The topics discussed were:
In order to help us get a better idea of what he meant by coordinate reference variables, John gave us an example of a staggered grid with N+1 coordinates for a grid dimension of N. We agreed that we'd think about this problem some more, and hope to get more examples and input from John on this subject.
John also raised other issues, at the meeting and in an email message to the NUWG group. Rather than summarize our discussions here, we point to an email thread which explores some of the issues in depth. All NUWG members and interested parties are encouraged to contribute their thoughts on this topic to the nuwg mailing list.
(1) x_dim and y_dim -- currently we have these dimensioned by nav, and
yet they are independant of the navigation but are dependent rather
on the items that are being navigated. I think we need another way
to do this, although I'm not even totally convinced on it's
usefullness.
(2) dimensioning the calibration information was also problematic for me
since it is really only a function of the scan line and not
a function of both the scan line and the calibration dimension.
In response to question 1, we decided that two sets of dimension naming
variables should be used in Andrew's CDL. One set, line_dim/elem_dim is dimensioned
by "satnav" and corresponds to the line and element dimensions used in the
satellite navigation parameterization, and a second set, x_dim/y_dim is dimensioned
by "gridnav" and corresponds to the x and y dimensions used in the remapped
grid navigation parameterization. Both the "gridnav" and "satnav" pseudo-structures
require information about the names of the dimensions. In the case of Andrew's
sample satellite CDL:
char x_dim(gridnav,dimNameLength);
x_dim:long_name = "x dimension name";
char y_dim(gridnav,dimNameLength);
y_dim:long_name = "y dimension name";
char line_dim(satnav,dimNameLength);
line_dim:long_name = "Line dimension name";
char elem_dim(satnav,dimNameLength);
elem_dim:long_name = "Element dimension name";
data:
x_dim = "x";
y_dim = "y";
line_dim = "y";
elem_dim ="x";
In response to question two, we discussed the dependence on the line dimension
(line_dim = y)of the calibration variables and the time associated
with each scan line. We decided that the most explicit way to document the
relationship was to dimension the four calibration variables and the time
by the line dimension. This scheme allows us to explicitly link the calibration
and time information for each line in the image. This will be very important
if the resolution of the image is reduced (from 4km to 20km resolution, for
example).
Andrew Sundelin has provided an updated sample satellite CDL to serve as an example. Please note that this sample CDL contains two navigation parameterizations, gridnav & satnav. Satnav represents the GOES-8 Kamel satellite navigation model, and has a corresponding list of navigation parameters all dimensioned by the variable "satnav". Gridnav represents the WMO GRIB Ed 1 orthographic projection, and has a corresponding list of navigation parameters all dimensioned by the variable "gridnav".
Andrew has provided the first draft of a document describing the NUWG conventions for un-remapped satellite data. The NUWG believes that these conventions are nearly complete.