Pathdistance problem

[JohnKantner]

Hello:

I'm trying to use Pathdistance and a custom vertical factor table to create a cumulative cost-surface that represents travel time away from the source point. It works just fine with a 30-m DEM and/or a small extent, but if I try a large 10-m DEM, I get "holes" and "ridges" in the resulting cost-surface in which the accumulated cost begins to decrease as you move away from the source -- and then starts to increase again. See attached image for an example (blue point is source; note decreasing accumulated cost to west).

I've tried all kinds of trouble-shooting -- rebuilding the DEM, reprojecting it, using source from a raster, using source from a shapefile -- all to no avail. I suspect there must be computational limitations in Pathdistance??

All thoughts on this welcome!

[JohnKantner]

Thanks for your interest in my question. I am implementing Tobler's hiking function using the vertical factor table created by Nico Tripcevich, something I first did in the mid-1990s in Arc/Info using a 30-m DEM (output attached). I wanted to update that earlier analysis using a 10-m DEM and with the full anisotropic implementation that I couldn't figure out how to do 15 years ago. I'm attaching an image of the base DEM, as well as a larger image of the whole 10-m cost-surface showing the "holes"/"ridges" I refer to -- if I try to generate cost-paths away from the cyan dot to, e.g., a point on the north end of my cost-surface, ArcGIS creates short cost-paths from the center of each "hole" to the nearby ridges (no image available at the moment), and does not connect my source with destinations; i.e., there's no way "around" the ridges.

[JohnKantner]

Sorry for not providing more detail earlier. Pathdist.pdf is a large region of New Mexico where I do my work as an archaeologist. All of the other maps are smaller sections from the southeastern corner of the larger map, and the dots are archaeological sites. The problem is when I try to implement Tobler's hiking function in Pathdistance using the vertical factor table that Nico Tripcevich generated from Tobler's formula (see <his page here>. It works just fine with a 30m DEM, but at at 10m, I get the results seen in pathdist.pdf. The Pathdistance implementation is pretty basic, using a DEM as input with no additional friction surface beyond the vertical factor table.

Your suggestion to try a simpler landscape is a good one, and I'll try that next. Thanks for the advice...