Conditional simulation is an advanced interpolation technique whereby Z estimates are based on a form of stochastic simulation in which measured data values are honored at their locations. Other interpolation methods, including kriging and IDW, will smooth out local details of spatial variation, especially as interpolated locations become more distant from measured locations. This can be a problem when you are trying to map sharp spatial discontinuities such as contamination hotspots or fault lines. GS+ uses a sequential gaussian simulation method. Note that output is sensitive to the number of simulations used to produce the interpolated values.
The Simulate tab is part of the larger Interpolation Window:
Variogram Model
Variogram models for isotropic and anisotropic variograms are chosen with the Model command in the Semivariance Analysis window. Here in the Simulate window you can specify whether to use the isotropic or anisotropic model for the variogram used in the kriging system.
Output
Two types of output are available with conditional simulation: either a) the estimated Z value and its standard deviation, or b) the probability that the estimated value for that location is greater than some threshold value t.
• An individual Z estimate and its standard deviation is the mean and standard deviation of n simulations for a specific location. Thus the number of simulations will affect both of these values.
• Probability is the proportion of simulations for a specific location for which the estimated value is greater than t. Thus this value will also be affected by the number of simulations.
Analysis tab
• Number of simulations. You may choose any number of simulations >0. Keep in mind that choosing too few simulations will produce a map that is very rough, although a large number of simulations will require a lot of cpu time. See below for examples. The default number of simulations is 1000.
• Use different seed. If this box is checked each simulation will use a different random number seed. This will slow the analysis somewhat and make each simulation unique to a specific run. The default is to use the same random number seed. In either case a different random path is used for each simulation.
• Multigrid refinements. Checking this box forces the analysis to follow a stepwise procedure when simulating interpolation nodes. In the first step a coarse grid is used to allow the influence of large-scale variogram structure; in subsequent steps the search neighborhoods are smaller. This avoids the need for extensive conditioning. The default is 3 refinements.
Secondary data tab
Collocated secondary data can be used to refine estimations. Collocated means that a secondary data value is available for every grid node to be estimated (not simply for every primary data value).
• Residuals. This is the same as simple kriging with a locally varying mean. Press the Define button to define the residuals to be used.
• External drift. Press the Define button to define the drift values.
• Collocated cokriging. Press the Define button to define the covariate values.