I'm working on a least-cost path project for part of my history dissertation, but I'm uncertain about one stage of the data preparation and I thought I might be able to find some thoughts on the issue here. I'm essentially doing what Rosenswig and Martínez Tuñón are doing in their June 2020 Journal of Archaeological Science article: creating a cost surface from elevation and hydrology data and then using Tobler's Hiking Function to come up least-cost paths and contour lines that tell me approximately how long it took to move across a certain distance on foot. My specific workflow is very close to 11.2B of the Berkeley Geospatial Archaeology course. I've run a few tests on a small area and so far things work out just fine when I'm using either a DEM or a slope generated from my DEM as the input surface raster in the path distance tool. The results are what I expect them to be.
So, two questions:
1. DEM vs. slope? The results are very similar in my 75km test zone, but the analysis I'm performing will extend out to about 800 kilometres, so even small variations will add up. Is one of these methods more acceptable?
2. How do I go about creating my input surface raster? I'd like to add a layer of hydrology. I've created weighted cost rasters in the past, but Rosenswig and Martínez Tuñón had a different solution: give the water bodies slope values based on how easy they think it is to traverse. I tried reclassifying my hydrology rasters and then combining them with the slope from the DEM, but path distance does not like the resulting raster, presumably because when I used the combine tool, the results were not in degrees. Could anyone point me towards what I should be doing to make this work?
My preference would be to use slope. Just because it would make more sense to tell the optimal path analysis "don't go here if slope >5" instead of "100 meters above sea level". If I used DEM instead of slope, the optimal path tool may arbitrarily avoid places with a specific elevation even though the slope isn't really that bad. But if you are getting pretty much the same answer along the same paths, it may not be such a big deal in your case. I would just test it every time you do it.
How is your hydrology data set up? Thinking off the top of my head, you should be able to assign costs to your hydrology data like your slope data. Meaning if current is >5 assign a more costly value. But that really depends on what data you have for your rivers. I honestly have not tried inputting hydrology into an optimal path analysis, so it may be that the programming just doesn't like it and so that's why inputting data like it is slope is recommended.