At least one polygon or raster layer must be loaded into ArcMap in order to use this
Preventing points from occurring in NoData cells.
The greater the proportion of NoData cells in a raster, the longer this tool will
take to run. This program generated a true random point within the extent of the raster
and THEN tests to see if occurs in a NoData cell. If there are many NoData cells
this will clearly result in the rejection of a high proportion of random points,
and more time will be needed to meet the quota the user defines. Warning: if your
grid is 100% NoData then this program will enter an infinite loop (to force the
shutdown of ArcMap, use Task Manager to end the application).
Enforcing a minimum distance between generated points
This option is available with all raster and polygon options (including the polygon statified random sampling options), but the implications of using this option can be complex depending on what other options you have switched on. Improperly using this option could result in a biased random sample.
The minimum distance value you supply must be in coordinate system units (e.g. meters for UTM, feet for SPC, etc). This option will simply prevent random points from being placed within this distance of another random point. In the case of stratified random samples, users have the option to enforce the rule across all points (e.g. including across adjacent polygon boundaries), or enforcing the rule only within each strata (i.e. each polygon). In the latter case, it is possible for two random points to be closer than the minimum distance you specify provided they occur in different strata (e.g. adjacent polygons).
There are a number of ways this option can create problems: (1) it may be impossible to generate a certain number of points if the minimum distance you specify is too large relative to the area of the raster/polygon(s), (2) this option can counteract or conflict with the weighted probability distribution raster option. The tool will generate as many points as it is able and then stop if, after 100,000 attempts, it has been unable to generate all the random points under the conditions you have specified. Therefore if you use the minimum distance option it is imperative that you take great care to check that the program has been successful by examining the output.
Note that you can use the Distance Between Points (Within Layer) tool in this toolset to check that the minimum distance has been enforced.
Using a raster layer as a weighted probability distribution.
If you check the box "Use raster as a weighted probability distribution", then the values in the raster are used as probabilities of placement such that raster cells with larger values are more likely to have a point placed in them. The probability for each cell is calculated by: (value of cell) / (maximum value). As such, the raster can represent true probabilities (range: 0-1), or can contain any range of numbers that represents the relative probability of occurance between cells. Note that you can use the other options with this tool, including: Prevent points from being located in NoData cells, and Prevent points from occuring in the polygons of the following layer.
Preventing points from occurring in the polygons of another
The greater the area in which you prevent points from being placed, the longer
the program will take to run. If the polygon layer specified to prevent points
from being located in certain areas completely covers the reference polygon layer,
then this program will enter an infinite loop.