I have square locations (see image). For my SDM, I need some measure of isolation of a square (related to the other squares or a particular specified set of squares). The most simple one would be to take the distance to the closest square, but that's not the best, because it might be like 20 other squares at that distance, or it might be just that single square and then vast area of nothing around... Taking just like average distance of 5 closest squares seems kind of hoky poky... Isn't there a better and statistically more robust measure of isolation?
The purpose of this measure: I want to see the distribution of isolation of the squares which need to be predicted with the SDM and compare it to the distribution of the isolation of the clustered cross-validation folds (in the picture), to see if the characteristics of isolation are similar.