You might glean something from a few of the papers I have On Researchgate. The 2015 paper with Andy Maceyka (GIS, LIDAR Specialist) explained the need to burn in channels, and also add the auxiliary channels in braided systems as the hydrologic models typically only place one channel. When the DEM is created, typically depression filling is done, but as suggested, this can create some flow path issues that we found if something diverts the channel to the floodplain, the flow path stayed linear on the floodplain until diverted back. Some tributary channel sections also became linear for some sections for no apparent reason except probably depression filling creating ultra flat or smooth gradient sections. We found that the LiDAR level of detail enabled much improved georeferencing of channels in dense forest, but with still some need to hand digitize and burn in channels when large logs, beaver dams, old roads or other features divert the hydrologic model from the channel flow path. Before we had LiDAR, we had issues with flow modeling when the pixels were 30m and 10m, and the hydrological model from pixel to pixel had to move in one of 8 directions. Another researcher thought this flow path could be improved by allowing the channel to move in any direction, but I think we felt that digitizing or burning in the channel produced better results. The 2007 paper with Dr Alan James also brought out benefits from LiDAR finding some stream branching issues present in past contour topographic map DEMs that lack enough detail to properly locate some channel branching properly.

For bridges, roads, they need to be removed or channel burned in to pass.

disclaimer: i have not done this before..

However, as you mention, in reality such raised transportation avenues have culverts to allow water to pass underneath. If you know where the culverts are, you could artifically edit the pixel elevation on the road (on a saved copy of the DEM), making the pixel lower than the surrounding ground to create small preferential drainage paths. Now if you are working on a regional scale, then you can try and locate some information as to culvert whereabouts, or design considerations of culvert citing under highways and roads in the area.

Dear Laura,

Before any hydrological modeling can be applied to a DEM, it needs to be adjusted in order to be hydrologically correct. You may find useful below links:

https://onlinelibrary.wiley.com/doi/pdf/10.1002/hyp.11385

https://hess.copernicus.org/articles/17/495/2013/hess-17-495-2013.pdf

Cheers

Ankur