Note: this vignette is far from complete and only contains a brief discussion (originally from an email) of the topic in the context of rasters. It will be rewritten and expanded in future releases.
Imagine you have a raster with 2 islands, and that water separating the islands in the raster is represented with
NA’s. These two islands would be disconnected and will result in
"Warning: Input contains disconnected regions. This does not work with the cond_passage() metric.". The warning isn’t an issue if you don’t need the
Now, each island (aka region) must have at least one pixel with a non-zero absorbing value. The math algorithms don’t work without that. If a region does not have at least one non-zero absorbing value, then the
samc() function will result in
Error: All disconnected regions must have at least one non-zero absorption value
Now the tricky part; an island/region can be as small as a single pixel. So, if you’re not careful with your raster, you might have very small, isolated patches as small as a pixel floating around somewhere. If you’re using just 4 directions for the transition function, they could even be immediately diagonal to another patch.
The easiest way to spot these is to pick one of your rasters and set all your non-
NA values to the same value, like 1. Then turn all the
NA’s to a different value, like 0. Then plot it. Code:
!is.na(raster)) <- 1 raster[is.na(raster)) <- 0 raster[plot(raster)
If your raster is particularly large, then you might need to save it out to a high-res image to spot solo pixels