I am planning to work with a LiDAR dataset for Western NC (working on just a subset now to learn) to complete a riparian forest analysis and I'm trying to do this at a very fine-scale resolution. Using LiDAR to get characteristics of the vegetation is something I am familiar with, but this issue with riparian vegetation is knowing where the actual streambanks are. All of the existing datasets for stream centerline (polyline) and stream area (polygons) are pretty inaccurate, sometimes over 100 meters off from where the stream really is. This is most likely due to anthropogenic alterations to the network, but nonetheless, these datasets don't produce an accurate assessment of riparian vegetation. So, I'm trying to use LiDAR data from recent years to determine where the streambanks are. This is easy enough on a small scale, but I'm at a roadblock trying to do this at a landscape scale. I think the solution may be some kind of machine learning, but I'm very unfamiliar with how that would work.
So far, I have tried using a fine-scale DEM to recreate the stream network, but this still has some pretty evident inaccuracies and only identifies stream centerline instead of streambanks. I am aware of tools like the River Bathymetry Tools, but I haven't been able to find anything in that to determine the location of streambanks at a landscape scale.
Any advice out there?
