2d estimation

2d estimation performs parameter estimation using kriging and other methods to map 2D analytical data onto surface grids defined by the limits of the data set as rectilinear or convex hull extents of the input data.

Its Adaptive Griddingfurther subdivides individual elements to place a "kriged" node at the location of each input data sample. This guarantees that the output will accurately reflect the input at all measured locations (i.e. the maximum in the output will be the maximum of the input).

The DrillGuide functionality produces a new input data file with a synthetic boring at the location of maximum uncertainty calculated from the previous kriging estimates, which can then be rerun to find the next area of highest uncertainty. The naming of the "DrillGuide©" file which is created when 2d estimation is run with all types of analyte (e.g. chemistry) files ends in apdv1, apdv2, apdv3, etc. the output file name will be .apdv2, apdv3, apdv4.... There are no limits to the number of cycles that may be run.

The use of 2d estimation to perform analytically guided site assessment is covered in detail in Workbook 2: DrillGuide© Analytically Guided Site Assessment.

This process can be continued as many times as desired to define the number and placement of additional borings that are needed to reduce the maximum uncertainty in the modeled domain to a user specified level. The features of 2d estimation make it particularly useful for optimizing the benefits obtained from environmental sampling or ore drilling programs. 2d estimation also provides some special data processing options that are unique to it, which allow it to extract 2-dimensional data sets from input data files that contain three-dimensional data. This functionality allows it to use the same .apdv files as all of the other EVS input and kriging modules, and allows detailed analyses of property characteristics along 2-dimensional planes through the data set. 2d estimation also provides the user with options to magnify or distort the resulting grid by the kriged value of the property at each grid node. 2d estimation also allows the user to automatically clamp the data distribution to a specified level along a boundary that can be offset from the convex hull of the data domain by a user defined amount.

Module Input Ports

• Input External Grid [Field / minor] Allows the user to import a previously created grid. All data will be kriged to this grid.
• Input External Data [Field / minor]   Allows the user to import a field contain data. This data will be kriged to the grid instead of using file data.
• Filename [String / minor] Allows the sharing of file names between similar modules.

Module Output Ports

• Output Field [Field] Outputs a 3D data field which can be input to any of the Subsetting and Processing modules which have the same color port
• Filename [String / minor] Allows the sharing of file names between similar modules.
• Status Information [String / minor] Outputs a string containing module parameters. This is useful for connection to write evs field to document the settings used to create a grid.
• Surface [Renderable] Outputs the kriged surface to the viewer

Properties and Parameters

The Properties window is arranged in the following groups of parameters:

• Grid Settings: control the grid type, position and resolution
• Data Processing: controls clipping, processing (Log) and clamping of input data and kriged outputs.
• Time Settings: controls how the module deals with time domain data
• Krig Settings: control the estimation methods
• Data To Export: specify which data is included in the output
• Display Settings: applies to maximum uncertainty sphere
• Drill Guide: parameters association with DrillGuide computations for analytically guided site assessment

 

Variogram Options: 

There are three variogram options:

1. Spherical: Our default and recommended choice for most applications
2. Exponential: Generally gives similar results to Spherical and may be superior for some datasets
3. Gaussian: Notoriously unstable, but can "smooth" your data with an appropriate nugget.

I specifically want to discuss the pros and cons of Gaussian. Without a nugget term, Gaussian is generally unusable. When using Autofit, our expert system will apply a modest nugget (~1% of sill) to maintain stability. If you're committed to experimenting with Gaussian, it is recommended that you experiment with the nugget term after EVS computes the Range and Sill. Below are some things to look for:

• If you find that Gaussian kriging is overshooting the plume in various directions, your nugget is likely too small.
• However, if the plume looks overly smooth and is too far from honoring your data, your nugget is likely too big.