Data & Grid Tab

This is where you choose what to interpolate and where the answer will be written. Everything downstream - the variogram, the run, the quality checks - works on the samples selected here.

Choosing the Data

Dataset lists every downhole table in the project. When more than one drill group exists, entries are prefixed with the group name.

Attribute lists the numeric columns of that table. Only floating point columns can be interpolated, so codes, names and text logs do not appear here. If the list is empty the dataset has nothing numeric in it and a warning says so.

Domain filter restricts the run to rows where some other column equals a chosen value - a lithology code, a weathering state, an oxide flag, a wireframe domain. Choose the column, then the value.

Domaining matters more than any parameter on the Variogram tab. Kriging assumes the values you feed it come from one population with one spatial behaviour. Mixing supergene and primary mineralisation, or ore and waste, produces a variogram that describes neither, and an estimate that smears grade across the contact between them. If you have a domain column, use it, and run each domain separately.

Depth range keeps only samples whose interval midpoint depth falls inside the range you set. Use it to cut off a barren cover sequence, or to restrict a run to one level of a mine.

Any change here re-extracts the samples immediately, with a progress bar. A change made while extraction is running cancels it and starts again with the new selection.

How Samples Are Built

Each interval in the table becomes one sample, placed at the desurveyed midpoint of that interval - the position along the actual hole trace, not the collar. Point tables give one sample per row at the logged position. Rows with no usable value are counted as nulls and dropped; rows failing the domain filter are excluded entirely and are not counted at all.

Samples carry the coordinate system of the drill group. If that system is geographic (latitudes and longitudes), the samples are reprojected into an appropriate UTM zone chosen from the centre of the data, because ranges and cell sizes only make sense in metres. The result grid is written with that projected system.

Sample Statistics

The statistics block describes exactly the samples that were selected. It is worth a look before going any further.

The declustered mean deserves an explanation. Deposits are not drilled evenly - the good parts get drilled out and the barren parts get one hole, so a plain average of all samples is biased upward. The declustered mean weights samples by how crowded their neighbourhood is, which corrects for that. If the two means differ substantially, your drilling is strongly clustered, and the declustered figure is the one your grid mean should end up near. The Quality tab makes exactly that comparison.

The two spacing figures are the most useful numbers on the tab, because nearly every other setting is derived from them. The dialog uses them to pick its own defaults.

Grid Definition

The grid is a regular lattice of nodes aligned with the coordinate axes, defined by three fields.

• Origin - the position of the first node, at the minimum corner.
• Cell size - the node spacing on each axis. The chain button links the three axes so they are edited together. Editing the cell size keeps the extent covered and recomputes the dimensions, so a finer cell gives you more nodes rather than a smaller grid.
• Dimensions - the node count on each axis. Editing these grows or shrinks the extent.

Below them the extent is shown as a readout: minimum, maximum and span for each axis.

Choosing a Cell Size

The cell size is a display and resolution decision, not an accuracy one. A grid finer than your data cannot manufacture detail - it interpolates the same smooth model onto more nodes. A useful starting point is about a third of the mean hole spacing, which is what Reset to data defaults uses, along with a bounding box padded by one downhole spacing.

Then watch the Nodes count, which the dialog colours as it grows:

• Under 10 million - comfortable.
• Amber, over 10 million - interpolation will take a while.
• Red, over 50 million - this will take a very long time and a serious amount of memory.

The estimated output size shown beside the count covers both arrays written to the file: the interpolated value and the distance to the nearest sample.

Node count grows with the cube of resolution, so halving the cell size multiplies the work by eight. When experimenting with variogram parameters, work on a coarse grid and only go fine once you are happy with the model.

Output

The output path is the .vti file the result is written to. It defaults to <dataset>_<attribute>.vti beside the project file, and you can type a path or browse for one. A file extension is added if you leave it off.

.vti is the standard VTK image data format. The file holds the interpolated attribute and a second array named distance - the distance from each node to the nearest sample, which the Quality tab uses to identify extrapolation. Other software that reads VTK files can open it, and it loads into the scene as an ordinary volume.

Reset to data defaults rebuilds the grid from the sample bounding box and hole spacing, and resets the output path. It is the quickest way back to a sane starting point after experimenting.

Batch

If the dataset has more than one numeric column, a Batch section appears. See Interpolation for how it behaves.

Also See

• Variogram tab - the next step
• Kriging Concepts
• Drill Manager - loading the drill data in the first place