Kriging Neighbourhood Analysis (KNA) in K-MINE Optimizing Parameters for Accurate Geological Estimation

What KNA Does in K-MINE

Kriging Neighbourhood Analysis (KNA) in K-MINE is a targeted workflow that tunes key estimation parameters to minimize conditional bias and improve the accuracy of block model grades. By systematically testing discretization, block size, and the number and configuration of samples, KNA helps engineers select settings that deliver stable, defensible estimates for resource evaluation.

Prerequisites: Statistics, Anisotropy, and Variograms

Before running KNA, users complete a preliminary statistical review, map anisotropy, and build directional variogram models along the three axes of the search ellipsoid. These inputs define the spatial continuity that kriging will honor during estimation.

Discretization: Setting Virtual Blocks or Using an Existing Model

During discretization testing, users can choose an existing block model from the current project or specify virtual block dimensions along X, Y, and Z. Each block is uniformly subdivided into discretization points; the density of these points is varied to assess its effect on estimation stability and bias.

Optimization Metrics: Conditional Bias Measures

K-MINE evaluates candidate settings using conditional bias indicators—primarily the slope of regression (trend toward under/overestimation) and kriging efficiency (information gain vs. simple averaging). These metrics guide users to an optimal discretization density that balances accuracy with computational efficiency.

Block Size Optimization

After discretization is set, users test multiple block-size scenarios. K-MINE compares outcomes to identify the block dimensions that maintain representativity of geology and orebody geometry while preserving estimation precision.

Sample Search Strategy and Negative Weights Control

Next, K-MINE analyzes the number of samples, their search distances, and neighborhood configuration. Users can constrain or filter negative weights to prevent instability in estimates. Adjusting these parameters directly influences conditional bias measures and helps achieve a well-behaved slope of regression and robust kriging efficiency.

Supported Kriging Methods

K-MINE supports Simple Kriging (SK), Ordinary Kriging (OK), Kriging with External Drift (KED), Cokriging, and other variants. Regardless of method, KNA determines how many samples to include, how far to search in each direction of the anisotropy ellipsoid, and how to weight neighbors appropriately.

Final Outcome: Defensible, Realistic Estimates

The KNA process delivers a set of optimized estimation parameters—discretization density, block size, sample count, search radii, and weighting controls—that reduce conditional bias and produce geologically realistic block model grades ready for reporting and design.