SRTM for South America

[Guest vlada stoje]

Hi Holger,the weekend is finally here; I have the time for the hobby and trying read the article (on http://www.elsevier.com/locate/isprsjprs ). Skipping the long complicated math formulas (definitely not for me), grateful for your simile to understand the principles. Still didn

[Guest]

I think one way the holes in the 3 arc second SRTM data could be fixed is by merging LOD9 and LOD5 dataResampling the SRTM at LOD5 seems to remove most holes.So if that LOD5 could be used to "plug up" the holes in the LOD9 data then it would be much better.My question is this :-How can I use RESAMPLE.EXE to make a single .TMF which merges LOD5 and LOD9 data? Is it possible? The PDF manual seems to indicate that it is possible (they talk about the Mt Rainer example).Any ideas?Matthew.

[Guest]

Uploaded v1.02 of SRTM_3ARCSEC_TO_BGL.It now interpolates to fix the missing SRTM data.So no more SRTM meshes with

[rhumbaflappy 113]

Hi Matthew.Your tools work fine for me. Thanks!Whitewater Lake, Wisconsin DEFAULT:http://forums.avsim.com/user_files/11853.jpgWhitewater Lake, Wisconsin Improved with Ground2K and SRTM data:http://forums.avsim.com/user_files/11854.jpgDick

[Guest vlada stoje]

http://www.volny.cz/stoje/scen/aconc01.jpgnice nice! The interpolator works funny but better then nothing, see the Aconcagua (made quickly with the version 1.02 from the SRTM tiles s33w071 and s33w070)Cheers

[Guest]

Thanks... I'll work on the 1ARC version today sometime.Matthew.

[Guest]

Looks good.Actually the interpolator just derives the missing data from the average of the surrounding data values.Do you have any suggestions as to a better method of interpolation? How does MicroDEM do it... any ideas?Matthew

[gaugi1 0]

For your information, I have compared three interpolation methods to correct a big hole in SRTM file.http://forums.avsim.com/user_files/11897.jpg

[Guest]

Thanks... I obviously need to work on the interpolation. I'll use your tile as a reference.Matthew.

[Guest vlada stoje]

Hi Matthew,as Holger Sandmann mentioned sooner, in the MicroDEM there is used the method Inverse Distance to power. This method is described in the Help for the Surfer ( http://www.goldensoftware.com ) as the most simple and quick to calculate. The value in the point is the weighted average of the valid points in fixed surrounding, where the weight depends on the distance, it is proportional to the inverse distance to a power, and the progress of the weight can be controlled by the value of the power (2 is default in Surfer). The sum of all weights must be one. This method creates the typical unreliable

[Holger 0]

Hi everyone:Giles, nice job on the comparison! It really shows that interpolation doesn't really work with big holes, other than to fill them with "some" data.In fact, I don't think that any interpolation method will bring about significant improvements over merging with low-res meshes - though I'd love to be proved wrong and I do appreciate Matthew's utility. Even the more complex methods, like the ones Vlada mentioned, require some data to work with. As far as I know, kriging and the like are all built around the premise of sparse/irregular data being available. They may fill a large hole in a raster grid with elevations that look similar to the surrounding area (and even include some valley-ridge systems of the same height and distance) but the results won't have any bearing of what, in reality, is situated in the hole. On the other hand, I suppose the issue here is not really whether the result is fully accurate but whether it looks reasonable in FS. One possible (slight) improvement to the GTOPO30 merging method is to use a slightly larger scale factor (in bigbsq.exe) with the low-res terrain. I've had pretty good results with 1.05 or 1.1; it adjusts for the fact that GTOPO30 data are usually lower in elevation (due to the 1-km interpolation), which leads to seams (cliffs) along the edges of the filled holes.Cheers, HolgerP.S. Matthew, have you any images to show the results of using LOD5 resamples for filling? That sounds like an interesting idea.About your question - I never had much luck with multisource/LOD compiling with the .inf file; I do all those things in MicroDEM.