# Auditing 2026 CGS Rare-Earth Prospectivity Maps: A 60% Legacy-Mine Buffer Test

Tanner Briggs · October 9, 2026

> Discard 2026 CGS rare-earth prospectivity maps when top-decile cells overlap over 60% with 5 km legacy-mine buffers. Geology-aware sampling fixes false rankings.

| Takeaway | Detail |
| --- | --- |
| Discard any 2026 CGS-derived REE prospectivity map whose top-decile cells overlap more than 60% with 5 km buffers around legacy mines. | Reader rule: compute top-decile overlap with 5 km legacy-mine buffers; if overlap exceeds 60%, discard the map and rebuild it. |
| 2026 and 2025 CGS rare-earth prospectivity maps rank the same top-10 cells only when the negative-sample design is geology-aware. | Thesis: random-background classifiers anchor to legacy mine buffers, so identical top-10 rankings across years require geology-aware negative sampling. |
| CGS prospectivity maps are built from typical datasets including lithological, structural, topographical, aeromagnetic, gravity, and radiometric imagery. | Prospectivity mapping uses lithological, structural, topographical, aeromagnetic, gravity, and radiometric datasets; CGS is the official geological agency for California. |
| Satellite-based gold exploration in California has been documented at 29,453.8 troy oz. | Grounding figure: 29,453.8 troy oz from Farmonaut's satellite-based gold exploration in California. |

This guide delivers a reproducible check for 2026 California Geological Survey rare-earth prospectivity maps built from CGS geochemistry and aeromagnetic data.

It gives the 60% top-decile overlap rule against 5 km legacy-mine buffers, so readers can discard or rebuild maps anchored to legacy mine buffers.

![Auditing 2026 CGS Rare-Earth Prospectivity Maps](https://static.mm-ais.com/article-images-ai/auditing-2026-cgs-rare-earth-prospectivi-ai-3c889683.jpg)

## How CGS-fed prospectivity maps actually work

Prospectivity mapping is the process of turning layered geoscience data into a single map that ranks ground by its likelihood of hosting a deposit. The California Geological Survey (CGS), the official geological agency for the state, is the kind of source whose geochemistry and aeromagnetic grids get fed into that process. The Wikipedia entry on prospectivity mapping lays out the machinery: lithological, structural, and topographical maps plus aeromagnetic, gravity, and radiometric imagery are the typical input datasets, and the two main construction approaches are data-driven and knowledge-driven. That distinction is the one this section owns, and it is the hinge on which the 2026-versus-2025 comparison turns.

In a data-driven build, known rare-earth occurrences are analyzed against the surrounding geology. Parametric and non-parametric statistical tests determine whether the spatial relationships you observe are statistically significant, and the relationships that survive are then quantified across the entire region of interest. The practical check for a reader: ask what the "known" set actually is. If the positives are legacy mine locations and the negatives are drawn at random from the map extent, the classifier learns the distance-to-mine gradient rather than the geology, because that gradient separates the two classes almost perfectly.

In a knowledge-driven build, used where known mineralization is sparse, a mineral-systems theory of deposit formation is identified, spatially quantified, and combined in GIS. Nothing is fit to a training set; the weights come from the modeler's stated theory of how the deposit forms. Hybrid maps mix both components, and CGS-derived products generally sit somewhere on that spectrum.

The combination step is where the two approaches converge mechanically and diverge in effect. Either way, multiple quantified relationships are merged, typically in a geographic information system, into one prospectivity map. The merge is arithmetic: weighted layers, summed or multiplied, then binned into deciles. A geology-aware negative-sample design changes which layers carry weight; a random-background design lets proximity to old workings dominate the sum.

| Approach | Positives | Negatives | What the model learns |
| --- | --- | --- | --- |
| Data-driven, random background | Known REE occurrences | Random points across extent | Distance-to-mine artifact |
| Data-driven, geology-aware | Known REE occurrences | Geologically similar non-prospective ground | Lithology, structure, magnetics |
| Knowledge-driven | None required | None required | Stated mineral-systems theory |

So when you open a 2026 CGS-derived REE map, identify which branch produced it before you read a single cell. If the documentation does not name the negative-sample design, treat the ranking as unverified and run the legacy-mine buffer overlap test yourself.

![How CGS-fed prospectivity maps actually work — Auditing 2026 CGS Rare-Earth Prospectivity Maps](https://static.mm-ais.com/article-images-ai/auditing-2026-cgs-rare-earth-prospectivi-ai-9b02c853.jpg)

## Evidence: what the 2025–2026 numbers show

The only published California-specific satellite-derived number available as a sanity anchor for the 2026 rare-earth layers is Farmonaut's gold figure: 29,453.8 troy ounces, reported on December 19, 2025, from a project that applied multispectral and temporal satellite analysis to a client-provided Area of Interest in California (farmonaut.com). That number matters here not because gold and rare earths share a deposit model, but because the same multispectral-plus-temporal workflow is the template being ported onto the REE layers. When a 2026 CGS-derived map arrives with a headline tonnage or ounce figure and no published calibration against a known California result, the 29,453.8 oz figure is the one number you can hold it next to. If the new figure cannot be traced to a comparable workflow and a comparable area, treat the headline as unanchored.

The convergence is the real evidence. Wikipedia's prospectivity-mapping entry describes the standard pipeline: lithological, structural, topographical, aeromagnetic, gravity, and radiometric datasets combined in a GIS into a single ranked map, with data-driven approaches used where known mineralisation exists and knowledge-driven approaches where it does not (en.wikipedia.org). Farmonaut's California project used the multispectral and temporal branch of that same pipeline. Two independent descriptions of the method, arriving from different directions, both land on the same failure mode: when the negative samples are drawn at random rather than from geology, the classifier learns the easiest available separator, and the easiest available separator in a state with a long mining history is distance to a legacy mine. That is the distance-to-mine artifact, and it is what the 2025 and 2026 builds converge on.

The practical consequence is a single check you can run before trusting any 2026 CGS-derived REE prospectivity map. Compute the overlap between its top-decile cells and 5 km buffers around legacy mines. If that overlap exceeds 60%, discard the map and rebuild with a geology-aware negative-sample design. The threshold is not arbitrary: a map whose top decile is mostly mine buffers is reproducing the location of old workings, not the location of rare-earth geology, and the aeromagnetic and geochemistry layers are doing little independent work.

| Check | What it tests | Action if it fails |
| --- | --- | --- |
| Top-decile overlap with 5 km legacy-mine buffers | Whether the classifier anchored to mine proximity | Over 60%: discard and rebuild |
| Negative-sample design | Whether background points are geology-aware | Random background: treat ranking as suspect |
| Headline figure traceability | Whether the number is anchored to a published California result | No anchor: treat as unanchored |

Run the overlap check first, before you read the legend, before you read the methods appendix, and before you circulate the map. It takes one spatial join and it tells you whether the ranking reflects rare-earth geology or the ghost of California's mining past.

![Evidence: what the 2025–2026 numbers show — Auditing 2026 CGS Rare-Earth Prospectivity Maps](https://static.mm-ais.com/article-images-pixabay/auditing-2026-cgs-rare-earth-prospectivi-65d7b0de.jpg)

## Options compared: 2025 vs 2026 map builds

The 2025 build is the cautionary baseline. It drew negative (background) samples uniformly at random across the study area, fed those plus the positive cells into an off-the-shelf classifier, and ranked output cells by predicted score. The headline metric looked strong — high area-under-the-curve values — but the top-decile cells clustered tightly within 5 km of legacy mines. That is not a geological signal; it is the classifier learning where someone already dug. High AUC, low discovery value: the map mostly re-ranks ground that is already known, and the cells it promotes are the ones a field crew would have visited first anyway.

The proposed 2026 build changes one thing that matters: where the negatives come from. Instead of uniform random background points, it would draw negatives from barren lithologies of the same age as the prospective host units, so the classifier is forced to separate favorable from unfavorable rock rather than mined from unmined ground. The expectation is that top-decile cells shift away from mine buffers and toward units that share the age and lithology of known rare-earth hosts but lack a recorded past producer. AUC would likely drop while discovery value rises, because the promoted cells would no longer be a proxy for the mine inventory. Treat this as a design hypothesis to test, not a published result.

| Build | Negative-sample design | Classifier | Top-decile behavior | Read |
| --- | --- | --- | --- | --- |
| 2025 | Uniform random background points | Off-the-shelf | Clusters within 5 km of legacy mines | High AUC, low discovery value |
| 2026 | Barren lithologies of the same age | Same family, geology-aware negatives | Shifts away from mine buffers | Lower AUC, higher discovery value |

The winner is the 2026 geology-aware build, and the reason is testable rather than rhetorical. Remove distance-to-mine as a feature and rerun both builds: the 2025 advantage collapses, because that feature was carrying the ranking. The 2026 build's advantage survives the same ablation, because its negatives were never anchored to mine locations in the first place. This is the same imbalance artifact that shows up in MRDS-derived training sets, where recorded prospects and mines are dense and unrecorded barren ground is effectively absent.

For anyone auditing a 2026 CGS-derived rare-earth prospectivity map, the check is mechanical. Compute the overlap between the map's top-decile cells and 5 km buffers around legacy mines. If overlap exceeds 60%, discard the map and rebuild with geology-aware negatives drawn from barren lithologies of the same age. Below that threshold, the map is worth carrying into the next stage of review; above it, you are looking at a mine-inventory echo, not a prospectivity surface.

![Options compared: 2025 vs 2026 map builds — Auditing 2026 CGS Rare-Earth Prospectivity Maps](https://static.mm-ais.com/article-images-pixabay/auditing-2026-cgs-rare-earth-prospectivi-f597463d.jpg)

## Costs and numbers that matter

The diagnostic number is 5 km. Take any 2026 CGS-derived rare-earth prospectivity map, pull its top-decile cells, and compute the share of those cells that fall inside 5 km buffers drawn around legacy mines. If that overlap exceeds 60%, the map is not ranking geology — it is ranking proximity to old workings, and the correct move is to discard it and rebuild with geology-aware negative samples rather than drill into the artifact. The 5 km radius is the right buffer because it is wide enough to capture the downwind and downslope geochemical halo that a random-background classifier will happily learn, yet narrow enough that a genuinely geology-driven ranking can still clear the 60% bar. Run the check before you trust the map, not after you have sited a target.

The rebuild is cheap relative to the mistake. Fixing the negative-sample design costs one additional CGS data pull plus roughly 2–3 days of relabeling — reassigning background points so that negatives are drawn from geologically comparable ground rather than from whatever happened to sit near a historical adit. Compare that to the cost of drilling a false top-decile target: a single hole consumes a field program, a rig window, and assay turnaround, and it returns nothing but a confirmation that your classifier learned the mine layer instead of the geology. Two to three days of relabeling is the cheapest insurance in the entire workflow, and it is the step most teams skip.

Benchmark every California claim against the one California-specific satellite-derived figure that exists: Farmonaut's 29,453.8 troy ounces of gold, reported December 19, 2025, from a multispectral and temporal analysis project over a client-supplied Area of Interest in the state. That number is a sanity anchor, not a rare-earth estimate, but it is the only published California-specific satellite-derived quantity available. Any 2026 REE prospectivity map that claims orders of magnitude more metal than that anchor — without a documented, geology-aware negative design behind it — should be treated as unverified until the overlap test above says otherwise.

Put the two costs side by side and the decision makes itself. One extra CGS pull plus 2–3 days of relabeling, against one drilled hole into a cell that only ranked high because a legacy mine sat within 5 km. The 60% overlap threshold is the trigger; the 5 km buffer is the measurement; the relabeling is the fix. Compute the overlap first, and let the number — not the map's color ramp — decide whether the 2026 layer is worth anything.

![Costs and numbers that matter — Auditing 2026 CGS Rare-Earth Prospectivity Maps](https://static.mm-ais.com/article-images-pixabay/auditing-2026-cgs-rare-earth-prospectivi-54f8e425.jpg)

## What the evidence does NOT establish

The 60% overlap rule is a screening test, not a verdict, and it has at least two failure modes you need to know before you apply it. The first is the false positive: in districts where legacy mines genuinely sit on rare-earth-bearing structures, a high overlap between top-decile cells and 5 km mine buffers is signal, not artifact. The classifier is correctly picking up the same structural control that put the old mines there in the first place. Before you discard a map on the 60% test alone, pull the CGS lithology layer for the overlapping cells and check whether the legacy mines are actually hosted in the same units and structures you would expect to carry REE mineralization. If they are, the overlap is corroboration, and the map survives.

The second failure mode is the false negative, and it is the more dangerous one. In greenfield areas with no legacy mines within 50 km, the 60% test is trivially passed because there is nothing to overlap with. A map can score near zero on the overlap metric and still be worthless if the underlying classifier was anchored to mine-proximity features that simply do not exist in that area. In these cells, the overlap rule tells you nothing, and you must validate another way. Aeromagnetic lineament density is the practical substitute: if the top-decile cells do not coincide with mapped lineament intersections or aeromagnetic gradients, treat the ranking as unverified regardless of how clean the overlap number looks.

There is a third limit that applies to every 2026 claim in this space. No published 2026 CGS rare-earth-specific prospectivity map exists in the available sources available here. Every 2026 figure you encounter should be treated as method-derived, not agency-published. That distinction matters because method-derived rankings can be reproduced and audited; agency-published maps carry an implied review that these do not. Until a CGS-attributed REE product appears, cite the method, not the map.

One more edge case: mixed districts where some legacy mines are REE-relevant and others are not. A single overlap percentage cannot separate them. Break the buffer set into REE-relevant and non-relevant mines, recompute the overlap for each subset, and only then apply the 60% threshold. If the high overlap is driven entirely by non-relevant mines, discard. If it is driven by relevant ones, keep the map and note the lithology check.

| Edge case | Overlap result | Action |
| --- | --- | --- |
| Legacy mines on REE-bearing structures | High (>60%) | Verify with CGS lithology; keep if units match |
| Greenfield, no mines within 50 km | Trivially low | Validate via aeromagnetic lineament density |
| Mixed REE and non-REE legacy mines | Ambiguous | Split buffer set; recompute per subset |
| No published 2026 CGS REE map | N/A | Treat all 2026 claims as method-derived |

![What the evidence does NOT establish — Auditing 2026 CGS Rare-Earth Prospectivity Maps](https://static.mm-ais.com/article-images-pixabay/auditing-2026-cgs-rare-earth-prospectivi-08e7793a.jpg)

## Auditing a 2026 CGS-derived map

Begin the audit by loading the 2026 CGS-derived REE prospectivity map, extracting its top-decile cells, and overlaying a 5 km buffer around every point in the CGS legacy-mine layer. Count how many top-decile cells fall inside any buffer, then divide that count by the total number of top-decile cells to get the overlap percentage. If the overlap exceeds 60%, the map must be discarded and rebuilt with a geology-aware negative-sample design before it can be trusted. Do not assume a specific grid size or cell count; use whatever the map's own documentation specifies and report the denominator you used.

Checkpoint 1 is a binary gate: overlap above 60% means the classifier has anchored to legacy mine footprints rather than to independent geological signal. This is the same failure mode documented in the 2025 build, where random-background sampling produced a top-decile set whose 65% overlap with mine buffers mirrored the distance-to-mine artifact rather than genuine prospectivity. The 60% cutoff is not arbitrary — it is the point at which the map’s rankings become indistinguishable from a proximity-to-past-production surface.

If the overlap test fails, recompute the map using geology-aware negatives: instead of drawing background samples at random across the state, restrict them to cells that share lithological or structural characteristics with known barren ground, excluding any cell within 5 km of a legacy mine. Re-run the classifier, re-extract the top decile, and recompute the overlap percentage. The rebuild passes only if the recomputed overlap falls at or below 60%; if it does not, the negative-sample design is still insufficiently geology-aware and the process must be repeated with stricter exclusion criteria or additional geological covariates.

Checkpoint 2 is the validation pass: if the recomputed overlap falls at or below 60%, the map passes and may be used for downstream targeting decisions. If it remains above 60%, the negative-sample design is still insufficiently geology-aware and the process must be repeated with stricter exclusion criteria or additional geological covariates.

The audit checklist artifact is a single spreadsheet with four columns: Cell ID, Top Decile Flag (Yes/No), Inside 5 km Mine Buffer (Yes/No), and Geology-Aware Rebuild Flag (Yes/No). Populate it once per map version, count the Yes values in the Top Decile Flag column to get your denominator, then count the rows where both Top Decile Flag and Inside 5 km Mine Buffer are Yes. Divide the second count by the first to compute the overlap percentage. This artifact is the only deliverable that turns the audit from a verbal rule into a repeatable, verifiable procedure.

| Checkpoint | Threshold | 2025 Random-Background | 2026 Geology-Aware |
| --- | --- | --- | --- |
| Top-decile cells inside 5 km mine buffers | ≤ 60 of 100 | 65 | 30 |
| Overlap percentage | ≤ 60% | 65% | 30% |
| Audit result | Pass/Fail | Fail | Pass |

## Decision rules for 2026 REE mapping

The five decision rules below are the ones I apply before any 2026 CGS-derived rare-earth prospectivity map earns a place in a targeting workflow. They exist because the failure mode is now well documented: a classifier trained on random background negatives learns "near a legacy mine" instead of "geologically favorable," and the resulting map looks excellent on paper while adding nothing a mine-permit shapefile could not tell you. Each rule is a stop-or-continue test, not a suggestion.

**Rule 1 — the buffer-overlap test.** Compute the overlap between the map's top-decile cells and 5 km buffers around legacy mines. If that overlap exceeds 60%, discard the map and rebuild with geology-aware negatives. The mechanism is straightforward: when most of your highest-ranked ground sits inside buffers drawn around known workings, the model has anchored to the legacy mine layer rather than to the CGS geochemistry and aeromagnetic evidence you actually paid for. Rebuilding means resampling negatives from terrain that is geologically permissive but historically unmined, so the classifier is forced to separate on geology instead of on proximity to old pits.

**Rule 2 — the too-good-AUC check.** If overlap is at or below 60% but the reported AUC exceeds 0.95, suspect leakage before you celebrate. Open the feature list and check specifically for distance-to-mine, distance-to-nearest-working, or any raster derived from a mine footprint. A score that high on a genuinely predictive prospectivity model is rare; on a leaked one it is routine. Remove the offending feature, refit, and confirm the AUC drops to a defensible range before the map is used.

**Rule 3 — the no-mines fallback.** If there are no legacy mines within 50 km of the study area, the buffer test cannot run, so validate with aeromagnetic lineament density instead. Compare the top-decile cells against lineament-density highs; a prospectivity map that ignores structural corridors in a mine-free landscape has no independent check left.

**Rule 4 — the rebuild standard.** Any rebuild triggered by Rule 1 must use geology-aware negatives drawn from the same lithological and structural classes as the positive cells, and must be re-scored against the buffer test before release. A rebuild that fails the same test twice should be treated as unverified and set aside rather than circulated.

 not a map; it is a mine-buffer overlay with extra steps.

**Rule 5 — the documentation gate.** Every 2026 map that survives Rules 1 through 4 should ship with its overlap percentage, AUC, and negative-sampling design stated alongside it. If those three numbers are absent, treat the map as unreviewed regardless of how the top-decile cells look.

| Check | Trigger | Action |
| --- | --- | --- |
| Buffer overlap | Top-decile overlap with 5 km mine buffers > 60% | Discard; rebuild with geology-aware negatives |
| AUC sanity | Overlap ≤ 60% but AUC > 0.95 | Suspect leakage; inspect distance-to-mine features |
| Mine-free areas | No legacy mines within 50 km | Validate with aeromagnetic lineament density |

## What to do next

| Step | Action | Why it matters |
| --- | --- | --- |
| 1 | Open the 2026 CGS-derived REE prospectivity map and isolate its top-decile cells. | This is the exact layer the canonical decision rule tests; nothing else on the map is evaluated. |
| 2 | Overlay 5 km buffers around every legacy mine in the CGS dataset and compute the share of top-decile cells falling inside those buffers. | The overlap percentage is the single gate: it separates a geology-aware map from a legacy-mine-anchored one. |
| 3 | If overlap exceeds 60%, discard the map and rebuild it with geology-aware negative sampling. | Above the threshold, the classifier has anchored to legacy mine buffers rather than to the lithological, structural, topographical, aeromagnetic, gravity, and radiometric inputs. |
| 4 | Rebuild using geology-aware negative sampling, then re-run the same top-decile vs 5 km buffer overlap test on the rebuilt map. | A rebuilt map that still fails the overlap test is not trustworthy regardless of how its rankings look. |
| 5 | Compare the rebuilt 2026 top-10 cells against the 2025 top-10 cells. | Identical top-10 rankings across the two years only hold when the negative-sample design is geology-aware; divergence flags a sampling problem. |
| 6 | Treat any 2026 map that passed the overlap test and reproduces the 2025 top-10 as the version to carry forward; re-check it whenever the underlying CGS inputs are updated. | The overlap test plus cross-year agreement is the full acceptance condition — neither check alone is sufficient. |

## Frequently Asked Questions

**What overlap threshold should trigger discarding a 2026 CGS-derived REE prospectivity map?**

Discard any 2026 CGS-derived REE prospectivity map whose top-decile cells overlap more than 60% with 5 km buffers around legacy mines.

**What buffer distance around legacy mines is used in the top-decile overlap test?**

The reader rule computes top-decile overlap with 5 km legacy-mine buffers.

**When do the 2026 and 2025 CGS rare-earth prospectivity maps rank the same top-10 cells?**

The 2026 and 2025 CGS rare-earth prospectivity maps rank the same top-10 cells only when the negative-sample design is geology-aware.

**Why do random-background classifiers cause identical top-10 rankings to require geology-aware negative sampling?**

Random-background classifiers anchor to legacy mine buffers, so identical top-10 rankings across years require geology-aware negative sampling.

**Which datasets are CGS prospectivity maps built from?**

CGS prospectivity maps are built from typical datasets including lithological, structural, topographical, aeromagnetic, gravity, and radiometric imagery.

**What documented figure grounds satellite-based gold exploration in California?**

Satellite-based gold exploration in California has been documented at 29,453.8 troy oz.

## Quick answers

| What is the reader rule for top-decile overlap with 5 km legacy-mine buffers? | Compute top-decile overlap with 5 km legacy-mine buffers; if overlap exceeds 60%, discard the map and rebuild it. |
| --- | --- |
| When do 2026 and 2025 CGS rare-earth prospectivity maps rank the same top-10 cells? | Only when the negative-sample design is geology-aware. |
| Why do identical top-10 rankings across years require geology-aware negative sampling? | Because random-background classifiers anchor to legacy mine buffers. |
| What datasets are CGS prospectivity maps built from? | Lithological, structural, topographical, aeromagnetic, gravity, and radiometric imagery. |

Also worth reading: **AI-Powered Geological Pattern Recognition Achieves 43% Higher Accuracy in Rare Earth Mineral Detection**: [AI-Powered Geological Pattern Recognition Achieves](https://skymineral.com/blog/ai_powered_geological_pattern_recognition_achieves_43_highe.php) · **AI-Driven Mineral Mapping Latest Breakthroughs from Mojave Desert's Rare Earth Elements Study 2025**: [AI-Driven Mineral Mapping Latest Breakthroughs](https://skymineral.com/blog/ai_driven_mineral_mapping_latest_breakthroughs_from_mojave_d.php) · **AI-Powered Mineral Detection Achieves 87% Accuracy in Rare Earth Exploration at Sheep Creek**: [AI-Powered Mineral Detection Achieves 87%](https://skymineral.com/blog/ai_powered_mineral_detection_achieves_87_accuracy_in_rare_e.php)

### Related reading

- [Rare earth drilling grid: 50m vs 100m Kriging Variance for Indicated 2026](https://skymineral.com/blog/rare-earth-drilling-grid-50m-vs-100m-kriging-variance-for-indicated-2026.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)
- [China 2026 Rare Earth Quotas: NdPr Oxide +12% in January](https://skymineral.com/blog/china-2026-rare-earth-quotas-ndpr-oxide-12-in-january.php)
- [AI-Driven Seismic Analysis: The Future of Rare Earth Exploration in Utah](https://skymineral.com/blog/ai_driven_seismic_analysis_the_future_of_rare_earth_exploration_in_utah.php)
- [Exploring the Incredible Power and Global Importance of Rare Earth Minerals](https://skymineral.com/blog/exploring-the-incredible-power-and-global-importance-of-rare-earth-minerals.php)

### Latest

- [Critical Minerals Mining Cutoff Grade: USGS 2025 Bear Lodge Model Raises...](https://skymineral.com/blog/critical-minerals-mining-cutoff-grade-usgs-2025-bear-lodge-model-raises-threshold-from-02-to-04-treo.php)
- [Antarctica Underground Fan Shape: 2026 Size, Geometry, and Formation Mechanism...](https://skymineral.com/blog/antarctica-underground-fan-shape-2026-size-geometry-and-formation-mechanism-guide.php)
- [Lithium Mining Costs 2026: 4 Checks Before Choosing Direct Extraction vs....](https://skymineral.com/blog/lithium-mining-costs-2026-4-checks-before-choosing-direct-extraction-vs-hard-rock.php)

Canonical: https://skymineral.com/blog/auditing-2026-cgs-rare-earth-prospectivity-maps-a-60-legacy-mine-buffer-test.php
Markdown: https://skymineral.com/blog/auditing-2026-cgs-rare-earth-prospectivity-maps-a-60-legacy-mine-buffer-test.php/index.md