| Re: Gridding-Interpolation of satellite data [message #38120] |
Mon, 23 February 2004 12:43  |
mmccabe
Messages: 4
Registered: May 2003
|
Junior Member |
|
|
Ben Tupper <btupper@bigelow.org> wrote in message news:<c1d1ad$1e5jka$1@ID-189398.news.uni-berlin.de>...
> I think you want to look into the FAULT_POLYGONS and FAULT_XY keywords.
> You should be getting rid of the missing data, before using
> GRIDDATA, by using the GRID_INPUT routine with the EXCLUDE keyword set
> to the indices of you NODATA values.
>
> Ben
Thanks guys (although MISSING wasn't my only problem). GRID_INPUT
certainly assisted, but the trick (as you suggested) was just to get
rid of the missing data completely before doing the triangulation,
using a simple search. I was able to get rid of most of the 'edge'
effects using combinations of SEARCH_ELLIPSE, SECTOR and
EMPTY_SECTORS. The result wasn't too bad. I wanted to affect the data
as little as possible - so limiting the amount of interpolation was a
priority.
Thanks to Dan Bergman for a funky little bit of code from way back in
1994!!! that I found in a posting. This is an intuitive (in that I can
actually read and understand what it is doing) approach to
interpolation - allowed me to customise for my own particular needs.
Cheers,
Matt
|
|
|
|
| Re: Gridding-Interpolation of satellite data [message #38128 is a reply to message #38120] |
Mon, 23 February 2004 06:05  |
btt
Messages: 345
Registered: December 2000
|
Senior Member |
|
|
Matt McCabe wrote:
> I always get nervous posting to this site for fear of asking a
> particularly stupid question....well, here goes.
>
> I'm using some satellite data and need to regrid the standard EASE
> projection (around 25km resolution) onto a regular grid of my own
> design - a 1/8th degree grid over the US. Previous postings have been
> very helpful in assisting me in this.
>
> I realise GRIDDATA and TRIANGULATE would do this no
> problems...however, I have gaps in the global/continental data from
> the overpasses which are causing me issues. The satellite scans leave
> no-data values (assigned 8888) in between overpasses. However, I don't
> see that GRIDDATA allows me to ignore these areas without affecting
> the interpolated values around the 'edges' (areas between data and
> no-data) i.e. values near the boundary of these, depending on the
> interpolation routine employed, will be affected by the no-data
> values. A post-interpolation mask, while getting rid of most of the
> affected data - will still leave numerous edge effects significantly
> different to the 'true' value.
>
> I thought I might be able to set no-data values to !VALUES.F_NAN - but
> GRIDDATA and TRIANGULATE don't handle this....basically I just want to
> interpolate within those areas that have data and exclude from the
> interpolation values I don't want so I don't significanly influence
> the 'real' data. My aim is not so much to smooth or massage the raw
> data - just regularly grid it.
>
> Have I missed something completely fundamental or is it actually a bit
> tricky.
>
> Thanks in advance,
> Matt
Hi,
I think you want to look into the FAULT_POLYGONS and FAULT_XY keywords.
You should be getting rid of the missing data, before using
GRIDDATA, by using the GRID_INPUT routine with the EXCLUDE keyword set
to the indices of you NODATA values.
Ben
|
|
|
|
|
|
| Re: Gridding-Interpolation of satellite data [message #38183 is a reply to message #38120] |
Tue, 24 February 2004 20:26 |
eoraptor
Messages: 5
Registered: January 2004
|
Junior Member |
|
|
>
> Thanks to Dan Bergman for a funky little bit of code from way back in
> 1994!!! that I found in a posting. This is an intuitive (in that I can
> actually read and understand what it is doing) approach to
> interpolation - allowed me to customise for my own particular needs.
>
> Cheers,
> Matt
Here's a link to see the interpolation code mentioned above.
http://tinyurl.com/2js92
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
