Rob Cermak wrote:
I am pleased that this has raised this question of Gaussian grids! We have rectified the
grid in X-Y for our datasets, but we have abused the describeCoverage XML to explain the
Z dimension as follows
And, not sure if this is the same issue, rotated grids? Grids like #203
and also RUC2 grid #252. No where near monotonic for lat and lon and
typically the only georeference information given is a reference lat/lon
and reference lon as below:
lat1: 16.281
lon1: 233.862
lov: 265.000
RUC2 #252:
rec 1:0:date 2005062312 VGRD kpds5=34 kpds6=105 kpds7=10 levels=(0,10)
grid=252 10 m above gnd anl:
VGRD=v wind [m/s]
timerange 0 P1 0 P2 0 TimeU 1 nx 301 ny 225 GDS grid 3 num_in_ave 0 missing 0
center 7 subcenter 0 process 105 Table 2
Lambert Conf: Lat1 16.281000 Lon1 233.862000 Lov 265.000000
Latin1 25.000000 Latin2 25.000000 LatSP 0.000000 LonSP 0.000000
North Pole (301 x 225) Dx 20.318000 Dy 20.318000 scan 64 mode 8
min/max data -12.4 12.2 num bits 8 BDS_Ref -124 DecScale 1 BinScale 0
Regridding is fine, but the preference is to maintain the integrity of the
data in its original form.
Hi Rob:
Generally, projection grids are curvilinear in lat, lon, so what you are
saying applies to all those, including the RUC.nc example. However,
these are almost always regularly spaced in x,y on the projection plane.
We do seem to have a problem in that GML does not appear to have a clean
way to specify these projections, unless they are in the EPSG database.
I agree with you about serving data in the original form, its what we
are hoping we can do with WCS.