Re: viewing large data sets

[Steve Hankin <hankin@xxxxxxxxxxxxx>]

Hi Mark,

The solution that should be used is to tell Ferret to subsample the data --
reducing the number of points to a value that is in keeping with the resolution
of the output window.  Typically this means, say 300x200 points or so. Ferret
can do this very straightforwardly with a command like

    SHADE my_var[i=lo:hi:del, j=lo:hi:del]

replacing the standard "SHADE my_var" in the plotting scripts.

Jon Callahan has already developed scripts that compute and apply the
subsampling in LAS. Unfortunately, Jon is away until late next week. Can you
wait that long?

When it comes to making fast images from satellite data Ferret is not the
optimal application.  Ferret uses scalable, geometric graphics, rather than
raster graphics, which makes it slower on this type of plot. If your users
require full reslution raster graphics you may want to consider using a
different back-end application for those data sets. You've probably seen
several emails in the recent past about the use of multiple back-end
applications in LAS.

A final note: If your data set is globe-encircling and you want to offer the
"modulo" capability of generating contiguous images that cross the branch point
of the data storage with Ferret (e.g. a plot spanning the International
dateline on a data set that is on a -180 to 180 axis), then you will need the
latest beta version of Ferret. Please get in touch with me directly if that is
the case.

    - steve

========================

Mark Rehbein wrote:

> Hi LAS people,
>
> I've just finished setting up our LAS server to allow users to view our
> satellite imagery.  It is currently behind our firewall so it is not
> accessible to the world as yet.
>
> Each image is 1600x1200 pixels for each day in time (1km resolution).
>
> A problem occurs when a large lat/lon area is chosen.  LAS reports an
> error from ferret:
>
> dding an acceptable error string: "*** NOTE: ".
>         NOAA/PMEL TMAP
>         Program FERRET
>         Version 5.22 - 07/27/00
>         15-Jun-01 11:46
>
> yes? cancel mode verify
> yes? cancel mode interp
> yes? canc data/all
> yes? go std_initialize
> "/u02/risdev/ferret/data/rs_sst.nc" "1" "1" "sst_ave"
> yes? set
> region/x="142.0":"155.0"/y="-27.0":"-7.0"/k=1/t="18-Apr-2001":"18-Apr-2001"
> yes? set win/size=.5
> yes? set win/asp=2
> yes? GO "/tmp/lasgo22799352674.jnl"
>  **ERROR: insufficient memory: 2603301 words were requested.
> CONTOUR/FILL/levels=30/set sst_ave
> Command file, command group, or REPEAT execution aborted
>
> This is the same/similar error when you plot too many data points while
> using ferret directly even after doing the set memory/size=xxx command.
>
> I would like to ask if anybody has overcome this problem as the last thing
> I want is for users to get freaked out simply because they chose such a
> big area.
>
> A solution might be to automatically reduce the resolution (ie pick every
> 4th pixel) when an arbitary large number of data points are chosen.  Not
> sure how to do this though using LAS.
>
> Any help/suggestions would be appreciated.
>
> cheers
>
> mark
>
> Mark Rehbein
> Intranet Developer
> Australian Institute of Marine Science

--

                |  NOAA/PMEL               |  ph. (206) 526-6080
Steve Hankin    |  7600 Sand Point Way NE  |  FAX (206) 526-6744
                |  Seattle, WA 98115-0070  |  hankin@pmel.noaa.gov