Generating a Viewshed with LiDAR data 'as obstructions'

BillB

Hi folks.  I have an area in the UK which is covered by downloadable LiDAR data, 50cm accuracy.  The data sets are DTM (ground only) and DSM (ground, trees, buidlings etc).
I tried generating a Viewshed using the DSM, which was quite efficient, but the problem was that it regarded the trees and buildings as part of the base terrain to be assessed, and therefore showed parts of the viewshed as occuring on tops of trees and buildings.  Technically, that's correct - if there was anyone up there - but this needs to be a ground-level assessment, with the trees and buildings as obstructions, not receptors.
So I tried subtracting the DSM from the DTM (Analysis > Combine terrain > Subtraction (Difference) Signed) to obtain a new 'combined terrain' representing the difference between the DTM and DSM.  I reasoned that if I could then generate vector objects from the resulting terrain, these could be used as obstructions which sit 'upon' the DTM in the Viewshed process.
I tried two main options:
1) from the 'combined terrain', generate contours and export to SHP, with 'Generate side areas for extruded3d areas (Buildings)' checked. Re-import SHP file.  That works quite well, apart from the contours being rather funky triangles.  In addition, when I view the model in the 3d window, the tops of the DTM plus contour SHP vectors is lower than the DSM.
2) Layer > Create Area features from Equal Values. Use new 'combined terrain'.  Allowed elevation fuzziness: 0.5m
This also works reasonably well, although everything it creates looks like it is made from lego - based on the grid of the original LiDAR (0.5m) and the height intervals determined by the elevation fuzziness.  However, the elevation grid area+DTM aslo look lower than the DSM in the 3d window (due to redraw times in the 3d window, I Save Image to JPGs, so I can flip between them for comparison).
So, the upshot is that I do not appear to have a reliable method to obtain vectors from the DSM to act as obstructions in the viewshed dilaog  - or at least, not to the correct height as referenced to the DSM.  Can anyone suggest a more accurate workflow to achieve this?
PS, I don't have the LiDAR module!
Thanks
Bill

macsmith

did you check the "Obstruction Area Features Always Hidden" option, down the bottom right of the viewshed options?

I've used straight DSMs from UK Lidar without the Lidar module successfully..

Ian

BillB

Ian, thanks for your suggestion.  The problem is that when using a DSM to generate the viewshed, the obstructions, in the form of trees and buildings, are an integral part of the terrain model that GM is using.  They are not separate vector areas, so "Obstruction Area Features Always Hidden" does not apply, as they are not area features. Hope that makes sense.