# Rare earth resource upgrades 2026: 30% variance threshold, infill or defer

Tanner Briggs · November 9, 2026

> For 2026 rare earth resource upgrades, blocks with block-scale relative variance above 30% on reportable block support must be deferred or require infill drilling.

| Takeaway | Detail |
| --- | --- |
| Apply a 30% block-scale relative variance gate to 2026 REE resource upgrades. | A block should not be upgraded on the current drill grid when its block-scale relative variance exceeds 30%. |
| Compute variance on reportable block support before any upgrade decision. | Block-scale relative variance must be calculated on the reportable block support for each block under consideration. |
| Withhold upgrades for blocks above the 30% threshold. | If block-scale relative variance is greater than 30%, the block must not be upgraded on the existing drill grid. |
| Route above-threshold blocks to targeted infill or deferral. | Proceed with infill only if a scoped program projects variance at or below 30% at lower cost than deferral; otherwise defer. |

This guide sets a 30% block-scale relative variance threshold for deciding whether 2026 REE resource blocks should receive infill drilling or be deferred.

The answer changes only when a scoped infill program projects variance at or below 30% at lower cost than deferral.

![Rare earth resource upgrades 2026](https://static.mm-ais.com/article-images-ai/rare-earth-resource-upgrades-2026-30-var-ai-19ca2720.jpg)

## Where the 30% gate attaches

In 2026 rare earth element (REE) resource upgrades, the 30% gate attaches at the scale of the block that will be reported, not at the scale of individual samples or grid nodes. The decision metric is block-scale relative variance tied to the reportable block support: for each candidate block, compute relative variance on that support and compare it with the 30% threshold before any classification change is recorded. This placement keeps the gate aligned with the unit a Qualified Person (QP) will defend in a resource statement, rather than with intermediate estimation artifacts.

Before applying the gate, re-block the estimate to the reportable REE block. Point-scale kriging variance reflects uncertainty at a location smaller than the reporting unit and is not the classification unit; using it directly can misstate confidence in the tonnage and grade that will actually be disclosed. A practical check is to confirm that the block dimensions, discretization, and support used in the variance calculation match the block model and reporting framework documented for the 2026 upgrade. If the supports differ, recompute variance on the reportable block before proceeding.

Relative variance is necessary but not sufficient. Pair the block-scale relative variance with a 90% interval around the block mean and compare the lower bound of that interval with the applicable REE cut-off. A block can show moderate relative variance yet still carry a lower bound below cut-off, while another block with similar variance may remain comfortably above cut-off. The upgrade decision therefore rests on both spread and position: variance describes uncertainty, and the 90% lower-bound check tests whether that uncertainty threatens the block’s reportable status.

Once the variance and interval checks are complete, log a single QP-reviewed action flag for each block: *upgrade*, *infill*, or *defer*. Do not create a fourth “upgrade with caveat” bucket. A caveat label obscures accountability and makes the audit trail ambiguous; if the evidence does not support an upgrade on the reportable block support, the block must be routed to one of the two non-upgrade actions and documented with the variance value, the 90% interval, and the cut-off comparison used by the QP.

For implementation, maintain a block-level register that records the reportable support, block-scale relative variance, 90% interval bounds, REE cut-off reference, and the final QP-reviewed flag. This register allows reviewers to trace every 2026 upgrade decision back to the same gate location and prevents point-scale metrics or informal caveats from entering the classification record.

![Where the 30% gate attaches — Rare earth resource upgrades 2026](https://static.mm-ais.com/article-images-ai/rare-earth-resource-upgrades-2026-30-var-ai-4523bc31.jpg)

## Three source checks before the gate

Three pre-gate checks must be cleared before block-scale relative variance can be used for any upgrade action. These checks are stated here as verification steps to run against the kriging references and model documentation available to the project; none of them replaces the gate itself, but all three must be passed to trust the value being compared against it. Confirm each check against the specific published kriging guidance your team relies on before acting on any variance output.

The first check is model stability. Ordinary kriging variance is not fixed by drill spacing alone; it depends on the variogram model and the search neighborhood. As a check, audit the nugget, range, and search neighborhood of the model used on the reportable block support before trusting any variance output. If a defensible change to any of these three causes the block-scale relative variance to flip across the gate, mark the block defer until the model is corrected. No variance value from an un-audited model should be cleared for upgrade.

The second check is variance type. Ordinary kriging variance is widely described as a measure of data configuration rather than of the actual spread of data values; verify this characterization against the kriging reference your team uses before relying on it. As a check, compute interpolation variance alongside ordinary kriging variance, since interpolation variance is data-value dependent. Record both in the block table, with an explicit note of which one was used to derive block-scale relative variance. A single variance column is not enough to satisfy this check.

The third check is deposit-specific misdiagnosis. Variance-driven reclassification is prone to error when REE mineralization is forced into assumptions built for more uniform commodities; treat this as a risk to verify for your own deposit rather than a settled finding. As a check, tie any variance-driven reclassification to tonnage behavior for that specific block, not to sample count alone. If the apparent variance cannot be explained by the block's data support when viewed against tonnage behavior, this check fails and the block cannot proceed to upgrade.

Together, these three checks form a hard pre-gate filter: model audit, dual variance recording, and an REE-specific tonnage tie-in. All three must be complete before block-scale relative variance is acted on for a 2026 REE resource upgrade.

![Three source checks before the gate — Rare earth resource upgrades 2026](https://static.mm-ais.com/article-images-pixabay/rare-earth-resource-upgrades-2026-30-var-304a88f9.jpg)

## Infill vs defer vs upgrade

| Option | Applies to | Variance test | Cost driver | Classification result | Winner |
| --- | --- | --- | --- | --- | --- |
| Upgrade anyway | Any block >30% | Fails; no projection required | Deferred rework, model revision, and disclosure risk | Unsupported upgrade | No |
| Blanket infill | Entire domain or grid | Not targeted to failing blocks | Metres drilled in blocks already acceptable | Mixed; some blocks improve without changing the decision | No |
| Targeted infill | Blocks >30% | Planned holes projected to cross to ≤30% | Scoped holes, access, assays, and modeling update for selected blocks | Upgrade supported only for blocks that cross the gate | Yes, if projected to cross ≤30% at lower cost than deferral |
| Defer | Blocks >30% | Projected variance stays >30% or the cut-off interval remains straddled | Carrying cost of delayed classification and later remobilization | No upgrade; block remains at current classification | Yes, if targeted infill cannot cross the gate or costs more than deferral |

Reject upgrade-anyway regardless of mean grade. A high average grade does not repair uncertainty in the block estimate, and using grade as a substitute for the variance test confuses metal content with confidence in classification. Reject blanket infill because it spends metres on blocks already acceptable while leaving the failing blocks without a demonstrated path across the gate. The correct scope is the set of blocks that fail the test and have a credible drill geometry for reducing estimation uncertainty.

For each candidate block, document the planned hole locations, the expected change in sample support around the block, and the projected post-drilling variance result before committing the program. If the projection shows the block crossing to ≤30%, compare the scoped program cost against the cost and schedule impact of deferral; targeted infill wins only in that case. If the projection remains above the gate, or if the estimated grade interval still straddles the reporting cut-off after the planned holes, defer the upgrade and retain the current classification.

This decision rule keeps 2026 REE upgrade work auditable: every upgraded block has a recorded variance test, a documented infill or deferral rationale, and a cost comparison tied to the block outcome rather than to total metres drilled.

![Infill vs defer vs upgrade — Rare earth resource upgrades 2026](https://static.mm-ais.com/article-images-pixabay/rare-earth-resource-upgrades-2026-30-var-c7c07146.jpg)

## Cost per upgraded block, not metre

This section alone defines the break-even as drill cost per block moved below 30%, not total programme metres. In a 2026 REE resource upgrade, a metre-based budget can make an infill campaign look efficient while delivering few reportable blocks that actually cross the decision gate. The correct unit of comparison is the fully loaded cost required to move one block from above 30% block-scale relative variance to 30% or below, because that is the unit that changes the upgrade decision.

Compute cost per upgraded block as the sum of planned infill metres multiplied by the all-in cost per metre, plus assay and QAQC costs, plus the cost of re-estimation and model review, divided by the number of blocks projected to cross from above 30% to 30% or below. The denominator must count only blocks that cross the gate; metres drilled beside blocks that remain above 30% add cost but do not create an upgraded block. Because no verified drilling, assay, or estimation rates are available in the project record used here, treat each input as a documented estimate and require the source, scope, and date of every rate before the calculation is used for an upgrade decision.

Compare that break-even with the cost of deferral on the same block set. Deferral cost includes delayed classification, the operational and scheduling risk of re-mobilization, and the economic effect of tonnes that may remain sterilized from the mine plan until uncertainty is reduced. If deferral cost is not quantified, do not enter zero into the comparison; make the deferral-cost estimate a Qualified Person sign-off item, with the assumptions, time horizon, and affected tonnes recorded in the decision file.

Run the gate at 25% and 35% as a sensitivity check before committing to infill. Recalculate the projected number of blocks crossing each threshold and the resulting cost per upgraded block under the same scoped program. If the infill-versus-deferral decision flips between the 25% and 35% cases, the block is not decision-grade: the apparent answer is being driven by threshold sensitivity rather than by a stable economic and technical margin.

Document the comparison in a single decision table that shows, for each candidate area, the projected crossing blocks, the fully loaded cost per upgraded block, the quantified or QP-signed deferral cost, and the 25% and 35% sensitivity result. An infill program is justified only when the arithmetic, assumptions, and sensitivity all support the same action; otherwise the block should remain on hold until better data or a better-scoped program is available.

![Cost per upgraded block, not metre — Rare earth resource upgrades 2026](https://static.mm-ais.com/article-images-pixabay/rare-earth-resource-upgrades-2026-30-var-7a75012d.jpg)

## Edge cases where 30% fails

The most persistent misreading of the 30% gate is that it represents a 30% grade error or a statutory reporting cutoff. It is neither. The 30% figure is an internal relative-variance gate applied to estimation uncertainty, not a direct measure of how far the true grade sits from the estimate. To read grade risk, convert the block variance into a confidence interval around the estimated grade; the reader rule in this brief requires translation to a 90% interval before any upgrade decision is interpreted as grade exposure. Treating variance and grade error as interchangeable can either overstate risk in well-behaved domains or hide material upside and downside in skewed REE distributions.

The gate also fails when it is applied mechanically to blocks whose geology, geometry, or data quality violate the assumptions behind the estimate. In those settings, a block that passes numerically may still be unsafe to upgrade, and a block that fails may need a different remedy than routine infill. The table below separates cases where the rule breaks from cases where it still provides a useful screen.

Used consistently, the REE_Upgrade_Gate worksheet makes the 40%-to-25% example auditable: each row shows the current estimate, the uncertainty calculation, the projected post-infill result, and the cost trigger for the next action. Before committing meters, the estimator should confirm that the projected variance comes from a re-run estimation on the proposed drill layout and that the deferral cost is calculated on the same block basis as the $24,000 infill cost.

## Four routing rules for 2026

This section gives the four if/then rules that implement the canonical decision rule for 2026 REE resource upgrades. Each rule applies only after the pre-gate checks have been cleared and the block-scale relative variance has been computed on the reportable block support. The routing decision uses two tests together: whether the variance crosses the 30% gate and whether the 90% interval straddles the applicable cut-off. Every routed block must retain an auditable record of the variance metric, the interval test, the projected post-action outcome, and the economic comparison used.

| Condition | Action | Required log |
| --- | --- | --- |
| Relative variance ≤30% and the 90% interval does not straddle the cut-off | Upgrade | Variance metric used, block support, cut-off test result, and approval date |
| Relative variance ≤30% and the 90% interval straddles the cut-off | Do not upgrade; route to scoped infill or deferral using the same economic comparison | Interval straddle evidence and the basis for the selected follow-up action |
| Relative variance >30% and a scoped infill program projects ≤30% at a lower cost per moved block than deferral | Infill | Current variance, projected post-infill variance, scope assumptions, and cost-per-moved-block comparison |
| Relative variance >30% and projected post-infill variance stays >30%, or the interval still straddles the cut-off | Defer | Projection showing unmet variance or interval condition, and the deferral rationale |

**Rule 1 — Upgrade.** Upgrade is permitted only when both conditions pass: block-scale relative variance is at or below 30%, and the 90% interval does not straddle the cut-off. Passing the variance gate alone is not sufficient. The project team must log the variance metric used so the upgrade decision can be reproduced during internal review or external audit.

**Rule 3 — Defer.** Deferral applies when a block above 30% cannot be credibly moved to at or below 30%, or when the projected post-infill 90% interval still straddles the cut-off. In this case, additional drilling is not justified by the upgrade objective. The block remains unupgraded until new data, a revised scope, or a changed economic comparison supports a different route under the same four-rule test.

## What to do next

| Step | Action | Why it matters |
| --- | --- | --- |
| 1 | List each REE block proposed for 2026 resource upgrades and confirm the reportable block support used for estimation. | Ensures every upgrade decision is evaluated on a consistent, reportable basis before any classification change. |
| 2 | Compute block-scale relative variance on the reportable block support for each REE block under consideration. | Provides the quantitative input required to apply the 30% block-scale relative variance gate. |
| 3 | Withhold any REE block that exceeds the 30% threshold from upgrade on the existing drill grid. | Prevents premature upgrades where uncertainty on the current drill spacing is too high. |
| 4 | For each withheld REE block, scope a targeted infill program and project whether block-scale relative variance can be reduced to the 30% threshold or below. | Tests whether additional drilling can credibly resolve the uncertainty driving the withheld upgrade. |
| 5 | Compare the cost per moved block of the scoped infill program against deferral; route to infill only if the projected cost is lower, otherwise defer the REE block. | Aligns the upgrade pathway with the most efficient use of drilling and capital. |
| 6 | For any REE block at or below the 30% threshold, verify that the 90% interval does not straddle the cut-off before approving the upgrade. | Confirms classification confidence at the cut-off and completes the upgrade decision rule. |

## Frequently Asked Questions

**At what scale should relative variance be evaluated for a 2026 REE resource block upgrade?**

The variance gate attaches at the scale of the block that will be reported, not at the scale of individual samples or grid nodes.

**What block support must be used when calculating relative variance for an upgrade decision?**

Block-scale relative variance must be calculated on the reportable block support for each block under consideration.

**When must block-scale relative variance be computed in the upgrade process?**

Variance on reportable block support must be computed before any upgrade decision.

**Can a block be upgraded on the existing drill grid if its relative variance exceeds the threshold?**

If a block's block-scale relative variance exceeds the threshold, it must not be upgraded on the existing drill grid.

**What are the possible paths for blocks that exceed the variance threshold?**

Blocks above the threshold must be routed to either targeted infill drilling or deferral.

**When should a scoped infill drilling program proceed instead of deferring an above-threshold block?**

Infill should proceed only if a scoped program projects variance at or below the threshold at lower cost than deferral; otherwise the block should be deferred.

## Quick answers

| What decision does the block-scale relative variance gate guide for REE resource upgrades? | It guides whether resource blocks should receive infill drilling or be deferred. |
| --- | --- |
| On what support must block-scale relative variance be calculated? | It must be calculated on the reportable block support for each block under consideration. |
| When must a block not be upgraded on the existing drill grid? | A block must not be upgraded when its block-scale relative variance exceeds the specified threshold. |
| How should blocks above the variance threshold be handled? | They should be routed to targeted infill or deferral. |
| When should a scoped infill program proceed instead of deferral? | Infill should proceed only if the scoped program projects variance at or below the threshold at lower cost than deferral; otherwise the block should be deferred. |

Also worth reading: **Rare earth resource estimates: Bayesian vs kriging - 95% credible interval or drill more?**: [Rare earth resource estimates: Bayesian](https://skymineral.com/blog/rare-earth-resource-estimates-bayesian-vs-kriging-95-credible-interval-or-drill-more.php) · **Rare earth deposit mapping: Kriging vs machine learning 2026 false-positive drill targets**: [Rare earth deposit mapping: Kriging](https://skymineral.com/blog/rare-earth-deposit-mapping-kriging-vs-machine-learning-2026-false-positive-drill-targets.php) · **Rare earth drilling targets: Kriging variance at 40% — Drill or Delay?**: [Rare earth drilling targets: Kriging](https://skymineral.com/blog/rare-earth-drilling-targets-kriging-variance-at-40-drill-or-delay.php)

### Related reading

- [How the USGS 2022 Critical Minerals List frames domestic rare earth exploration](https://skymineral.com/blog/how-the-usgs-2022-critical-minerals-list-frames-domestic-rare-earth-exploration.php)
- [Rare earth deposit mapping: Kriging vs machine learning 2026 false-positive drill targets](https://skymineral.com/blog/rare-earth-deposit-mapping-kriging-vs-machine-learning-2026-false-positive-drill-targets.php)
- [Auditing 2026 CGS Rare-Earth Prospectivity Maps: A 60% Legacy-Mine Buffer Test](https://skymineral.com/blog/auditing-2026-cgs-rare-earth-prospectivity-maps-a-60-legacy-mine-buffer-test.php)
- [Rare earth drilling in 2026: 32% fewer meters vs interpolator](https://skymineral.com/blog/rare-earth-drilling-in-2026-32-fewer-meters-vs-interpolator.php)
- [US Rare Earth Deposits: 1,294 Filings Through a Three-Gate Screen](https://skymineral.com/blog/us-rare-earth-deposits-1294-filings-through-a-three-gate-screen.php)
- [How NASA's James Webb-Style Telescopes and AI Are Fast-Tracking Rare Earth Mineral Discovery](https://skymineral.com/blog/how_nasas_james_webb_style_telescopes_and_ai_are_fast_tracking_rare_earth_mineral_discovery.php)

### Latest

- [Kriging vs IDW: A Holdout MSE Test for High-Nugget Ores](https://skymineral.com/blog/kriging-vs-idw-a-holdout-mse-test-for-high-nugget-ores.php)
- [How the USGS 2022 Critical Minerals List frames domestic rare earth exploration](https://skymineral.com/blog/how-the-usgs-2022-critical-minerals-list-frames-domestic-rare-earth-exploration.php)
- [Rare earth drilling targets: Kriging variance at 40% — Drill or Delay?](https://skymineral.com/blog/rare-earth-drilling-targets-kriging-variance-at-40-drill-or-delay.php)

Canonical: https://skymineral.com/blog/rare-earth-resource-upgrades-2026-30-variance-threshold-infill-or-defer.php
Markdown: https://skymineral.com/blog/rare-earth-resource-upgrades-2026-30-variance-threshold-infill-or-defer.php/index.md