Dear Paul,
I have found your PhD thesis excerpt very interesting: thanks for sharing it.
Date: Tue, 12 Feb 2008 21:31:44 +0000 (GMT)
From: Paul Kelly <paul-grass@stjohnspoint.co.uk>
Subject: Re: [GRASS-user] strange behaviour of v.surf.idw
To: "G. Allegri" <giohappy@gmail.com>
Cc: grassuser <grass-user@lists.osgeo.org>
Also, have you tried running using the -n flag (no indexing) and comparing the results. I think they should be the same for all regions then but might turn up something else interesting.
I started using v.surf.idw (for irregularly spaced points) in grass6.0 with no -n flag, then in 6.1 I noticed that the output raster changed from what it used to be (i.e., some stair-like artifacts appeared) and it was restored to the original look (i.e., a much smoother look) by adding the -n flag. By looking at your PhD thesis excerpt, it looks like the -n flag should only make the interpolation faster: I wonder if you have any idea why there may be differences, as you also say above.
Kind regards,
Luigi
I used -n flag. The surface is now more realistics and appears allineated to the narrower IDWs results: http://www.geospatial.it/allegri/IDW5.png .
I can’t access your PhD thesis…
Giovanni
2008/2/13, Luigi Ponti <lponti@infinito.it>:
Dear Paul,
I have found your PhD thesis excerpt very interesting: thanks for
sharing it.
Date: Tue, 12 Feb 2008 21:31:44 +0000 (GMT)
From: Paul Kelly <paul-grass@stjohnspoint.co.uk>
Subject: Re: [GRASS-user] strange behaviour of v.surf.idw
To: “G. Allegri” <giohappy@gmail.com>
Cc: grassuser <grass-user@lists.osgeo.org>
Also, have you tried running using the -n flag (no indexing) and comparing
the results. I think they should be the same for all regions then but
might turn up something else interesting.
I started using v.surf.idw (for irregularly spaced points) in grass6.0
with no -n flag, then in 6.1 I noticed that the output raster changed
from what it used to be (i.e., some stair-like artifacts appeared) and
it was restored to the original look (i.e., a much smoother look) by
adding the -n flag. By looking at your PhD thesis excerpt, it looks like
the -n flag should only make the interpolation faster: I wonder if you
have any idea why there may be differences, as you also say above.
Kind regards,
Luigi
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user
Hello Luigi
On Wed, 13 Feb 2008, Luigi Ponti wrote:
Paul Kelly wrote:
On Wed, 13 Feb 2008, Luigi Ponti wrote:
I started using v.surf.idw (for irregularly spaced points) in grass6.0 with no -n flag, then in 6.1 I noticed that the output raster changed from what it used to be (i.e., some stair-like artifacts appeared) and it was restored to the original look (i.e., a much smoother look) by adding the -n flag. By looking at your PhD thesis excerpt, it looks like the -n flag should only make the interpolation faster
It might do, for a small number of points. It certainly uses less memory - that is clear. But for a large number (I'm talking many hundreds of thousands) the overhead of implementing the search radius method is well outweighed by the time saving in not having to search through hundreds of thousands of points for every cell in the output raster.
How many points are typically in the maps you used it with? I really can't see any changes between 6.0 and 6.1 that would have affected the output - it's quite intriguing - you can look at the WebCVS yourself: http://freegis.org/cgi-bin/viewcvs.cgi/grass6/vector/v.surf.idw/main.c and see if you can see anything.
Paul
True: the code did not change much from 6.0 to 6.1. Maybe I just happened to find a peculiar combination of data that revealed the possible output difference with and without the -n flag (I use v.surf.idw in a script to routinely map output of ecological models). Typically, the input vector has 108 input points and the raster output is 1207 rows by 1350 columns. Is this considered a small number of points?
Yes, I would consider that very small. As you can see in the thesis excerpt, the size of dataset I had in mind when I implemented the new algorithm was hundreds of thousands of points.
Did you avoid replying also to the list on purpose (this is just for me to understand mailing list etiquette -- I have limited experience)?
Oops - for some reason I thought you had replied to me off-list, and so I sent my reply off-list as well. Sorry for the confusion. Back on-list now.
Paul
Paul Kelly wrote:
Hello Luigi
Yes, I would consider that very small. As you can see in the thesis excerpt, the size of dataset I had in mind when I implemented the new algorithm was hundreds of thousands of points.
Paul
OK, I got it -- input points are very few: the algorithm was not designed for that and there is little performance gain. Thanks for the explanation.
Luigi
