3D Analyst - Get TIN or Terrain points of varying resolutions

TimHoward1 · ‎05-25-2010

All -
I'm using a method that creates then exports TIN edges and nodes to apply a suite of graph analysis routines in the python package NetworkX. My problem is that my analysis area has grown so much as to make it cumbersome for the level of resolution I'd like to have the TIN. I realized I only need this higher TIN resolution in some parts of the study area, not others. Enter the Terrain dataset, or so I thought. I still can't seem to export nodes and edges at different locations in the study area with different (specified) z-tolerance. I'm using python scripts in Arc 9.3 and can't expect an upgrade to 10 for a few more months.

My question: Given a TIN or Terrain, and two areas of interest: A and B. These areas of interest do not overlap but might have coincident boundaries. How do I extract the nodes and edges from area A at one z-tolerance and the nodes and edges from B at a different z-tolerance?

I got a terrain built and like the application of the multiple pyramid levels, but it looks like I can only export a complete TIN (or a portion defined by a rectangle - not a polygon). It looks like "Terrain to Points", a new addition for Arc 10 may help, but still won't get me all the way there.

Looking forward to any thoughts or suggestions.
Tim Howard

JakubSisak · ‎06-02-2010

I would use TIN not a Terrain dataset.

Create 2 TINS - 1 in full resolution, 2nd generalized.
Convert both to nodes.
Delete all nodes from the 2nd "generalized" layer where you need high resolution data.
Select nodes by location from 1st "high resolution" copy and paste into the 2nd "generalized" layer
You will now have a layer containing low resolution nodes and high resolution nodes in some areas.
Create a new TIN from this layer.
If you need to add a high resolution edge to this dataset i would create a 3D outline using the high res TIN then add this to the final "Combined" TIN as a breakline.

TimHoward1 · ‎06-03-2010

Thanks for your reply. I've settled on another, similar way that seems to get what I want. The original TIN is based on a continuous raster. I keep the 'high-res' portions of the raster as-is, but run a roving window analysis (calculating a mean value) for the remainder of the raster. These portions are in effect smoothed by the calculations of mean value. I then convert this raster to TIN.

Best,
Tim