How does work z tolerance in "raster to TIN" tool ?

MangeantBaptiste · ‎07-03-2013

Hi,
i'm using ArcGIS 10.1, and i study the influence of the z tolerance of the "raster to TIN" tool. I proceed in this way :
1) i build a TIN from an initial 1-meter raster grid, using the "raster to TIN" tool with a z tolerance of 1 (the number of nodes of this TIN do not exceed the maximum number of point...)
2) i convert this TIN into a raster with the "TIN to raster" tool, using a "cellsize" of 1.
3) i calculate the residual between my initial raster and the new raster obtained from the TIN (itself obtained from the initial raster grid) using the "raster calculator" tool.

When i visualize the results of this residual raster, i realize that some cells exceed 1 meter of residual value (basically around 1,20 meters, but some cells raise to 4 meters). How can it be possible, with the z tolerance of 1 meter ? I read this help of ArcGIS which describe the z tolerance method in the "raster to TIN" tool : "http://resources.arcgis.com/en/help/main/10.1/index.html#//00q90000008r000000"

Thank you in advance for your help.

MattMead · ‎07-05-2013

I'll let someone else get technical on Z tolerance as it was always explained to me that the Z tolerance was the max difference allowed during the processing.

As for the potential cause for a greater that 1m difference, it is likely due to your resultant raster not having the same extents as your original. When you convert your raster to a TIN I believe it takes the centerpoint of each cell and uses that location as the "known" elevation for use in triangulating the nodes. When you take the TIN and convert it back to a raster it may not line those centerpoints back up exactly causing your elevation data to appear to shift.

Technically your data will still be within your Z tolerance of 1, its just that the new raster is using an elevation that is based on a location interpolated between your previous data points. Check out "snap raster" in the help files. This forces your DEM/raster files to use a specific grid starting point.

MangeantBaptiste · ‎07-07-2013

Thank you Matt for your explanation. I'm a little bit more confortable to know that the z tolerance works well. However, the trick with "snap raster" does not provide me residuals lower than 1... If anybody has an idea on the subject, please tell me.
I am now quite sure that the problem is effectively a shift between my two rasters to compare, but i still don't know how to handle with it.
Thank you for your attention.