editing hxn file to remove artificial land bridges?

[Ron]

So, I'm running into an interesting problem where the large hexagons I am using in CAT (~1km wide for fast pilot analyses of a broad study area) are sufficiently blurring the coastline of states like North Carolina, such that artificial landbridges are created. These artifacts allow current/shortest paths to arc across water in places where they really shouldn't flow, bypassing the longer route that most animals would have to take to get from one peninsula to the next. This is happening even when the base conductance values are set to noData for the major water bodies (e.g. estuaries) at the original scale of the conductance/suitability surface. What I think is happening is that hexagons that intersect the shoreline are all getting a non-zero/non-noData value for conductance, and so when two such hexagons are across from each other the water gets bridged. So the minimum width of a noData water barrier is something like just over twice the diameter of the hexagons used in CAT, or otherwise it can get smoothed into a land bridge.

If this seems like an accurate assessment, is there an easy fix to prevent this problem or remove the land bridges? It is easy of course to edit the hexagon shapefile that CAT produces, but that doesn't change the .hxn file that CAT uses for the graph creation, right? Or should I just bank on the fact that at higher resolution the potential for land bridges will disappear, at least until you get to waterbodies narrow enough to be swimmable by most terrestrial vertebrates?

Thanks!

Ron

 

[carlos]

Ron - As a short-term fix, could you set in ArcGIS the value of the
background area (water) to have high resistance/low conductance?
I believe that currently the CAT hxn generation process will ignore
nodata in calculating a hex's value so as you say an area that is half
water and half high quality habitat might appear as a potential
linkage area.
Additionally, asc import in the next release of the CAT may be
somewhat improved, but I'm not sure if any change that affects this
issue will be implemented.
It is possible to edit hxn values using Hexsim (www.hexsim.net), but
that is likely more laborious than you want to attempt.