Domain definition is arguably the most consequential decision in any resource estimation workflow, yet it receives far less attention in regulatory guidance than the interpolation method that follows. Choose the wrong domain boundaries and no amount of kriging sophistication will rescue your estimate.
Over the course of several dozen NI 43-101, SK-1300 and JORC-compliant resource estimates, I have seen domain errors manifest in every possible way: grade smearing across lithological contacts, inflated continuity in structurally complex systems, and variograms that simply refuse to model because the data belong to two populations masquerading as one. This post captures the framework I now apply at the outset of every engagement.
What Is a Domain and Why Does It Matter?
A domain — sometimes called an estimation domain, mineralisation envelope, or constraining solid — is a three-dimensional volume within which the geologist asserts that grades are drawn from a single statistical population and display a consistent spatial continuity structure. The key word is asserts: domain definition is a geological interpretation, not a mathematical optimisation.
The practical consequence is profound. All geostatistical interpolation methods — ordinary kriging, indicator kriging, multiple indicator kriging, sequential Gaussian simulation — assume stationarity within the domain. That is, the mean, variance and spatial covariance structure are assumed constant throughout the volume being estimated. If that assumption is violated, the resulting block model will contain systematic bias that no subsequent validation exercise can correct.
"The domain is your geological hypothesis. Every assay inside it is a test of that hypothesis. If you can't articulate the geological argument for your boundary, you don't yet have a domain — you have a line on a map."
Hard vs. Soft Boundaries: The Regulatory and Practical Difference
NI 43-101 Form 43-101F1 (Item 14) and JORC Table 1 (Section 4, Estimation and Modelling Techniques) both require the QP/Competent Person to disclose whether domain boundaries were treated as hard or soft during interpolation. This is not a bureaucratic checkbox — it is a material disclosure that affects resource classification confidence and the reliability of any downstream economic analysis.
Hard Boundaries
A hard boundary is a contact across which no grade information passes during interpolation. Data on one side of the boundary are not used to estimate blocks on the other side. Hard boundaries are appropriate where there is a strong geological reason to expect an abrupt, geologically controlled grade change: a fault plane separating oxidised and fresh material, a lithological contact between a mineralised intrusive and a barren host, or a sharp redox front in a sediment-hosted deposit.
The validation test for a hard boundary is simple: do grade histograms and probability plots for samples on each side of the contact show statistically distinct populations? A Kolmogorov-Smirnov test or a simple Q-Q plot comparison will usually settle the question.
Soft Boundaries
A soft boundary allows a limited amount of grade information to pass across the contact during search. This is appropriate where the geological transition is gradational — for example, the outer margin of a porphyry halo where copper grades decline progressively rather than dropping to background instantaneously. Soft boundaries require careful documentation of the search strategy and the rationale for the distance over which cross-boundary data are permitted to influence estimates.
In my experience, regulators and technical reviewers are far more likely to challenge undocumented soft boundaries than well-argued hard ones. If you use a soft boundary, explain it — in the report and in your estimation parameters table.
How I Approach Domain Construction Under NI 43-101 and JORC
My domain-building workflow has four phases, and none of them involve opening a block modelling package first.
- Geological framework review. Before touching the drill database, I review all available geological maps, cross-sections, structural interpretations and deposit model analogues. The deposit model — IOCG, orogenic gold, VMS, porphyry, sediment-hosted stratiform — immediately constrains the plausible domain geometry.
- Exploratory Data Analysis (EDA) by lithology and alteration. I run grade histograms, log-probability plots and scatter plots broken out by lithological code, alteration type and structural domain. Population mixing shows up clearly as inflection points on log-probability plots.
- Contact analysis. Using the drill-hole database, I extract assay pairs that straddle interpreted geological contacts and test whether the contact is statistically significant. A contact with a mean grade ratio greater than 3:1 across it is almost always a hard boundary candidate.
- Solid modelling and review. Once the geological case is assembled, I build wireframe solids in Leapfrog, GEMS or Micromine (depending on client preference) and back-validate by checking that drill intercepts that define domain margins are correctly honoured.
Common Mistakes Junior Teams Make
The most prevalent error I encounter in technical due diligence engagements is using a single grade shell — typically a >0.3 g/t Au or >0.1% Cu cut-off envelope — as both the domain boundary and the reporting envelope. This conflates two separate concepts. The domain exists to honour geological control; the reporting envelope exists to define economically relevant mineralisation. They frequently coincide, but assuming they always do is geologically naive and can produce a resource estimate that fails review.
A second common error is defining domains at a scale too fine for the available data. I have reviewed resource models with twenty-seven separate estimation domains for a deposit drilled on 40-metre centres. Each domain contained fewer than thirty composites — not enough to build a reliable variogram or validate the estimate. Domain definition must be commensurate with data density.
If the number of samples per domain is less than 25–30 composites, the variogram is unreliable and the domain should be merged with a geologically similar neighbour or the estimate should be classified as Inferred regardless of drill spacing.
A Practical Domain-Review Checklist for QPs
When I receive a resource model for independent review — whether for a NI 43-101 technical report or a pre-transaction due diligence exercise — here is the minimum domain documentation I expect to find:
- A geological rationale for each domain (lithology, alteration, structure, or grade shell — and which one controls mineralisation)
- Grade population statistics (mean, median, CV, P90) for each domain, demonstrating internal homogeneity
- Contact analysis across key domain boundaries
- A clear statement of whether each boundary is hard or soft, and why
- Variogram models fit within each domain, with nugget-to-sill ratios and range parameters
- Sample count per domain, with justification where n < 50
- Swath plots demonstrating that the estimate honours the input data within each domain
If any of these elements are absent, I flag them as disclosure deficiencies in the technical report or due diligence memorandum. A geologically defensible domain framework is not bureaucratic box-ticking — it is the foundation on which every subsequent modelling decision rests.
Questions about domain construction for your project? I work with exploration companies and capital market participants across NI 43-101, SK-1300 and JORC jurisdictions. Get in touch to discuss your resource estimation needs.