CoreElement.AIBlog › Soviet ГКЗ to JORC
METHODOLOGY

Soviet ГКЗ to JORC: a practical conversion guide

May 6, 2026 · Daniel Tonkopiy · 14 min read

Geologists working in Kazakhstan, Russia, Uzbekistan and other former Soviet republics inherit a peculiar problem: vast archives of high-quality exploration data, classified under a Soviet-era resource categorisation system that no Western exchange will accept. Re-classifying that data into JORC, NI 43-101 or KAZRC is straightforward in principle and treacherous in practice. This is what we have learned from 4,859 such reports digitised across CoreElement.AI deployments.

The Soviet system in one paragraph

The USSR State Committee on Mineral Reserves (Государственный комитет по запасам, "ГКЗ" / "GKZ") classified deposits on two orthogonal axes. The geological-confidence axis ran A → B → C₁ → C₂ → P₁ → P₂ → P₃, with confidence decreasing as the letter increases (and prediction levels P₁/P₂/P₃ being below "resource" grade by modern standards). The economic-feasibility axis separated balance reserves (балансовые) from off-balance reserves (забалансовые) and predicted resources (прогнозные ресурсы). Crucially, these axes are independent: balance reserves of category C₁ might map to a Mineral Resource or a Mineral Reserve under JORC, depending on whether a modern feasibility study is attached.

Geological-confidence axis: the conversion table

GKZ categoryCyrillicDrill spacingJORC equivalentNotes
AА≤25 mMeasuredMining-grade detail. Rare outside producing mines.
BБ~50 mMeasuredProduction-planning grade.
C₁С₁100–200 mIndicated (default) or Measured (with strong continuity evidence)The workhorse category. See traps below.
C₂С₂≥200 mInferredDirect equivalent.
P₁П₁Predicted, near-depositNot reportable as ResourceExploration target only.
P₂П₂Predicted, district-scaleNot reportable as ResourceExploration target only.
P₃П₃Predicted, regionalNot reportable as ResourceExploration target only.

The four traps

1. C₁ is not always Measured

The most common mistake we have seen in early Kazakhstan-to-JORC translations was treating C₁ as a wholesale equivalent of "Measured." Under JORC 2012, Measured requires not just tight drill spacing but documented sampling-method appropriateness, demonstrated geological continuity, and QA/QC evidence at the level expected by an experienced Competent Person. Soviet C₁ studies often satisfy the spacing requirement but rarely document QA/QC to the same standard.

The safe default: map C₁ to Indicated. Upgrade to Measured only when you can demonstrate the additional evidence — usually by re-sampling, adding QA/QC standards and blanks, and re-validating continuity at infill spacing.

2. Balance reserves are not always Reserves

Soviet "balance reserves" embed a USSR-era pricing and processing-cost assumption from the year the report was filed. Re-mapping to a JORC Reserve always requires a fresh modifying-factor analysis at current commodity prices, current processing technology, current environmental and social inputs, and current infrastructure costs.

In our deployments roughly two thirds of historical balance reserves survive a current-price modifying-factor study; the other third either upgrade (low-grade tonnage that was uneconomic in 1985 but is mineable today, especially for Cu and U) or downgrade (high-grade tonnage that was economic under USSR planning prices but does not pencil at current operating costs).

3. Density assumptions hide a lot of variance

Soviet reports frequently use a regional bulk-density average rather than deposit-specific measurements. JORC 2012 requires density determination with QA/QC at the deposit level. A 5 % difference in bulk density propagates linearly into tonnage and therefore into contained metal — the difference between a 1.5 Moz deposit and a 1.43 Moz deposit, which materially affects valuation.

If you inherit a Soviet report with regional-average density, schedule deposit-specific density testwork before signing anything that goes to an exchange.

4. Sampling intervals: read the fine print

The Soviet system permitted wider channel-sample intervals for some commodity types and deposit styles — particularly placer gold and certain coal seams. JORC requires explicit demonstration that the sampling method is appropriate for the deposit style. A historical channel-sample dataset that satisfied the Soviet code may still be appropriate for JORC, but the Competent Person must document the justification rather than rely on the legacy approval.

Pulkovo 1942 and the reprojection problem

Almost every Soviet-era report uses the Pulkovo 1942 datum (also known as СК-42, "система координат 1942 года") with the Gauss-Krüger projection on a 6-degree zone system. Modern reporting and modern mining software requires WGS 84 or one of its UTM zones. The conversion looks like a one-liner in any GIS tool — and produces wrong answers if you take the default parameters.

The correct seven-parameter Helmert transformation for Russia and Kazakhstan, registered in the EPSG database:

From: Pulkovo 1942
To:   WGS 84

dx =  +23.92 m
dy = -141.27 m
dz =  -80.91 m
rx =   +0.00 ″
ry =   +0.35 ″
rz =   +0.82 ″
ds =   -0.12 ppm

These values are nationwide; they will land you within a few metres of correct anywhere in Russia or Kazakhstan. For mine-site work — where drillhole collars, surface infrastructure, and historical pit boundaries all need to align with current survey — you need a local Helmert tuned with on-site ground control points. Without it, your historical drillholes can sit 5–10 metres off the modern surface model, and your Competent Person will rightly refuse to sign.

RULE OF THUMB

If your Pulkovo-to-WGS conversion error is >1 m on the deposit, something is wrong. If it is >5 m, something is very wrong. Always verify with at least three known points before trusting the transformation across the dataset.

What CoreElement.AI's Module 19 does

Module 19 of the CoreElement.AI platform handles the conversion automatically, end-to-end:

  1. OCR. PaddleOCR PP-OCRv5 with Russian + Kazakh + technical-symbol vocabulary. Handles scanned reports of varying age and image quality. Recognises hand-corrected typewritten text from the 1960s–1980s alongside cleaner 1990s–2000s reports.
  2. Structured extraction. Recognises A/Б/С₁/С₂/П-categories in tables, narrative text, and figure captions. Parses tonnage and grade rows. Captures cut-off statements, density assumptions, and QA/QC declarations.
  3. Tonnage and grade re-tabulation. Converts to modern units (tonnes, ppm, %) and applies a sanity envelope based on deposit-style averages.
  4. Density flagging. Where density is regional-average, flags the dataset and suggests deposit-specific testwork before any reporting.
  5. Reprojection. Pulkovo 1942 → WGS 84 with site-local Helmert tuning when ground-control points are available.
  6. CRIRSCO mapping. Emits draft Mineral Resource and Mineral Reserve tables under the chosen target standard (JORC, NI 43-101, KAZRC, SAMREC, PERC) with provenance back to source-report page numbers.
  7. Modifying-factor re-evaluation. Re-runs at current commodity prices, current processing technology, and current ESG inputs.
  8. CP / QP review. Every conversion is queued for human Competent Person sign-off before publication. Grant — the AI Geologist — drafts; the Competent Person signs.

As of May 2026 the platform has digitised 4,859 documents from this corpus across deployments in Kazakhstan and Russia. The catalogue grows by 50–200 reports per week.

Why this matters

The CIS region holds tens of thousands of Soviet-era exploration reports. Many describe deposits or exploration targets that have not been touched since the 1980s — economically uninteresting under USSR planning prices, but potentially viable under current commodity prices and modern processing technology. Most of this archive is undigitised, untranslated, and inaccessible to international capital.

Converting it to CRIRSCO-compliant categories is therefore a strategic unlock for explorers, miners and capital allocators alike. It is the single largest pool of "untouched" exploration data on the planet — and the one we have specialised in.

References

Daniel Tonkopiy
CEO and Product Architect, CoreElement.AI. 15+ years building enterprise SaaS and AI/ML systems. Three prior exits. Based in the San Francisco Bay Area.