I'm processing a LiDAR survey along the Connecticut River. The LiDAR spacing is around 7 feet. The survey follows the river but is only 1-6 miles wide.
The hydrologists want to use it for flood inundation mapping, but need the DEM raster expanded to the entire hydrologic basin, thirty or more miles wide.
Conn. has a state-wide LiDAR survey available at ~28 foot resolution, including and surrounding my area. I've got ASCII XYZ sets for both surveys, broken into tiles.
I've read Clayton Crawford's excellent series on LiDAR Solutions in ArcGIS. Also the Basic Terrain Modeling outline from Harvard. I still have a few questions.
Q: how best to embed the hi-res points in the low-res?
Q: The final product should be a DEM raster, so they can generate flow-direction grids.
What is the highest raster resolution I can push out of this data? Can I get a 10-foot DEM that is detailed along the river and blurrier elsewhere?
Q: Should the processing go {points, 2-D breaklines} -> {Terrain} -> DEM?
Is TIN involved at any point?
Q: Most of the river and lake linework don't have elev or have a field ELEVATION=0.
Can/should elev be set? Some provide river channels, others just centerlines. How should SFType be set in terrain builder?
Q: Should the low-res background and the hi-res river area be processed separately,
then merged as terrains or rasters? Or just dump all the points into one big mixed-res multipoint file and process that from start to finish?
Q: The high-res is too dense to process all at once and presumably there are
similar limits on the low res. How best to process tiles and match and merge? This is for hydrology, so a 6-foot difference between adjoining tiles amounts to a dam (or a waterfall).
Have you compared the positional and elevation relationship between the 2 datasets? You could try exporting both datasets to points then replace the points in the low resolution dataset with points from the high resolution one and run TopoToRaster to create a hydrologically correct raster.
You should not have to need to use TIN unless you would like to manually correct the surface. (I find the TIN editing allow me to visually correct issues and then i can convert the TIN to Raster or to nodes then interpolate surface from nodes)
I like your idea about raster to point then use the topo to raster tool in order to merge two separate point elevation layers into one elevation raster.
If we created an elevation raster with 10 sparse / precise points and 500,000 dense / imprecise points, wouldn't we end up with small peaks or depressions where the 10 points occur? We are improving the accuracy at 10 locations but we still have many points that are imprecise.
