal of agreeing on a "basic" set of metadata
> and structure. It is an example of a multi-banded image structure
> that contains navigation on a point-by-point basis, and provides for
> linear calibration. The variable "names" are not critical, but the
> "standard_name" should be used by applications to identify the
> quantity.
>
> (In the following, most of the numbers are just samples....to make it
> more readable...)
>
> dimensions:
> x = 5304 ;
> y = 2438 ;
> band=16
>
> variables:
> float lat(y, x);
> lat:standard_name = "latitude" ;
> lat:units = "degrees_north" ;
> lat:valid_range = -90.f, 90.f ;
>
> float lon(y, x) ;
> lon:standard_name = "longitude" ;
> lon:units = "degrees_east" ;
> lon:valid_range = -180.f, 180.f ;
>
> int band(band);
>
> double time;
> time:long_name = "Nominal time of image";
> time:units="s since 1990-1-1 0:0:0";
>
> string bandname(band);
> bandname:standard_name = "band_name" ;
>
> float offset(band) ;
> offset:standard_name="linear_calibration_offset";
>
> float scale(band) ;
> scale:standard_name="linear_calibration_scale_factor";
>
> float wavelength(band);
> wavelength:standard_name="radiation_wavelength";
> wavelength:units="micron";
>
> string units(band);
> units:standard_name="band_units";
>
> short ImageData(band, y, x) ;
> ImageData:coordinates = "band lat lon" ;
> ImageData:valid_min = 1s ;
> ImageData:_FillValue = 0s ;
> ImageData:missing_value = 0s ;
>
> and for people who want/need to use a table look-up for calibration --
> instead of the "linear_calibration..."
> variables, we would use:
>
> float lookup(band, 256);
> lookup:standard_name="calibration_lookup_table";
>
>
> As Russ Rew pointed out, if there is only one band, it would be
> possible to dimension the Data as (y,x) but have the coordinates
> attribute contain "band".
>
>
> tom
> (with help from Tom Rink and Kaba Bah)
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/
iQEcBAEBAgAGBQJMc2/IAAoJEBdvyODiyJI4w0wIAIw5QsZO9r3qmDnZLCz1ot7h
+ws9qbrxklIfIxN7vuaSxTedonv++ulc9E90LVqyxKcOn/e8GM5XGSREjw8CAVhk
8PynsvvfHnjI5y3yZchoIy/znIt7bgTtGtTA6Ll+zgk3M5QxuFSC+jonfLk1EWzJ
KFR6+Vtb64aztmBFYaKkvvOyUC0Xr/kVX7y+5K4sclHbQQaOSxO67yX1BdpWWPAX
hXWWuLuLUZQNNwzi5o6FgQ2jom4mUxhZu/xbjeSl4fyehCgcw1TnE9BQ1aikTSaZ
dIxTGI7gbHNARglIp2f7npaR6/yvCx5jQKf5SqgHQBSmyG2TTmzsxbokLACrsd8=
=w/WO
-----END PGP SIGNATURE-----
begin:vcard
fn:Martin Raspaud
n:Raspaud;Martin
org:SMHI
adr;quoted-printable:;;;Norrk=C3=B6ping;;60176;Sweden
email;internet:martin.raspaud@xxxxxxx
tel;work:+46 (0)11 495 8261
tel;cell:+46 (0)11 495 8261
x-mozilla-html:FALSE
version:2.1
end:vcard
