CF-ers,
I can certainly see that not all model metadata can be described
in CF, but it would be very useful if CF could provide some
guidance for people who would like to put georeferencing
information (GIS-required info) in their CF files. It seemed like
we were close to a solution earlier this year.
The last posting on the georeferencing issue suggested to allow
(*not* require) the use of an OGC-compliant Well-Known Text
(WKT) format to specify the coordinate reference system, so
that existing openGIS libraries could parse and utilize the
georeferencing information. Could this be done as a via the
existing "grid_mapping_name"?
This would have the side benefit that we wouldn't have
to keep expanding the CF Appendix F on Grid Mappings.
For the "grid_mapping_name", people could just specify
"SpatialReferenceSystemWKT" to allow complete specification of
the spatial reference system (vertical as well as horizontal).
So a CDL snippet might look like:
dimensions:
i = 228;
j = 306;
lev = 10;
time = 21;
float T(time,lev,j,i)
T: long_name = "temperature" ;
T: coordinates = "x y";
T: grid_mapping="SpatialReferenceSystemWKT"
double y(j,i)
double x(j,i)
x: units = "m";
x: standard_name = "projection_x_coordinate"
double y(j,i)
y: units = "m"
y: standard_name = "projection_y_coordinate"
char SpatialReferenceSystemWKT =
"PROJCS[\"Albers_Equal_Area_Conic\",GEOGCS[\"GCS_North_American_1927\",DATUM[\"D_North_American_1927\",SPHEROID[\"Clarke_1866\",6378206.4,294.9786982]],PRIMEM[\"Greenwich\",0],UNIT[\"Degree\",0.017453292519943295]],PROJECTION[\"Albers\"],PARAMETER[\"False_Easting\",0],PARAMETER[\"False_Northing\",0],PARAMETER[\"Central_Meridian\",20],PARAMETER[\"Standard_Parallel_1\",21],PARAMETER[\"Standard_Parallel_2\",-19],PARAMETER[\"Latitude_Of_Origin\",1],UNIT[\"Meter\",1]]";
In the fictitious case above, I was thinking of a case where
only the x,y projected coordinates are supplied and lon,lat
are not actually stored. In this case, could "coordinates"
point to the projected coordinate auxilliary variables (x,y)
and lon/lat be calculated by clients via the WKT string?
Could we allow "coordinates" to be used in this way, or do
we want to insist that "coordinates" always point to lon/lat,
and that lon/lat be supplied?
The WKT standard is described in the OpenGIS at:
<
http://geoapi.sourceforge.net/snapshot/javadoc/org/opengis/referencing/doc-files/WKT.html>
The objection was raised that CF would then become dependent
on a standard external to CF instead of being self-contained.
Wouldn't this be ok as long as CF didn't *require* the
use of OGC WKT? This seems to me a practical solution,
as open source software to parse these strings already exists
(e.g.
http://www.remotesensing.org/gdal/ogr/index.html, ArcGIS,)
and could be put to immediate use to serve CF-compliant
NetCDF data through OpenGIS Web Services or to injest into
GIS applications.
I hope we figure out something to do here, because we
(the USGS) are using CF for nested coastal ocean/estuary modeling
simulations where the grid spacing on the innermost mesh is
typically several *meters* to tens of meters. These simulations
are being used to help assess/predice beach erosion and
absolutely need to be properly georeferenced to compare with
shoreline change data.
Some other useful georeferencing links:
OGC specification for spatial reference systems:
(
https://portal.opengeospatial.org/files/index.php?artifact_id=6716)
Java libraries and tools using WKT: (
http://www.geotools.org) and
Geoserver (
http://geoserver.sourceforge.net/html/index.php)
-Rich
--
Richard P. Signell rsignell at usgs.gov
U.S. Geological Survey Phone: (508) 457-2229
384 Woods Hole Road Fax: (508) 457-2310
Woods Hole, MA 02543-1598
Received on Wed Apr 06 2005 - 15:05:51 BST
