Label  Explanation  Data Type 
Input Inclusion Probability Raster  Defines the inclusion probabilities for each location in the area of interest. The location values must range from 0 (low inclusion probability) to 1 (high inclusion probability).  Raster Layer; Mosaic Layer 
Number of Output Points  The number of sample locations that will be created.  Long 
Output Point Feature Class  The output feature class containing the selected sample locations and their inclusion probabilities.  Feature Class 
Summary
Creates a set of sample points based on inclusion probabilities, resulting in a spatially balanced sample design. This tool is typically used for designing a monitoring network by suggesting locations to take samples, and a preference for particular locations can be defined using an inclusion probability raster.
Usage
The input probability raster must contain only values between 0 and 1. The higher the value, the more likely the cell will be included in the sample design.
All values in the study area should have inclusion probabilities >= 0, while all areas outside the study area should have Null values.
The cell size of the inclusion probability raster determines the finest resolution at which samples will be generated. In other words, the points the tool creates will always be located at the centers of the raster cells. Using a smaller cell size for the inclusion probability raster will result in more possible locations for the points to be created.
When point, line, or polygon features are converted to raster (to obtain the input probability raster), the following should be considered:
 The cell size (resolution) should be fine enough to distinguish all the important features in the population. To accomplish this, the cell size can be set to less than half the minimum distance between features. This distance can be calculated with the Generate Near Table tool.
 For line and polygon features, the cell size should be set so features (such as meandering streams) are adequately represented in the resulting raster. For example, you may not be able to represent a complex river with a large raster cell size; curves in the river may be smoothed over if the cell size is too large.
 The precision with which sample locations can be located in the field should also be considered. For example, if locations are to be found using a GPS with a positional accuracy of 10 meters, the cell size should be 10 meters.
 Be mindful of the size of the inclusion probability raster, because as the number of cells increases, the processing time also increases.
To avoid outputs that appear spatially unbalanced, it is recommended that the number of sample locations be less than 1 percent of the number of cells in the inclusion probability raster.

In the Random number generator environment, only the Mersenne Twister option is supported. If other options are chosen, Mersenne twister will be used instead.
Parameters
arcpy.ga.CreateSpatiallyBalancedPoints(in_probability_raster, number_output_points, out_feature_class)
Name  Explanation  Data Type 
in_probability_raster  Defines the inclusion probabilities for each location in the area of interest. The location values must range from 0 (low inclusion probability) to 1 (high inclusion probability).  Raster Layer; Mosaic Layer 
number_output_points  The number of sample locations that will be created.  Long 
out_feature_class  The output feature class containing the selected sample locations and their inclusion probabilities.  Feature Class 
Code sample
Create a set of spatially balanced points based on an input inclusion probability raster.
import arcpy
arcpy.env.workspace = "C:/gapyexamples/data"
arcpy.CreateSpatiallyBalancedPoints_ga("ca_prob", "10", "C:/gapyexamples/output/csbp")
Create a set of spatially balanced points based on an input inclusion probability raster.
# Name: CreateSpatiallyBalancedPoints_Example_02.py
# Description: This tool generates a set of sample points based on inclusion
# probabilities. The resulting sample design is spatially balanced, meaning
# that the spatial independence between samples is maximized, making the
# design more efficient than sampling the study area at random.
# Requirements: Geostatistical Analyst Extension
# Import system modules
import arcpy
# Set environment settings
arcpy.env.workspace = "C:/gapyexamples/data"
# Set local variables
inProb = "ca_prob"
numberPoints = 10
outPoints = "C:/gapyexamples/output/csbp"
# Execute CreateSpatiallyBalancedPoints
arcpy.CreateSpatiallyBalancedPoints_ga(inProb, numberPoints, outPoints)
Environments
Licensing information
 Basic: Requires Geostatistical Analyst
 Standard: Requires Geostatistical Analyst
 Advanced: Yes