ArcGIS Geostatistical Analyst

ArcGIS Geostatistical Analyst is an extension to ArcGIS Desktop that provides a powerful suite of tools for spatial data exploration and surface generation. It effectively bridges the gap between geostatistics and GIS analysis by enabling you to model spatial phenomena, assess risk, and accurately predict values within your study area.

With ArcGIS Geostatistical Analyst, you can create surfaces from data measurements taken over areas where collecting information for every location would be impossible or cost prohibitive. You can fully examine sample data, evaluate uncertainties, generate unique insights, and create customized interpolation surfaces for more informed decision making.

With ArcGIS Geostatistical Analyst, you can

• Visualize, model, and predict spatial relationships.
• Link data, graphs, and maps dynamically.
• Perform deterministic and geostatistical interpolation.
• Evaluate models and predictions probabilistically to assess risks.

ArcGIS Geostatistical Analyst helps you cost-effectively probe real-world issues in

• Atmospheric data analysis
• Petroleum and mining exploration
• Environmental analysis
• Precision agriculture
• Fish and wildlife studies

With online support, an included tutorial, and an interactive wizard to guide you through the interpolation process, it's easy to add geostatistical analysis to your GIS capabilities.

Explore your data using the following exploratory spatial data analysis (ESDA) tools:

• Histogram and summary statistics
• Normal quantile–quantile plot
• Trend analysis
• Semivariogram/Covariance cloud and map
• Voronoi map
• General quantile–quantile plot
• Cross-covariance cloud and map

Model

Interpolation

Create surfaces from sample data using these interpolation methods:

• Inverse distance weighted
• Radial-based functions, which include the following kernels
  • Thin plate spline
  • Spline with tension
  • Multiquadratic
  • Inverse multiquadratic
  • Completely regularized spline kernels
• Global and local polynomials
• Kriging for exact data and for error-contaminated data
  • Ordinary, for data with unknown constant mean value
  • Simple, for data with known mean value
  • Universal, for data with mean value as a function on coordinates
  • Indicator, for discrete data or data transformed to discrete
  • Probability, for discrete data as primary variable and continuous data as secondary variables
  • Disjunctive, for nonlinear predictions
• Cokriging (multivariate version of the above-mentioned kriging models)
• Isotropical or anisotropical models

Kriging Output Surface Types

• Prediction
• Prediction standard error (measure of the prediction quality)
• Probability map (probability that specified threshold value is exceeded)
• Error of indicators (measure of the probability map uncertainty)
• Quantile map (over- and underpredicted values)

Modeling Tools for Kriging

• Data transformations
  • Box–Cox
  • Logarithmic
  • Arcsine
  • Normal score
• Data detrending
  • Global polynomial
  • Local polynomial

• Variography
  • Models (four can be used simultaneously)
  • Nugget
  • Circular
  • Spherical
  • Tetraspherical
  • Pentaspherical
  • Exponential
  • Gaussian
  • Rational quadratic
  • Hole effect
  • K-Bessel
  • J-Bessel
  • Stable
  • Semivariogram/Covariance surface
  • Anisotropy
  • Specifying or estimating the proportion of measurement error in the nugget
  • Cross-covariance option for shift between variables
  • Estimation of all or part of the model parameters by a modified weighted least squares algorithm
• Declustering
  • Cell
  • Polygonal
• Checking for data bivariate distribution

Searching Neighborhood

To select neighboring data to predict the value for the target point

• Ellipse with one, four, or eight angular sectors
• Minimum and maximum number of points in each sector

Evaluate

Evaluate how your models perform using the following diagnostics:

• Cross-validation for checking the model's quality
• Validation for checking prediction quality
• Compare cross-validation results of several models
• Show predicted value at cursor (MapTips)

Simulate

Interpolation methods produce one value for every location. In reality, there are many, equally probable values that could occur at each unsampled location (the true value for each unsampled location is unknown). Geostatistical simulation produces multiple surfaces that mimic the real phenomenon and provide these possible values, giving a basis for risk analyses, economic decision making, and other estimations involving uncertainty, allowing analysts to make more informed decisions.

• Conditional Simulation
• Unconditional Simulation
• Postprocessing (clipping, summary statistics for cells and for polygonal areas of interest)

Refine

Renderers

• Contours (isolines)
• Filled contours
• Regular grid (All models allow data averaging in each cell; block interpolation.)
• Hillshading
• Combination of several options above

Export Result of Predictions to

• Contour lines
• Polygons
• Raster
• Specified point locations
• Geostatistical layer that stores the model parameters for the renderers used