Raster to Weighted Points (Spatial Analyst Tools)

Summary

Generates representative, value-proportional points from continuous surfaces, preserving spatial patterns and total magnitudes.

Learn more about how Raster to Weighted Points works

Illustration

Usage

Several parameters are available that control the total number of output points and how those points will be distributed across the input surface. The Maximum Number of Points (max_number_of_points in Python) parameter defines the maximum number of output points. The Limit One Point per Cell (single_point_per_cell in Python) parameter restricts allocation to at most one point per cell. The Point Distribution Method (point_distribution_method in Python) parameter determines how multiple points are positioned within cells.
The Maximum Number of Points parameter specifies the total number of points that will be generated from the entire raster. The number of points allocated to each cell is proportional to the cell values. Cells with larger values contribute more points and cells with smaller values contribute fewer points. This allows the output to preserve the spatial pattern and relative magnitude of the input surface.
Use the Limit one point per cell parameter to prevent more than one point from being assigned to the same cell. This is useful when a uniform spatial distribution is required. Using this parameter may result in lesser number of points than the maximum number of points specified.

When the parameter is unchecked (single_point_per_cell = 'DO_NOT_LIMIT' in Python), multiple points maybe assigned to a single cell if the cell's value justifies additional representation.

If checked (single_point_per_cell = 'LIMIT' in Python), the output point will be placed at the center of the cell.
The distribution pattern of output points is determined by the Point Distribution Method parameter. Placement can occur at the cell center, randomly within the cell, or in an even arrangement when multiple points are present. Random placement is useful for simulation workflows, uncertainty assessment, and generating anonymized representations of sensitive surfaces. Deterministic placement methods produce consistent point locations across repeated runs and are appropriate for reproducible spatial analysis.
- The Circular method places multiple points in a radial configuration around the cell center. The largest distance from the cell center should be less than half of cell radius for accurate placement.
- The Equal Area Voronoi method subdivides the cell into regions of equal area and places one point per region to maximize spatial separation.
- The Random method places points randomly within the cell footprint.
- The Fibonacci Lattice method distributes multiple points in an isotropic pattern suitable for uniform sampling.
- The Fibonacci Spiral method arranges points along a deterministic spiral that produces even spacing.
If the Create multipoint output parameter is checked (create_multipoint_output = `MULTIPART`` in Python), all points generated from a single cell are stored as a multipoint feature geometry. This option reduces the number of output records and is useful for compact geometry storage. The output multipoint features can be used in downstream workflows that treat each cell as a single logical unit.
If the Normalization Method parameter value is set to None, no normalization will be applied and the original values from the input are used directly for point allocation. This is appropriate for surfaces where the magnitude of values already carries semantic meaning, such as kernel density estimates or physically measured environmental surfaces.
The available normalization methods are:
- Min–Max—The values will be scaled linearly between 0 and 1, preserving relative differences.
- Log—Large numeric ranges will be compressed to reduce the influence of extreme high values while enhancing distinctions among low values.
- Sigmoid—A logistic transform will be applied to emphasize midrange differences and compress extremes.
- Softmax—The raster will be converted into a probability distribution where each cell's normalized value reflects its relative magnitude compared to all other cells.
- Sum-to-1—All values will be scaled so that their sum equals exactly 1, enabling probability-based sampling workflows.
- Z-Score—The values will be standardized by subtracting the mean and dividing by the standard deviation.
- None—The original raster values will be used without modification.
If the Normalize values before point generation parameter is checked (normalize_before_point_generation = `NORMALIZE`` in Python), the raster values are transformed before allocation using the selected normalization method. Normalization adjusts the numeric scale and distribution of the raster while maintaining its spatial structure. This is useful when raster values span wide numeric ranges, contain extreme outliers, or exhibit skew that could otherwise dominate the allocation. When normalization is not applied, the raw raster values are used directly in the point allocation process.
All transformation, normalization, allocation, and point distribution operations occur in memory.
Points are not created for NoData cells from the input surface.
The output feature class will include attributes for the original cell value, a normalized value (if a Normalization Method is used), and a distributed value when multiple points originate from the same cell.
See Analysis environments and Spatial Analyst for additional details on the geoprocessing environments that apply to this tool.

Label	Explanation	Data type
Input Surface Raster	The continuous raster for which proportional points will be generated. Each input cell value represents the relative magnitude or intensity to be converted into one or more points.	Raster Layer
Output Point Features	The output feature class that will contain the generated points.	Feature Layer
Maximum Number of Points	The maximum number of points to generate from the input raster. Use this parameter to limit the output density of a large raster.	Double
Limit one point per cell (Optional)	Specifies whether only one point will be generated per raster cell. Checked—A maximum of one point will be created for each raster cell, regardless of its value. Unchecked—Multiple points may be generated within a cell, proportional to the cell's value. This is the default.	Boolean
Create multipoint output (Optional)	Specifies whether points with the same value will be preserved in the output or be aggregated. An aggregated output reduces feature count and improves performance when generating large numbers of points. Checked—Points that would share the same attribute value will be aggregated into a single geometry record in the output feature dataset. Unchecked—Every selected point will contain its own distinct feature record in the output multipoint dataset. This is the default.	Boolean
Point Distribution Method (Optional)	Defines how points will be spatially distributed within each raster cell. By default, a point is placed at the geometric center of each cell. This parameter is unavailable when the Limit one point per cell parameter is checked. Circular—The points will be distributed randomly within a circular area centered on each cell. This is the default. Equal Area Voronoi—The points will be positioned to ensure equal area representation across the raster surface, minimizing clustering. Random—The points will be placed randomly within each cell according to the proportional weight of the cell's value. Fibonacci Lattice—The points will be distributed in a nearly uniform pattern within each cell, creating an even and natural-looking arrangement without visible grid patterns. Fibonacci Spiral—The points will be placed in a smooth spiral pattern that creates a gradual, outward flow from the center of each cell, producing an even and visually continuous distribution.	String
Distance from Cell Center (Optional)	The maximum distance, in map units, that points may be offset from the cell center when a non-center placement method is used. This parameter is only available if the Point Distribution Method parameter value is set to the Circular or Equal Area Voronoi options. The maximum value allowed is 0.45. If 0 is used, all the output points will placed in cell center.	Double
Normalization Method (Optional)	Specifies how raster values will be scaled or transformed before proportional point generation. Scaling methods only change the values assigned to output points, while transformation methods also affect how many points are generated in each cell by changing the relative weighting of cell values. Min–Max—The values will be normalized linearly between 0 and 1. This preserves relative differences and leaves the number of points per cell unchanged. Log—A logarithmic transformation will be applied that compresses large ranges and reduces the effect of extremely high values. This produces fewer points in peak areas and more balanced coverage elsewhere. Sigmoid—Normalizes the values by standard z-score and then applies a logistic sigmoid transformation to convert those scores into a probability weight between 0 and 1. Softmax—The values will be normalized by applying an exponential scaling and converting all values into probability distribution. This amplifies high-intensity regions while ensuring sampling reflects relative importance. Sum-to-1—All values will be scaled so that their total equals 1. This produces point values that represent probabilities or proportions without changing point density patterns. Z-Score—Tne values will be standardized by their mean and standard deviation (mean = 0, SD = 1). This increases the contrast between areas above and below average and shifting point density toward cells that differ most from the mean. None—No normalization will be applied; original raster values are used. This is the default.	String
Normalize before point generation (Optional)	Specifies whether the input surface raster will be transformed using the chosen normalization method before point allocation. This parameter is only available if the Normalization Method parameter value is set to the Min–Max, Log, Sigmoid, or Softmax option. Checked—Points will be normalized before point allocation. Unchecked—Normalization will not influence the point allocation. This is the default.	Boolean

RasterToWeightedPoints(in_surface_raster, out_point_features, max_number_of_points, {single_point_per_cell}, {create_multipoint_output}, {point_distribution_method}, {distance_from_cell_center}, {normalization_method}, {normalize_before_point_generation})

Name	Explanation	Data type
in_surface_raster	The continuous raster for which proportional points will be generated. Each input cell value represents the relative magnitude or intensity to be converted into one or more points.	Raster Layer
out_point_features	The output feature class that will contain the generated points.	Feature Layer
max_number_of_points	The maximum number of points to generate from the input raster. Use this parameter to limit the output density of a large raster.	Double
single_point_per_cell (Optional)	Specifies whether only one point will be generated per raster cell. `LIMIT`—A maximum of one point will be created for each raster cell, regardless of its value. `DO_NOT_LIMIT`—Multiple points may be generated within a cell, proportional to the cell's value. This is the default.	Boolean
create_multipoint_output (Optional)	Specifies whether points with the same value will be preserved in the output or be aggregated. An aggregated output reduces feature count and improves performance when generating large numbers of points. `MULTIPART`—Points that would share the same attribute value will be aggregated into a single geometry record in the output feature dataset. `SINGLE_PART`—Every selected point will contain its own distinct feature record in the output multipoint dataset. This is the default.	Boolean
point_distribution_method (Optional)	Defines how points will be spatially distributed within each raster cell. By default, a point is placed at the geometric center of each cell. This parameter is not supported when the `single_point_per_cell` parameter is set to `LIMIT`. `CIRCLE`—The points will be distributed randomly within a circular area centered on each cell. This is the default. `EQUAL_AREA_VORONOI`—The points will be positioned to ensure equal area representation across the raster surface, minimizing clustering. `RANDOM`—The points will be placed randomly within each cell according to the proportional weight of the cell's value. `FIBONACCI_LATTICE`—The points will be distributed in a nearly uniform pattern within each cell, creating an even and natural-looking arrangement without visible grid patterns. `FIBONACCI_SPIRAL`—The points will be placed in a smooth spiral pattern that creates a gradual, outward flow from the center of each cell, producing an even and visually continuous distribution.	String
distance_from_cell_center (Optional)	The maximum distance, in map units, that points may be offset from the cell center when a non-center placement method is used. This parameter is only valid if the `point_distribution_method` parameter value is set to the `CIRCLE` or `EQUAL_AREA_VORONOI` options. The maximum value allowed is 0.45. If 0 is used, all the output points will placed in cell center.	Double
normalization_method (Optional)	Specifies how raster values will be scaled or transformed before proportional point generation. Scaling methods only change the values assigned to output points, while transformation methods also affect how many points are generated in each cell by changing the relative weighting of cell values. `MIN_MAX`—The values will be normalized linearly between 0 and 1. This preserves relative differences and leaves the number of points per cell unchanged. `LOG`—A logarithmic transformation will be applied that compresses large ranges and reduces the effect of extremely high values. This produces fewer points in peak areas and more balanced coverage elsewhere. `SIGMOID`—Normalizes the values by standard z-score and then applies a logistic sigmoid transformation to convert those scores into a probability weight between 0 and 1. `SOFTMAX`—The values will be normalized by applying an exponential scaling and converting all values into probability distribution. This amplifies high-intensity regions while ensuring sampling reflects relative importance. `SUM_TO_ONE`—All values will be scaled so that their total equals 1. This produces point values that represent probabilities or proportions without changing point density patterns. `ZSCORE`—Tne values will be standardized by their mean and standard deviation (mean = 0, SD = 1). This increases the contrast between areas above and below average and shifting point density toward cells that differ most from the mean. `NONE`—No normalization will be applied; original raster values are used. This is the default.	String
normalize_before_point_generation (Optional)	Specifies whether the input surface raster will be transformed using the chosen normalization method before point allocation. This parameter is only valid if the `normalization_method` parameter value is set to the `MIN_MAX`, `LOG`, `SIGMOID`, or `SOFTMAX` option. `NORMALIZE`—Points will be normalized before point allocation. `DO_NOT_NORMALIZE`—Normalization will not influence the point allocation. This is the default.	Boolean

Code sample

RasterToWeightedPoints example 1 (Python window)

This example demonstrates using this tool in the Python window.

import arcpy
from arcpy import env
from arcpy.sa import *

env.workspace = "C:/sapyexamples/data"
OutPoints = RasterToWeightedPoints("CrimeDensity.tif", 5000,
                                   "DO_NOT_LIMIT","RANDOM", "NONE",
                                   "NORMALIZE", 0.25, "SINGLE_PART")
OutPoints.save("out_WeightedPoints_ex1.shp")

RasterToWeightedPoints example 2 (stand-alone script)

This example demonstrates using this tool in a stand-alone script.

## Name: RasterToWeightedPoints_Ex_standalone.py
## Description: Generates value-proportional points from a raster,
##               preserving spatial patterns and total magnitudes.
## Requirements: Spatial Analyst extension

## Import system modules
import arcpy
from arcpy import env
from arcpy.sa import *

## Set environment settings
env.workspace = "C:/sapyexamples/data"

# To allow overwriting outputs change overwriteOutput option to True.
env.overwriteOutput = False

## Set local variables
in_surface_raster="CrimeDensity.tif",
max_number_of_points=5000,
single_point_per_cell="DO_NOT_LIMIT",
point_distribution_method="EQUAL_AREA_VORONOI"
normalization_method="SIGMOID",
normalize_before_point_generation="NORMALIZE",
distance_from_cell_center=0.30,
create_multipoint_output="MULTIPART"

out_WeightedPoints = RasterToWeightedPoints(in_surface_raster,
                                            max_number_of_points,
                                            single_point_per_cell,
                                            point_distribution_method,
                                            normalization_method,
                                            normalize_before_point_generation,
                                            distance_from_cell_center,
                                            create_multipoint_output)
# Save the output
out_WeightedPoints.save ("C:/sapyexamples/output/out_WeightedPoints_ex2.shp")

Environments

Current Workspace, Extent, Geographic Transformations, Random number generator, Output Coordinate System, Scratch Workspace

Licensing information

Basic: Requires Spatial Analyst
Standard: Requires Spatial Analyst
Advanced: Requires Spatial Analyst