Geology

Geological Domaining: Splitting Your Deposit the Right Way

When to split domains, how to validate them, and the over-domaining trap. Indonesian epithermal case study showing why vein and halo must be estimated separately.

By Ghozian Karami · July 1, 2026 · 7 min read ·1379 words

The first question I ask when reviewing a resource estimate is: “Show me your domains.” The second question is: “Why did you split them that way?” The answer to the second question tells me everything I need to know about the quality of the estimate.

Domaining is the single most consequential decision in resource estimation. It determines what gets estimated together and what gets estimated separately. It determines which statistics apply to which volume. It determines whether your variography is meaningful or garbage. And it’s the decision that most geologists make by feel rather than by evidence.

What is a domain?

A domain is a volume of the deposit within which the grade distribution is statistically and geologically homogeneous enough to be estimated as a single population. “Homogeneous enough” is the operative phrase — no domain is perfectly homogeneous, but the variation within a domain should be significantly less than the variation between domains.

A domain can be defined by:

Lithology: different rock types (e.g., vein vs. host rock)
Weathering: oxide vs. transition vs. fresh
Grade: mineralized vs. waste, or high-grade vs. low-grade zones
Structure: fault-bounded blocks
Alteration: different alteration assemblages

Most deposits need a combination. A typical epithermal gold deposit might have domains defined by lithology (vein, breccia, halo), weathering (oxide, fresh), and structure (fault-bounded blocks).

When to split domains

Split when the populations are statistically different. Don’t split when they’re not. The decision is empirical, not aesthetic.

The statistical test

The standard test is the ANOVA (Analysis of Variance) F-test, which compares the variance between groups to the variance within groups. If the between-group variance is significantly larger than the within-group variance (p < 0.05), the groups are statistically distinct and should be split.

In practice:

Group your composite samples by the proposed domain assignment
Run a one-way ANOVA on the grade variable
If p < 0.05, the split is justified. If p > 0.05, the groups are not statistically different — don’t split

Also check: the t-test between pairs of domains. ANOVA tells you that at least one group is different, but not which pairs. The t-test tells you whether domain A is different from domain B specifically.

The geological test

Statistics are necessary but not sufficient. A statistical difference that has no geological explanation is a red flag. The split needs to make geological sense:

Vein vs. halo: different mineralization processes. Split. ✓
Oxide vs. fresh: different metallurgical behavior, different density, potentially different grade. Split. ✓
North side vs. south side of a fault: only split if the fault offsets mineralization or changes the grade distribution. If the grade is the same on both sides, the split adds domains without adding information. ✗
High-grade shoots vs. low-grade zones: split only if the high-grade zone is geologically defined (e.g., a specific vein structure), not just a grade threshold. Grade-based domaining without geological control is circular reasoning. ✗

The variogram test

If you’ve split domains correctly, each domain should have a definable variogram. If the variogram within a domain is pure nugget (no spatial structure), the domain might be too small, too heterogeneous, or wrongly defined. Variography is the validation tool for domaining — if the variogram works, the domain probably works.

Hard vs. soft boundaries

A hard boundary is one where the grade changes abruptly at the domain contact. The vein-wallrock contact in an epithermal system is typically hard: 8 g/t Au in the vein, 0.3 g/t in the wallrock, with the transition occurring over centimeters.

A soft boundary is one where the grade changes gradually. The oxide-transition-fresh boundary in a weathering profile is typically soft: grade and mineralogy change over meters to tens of meters.

The boundary type affects estimation:

Boundary type	Estimation approach	Why
Hard	Hard boundary — no samples cross the contact	Grade changes abruptly; mixing populations smears the estimate
Soft	Soft boundary — samples within a search distance can cross	Grade changes gradually; excluding cross-boundary samples creates artificial discontinuities

Common error: Using hard boundaries everywhere because the software default is hard. On a laterite nickel deposit, the limonite-saprolite boundary is gradational over 0.5-2m. A hard boundary at the contact creates a grade cliff in the block model that doesn’t exist in reality.

The over-domaining trap

More domains is not better. I’ve seen a 40-hole gold project split into 12 domains — 4 lithology types × 3 weathering zones. Several domains had fewer than 20 composites. The variograms were pure nugget. The estimates were unstable. The classification was Inferred across the board because the kriging variance was enormous in every domain.

Rule of thumb: Each domain needs a minimum of 30-50 composites to define a meaningful variogram. If a proposed domain would have fewer than 30 composites, merge it with the geologically closest domain.

The trade-off: Fewer domains means more homogeneous populations within each domain (good for variography) but less geological specificity (bad for mine planning). More domains means better geological resolution but worse statistics (bad for estimation). The optimal is usually 3-7 domains for a typical deposit.

Case study: Sumatran epithermal gold

A project I reviewed in North Sumatra had a classic epithermal vein system. The initial domain model had two domains: “mineralized” (>0.5 g/t Au) and “waste” (<0.5 g/t). The resource estimate came out at 2.1 Moz, mostly Indicated.

The problem: the “mineralized” domain included the vein, the breccia halo, and the stockwork zone. These are geologically distinct:

Domain	Mean Au (g/t)	Std dev	Samples	Boundary type
Vein (massive silica)	6.8	4.2	180	Hard with halo
Breccia halo	1.8	1.5	240	Hard with waste
Stockwork	0.9	0.7	160	Soft with waste

ANOVA p-value: <0.001. The populations are statistically distinct. The single-domain estimate was mixing a 6.8 g/t population with a 0.9 g/t population, producing a smoothed estimate that overstated the stockwork grade and understated the vein grade.

Re-domaining: Split into three domains. The vein estimate was re-run with a hard boundary against the breccia. The breccia was estimated with a hard boundary against the stockwork. The stockwork used a soft boundary with waste.

Result: Vein tonnage decreased by 15% (the original estimate had smeared vein grade into the breccia, inflating the vein volume). Breccia tonnage increased. Stockwork tonnage decreased. Total contained gold dropped by 8% — but the grade distribution was much more accurate. The compositing guide was then applied per-domain, which further improved the estimate.

Domain validation checklist

ANOVA F-test: p < 0.05 between each pair of domains
t-test: significant difference between each domain pair
Box plot: visually distinct distributions (overlap is OK, but medians should differ)
Geological rationale: can you explain why the domains are different in 2-3 sentences
Variogram per domain: definable structure (not pure nugget)
Sample count: ≥30 composites per domain (≥50 ideal)
Boundary type: hard vs. soft, justified by geology
Cross-section review: domain boundaries match geological contacts on sections

How Orebit GeoSuite helps

The domaining workflow in Orebit Assay (Phase 02) and Orebit Resource (Phase 03):

Domain assignment: Upload a domain code column in the geology CSV. The tool maps domains to intervals automatically.
Domain statistics: Mean, median, std dev, min, max, and sample count per domain. Box plots generated side-by-side for visual comparison.
ANOVA + t-test: One click. The tool runs the statistical tests and reports p-values. If p > 0.05, it flags the domain pair for review.
Variogram per domain: Each domain gets its own variogram fitting panel. If the variogram is pure nugget, the tool suggests merging with the nearest domain.
Hard/soft boundary control: Per-domain-pair boundary type. Default is hard; switch to soft for weathering boundaries.
Domain boundary validation: Cross-section viewer with domain overlays. You see where the boundaries sit relative to the geological contacts.

Try Core free → · Get the bundle — Rp 99K → (single modules Rp 49K · one-time · yours forever)

Bottom line

Domaining is where geology meets statistics. Get it right and the rest of the estimation flows. Get it wrong and no amount of fancy kriging will save you. Split when the statistics and geology both justify it, validate with ANOVA and variograms, and resist the urge to over-domain. Three to seven well-defined domains will almost always outperform twelve under-sampled ones.

Reviewing your domain model before estimation? Email hello@orebit.id with the domain count, sample counts, and ANOVA results. I’ll tell you what an auditor would question.

Tagged with

Part of the Orebit ecosystem — geological workflow tools for drillhole validation, resource estimation, and JORC/KCMI reporting.
→ Explore GeoSuite

Try it

Try the toolkit this article uses.

Orebit GeoSuite — single-file HTML, works offline, no install. From CSV to resource report in one afternoon.

Explore GeoSuite →

# From this article:
open geosuite.orebit.id
load(your_drillhole.csv)
apply(workflow_above)

# Done. Ship the report.