Kriging
Release 9.3 Last modified September 7, 2011	Print all topics in : "Interpolation (Spatial Analyst)"

Interpolates a surface from points using kriging.

Learn more about how Kriging works

Usage tips

Command line and Scripting

• Kriging is a processor-intensive process. The speed of execution is dependent on the number of points in the input dataset and the size of the search window.

• The Semivariogram Properties dialog box has several models to choose from. When the Kriging method is set to Ordinary, the available models are Spherical, Circular, Exponential, Gaussian, and Linear. When the Kriging method is set to Universal, the available models are Linear with linear drift and Linear with quadratic drift.

• The Universal Kriging types (Linear with linear drift and Linear with quadratic drift) assume that there is a structural component present and that the local trend varies from one location to another.

• The advanced parameters allow control of the semivariogram used for kriging. A default value for lag size is initially set to the default output cell size. For Major range, Partial sill, and Nugget, a default value will be calculated internally if nothing is specified.

• The optional output variance of prediction raster contains the kriging variance at each output raster cell. Assuming the kriging errors are normally distributed, there is a 95.5 percent probability that the actual z-value at the cell is the predicted raster value, plus or minus two times the square root of the value in the prediction raster.

• Some input datasets may have several points with the same X,Y coordinates. If the values of the points at the common location are the same, they are considered duplicates and have no affect on the output. If the values are different, they are considered 'coincident' points. Interpolation functions (other than IDW) may handle this data condition differently, in some cases using the first coincident point encountered for the calculation, or sometimes the last. This may cause some locations in the output raster to have different values than what you might expect. The solution is to prepare your data by removing these coincident points.

• Low values within the output variance of prediction raster indicate a high degree of confidence in the predicted value. High values may indicate a need for more data points.

• The following environment settings affect this tool:

  • General: Current Workspace, Scratch Workspace, Output Coordinate system, Output Extent, Snap Raster.
  • Raster Analysis: Cell size.
  • Raster Storage: See the Raster Storage settings for the particular output raster format selected.

Map Algebra

• Kriging is a compute-intensive process. The speed of execution is dependent on the number of points in the input dataset and the size of the search window. The GRAPH option is relatively quick and should be used before generating the final output surface to help determine the most appropriate options.

• GENERATE input files in the POINT format provide an easy method to create surfaces and two-dimensional coverages from ASCII files containing x,y coordinates and z-values. The x,y coordinates and z-values can be expressed as integers or single- or double-precision floating-point values. The x,y coordinates for the output raster are stored in DOUBLE precision. The z-values are stored in SINGLE precision. GENERATE files in the POINT format contain the feature ID, x,y coordinate, and z-values for point features. The ID values are not used by the kriging function. The format for GENERATE files containing points is:

       ID,X,Y,Z
       ...
       END

  The end of the file is indicated with the END keyword.A typical GENERATE file containing points is shown below. The file can be delimited by a comma or space.

       1,234.065,837.123,125.845
       2,735.774,729.345,93.739
       37,840.934,385.700,157.037
       5,385.888,286.602,25.764
       END

• The barriers option is used to specify the location of linear features known to interrupt the surface continuity. These features do not have z-values. Cliffs, faults, and embankments are typical examples of barriers. Barriers limit the selected set of the input sample points used to interpolate output z-values to only those samples on the same side of the barrier as the current processing cell. Input sample points that lie exactly on the barrier line will be included in the selected sample set for both sides of the barrier.

• The barriers options will significantly extend the processing time required to interpolate the output raster.

• Barrier features can be input as either a coverage or a GENERATE file in the LINE format. Kriging will look for the {barrier_cover} first, and if not found, it will look for the {barrier_file}. Kriging only uses the x,y coordinates for the linear feature; therefore, it is not necessary to provide z-values for the left and right sides of the barrier. Any z-values provided will be ignored. LINE input files contain the feature ID, followed by the x,y coordinates for each vertex in the file. Each line feature must have at least two, and no more than 20,000, vertices. The general LINE input file format is:

       ID
       X,Y,(Z)
       ...
       X,Y,(Z)
       END
       END

  The end of each line is indicated with the END keyword, and the end of the file is indicated with a second END keyword.The following is an example of a 3D GENERATE file in the LINE format. In this case, only the x,y coordinates are used by Kriging. The feature IDs and z-values are ignored.

       2
       345321.5  678543.1  123
       345879.0  689231.2  145
       345999.9  690579.3  157
       END
       2
       367809.0  712004.0  209
       368098.9  710776.8  214
       369890.6  709768.2  206
       370023.4  716098.3  198
       END
       2
       390879.0  602090.0  133
       391000.6  606668.2  169
       397547.3  609608.4  125
       END
       END

• The {output_variance} contains the kriging variance at each output raster cell. Assuming the kriging errors are normally distributed, there is a 95.5 percent probability that the actual z-value at the cell is the predicted raster value  ± two times the square root of the value in the {output_variance}, that is:

       <out_grid>  ± (2 * SQRT{output_variance})

• When using the <num_points> method, the resolution of the output raster and {output_variance} is determined by dividing the longest dimension of the extent of the input sample points by the specified number of sample points. The aspect ratio of the input point cover or file is automatically preserved by applying the mesh-point spacing derived along the longest side to determine the number of mesh points required on the shorter side. Therefore, unless the extent of the input sample points forms a square, the number of points in the x and y coordinates will be different. Low values within the {output_variance} raster indicate a high degree of confidence in the predicted value. High values may indicate a need for more data points.

• You can accept the input point cover or file XMIN,YMIN as the default origin and the XMAX,YMAX as the upper right corner of the extent to generate a raster covering the full extent of the input sample points. Alternatively, you can specify the out_grid origin, extent, and resolution or mesh-point spacing to create a raster covering only a portion of the input sample points.

• The semivariance values datafile is an INFO file with the specified output name followed by ".SVG". The semivariance values can be used to plot a semivariogram graph, which graphically illustrates how the actual semivariance and the predicted semivariance vary with the distance between the pairs of original sample points. Four items are written to the out_grid.SVG file. They are defined as follows:

      NAME            WIDTH    OUTPUT    TYPE    N.DEC
      METHOD          4         8        B       -
      TYPE            6         6        C       -
      DISTANCE        8        16        F       4
      SEMIVARIANCE    8        16        F       4

  • METHOD  — An integer code value specifying the type of mathematical function used to model the semivariance.
  The code values are:
  

      CODE    {method}
      1       SPHERICAL
      2       CIRCULAR
      3       EXPONENTIAL
      4       GAUSSIAN
      5       LINEAR

  
  
  • TYPE  — The type of data represented by the record.

  The semivariance INFO table contains two types of data: ACTUAL and FITTED. ACTUAL records contain the actual semivariance values. FITTED records contain the predicted semivariance values calculated by the specified modeling {method}.
  

  • DISTANCE  — The distance between pairs of input sample points at which the semivariance values have been measured.

  The actual and predicted semivariance is calculated at each interval of the mesh-point spacing. The maximum DISTANCE value is equal to twice the specified RADIUS {r} distance, or it is the maximum distance between pairs of input sample points.
  

  • SEMIVARIANCE  — The actual or predicted semivariance value at the current DISTANCE.

  Note that if a datafile with the same name already exists in the workspace, the new information will be appended to it; otherwise, a new file will be created.

• The graphing option is not available with the UNIVERSAL1 and UNIVERSAL2 kriging methods. Universal kriging assumes that there is a structural component present and that the local trend varies from one location to another. Plotting a semivariogram for each input sample point is not practical.

• The input feature dataset in the Raster Calculator can only be a point shapefile.

• Some input datasets may have several points with the same X,Y coordinates. If the values of the points at the common location are the same, they are considered duplicates and have no affect on the output. If the values are different, they are considered 'coincident' points. Interpolation functions may handle this data condition differently, in some cases using the first coincident point encountered for the calculation, or sometimes the last. This may cause some locations in the output raster to have different values than what you might expect. The solution is to prepare your data by removing these coincident points.

• Kriging does not respect the Mask setting.

• Learn more about how to specify the input feature dataset in the Map Algebra expression of Raster Calculator.

Command line syntax
An overview of the Command Line window
Kriging_sa <in_point_features> <z_field> <out_surface_raster> <SPHERICAL | CIRCULAR | EXPONENTIAL | GAUSSIAN | LINEAR | LINEARDRIFT | QUADRATICDRIFT | NONE> {Lag size} {Major range} {Partial sill} {Nugget} {cell_size} {search_radius} {out_variance_prediction_raster}

    "Spherical 0.4496 2.6185 542.65 0"

Command line example

Kriging_sa C:/data/pts.shp z_value C:/data/final_1 "Spherical 0.4496 2.6185 542.65 0" 1 "VARIABLE 12" ""

Scripting syntax
About getting started with writing geoprocessing scripts
Kriging_sa (in_point_features, z_field, out_surface_raster, semiVariogram_props, {Lag size} {Major range} {Partial sill} {Nugget}, cell_size, search_radius, out_variance_prediction_raster)

    "Spherical 0.4496 2.6185 542.65 0"

Script example

# Kriging_sample.py
# Description:
#   Interpolate a series of point features onto a rectangular raster using kriging.
# Requirements: None
# Author: ESRI
# Date: Sept 6, 2005
# Import system modules
import arcgisscripting

# Create the Geoprocessor object
gp = arcgisscripting.create()

try:
    # Set the input feature dataset
    inputFeatureDataset = "C:/data/pts.shp"

    # Set the output raster name
    outputRaster = "C:/data/final_1"

    # Don't write out the variance raster
    Output_variance_of_prediction_raster = ""

    # Set the output extent 
    gp.Extent = "0 0 6 6"

    # Define the semivariogram 
    Model = "Spherical "         #(Remember the space at the end)
    LagSize = "0.4496 "          #(Remember the space at the end)
    MajorRange = "2.6185 "       #(Remember the space at the end)
    PartialSill = "542.65 "      #(Remember the space at the end)
    Nugget = "0"
    Semivariogram = Model + LagSize + MajorRange + PartialSill + Nugget

    # Set the attribute field
    attributeName = "z_value"

    # Check out Spatial Analyst extension license
    gp.CheckOutExtension("Spatial")

    # Process: Kriging...
    gp.Kriging_sa(inputFeatureDataset, attributeName, outputRaster, Semivariogram, _
                  "1", "VARIABLE 12", Output_variance_of_prediction_raster)

except:
    # If an error occurred while running a tool, then print the messages
    print gp.GetMessages()

Map Algebra syntax
Kriging(<point_feature | point_file>, {spot_item}, {barrier_cover | barrier_file}, {BOTH | GRAPH | GRID}, {output_variance}, {method}, {SAMPLE, {num_points}, {max_radius}}, {cellsize}, {xmin, ymin, xmax, ymax})

Kriging(<point_feature | point_file>, {spot_item}, {barrier_cover | barrier_file}, {BOTH | GRAPH | GRID}, {output_variance}, {method}, {RADIUS, {radius}, {min_points}}, {cellsize}, {xmin, ymin, xmax, ymax})

Map Algebra example

kriging(cover)
kriging(file, #, #, graph, #, circular)
kriging(cover, item, #, #, out_var, linear, sample, 9)
kriging(cover, #, #, #, #, gaussian, #, #, #, 36)
kriging(file, #, #, grid, #, universal2, radius, 400)

ArcObjects syntax
See Krige and Variogram