Raster to TIN Conversion. Very erratic results depending on the Z tolerance

Question

I needed to convert a DEM to a TIN for use with HEC-GeoRAS and I was so surprised how sensitive this tool was to the Z-tolerance factor which is supposed to be an optional parameter. I first left it at the default value which was 18 feet. Needless to say the difference in point elevations between the DEM and the TIN were in the order of 2-3 feet and the results were off. I tried a Z-tolerance of 0.01, thinking that would reduce the difference but that made it worse. The elevation differences between the DEM and the TIN were now in the order of 6-8 feet! I consider myself lucky as when I tried a Z-tolerance of 0.25, the TIN elevations were pretty close (no more than 0.5 ft difference). All other values for the Z-tolerance: 0.1, 1, 0.01, 5, ... yielded a TIN that was way off the DEM.

So how could such a process be so contingent upon the proper selection of a Z-tolerance, yet this parameter is specified as optional?? How come there's very little guidance on how to select this parameter? And isn't it supposed to be the maximum allowable error?? So if I specify that the max Z-tolerance is 0.1, shouldn't I expect that the max difference in elevation at any given point between the DEM and the TIN is 0.1 feet?

Thanks for helping understand this process.

Joe Farah
Civil Engineer
SAIC

EricRice · Answer

So if I specify that the max Z-tolerance is 0.1, shouldn't I expect that the max difference in elevation at any given point between the DEM and the TIN is 0.1 feet?

Hi Joe,

The answer to that question is no.  What you should expect though, is that at the cell center the value of the input raster and output TIN will have similar values that are within the z tolerance. I used the default behavior (1/10 the z range of the DEM), which for me is over 300 ft and my TIN and raster values are nearly identical. For example, my TIN is 1477.044 and my DEM is 1477 (its an integer DEM).  Another area I spot checked had the TIN at 1545.104 and the DEM at 1545.

Raster To TIN first generates a  candidate TIN using sufficient input raster points (cell centers) to  fully cover the perimeter of the raster surface. It then incrementally  improves the TIN surface until it meets the specified Z tolerance. It  does so by adding more cell centers on an as-needed basis during an  iterative process.  It will not arbitrarily add more points to "improve" the surface given the input.

If you are using Identify to compare the raster and TIN, but not clicking on the cell center then you should not expect the values to be within the tolerance.  To check, you can go to the symbology tab of the TIN and click Add on the left side.  You want to add the renderer at the bottom called, "Nodes with the same symbol".  These nodes should align perfectly with the raster cell centers. Turn off the elevation renderer on the TIN so only the nodes are displaying. Overlay the nodes on top of the DEM and zoom in pretty far, until you begin to see individual pixels.  Using Identify, click on the point to compare the two surfaces.  Please let us know if you still see erratic  results.  You should also click in between nodes to see how the triangle faces are arriving at their interpolated values and note the values of the raster in between the nodes.

Best Regards,
Eric