Flow routing

How should we route flow through Arctic hillslopes, and what digital elevation models are best?

I've been routing flow using a d-infinity algorithm across 2m ArcticDEM topography. I noticed that in a lot of places the channel networks were actually distributary; depressions were filling up with water and then routing multidirectional flow downslope. This makes sense and probably happens in permafrost landscapes, but it's a thorn in my side since channel delineation relies on progressively higher flow accumulation downslope, so it's hard for me to define flowpaths.

A new problem: recently I got a snippet of lidar (collected with NCALM) and it gives a really different answer than ArcticDEM! Out of curiosity I (1) coarsened the lidar to 2 m and (2) routed flow through a Wiener-filtered ArcticDEM landscape via LSDTopoTools.

I run the following code for each DEM from pysheds

RichDEM gives me even weirder answers!

In case you wanted to know, the difference between ArcticDEM and lidar can be seen in the DEM of difference below. Unfortunately I can't figure out how to add a spatial reference to an xarray DataArray so you can't have a nice zoom box like the other figures. Also, colorbar is in meters.

Maybe the smoothness of the ArcticDEM product, which is a blend of their "best" surveys, allows more diffuse flow? Whereas even in the downscaled DEM lidar has crisper microtopography and therefore channels flow in fewer directions?