RE: Resampling from lat/lon to projection

Don -

> -----Original Message-----
> From: Don Murray [mailto:dmurray@xxxxxxxxxxxxxxxx]
> Sent: 20. mars 2003 15:48
> To: Kjell Røang
> Cc: visad-list@xxxxxxxxxxxxx; Richard D Gonzalez; Tor Christian Bekkvik
> Subject: Re: Resampling from lat/lon to projection
>
>
> Kjell-
>
> Kjell Røang wrote:
>
> > As mentioned elsewhere, we want to use another projections package (
> > http://www.geotools.org/ ). We plan to use completly other kind out data
> > that we need to transform. I can't find this MapProjectionDisplay in my
> > display.
>
> You could still use GeoTools on the VisAD adapter by creating
> a GeoToolsCoordinateSystem that would take a GeoTools object
> and use that to do the conversion.  I'll look at GeoTools when
> I get a chance (maybe next week).  MapProjectionDisplay
> is in the examples directory.  The examples do not have
> packages, so you would have to cd to the examples directory
> and run java MapProjectionDisplay.
>

That would be a good idea. The MapProjectionDisplay was not in the src I
downloaded from sourceforge (the latest from 2002) but I found it in the
source at ftp://ftp.ssec.wisc.edu/pub/visad-2.0/visad_src-2.0.jar . By the
way. Which version should I use, or should I use CVS at sourceforge?

>
> > It sounds pretty much how I will like to do it, but I might use
> Ricks mods
> > to do the transformations. (JGrib should also support new projections as
> > many will use JGrib without using Visad). In VisAd notation,
> the data would
> > look something like:
> >
> > time -> ((lat, lon, z?) -> (pressure, v-wind, u-wind, ...))
>
> It would be good to support new projections.  As I say above,
> if you are going to use GeoTools, then you could create a
> GeoToolsCS that could be used in the VisAD objects (instead of
> GRIBCoordinateSystem).  It would extend visad.georef.MapProjection.
>
> I would also suggest that you try to do lazy evaluation of t
> the GRIB file using FileFlatFields instead of FlatFields.  This
> will be a little more work, but will only cause memory to be used
> when a particular Field in requested.  So the structure would be:
>
> time -> (((lat, lon, z) -> pressure),
>           ((lat, lon, z) -> v-wind),
>           ((lat, lon, z) -> u-wind),
>            ....)
>
> The Vis5D and netCDF adapter (Plain) use this strategy.

JGrib use streams only (not random access), but I can keep a compressed GRIB
record in memory (LigthtRecord) and probably use that as input to a "lazy"
evaluation".

Kjell

>
> Don
> *************************************************************
> Don Murray                               UCAR Unidata Program
> dmurray@xxxxxxxxxxxxxxxx                        P.O. Box 3000
> (303) 497-8628                              Boulder, CO 80307
> http://www.unidata.ucar.edu/staff/donm
> *************************************************************
>
>