Estimation

Overview

Use estimation to populate block model variables from composites. The estimation workflow is configured per variable, so one block model can contain different estimation methods for different variables.

Open the Block Models list with B or from the Set panel, select a block model, then open Estimation > Estimation Settings…

The Estimation for block model name dialog lets you:

• choose which variables will be estimated,
• assign an estimation mode to each variable,
• open the detailed setup for the selected mode,
• copy settings between variables,
• copy settings to other block models.

Available modes are:

• Estimation: standard estimation methods such as inverse distance and nearest neighbor.
• Kriging: kriging-based estimation.
• Simulation: conditional simulation.
• None: do not estimate that variable.

If you select several variables before opening the setup, the same configuration can be applied to all selected variables. For panel block models, only Kriging is available.

Estimation menu commands

The Estimation menu contains several related commands:

• Estimation Settings…: define or edit the estimation method for each variable.
• Copy Estimation To…: copy selected parts of an estimation setup to other block models.
• Clear Estimation…: remove existing estimation settings.
• Estimate: run the estimation for the selected block model or block models.
• Test One Block…: test the setup on a single block and review the detailed log.
• Estimate Distances…: write the distance to the nearest selected object into a destination variable.
• Simulation Variable Settings…: choose how simulated variables are displayed after simulation.

Estimation settings dialog

The variable list on the left shows the estimable variables in the selected block model. Use it to select one or more variables, then choose the estimation mode.

Main actions

• Set Estimation… opens the detailed setup for the current mode.
• Copy To Variable copies the current variable settings to another variable in the same block model.
• Copy To Bm copies settings to matching variables in other block models.

The copy buttons are intended for a single selected source variable.

Standard estimation

Choose Estimation, then click Set Estimation… The standard estimation setup contains four tabs:

• Method
• Search
• Envelopes
• Variable

Method tab

The Method tab combines the estimation method and block discretization.

Available methods are:

1. User defined script Customize the estimation script directly.
2. Inverse weighted distance Define the inverse distance exponent. A value of 2 is the usual default. You can also choose to use ellipsoid-influenced distances instead of real distances for the calculation.
3. Nearest Neighbor Assigns the value from the closest retained sample.

Block discretization

Block discretization splits each block into sub-units in X, Y, and Z. Each sub-unit is estimated, then the results are averaged to produce the final block value. Higher discretization usually improves support representation but increases runtime.

Search tab

The Search tab contains the main neighborhood controls:

• Composites Set
• Pass
• Ellipsoid
• Ellipsoid Parameters
• Search Type

Composites Set

Select the composite set used for estimation. You can also choose the composite variable to use. By default, the estimated variable uses the composite variable with the same name, but you can override it when needed. Why override the variable name ? for example you may want to compare a kriging estimation with ID2. In this case you could add a second variable called ag_krig and estimate this variable with Ag since Ag_Krig is not a composite variable

Passes

Each pass is a separate estimation attempt. Passes are applied in order, and estimation only estimate an unestimated block.

This is useful when you want:

• a tighter search on the first pass,
• then a broader search on later passes.

Each pass has its own search parameters and ellipsoid. The composite set remains the same across passes.

Ellipsoid

The ellipsoid controls the search geometry. You can:

• select an ellipsoid from the project ellipsoid list,
• use Variable Ellipsoid when local azimuth, dip, and spin values exist and are populated,
• use Best Fit Ellipsoid to align the search orientation with the current block model orientation, usefull if you have multiple block model and dont want to create manually different ellipsoid orientation.

Ellipsoid Parameters

These settings control how many composites are accepted and how they are filtered:

• Minimum number of composites: estimate the block only if at least this many composites are found.
• Maximum number of composites: keep only the closest composites once this limit is reached.
• Limit number of composites per drillhole: restricts the number of retained composites from the same drillhole.
• Use ellipsoid-influenced distances in composite selection: uses relative ellipsoid position instead of only real distance for sample selection.
• Cap composites: caps composite values at a chosen threshold before using it for estimation.
• Cap when distance is greater than: applies the cap only when the retained composite is farther than the specified distance.

Search Type

Two search types are available:

• Regular Search
• Octant Search

Octant Search adds distribution rules:

• minimum number of diagonals containing at least one composite,
• minimum number of octants containing at least one composite,
• maximum number of composites selected from one octant.

Envelopes tab

The Envelopes tab can restrict either the blocks being estimated or the composites being used.

• Blocks Limit Envelope: estimate only blocks inside the selected envelope.
• Composites Screening Envelope: use only composites inside the selected envelope.

This is a legacy option, often unnecessary when the block model and composite set already represent the intended domain

Variable tab

The Variable tab is used to review variable-related limits and supporting fields. Depending on the estimation type, it can include:

• the variable being estimated,
• minimum and maximum accepted values,
• specific gravity selection,
• recovery selection.

A value range of 0 typically means no limiting value is applied.

This is a legacy tab, usually unnecessary.

Kriging

Choose Kriging, then click Set Estimation… The kriging setup contains five tabs:

• Kriging
• Search
• Variograms
• Envelopes
• Variable

The Search, Envelopes, and Variable tabs follow the same general logic as standard estimation. The main differences are the Kriging and Variograms tabs.

Kriging tab

The Kriging tab contains three main groups:

• Type
• Blocks Discretization
• Option

Type

Available kriging methods include:

• Ordinary Kriging
• Simple Kriging
• Indicator Kriging
• Median Indicator Kriging

Ordinary Kriging

This is the standard kriging method.

Simple Kriging

Simple kriging uses a known mean for the domain instead of forcing the weights to sum to 1. The influence of that mean increases when nearby samples are sparse or when the nugget effect is strong.

You can use:

• a local mean from file, or
• a fixed local mean.

When using a local mean file, it must contain block indices and the local mean value.

Indicator Kriging

Indicator kriging estimates a value between 0 and 1 relative to a cut-off. In practice, it can be interpreted as the proportion or probability above the selected cut-off.

Median Indicator Kriging

Median indicator kriging repeats indicator kriging over several cut-offs to build a grade distribution, then derives the final estimate from that distribution.

You can choose:

• User Defined CutOffs and define them manually, or
• Automatic Cutoffs generated from the sample data distribution.

You can also choose:

• Regular indicators, or
• Nested indicators.

Additional kriging options

The following options can be combined with the main kriging types when appropriate:

• Lognormal
• Irregular

Lognormal Kriging is intended for variables that are suitably lognormally distributed. The software handles the log transformation and back-transformation.

Irregular:

The theory of kriging applies indifferently to the shape of the block. It is thus possible to consider a complete block model as being a single block. In this case, the blocks themselves are considered as units of discretization. In the “Estimation Settings” menu, it is possible to choose the kriging sub-option “Irregular”. In this case, the discretization is disabled, and you must specify a “Maximal Distance” to the block model to select your composites because the ellipsoid is not used. This maximum distance should be at least half of the maximum diagonal of the block size of the model, to select at least all composites within the model.

Although the ellipsoid is not used for research, the search parameters of the ellipsoid (Minimum Composites per Block, Maximum Composites per Block, Limit Number of Composites per Drillhole, Cap Composites) are used. Normally the “Maximum Composites per Block” should correspond to the total number of composites available for kriging this block. On the other hand, if your number of composites exceeds several hundreds, it is advisable to test with a maximum of one hundred to check if the result is numerically stable.

Of course, you must also specify a variogram. After that you can run kriging by clicking on the “Estimate / Estimate” menu. After kriging, each of the blocks of the block model contains the same estimated grade, that of the irregular block

Blocks Discretization

Kriging discretization defines how many internal nodes are used inside each block in X, Y, and Z. Typical values are 2x2x2 or 3x3x3. A finer discretization usually gives a better block-support estimate but requires more processing time.

Option

The Option group includes:

• Zero Kriging Bypass: assigns zero directly when all retained samples around the block are zero.
• Removal of Negative Weights: removes negative kriging weights and rescales the remaining positive weights.
• Assign the Default Grade to Un-Estimated Blocks: fills blocks that cannot be estimated because of insufficient nearby data.

Variograms tab

Use the Variograms tab to choose the variogram model used for kriging.

You can select an existing variogram, modify one, or create a new one if needed.

Simulation

Choose Simulation, then click Set Estimation… The simulation setup is similar to kriging, but it uses a dedicated Simulation tab.

The simulation setup contains these tabs:

• Simulation
• Search
• Variograms
• Variable

Simulation tab

The Simulation tab contains these groups:

• Algorithm
• General
• Random Number Generator Settings
• Histogram to normalize
• Kriging Method

Algorithm

Two simulation engines are available:

• SGeMS
• GSLib

Both are supported by the software. Some labels and tail options change slightly depending on the selected engine.

General

General settings include:

• Number of simulations to carry out
• Maximum conditioning data or maximum number of simulated nodes to use, depending on the selected engine
• Assign data to grid

In practice, many users consider 30 simulations a minimum for a basic uncertainty review, while larger studies often use 100 or more.

Random Number Generator Settings

You can:

• use an automatic seed, or
• enter a specific seed to make runs reproducible.

Histogram to normalize

This section controls the back-transformation limits and tail behavior used for simulated values. It includes:

• minimum and maximum values for simulated nodes,
• a Reset option to return to the control point distribution limits,
• lower-tail and upper-tail behavior.

Depending on the selected engine, the available tail options can include:

• no interpolation or linear interpolation,
• power model,
• exponential or hyperbolic behavior.

Kriging Method

Simulation can use either:

• Simple Kriging, or
• Ordinary Kriging.

Simulation checks before running

Before simulation starts, the software can warn you if:

• the selected variogram is not normalized, or
• the variogram contains linear components not supported by the simulation.

These warnings should be reviewed carefully before continuing.

Simulation Variable Settings

After simulation, open Estimation > Simulation Variable Settings… to choose how simulated variables are displayed.

For each simulated variable, you can display:

• the Average, or
• the Probability Cut Off view.

When using Probability Cut Off, enter the cut-off value to display the probability that the simulated value is above that threshold.

Running the estimation

After the settings are complete, use Estimation > Estimate.

Genesis processes the selected block model or block models and applies the configured method to each estimable variable. If several passes are defined, they are run in order.

Test One Block

Use Estimation > Test One Block… to validate the setup before a full run or to understand how the estimation of one block is processed in details if you have any doubt.

You will be asked for the block indices:

• IX
• IY
• IZ

The software then runs the selected estimation in diagnostic mode and opens a detailed log. This is especially useful for checking:

• pass behavior,
• retained composites,
• search behavior,
• kriging or estimation diagnostics for one block.

Estimate Distances

Use Estimation > Estimate Distances… when you want to write the distance from each block to the nearest selected object into a variable.

This command works on a selected destination block model and lets you choose:

• the destination variable,
• one or more composite sets,
• block models,
• envelopes,
• intervals,
• holes,
• prisms,
• figures,
• geolines.

The selected variable is filled with the shortest distance from each block center to the nearest chosen object. This can be useful for proximity analysis, support review, or building secondary interpretation fields.

Panels Estimation

Overview

Panel estimation combines MIK (Multiple Indicator Kriging) and LIK (Localized Indicator Kriging / localized uniform conditioning style workflows) to produce block values that better reflect local variability while preserving larger-panel behavior.

In practical terms, panel estimation is used to generate SMU-scale values that are less smooth than a standard kriged block model, while still respecting the average behavior of the larger panel.

General concept

MIK estimates a local grade distribution for each unsampled location. Like simulation, it provides information about uncertainty and grade-tonnage behavior, but it is often easier to apply in deposits with strong geological controls.

LIK then converts that panel-scale distribution into a single localized value for each block. This is important because a distribution alone is informative for reporting, but it does not directly assign a practical SMU value to each location.

The result is a block model that preserves the panel average while increasing local variability inside the panel. In most cases, the final model is less smooth than a model estimated only by kriging.

MIK: Multiple Indicator Kriging

In this workflow, each panel is kriged as a single irregular block using composites located within a limited distance. The kriging weights are then used to build an indicator-based cumulative distribution for the panel.

This makes it possible to describe the probability of reaching different grades inside the panel. However, this distribution still represents the panel support rather than the individual block support, so an additional support adjustment is required.

Change of support

Grade variability changes with support size. As support increases, variability usually decreases and the distribution tends to become more symmetric, while the mean remains unchanged.

An important idea in this workflow is that variances are additive. In simplified terms, the composite variance inside a panel can be separated into:

• variance between blocks inside the panel, and
• variance inside the blocks themselves.

Because panel kriging provides information about these variance components, it becomes possible to estimate the block-scale variance needed for localization.

Several change-of-support approaches exist, including affine, indirect lognormal, and discrete Gaussian methods. In this workflow, the affine approach is generally appropriate because it reduces variance without forcing the distribution into an artificial normal shape.

Localization

The final localization step assigns values to individual blocks so that the panel average is preserved while the internal block distribution becomes more variable.

In simple terms, the blocks inside a panel are ranked and matched to the cumulative frequency distribution derived for that panel. If blocks were not previously estimated and therefore still contain null or default values, the localized values are assigned using the block ordering rules available in the workflow. If only a few blocks are missing, they are generally placed near the center of the distribution.

Create panels

To create panels, select the block model and use Edit > Create Panels… In the Panel Generator dialog, define the panel size along the three orthogonal axes of the model:

• X
• Y
• Z

The panel dimensions must correspond to an integer number of blocks in each direction, so each dimension must be a multiple of the block size. You must also select the variables to include in the panel workflow.

The software creates a new block model with the same name plus the Panel suffix. This panel model is highlighted in deep pink.

For each variable included in the panel workflow:

• a new variable with the same name and the Panel suffix is added to the parent block model,
• Tag_View_Panel is added to indicate which panel each block belongs to.

Estimate panels

Before estimating panels, the parent blocks should already be estimated, usually by kriging. Then select the panel model and open Estimation > Estimation Settings…

For each panel variable, configure irregular block kriging. This typically requires:

• a variogram, usually the same one used for the blocks,
• the maximum distance from the panel boundary used to retain composites,
• the minimum and maximum number of composites,
• optional drillhole-based composite limits.

Once the setup is complete, run the estimate from Estimation > Estimate. If the parent blocks are re-estimated later, the panels should also be re-estimated so they remain consistent.

When a block model is associated with panels, the block context menu includes Show Panel Graph. This displays the cumulative frequency curve for the panel containing the selected block.

Local Anisotropic Kriging (LAK)

LAK (Local Anisotropic Kriging) combines dynamic anisotropy concepts with kriging so that both the search neighborhood and the variographic directions can follow local anisotropy. This makes the workflow more adaptable in deposits where direction changes significantly from place to place.

Dynamic anisotropy

Open Estimation > Estimation Settings… for the selected block model, then go to the kriging settings for the relevant variable and enable Local Anisotropy Kriging (LAK).

When LAK is used, variable ellipsoids are automatically required, even if the variable ellipsoid option is not explicitly checked elsewhere. For that reason, the variable ellipsoid fields must already exist and be properly populated.

LAK does not depend on how the variable ellipsoids were generated. They can come from control geolines, block model geometry, interpreted continuity directions, or other supported orientation workflows.

Variogram behavior with LAK

When LAK is active, the directions of the variable ellipsoid are used as the variographic directions during kriging. In other words, the stored variogram directions are replaced locally by the orientation of each block’s variable ellipsoid.

The major, intermediate, and minor ellipsoid axes then act as the equivalents of the long, medium, and short variogram ranges for that block. This allows the kriging orientation to adapt locally instead of remaining fixed across the whole domain.