Hard, soft and semi-soft domain boundaries: how to treat grade across a contact
A hard boundary estimates each domain only from its own samples, a soft boundary lets samples from both sides inform the estimate, and a semi-soft boundary shares only the samples within a chosen distance of the contact.
The right treatment depends on how grade behaves across the contact. If grade changes abruptly, use a hard boundary. If it changes gradually, a soft boundary avoids creating an artificial step. In GEOREKA every domain boundary is a signed distance function, so you can select the samples within a threshold distance of the boundary and include them in the neighbouring domain’s estimate. A short threshold gives a semi-soft boundary.
What the choice means
| Boundary | Samples used for a domain | Typical when |
|---|---|---|
| Hard | Only samples inside the domain | Grade changes abruptly at the contact, such as a sharp vein wall |
| Soft | Samples from both sides, all used | Grade changes gradually, such as a diffuse alteration halo |
| Semi-soft | Inside samples plus those within a set distance of the contact | A transition zone of limited width |
How to decide
Plot grade against distance from the contact, using samples on both sides. A sharp step in the average means a hard boundary. A smooth ramp that crosses the contact means a soft one, and the width of the ramp suggests the distance for a semi-soft boundary. This contact analysis should be made on the composites, not the raw samples, so that sample length does not distort the plot.
Using a hard boundary where the real contact is gradational produces an artificial step in the estimate at the contact. Using a soft boundary where the contact is sharp smears high grade into waste and overstates tonnes of mineralisation.
Signed distance functions make it practical
In GEOREKA, each domain boundary is an implicit surface, the place where a signed distance function equals a threshold. That function gives every sample its distance to the boundary, positive on one side and negative on the other. See signed distance functions explained.
To build a semi-soft boundary, select the samples where the absolute signed distance is below your threshold and include them in the interpolation of the neighbouring domain, along with that domain’s own samples. A short threshold shares only the samples next to the contact. A longer threshold approaches a fully soft boundary.
There is no single switch for this. It is a manual workflow for each boundary, which keeps the decision explicit, and the settings stay with the workflow so they can be re-run when new drilling arrives.
Things to watch
- Include shared samples only for the estimate, not for the domain’s statistics, so that the population of each domain stays distinct.
- Different elements can need different treatment. A boundary that is hard for one element may be soft for another.
- Check the result with swath plots across the contact and cross-validation, see validating a resource estimate.
Let the data decide. Look at grade against distance from the contact, then choose hard, soft or a transition distance, and record why.
Apply it
Control boundaries on your domains
Domain modelling with implicit boundaries is included in Core, and interpolation with IDW, RBF and SVR is part of Core. Geological modelling in GEOREKA →
Common questions
Soft and hard domain boundaries FAQ
What is a hard boundary in resource estimation?
A hard boundary means each domain is estimated using only the samples inside it, so there is no mixing of grades across the contact.
When should I use a soft boundary?
When grade changes gradually across the contact, so that a hard boundary would create an artificial step in the estimate.
What is a semi-soft boundary?
A boundary where only the samples within a set distance of the contact are shared with the neighbouring domain, to represent a transition zone of limited width.
How do signed distance functions help?
They give every sample its distance to the boundary, so you can select the samples within a threshold distance and include them in the neighbouring estimate.