4 replies to this topic

[David Medeiros]

Terrain Sculptor continues to confound me. I have a GeoTIFF of Panama that I converted in ArcGIS to an ESRI ascii grid (using the [raster to ascii] tool). The output is a .txt file with a .prj file. TS reads the file ok, then begins the filtering process but just as it ends I get the following message: java.lang.OutOfMemoryError: Java heap space. The file is about 175mb, I have 4gb of ram installed.

I have also created ESRI GRID files from LandSerf and downloaded an SRTM tile directly to “ArcInfo ASCII” format (both end up with the .asc extension). Again, TS will read the file but just as the filtering ends i get the same message.

I have one smaller 18mb .txt elevation file that I created in Arc by using the right click > export raster to different format and simply changing the extension to .txt. This file will work in TS.

Any one here who is familiar with using these file formats and Terrain Sculptor have any idea what the issue might be?

thanks,
david

[David Medeiros]

Update: because the issue seems to be memory according to the error message (even though these are not huge datasets by normal; standards), I cropped the original data in LandSerf before again exporting to an ESRI GRID .grd format. This file was around 38mb and now opens and previews in TS just fine.

This suggest a major limitation to the Terrain Sculptor program. If it can't handle larger, regional data sets I'm not sure what use it is. Bummer. I'll have to dig some more.

Again, anyone who is successfully using this program feel free to chime in with your experiences.

[frax]

David - is that big GRID file with floating point data?

[David Medeiros]

David - is that big GRID file with floating point data?

Not sure, it's a single SRTM GeoTIFF tile converted to ESRI ASCII GRID format for Terrain Sculptor. Is floating point more compressed than integer?

Edited to add: I often work with up to 8 or 10 or more tiles in Natural Scene Designer with no problems.

[frax]

floating point data takes up much more space, and complicates processing. You likely don't need that precision. It should be easy to examine the files to see if it is fp or integer.