| Re: Data format for spherical data [message #44227] |
Wed, 25 May 2005 20:16 |
James Kuyper
Messages: 425
Registered: March 2000
|
Senior Member |
|
|
Wayne Landsman wrote:
> This is not really an IDL question, though I will be using IDL to convert to
> the new data format.
>
> I've only worked only within the provincial field of astronomy, and so I've
> never had to deal with data formats other than FITS. But our data center
> is now being asked to make our all-sky data available in a more popular (but
> still scientific) data format, where suggestions so far for the new format
> have included NetCDF and GeoTIFF. The data will consist of various
> quantities (e.g. flux, error, # of observations) attached to ~60,000
> longitude, latitude points covering the entire sphere.
>
> I suppose the most important criterion for the format is popularity (we only
> want to do this once) although an ability to naturally handle spherical data
> is a plus. Any suggestions?
NetCDF provides no special support for spherical data. You can store
such data without any problems, but NetCDF provides no special
interpretation of such data.
HDF-EOS is built on HDF, which is implemented in a fashion compatible
with NetCDF. HDF-EOS has as a fundamental concept that data fields and
geographic fields (latitude, longitude, and altitude) are linked
together. It provides three different formats for geographically
organized data.
The Point format is for an unorganized collection of individual data
points, and associates a latitude and longitude with each of those data
points. However, there's nothing in the Point interface which
interprets the latitude and longitude fields as latitudes and
longitudes. You could fill those field with temperatures and pressures,
and it would work the same.
The Swath format organizes the data and geographic fields as
2-dimensional arrays, which might have different resolutions. Typically
the dimensions of the data fields will represent image coordinates for
data collected from a satellite. It attaches HDF metadata to the file
documenting a simple linear mapping relationship between each dimension
of the data arrays and the corresponding dimension of the arrays
containing latitude and longitude. It's intended that the lower
resolution fields be interpolated to match the higher resolution fields
using that mapping, but the Swath interface doesn't actually provide a
routine for carrying out that interpolation. As a result, just like the
Point interface, the geographic fields don't have to actually contain
geographic data.
Only the Grid interface actually provides routines that interpret
latitude and longitude as geographic coordinates. HDF-EOS Grids are
2-dimensional arrays of data, arrays represent positions evenly spaced
on a rectangular grid in some map projection, and the grid has to be
oriented to the U-V coordinates in that projection. Unlike the Point
and Swath formats, the Grid format does not store latitude and
longitude directly in the file. It only stores information about the
map projection used, the position on that map projection of the corners
of the rectangular grid, and the number of rows and columns in the
grid. The grid interface includes a routine (represented by
EOS_GD_INTERPOLATE in the IDL interface) that takes a series of
latitude/longitude pairs, interprets them according to the map
projection, and interpolates specified data fields in that map
projection to those locations. Oddly enough, it doesn't provide
routines for calculating latitudes and longitudes that correspond to
the data grid points.
|
|
|
|
|
|
| Re: Data format for spherical data [message #44238 is a reply to message #44236] |
Wed, 25 May 2005 08:43  |
R.G. Stockwell
Messages: 363
Registered: July 1999
|
Senior Member |
|
|
"Wayne Landsman" <vze49sx6@verizon.net> wrote in message
news:1117031988.627310.122490@o13g2000cwo.googlegroups.com.. .
> Sorry if this posting appears twice. Also, this is not really an IDL
> question, though I will be using IDL to convert to the new data format.
>
> I've only worked only within the provincial field of astronomy, and so
> I've never had to deal with data formats other than FITS. But our
> data center is now being asked to make our all-sky data available in a
> more popular (but still scientific) data format, where suggestions so
> far for the new format have included NetCDF and GeoTIFF. The data
> will consist of various quantities (e.g. flux, error, # of
> observations) attached to ~60,000 longitude, latitude points covering
> the entire sphere.
>
> I suppose the most important criterion for the format is popularity (we
> only want to do this once) although an ability to naturally handle
> spherical data is a plus, as is the ease of the conversion within IDL.
> Any suggestions?
>
> Thanks, -Wayne Landsman
NetCDF is very nice. It is always my format of choice. The size of your
data
does not seem restrictive. What other languages do you anticipate needing to
supply?
netcdf has builds on a variety of platforms:
http://my.unidata.ucar.edu/content/software/netcdf/builds/in dex.html
Cheers,
bob
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
