Local Kriging Neighbourhood Optimisation (LKNO) is an alternative approach to optimising estimation parameters. The purpose of LKNO is to ensure that each block is estimated with the best possible combination of parameters presented. Block estimates are independent of one another and as such there is no reason to apply the same estimation parameters to each and every block in a resource model. The LKNO approach is to run the Kriging estimate using a range of parameter options and select the estimate that produces the highest Kriging Efficiency, Slope of Regression or a combination of both on a block by block basis.
The LKNO process is outlined in “I’d like to be OK with MIK, UC?” by Jacqui Coombes and is summarised below:
The first step is to identify a suitable range of settings for each of the search parameters which include:
- Minimum and maximum number of samples
- Search range for each of the three directions
- Maximum number of samples taken from a drillhole
- Maximum samples per Octant, if Octants are to be used
Each parameter set is assigned a reference number (1,2,3 …). The kriging process is run using each parameter set and if the Kriging Efficiency for a particular block is greater than the previous one then the estimated grade, Kriging Efficiency, parameter set number and other model variables are updated accordingly.
Every block in the resulting grade model will be based on the best combination of search parameters. The beauty of LKNO is that it eliminates the need for tedious analysis of “best parameter selection” per domain, and instead directly provides an estimate for each block based on what is appropriate for that block.
Viewing the parameter set variable as a 3D model colour-coded on parameter set informs which parameters sets have been used for each block in the resource model as shown below.
Of course, there are some underlying assumptions:
- The variogram model is assumed to be representative across the blocks being estimated. Since the LKNO process relies on the Kriging Efficiency, which is an expression of the quality of the estimate based on the pattern of continuity defined in the variogram model, the LKNO process is conditional to how well the variogram model reflects the underlying grade continuity.
- As always, there is an underlying assumption of data integrity.
LKNO will be available in Snowden’s Supervisor software (v8.11).
Click here to find out more or download a free trial.