# Rare Earth Deposits: Test a 2% Cutoff With Domain-Level Grade–Tonnage Curves

Tanner Briggs · September 17, 2026

> Test a 2% rare earth cutoff with domain-level grade–tonnage curves to reveal qualifying tonnes, average grades, and why deposit-wide means fall short.

| Takeaway | Detail |
| --- | --- |
| An average does not establish qualifying tonnes. | A deposit-wide mean above 2% does not reveal how much material exceeds that cutoff; the missing evidence is the grade distribution. |
| Define the threshold before testing it. | The supplied excerpts do not establish whether 2% means total rare-earth oxides, total rare-earth elements, or another grade measure. |
| Read the cutoff response by domain. | Test 2% against domain-level curves showing qualifying tonnage and its average grade. Orebit explains that raising the cutoff removes lower-grade blocks, trading tonnage for higher retained grade. |
| Validate the model and the economics separately. | Block-size mismatch, conditional bias, and grade smearing can distort the modeled response at 2%; the supplied excerpts also lack rare-earth recovery, price, and cost evidence supporting that threshold. |

A 2% cutoff has no reported rare-earth tonnage behind it in the supplied excerpts. That gap matters more than a headline average: even a deposit averaging above the threshold does not disclose the quantity of qualifying material. The economic-geology question is not whether the mean looks attractive, but whether the grade distribution supports a defensible inventory above a clearly defined cutoff.

Treat 2% total rare-earth oxides as a proposed test, not an established source definition or economic boundary. Request grade–tonnage curves for each geological estimation domain, with cutoff on the horizontal axis and qualifying tonnage and retained average grade on the vertical axes. Orebit’s explanation describes the trade-off: excluding lower-grade blocks reduces tonnes while increasing the average grade of retained material.

Then interrogate the model behind those curves. Orebit’s validation guidance flags block-size mismatch with mining selectivity, conditional bias, and high-grade smearing; matching a global mean does not establish correct local grades. Examine contained oxides alongside tonnes and grade, but keep modeled inventory separate from economic viability. Without rare-earth recoveries, prices, processing costs, and domain-level cutoff results, the supplied evidence cannot substantiate a resource estimate at the proposed threshold.

![Rare Earth Deposits](https://static.mm-ais.com/article-images-pixabay/rare-earth-deposits-test-a-2-cutoff-with-2b3c3519.jpg)

## Build the Curve

A grade–tonnage curve tests the distribution of block mass, not whether the deposit’s average clears a threshold. An average above the proposed cutoff does not establish that most resource tonnes survive selection: a smaller, richer population can lift the mean while substantial lower-grade mass is excluded.

Define the test variable as total rare earth oxides, or TREO, before selecting blocks. Record which oxides enter the total, their stoichiometric forms, and the dry-mass reporting basis. The unit identity is 1 wt% = 10 kg/t. Elemental REE assays require element-specific conversion using the oxide’s formula mass relative to its constituent rare-earth mass; directly relabeling elemental concentrations as oxides changes the variable without performing the conversion. Express block grades and cutoffs on the same basis.

Calculate retained dry tonnage as T(c) = Σ[t_i × I(g_i ≥ c)], where t_i is block i’s dry tonnage, g_i is its TREO mass fraction, c is the cutoff expressed as a mass fraction, and I is an inclusion indicator. The displayed convention includes blocks exactly equal to the cutoff. Document that convention and apply it to stored grades rather than rounded display values; otherwise, rounding can silently move blocks across the boundary.

Calculate retained mean grade as G(c) = Σ[t_i × g_i × I(g_i ≥ c)] / T(c), and contained oxide as M(c) = T(c) × G(c). With grade expressed as a mass fraction and tonnage in dry tonnes, M(c) is tonnes of contained TREO—not recovered product. If no blocks qualify, retained tonnage and contained oxide are zero, while retained mean grade is undefined.

Tonnage weighting is essential because unequal block masses make an arithmetic grade average physically incorrect: each block contributes oxide in proportion to its mass. Averaging drill assays instead measures the sampled observations, not the retained block inventory, and can overweight densely drilled areas. According to Orebit’s “Block Model Validation: The 7-Step Checklist Before You,” model-mean comparisons require volume or tonnage weighting. For contained-oxide accounting, use tonnage; volume weighting alone is insufficient when assigned densities differ.

Construct separate curves within geologically justified estimation domains before aggregation. Derive block mass from represented volume multiplied by assigned dry bulk density, respecting partial-block geometry. Distinct mineralized populations can have different grade distributions and spatial continuity. Pooling them during estimation can smooth across inappropriate boundaries; presenting only their combined curve can conceal which domain supplies qualifying tonnes even when the arithmetic is correct. Aggregate domain tonnages and oxide masses, then derive the combined mean from those totals—not from an unweighted average of domain means.

According to the JORC Code, 2012 Edition, geological evidence, resource classification, and reasonable prospects for eventual economic extraction constrain Mineral Resource reporting. A mathematical selection does not independently satisfy those requirements or upgrade confidence. Preserve classification and reporting constraints in each domain’s curve.

Require the curve deliverable to pair domain-resolved tonnes, grade, and contained oxide with explicit estimation uncertainty on the same reporting basis. State how uncertainty was assessed and how dependencies between domains affect aggregate ranges; a deterministic curve alone supplies no uncertainty interval. Treat the 2% TREO cutoff as unvalidated until that deliverable quantifies retained tonnes, contained oxides, and estimation uncertainty.

![Build the Curve — Rare Earth Deposits](https://static.mm-ais.com/article-images-pixabay/rare-earth-deposits-test-a-2-cutoff-with-ba5b8455.jpg)

## Read the Evidence

Mt Weld’s high reported grade and Songwe Hill’s lower reported grades are not opposing verdicts on the proposed cutoff. They describe resources assembled on their respective reporting bases. The useful comparison is therefore evidentiary: what population does each number describe, what oxide inventory follows from it, and what threshold question remains unanswered?

According to Lynas Corporation’s July 2018 Mt Weld resource update, the historical total Mineral Resource was 55.4 million tonnes at 5.4% TREO. That is evidence of a high-grade reported resource—not a measurement of tonnes retained at this guide’s proposed threshold. Its category label here is the reported total, not an individual confidence category. Preserve that distinction rather than treating the aggregate as uniformly classified material.

Using those rounded Lynas inputs, multiplying resource tonnes by the TREO mass fraction gives approximately 2.99 million tonnes of contained TREO. This is contained-oxide arithmetic for the historical total. It is neither recoverable production nor a threshold-retention estimate: multiplication carries forward the boundaries of the reported resource; it does not select material above a different cutoff.

According to Mkango Resources’ February 2019 Songwe Hill resource update, Indicated Resources were 21.0 million tonnes at 1.41% TREO, and Inferred Resources were 27.5 million tonnes at 1.33% TREO. Both were reported using a 1.0% TREO cutoff. That original cutoff belongs beside the tonnage and grade in the evidence note because it defines the selection basis of the cited resource. Keep the confidence categories separate; they are not interchangeable observations.

Using the rounded Mkango inputs for the Indicated category alone gives approximately 296,100 tonnes of contained TREO. Again, this is contained-oxide arithmetic, not recoverable production or an estimate of inventory surviving the proposed threshold. The category boundary matters: the calculation describes the Indicated inventory only and does not incorporate the separately reported Inferred material.

The interpretive asymmetry is important. Mt Weld’s cited high average does not disclose how much lower-grade material sits within its reported resource. Songwe Hill’s cited lower averages do not exclude higher-grade subsets. Neither announcement alone supplies the cumulative grade distribution needed for the proposed test. Ranking these averages cannot establish retained tonnes, and the high-grade example does not justify assuming that most reported tonnes survive a higher selection threshold.

For this 2026 guide, maintain an evidence note pairing each publication date with its resource category, tonnage-and-TREO basis, and original cutoff and domain conventions. Songwe Hill’s cutoff is explicit in the cited inputs; Mt Weld’s original reporting conventions must be carried over from the underlying update, not inferred from its average. Check subsequent technical disclosures before labeling either historical figure the current resource. Until a consistently defined, domain-resolved grade–tonnage analysis quantifies retained tonnes, contained oxides, and estimation uncertainty, leave the proposed cutoff unvalidated.

![Read the Evidence — Rare Earth Deposits](https://static.mm-ais.com/article-images-pixabay/rare-earth-deposits-test-a-2-cutoff-with-f34b1afc.jpg)

## Choose the Winning Evidence

Domain-resolved curves with validated uncertainty realizations win for testing the cutoff—not because simulation guarantees accuracy, but because this package can quantify retention while representing local estimation uncertainty. A headline average cannot establish how much material survives thresholding; a deterministic curve answers that question only for its particular model.

| Evidence package | Quantifies threshold retention? | Represents local estimation uncertainty? | Verdict |
| --- | --- | --- | --- |
| Headline tonnes and average grade | No | No | Insufficient |
| Single deterministic whole-deposit curve | Yes, for that model | No | Preliminary screening only |
| Domain-resolved curves with validated uncertainty realizations | Yes | Yes, subject to model assumptions | Winner for testing the cutoff |

Use the analyst-selected TREO sensitivity grid below as an experimental design, not as observed deposit data. Hold the modeled resource envelope constant across every step. Keep domain assignments, block support, density treatment, and oxide definitions consistent as well; otherwise, apparent threshold sensitivity can include changes in the population being tested.

Require this results table for each domain and the consistently aggregated deposit. Symbols are required outputs, not fabricated estimates: T is retained tonnage, G is its mass-weighted mean grade, and M is contained TREO, calculated using grade as a mass fraction. Each Δ is the current result minus the preceding step, reported in absolute and relative terms.

| Selected cutoff, TREO | Retained tonnes | Retained mean grade | Contained TREO | Adjacent-step changes |
| --- | --- | --- | --- | --- |
| 1.50% | T at cutoff | G at cutoff | M at cutoff | Baseline; no preceding step |
| 1.75% | T at cutoff | G at cutoff | M at cutoff | ΔT, ΔG, ΔM |
| 2.00% | T at cutoff | G at cutoff | M at cutoff | ΔT, ΔG, ΔM |
| 2.25% | T at cutoff | G at cutoff | M at cutoff | ΔT, ΔG, ΔM |
| 2.50% | T at cutoff | G at cutoff | M at cutoff | ΔT, ΔG, ΔM |

According to White Metal, Table 2 provides a grade–tonnage sensitivity analysis for Okohongo: a named example of sensitivity reporting, not validation of this candidate threshold. Here, compare adjacent tonnage losses across the equally spaced grid. Concentrated losses immediately below or above the candidate identify a locally steep segment; a rising retained grade does not cancel that loss.

Generate the uncertainty envelope with conditional geostatistical simulation, checking reproduction of domain grade distributions and spatial continuity, alongside conditioning to observations. Evaluate every realization on the same reporting support. Calculate deposit totals within each realization before taking quantiles; adding domain quantiles generally does not produce the corresponding total quantile.

Define Q(q) as the quantile with fraction q of realization outcomes at or below it. Its exceedance fraction is the complement, subject to ties—not q. Report the chosen quantile levels and model assumptions explicitly. This envelope describes uncertainty conditional on those assumptions, not protection against every modeling error.

Compare ordinary-kriged and simulated grade–tonnage responses directly. Kriging smooths estimated grades and can change modeled proportions above a cutoff; its block histogram need not reproduce the underlying distribution. Compare retained tonnes and oxides across the grid rather than assuming agreement from similar overall means.

| Decision branch | Concrete rule |
| --- | --- |
| Reject headline-only evidence | If only tonnes and average grade support 2.00%, retain “unvalidated”; require threshold-retention outputs. |
| Restrict deterministic evidence | If a single curve spans 1.50%–2.50% without local uncertainty, use it only for preliminary screening. |
| Return inconsistent comparisons | If the envelope changes between 1.50% and 2.50%, rerun on a fixed envelope before interpreting losses. |
| Flag threshold fragility | If losses across 1.75%–2.00% or 2.00%–2.25% dominate neighboring steps, flag the candidate as locally sensitive. |
| Select the winning package | For 2.00%, select domain-resolved curves only when retained tonnes, contained oxides, validated uncertainty, explicit quantiles, and kriging–simulation comparisons share a consistent basis; otherwise leave the cutoff unvalidated. |

![Choose the Winning Evidence — Rare Earth Deposits](https://static.mm-ais.com/article-images-pixabay/rare-earth-deposits-test-a-2-cutoff-with-c95f6ed0.jpg)

## What the Data Doesn't Tell You

A stable grade–tonnage curve can describe contained oxides without establishing either a mineable feed or a saleable product. Its limits begin with composition: identical total rare earth oxide grades can contain different proportions of neodymium, praseodymium, cerium, and lanthanum. Summing those oxides suppresses differences in product mix, separation requirements, and recoverable value. A geological cutoff expressed in total oxides therefore cannot establish equivalent value across deposits. Nor does an above-cutoff resource average establish that most tonnes survive the grade screen.

Mineralogy adds a separate constraint. Bastnäsite is a fluorocarbonate; monazite and xenotime are phosphates, but their shared mineral class does not imply interchangeable processing behavior. Host mineral chemistry, grain size, liberation, and associations with gangue influence beneficiation and chemical extraction. Even within one deposit, different domains may require different treatment. Representative metallurgical evidence must connect each retained material type to recovery and product specifications before contained oxides are interpreted as saleable output. Results from a favorable composite cannot establish performance for an untested domain.

Thorium- and uranium-bearing material can pass the total-oxide screen while introducing constraints absent from the curve. Processing can redistribute these elements into concentrates, process streams, or residues, changing containment, treatment, and disposal requirements. Permitting obligations depend on the material and jurisdiction, not merely its rare earth grade. The relevant evidence is therefore the distribution and processing fate of radionuclides in the retained domains, rather than a deposit-wide statement that radioactive constituents are present or absent.

Consider a counter-case in which every qualifying model block is correctly estimated, but those blocks occupy isolated pods smaller than a practical selective mining unit. Recovering a pod requires taking surrounding lower-grade material; excluding that material may instead leave part of the pod behind. Dilution reduces delivered grade, while ore loss reduces recovered tonnes and contained oxides. The modeled high-grade subset can consequently be physically unobtainable at its undiluted grade without any mathematical error in the curve.

Before interpreting such pods as genuine selectivity problems, distinguish geology from estimation artifacts. According to Orebit’s block model validation checklist, bullseye patterns around isolated high-grade composites can indicate an overly small search or an unapplied top-cut. That warning does not prove that a particular pod is spurious. It identifies a necessary diagnostic distinction: correcting an artificial concentration and testing the mineability of a real concentration are different tasks.

Economic assumptions remain independent of numerical stability. According to Mayfair Gold, Fenn-Gib cutoff grades use a gold price of US$1,765/oz. That is an example of an explicit commodity-price assumption, not support for a rare earth threshold. The supplied source excerpts provide no rare earth recovery rates, product prices, processing costs, or cutoff calculation supporting economic use of the proposed threshold. Compatible recovery, processing-cost, and capacity assumptions are essential: a capacity-constrained plant can favor a different feed selection than an unconstrained operation.

The defensible distinction is between validating a resource-estimation threshold and demonstrating an economically optimal mine plan. Require a reconciliation from retained domain tonnes and contained oxides to diluted mining feed, metallurgical products, constrained residues, and plant throughput, with uncertainty carried through those transitions. Missing downstream evidence limits the interpretation of a validated curve; it never substitutes for the domain-resolved, consistently reported grade–tonnage analysis required to validate the threshold itself.

![What the Data Doesn't Tell You — Rare Earth Deposits](https://static.mm-ais.com/article-images-pixabay/rare-earth-deposits-test-a-2-cutoff-with-1e4c51a8.jpg)

## Work the Nolans Case

Arafura Rare Earths' 2023 Annual Report reports the Nolans project at 56 million tonnes at 2.6% TREO. That is the published input for this worked example, and it is a rounded aggregate resource figure — one number for tonnes, one number for grade, no domain breakdown, no cutoff stated in the same breath. The calculation below is therefore restricted to that reported material population and nothing else. It is a mass-balance check on a rounded headline, not a re-estimation of the deposit.

Nominal contained TREO follows directly: 56,000,000 × 0.026 = 1,456,000 tonnes. According to Arafura Rare Earths' 2023 Annual Report figures, that product is the aggregate oxide inventory implied by the rounded inputs. What it does not supply is a block-by-block grade distribution. A single contained-oxide total is compatible with a deposit whose mass sits tightly around 2.6% TREO, and equally compatible with one where a high-grade core carries the average while a large low-grade halo drags beneath it. The arithmetic cannot distinguish those two worlds.

Now bound the loss. If a 2% TREO cutoff were applied, every tonne excluded must contain less than the threshold grade — that is the definition of exclusion. So the oxide removed is strictly less than 56,000,000 × 0.02 = 1,120,000 tonnes. This is deliberately loose: it charges every excluded tonne the maximum grade it could possibly carry, which no real distribution does. It is a ceiling, not an estimate.

Subtracting that ceiling from the aggregate inventory gives a nominal retained-oxide lower bound greater than 336,000 tonnes, equivalent to more than approximately 23.1% of the input inventory. Label this precisely for what it is: an arithmetic bound conditional on the rounded figures, not an estimated point on the deposit's actual grade–tonnage curve. The true retained oxide is higher than 336,000 tonnes and the true retained tonnage is unknown. The bound survives any distribution; it also constrains almost nothing.

That is where the worked decision stops. Retained tonnage and retained average grade remain undetermined, because many grade distributions satisfy the published aggregate — a narrow unimodal spread, a bimodal high-grade-plus-halo mix, a long right tail. All three reproduce 56 million tonnes at 2.6% TREO and all three produce different retained tonnages above the candidate threshold. A complete mass-balance check of this kind cannot validate the cutoff, and it cannot justify inventing a retained-tonnage figure to fill the gap.

One edge case sharpens the point. Cutoffs are not all defined on the same basis. According to Tudor Gold, the base case NSR cut-off value for Treaty Creek is $50 per tonne — an economic cutoff, not a grade cutoff. A 2% TREO threshold is a grade screen; an NSR screen folds in recovery, payability, and cost. Treating one as a proxy for the other is a category error, and it is the same error as reading a rounded average as a retained-tonnage answer.

| Quantity | Value | Status |
| --- | --- | --- |
| Reported resource (input) | 56,000,000 t at 2.6% TREO | Rounded aggregate, Arafura 2023 Annual Report |
| Nominal contained TREO | 1,456,000 t | Arithmetic on rounded inputs |
| Excluded-oxide ceiling below 2% TREO | Less than 1,120,000 t | Loose upper bound |
| Retained-oxide floor | Greater than 336,000 t | Arithmetic bound, not a curve point |
| Share of input inventory | More than approximately 23.1% | Conditional on rounded figures |
| Retained tonnage | Undetermined | Requires domain-resolved grade–tonnage curve |
| Retained average grade | Undetermined | Requires domain-resolved grade–tonnage curve |

![Work the Nolans Case — Rare Earth Deposits](https://static.mm-ais.com/article-images-pixabay/rare-earth-deposits-test-a-2-cutoff-with-482840b6.jpg)

## How to Choose Well

A reproducible extraction can still be an unacceptable estimate: reproducing a calculation does not resolve defective assays or missing uncertainty. An average above the proposed cutoff establishes neither the retained tonnage nor that most resource survives filtering. Keep the threshold explicitly unvalidated until a domain-resolved grade–tonnage analysis quantifies retained tonnes, contained oxides, and estimation uncertainty on a consistent reporting basis. Apply these decision gates in order; passing an earlier gate never substitutes for passing the remaining gates.

Rule one — Reject or proceed on assay quality. If blanks indicate unresolved contamination, certified reference materials fail acceptance criteria, or duplicates show unexplained discrepancies in the grade range controlling inclusion, reject the retained-inventory calculation pending documented investigation. Request a record linking each exception to affected assay batches, composites, and estimation domains, with the disposition of affected results. A laboratory-wide summary cannot establish that threshold-controlling samples are acceptable. Proceed only when the investigation supports their use or the affected inputs have been corrected and the estimate rerun.

Rule two — Mark unresolved or assign inclusion. If detection limits or reporting precision cannot distinguish material reliably around the proposed threshold, classify its inclusion status as unresolved. Request unrounded laboratory results where available, applicable detection limits, and the convention used for censored results. Extra decimal places generated downstream do not restore information absent from the assays. Keep the affected inventory identifiable rather than silently assigning it to either side. Proceed with a definitive assignment only when the underlying evidence supports that distinction, not merely because a displayed value appears to meet the cutoff.

Rule three — Separate classifications or withhold use. If an extraction combines Measured, Indicated, and Inferred material, require retained tonnes and contained oxides for each classification within the relevant domains. Preserve those labels through selection and aggregation; do not let the reporting operation erase geological confidence. Require the uncertainty statement to identify which classifications it covers and any differing treatment. Use the combined result only alongside those classification-specific inventories. A smooth aggregate curve is not evidence that every retained tonne has equal confidence.

Rule four — Reconcile or reject the extraction. Before accepting filtered results, require the zero-additional-cutoff selection to reproduce the stated model population’s tonnage and contained oxides within documented numerical tolerances. This baseline must preserve the population’s existing boundaries, exclusions, and reporting basis; it does not mean selecting every block in the database. Request residuals for both quantities and an explanation of any differences. If reconciliation fails, reject the extraction until the discrepancy is resolved rather than adjusting tolerances simply to make the comparison pass.

Rule five — Validate or retain “unvalidated.” Accept the threshold for the stated resource-estimation purpose only after an independent reviewer reproduces the domain- and classification-resolved inventories and uncertainty outputs from the identified model version, selection masks, and calculation specification. That specification must make units, oxide definitions, mass weighting, threshold equality, and missing-value treatment explicit. Agreement with a headline total alone is insufficient. If the reviewer cannot reproduce the result, or any preceding gate remains unresolved, retain “unvalidated” and name the outstanding defect instead of substituting a confident narrative.

## What to do next

| Step | Action | Why it matters |
| --- | --- | --- |
| 1 | Define the proposed 2% cutoff as TREO on a dry-mass basis; document included oxides, formulas, and element-specific assay conversions. | The supplied excerpts do not establish the threshold’s grade measure; elemental REE cannot simply be relabeled as oxides. |
| 2 | Request grade–tonnage curves for every geological estimation domain, plotting cutoff against qualifying tonnes and retained average TREO grade. | A deposit-wide mean above 2% does not establish qualifying tonnage. |
| 3 | At 2% TREO, tabulate each domain’s retained tonnes, average grade, and contained oxides on the same reporting basis. | This quantifies the retained inventory and the tonnage–grade trade-off described by Orebit. |
| 4 | Apply Orebit’s validation checks: compare block size with mining selectivity and investigate conditional bias and high-grade smearing. | A matching global mean can conceal incorrect local grades and distort selection at 2%. |
| 5 | Request domain-level uncertainty estimates for retained tonnes and contained oxides at 2% TREO, including sensitivity to estimation assumptions. | Treat the cutoff as unvalidated until the domain-resolved analysis quantifies inventory and estimation uncertainty consistently. |
| 6 | Require rare-earth recovery, price, and processing-cost evidence before treating 2% TREO as an economic boundary. | The supplied excerpts lack that support; modeled inventory alone does not establish economic viability. |

## Frequently Asked Questions

**Does a deposit-wide average above 2% prove that most tonnes qualify at that cutoff?**

An average above the proposed cutoff does not establish that most resource tonnes survive selection because a smaller, richer population can lift the mean while substantial lower-grade mass is excluded.

**What three modeling problems can distort the modeled response at a 2% cutoff?**

Block-size mismatch, conditional bias, and grade smearing can distort the modeled response at 2%.

**How should blocks exactly equal to the cutoff be handled?**

The displayed convention includes blocks exactly equal to the cutoff, and that convention should be documented and applied to stored grades rather than rounded display values.

**What happens if no blocks qualify at a cutoff?**

If no blocks qualify, retained tonnage and contained oxide are zero, while retained mean grade is undefined.

**What contained TREO does Mt Weld’s historical total imply using rounded Lynas inputs?**

Using those rounded Lynas inputs, multiplying resource tonnes by the TREO mass fraction gives approximately 2.99 million tonnes of contained TREO.

**What cutoff did Songwe Hill use for its Indicated and Inferred resources?**

Both were reported using a 1.0% TREO cutoff.

## Quick answers

| Why does a deposit-wide average above 2% not establish qualifying tonnes? | A deposit-wide mean above 2% does not reveal how much material exceeds that cutoff; the missing evidence is the grade distribution. |
| --- | --- |
| What should domain-level grade–tonnage curves show? | Request grade–tonnage curves for each geological estimation domain, with cutoff on the horizontal axis and qualifying tonnage and retained average grade on the vertical axes. |
| What happens when lower-grade blocks are excluded? | Orebit’s explanation describes the trade-off: excluding lower-grade blocks reduces tonnes while increasing the average grade of retained material. |
| Which model issues can distort the response at a 2% cutoff? | Orebit’s validation guidance flags block-size mismatch with mining selectivity, conditional bias, and high-grade smearing; matching a global mean does not establish correct local grades. |
| When should the proposed 2% TREO cutoff remain unvalidated? | Treat the 2% TREO cutoff as unvalidated until that deliverable quantifies retained tonnes, contained oxides, and estimation uncertainty. |

Also worth reading: **Essential predictive modeling types for smarter mineral resource management**: [Essential predictive modeling types for](https://skymineral.com/blog/essential-predictive-modeling-types-for-smarter-mineral-resource-management.php) · **Innovative artificial intelligence tools for modern mineral resource discovery**: [Innovative artificial intelligence tools for](https://skymineral.com/blog/innovative-artificial-intelligence-tools-for-modern-mineral-resource-discovery.php) · **How the rising legal tide for global climate justice is transforming the future of resource accountability**: [How the rising legal tide](https://skymineral.com/blog/how-the-rising-legal-tide-for-global-climate-justice-is-transforming-the-future-of-resource-accountability.php)

### Related reading

- [Unlocking Earth's Secrets Advanced Techniques for Finding Deep Mineral Deposits](https://skymineral.com/blog/unlocking-earths-secrets-advanced-techniques-for-finding-deep-mineral-deposits.php)
- [How satellite imaging helps professionals scout for rare mineral deposits](https://skymineral.com/blog/how-satellite-imaging-helps-professionals-scout-for-rare-mineral-deposits.php)
- [China's 2026 Dy Export Cuts Raise US Monazite Cutoff Grade](https://skymineral.com/blog/chinas-2026-dy-export-cuts-raise-us-monazite-cutoff-grade.php)
- [Grade Variability Challenges Ion-Clay REE Cutoff and Reporting](https://skymineral.com/blog/grade-variability-challenges-ion-clay-ree-cutoff-and-reporting.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)
- [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

- [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)
- [Rare earth drilling results: 19% Root Mean Square Error (RMSE) vs 31% overcall...](https://skymineral.com/blog/rare-earth-drilling-results-19-root-mean-square-error-rmse-vs-31-overcall-2026.php)
- [Rare earth drilling spacing: 50m vs 100m, 1% Total Rare Earth Oxides (TREO) 2026](https://skymineral.com/blog/rare-earth-drilling-spacing-50m-vs-100m-1-total-rare-earth-oxides-treo-2026.php)

Canonical: https://skymineral.com/blog/rare-earth-deposits-test-a-2-cutoff-with-domain-level-gradetonnage-curves.php
Markdown: https://skymineral.com/blog/rare-earth-deposits-test-a-2-cutoff-with-domain-level-gradetonnage-curves.php/index.md