LodeCraft insight · Data readiness
How to prepare exploration data for an independent evidence review.
A defensible review starts before modelling: with clear provenance, coordinate systems, coverage, QA/QC, transformations and known gaps.
Publication details & scope
Direct answer
Prepare the evidence so another person can understand what each file is, where it came from and what happened to it.#
An independent review should not begin with a folder of unexplained exports. The handover needs enough context to distinguish recorded observations from interpretations, original data from derived products and missing coverage from a genuinely quiet response.
1. Start with the decision—not the software.#
State the geological question, decision area, commodity context and intended use of the review. “Find targets” is too broad unless the team also defines what type of target, at what scale, for which stage of exploration and what decision will follow.
The same dataset may be adequate for regional screening and inadequate for prospect-scale ranking. That boundary should be explicit before a model is selected.
2. Provide an inventory with stable identities.#
For every supplied dataset, record its owner or publisher, product name, version or acquisition date, licence or permission basis, coordinate reference system, spatial extent, resolution or sample support and the file actually supplied.
Persistent identifiers and lineage metadata make later review materially easier. Geoscience Australia’s FAIR-data guidance describes the value of persistent identifiers, ISO-aligned metadata, lineage, legal constraints and source information.
3. Separate original, cleaned and interpreted material.#
Do not overwrite the original files. Keep a read-only source layer, a documented processing layer and a clearly labelled interpretation layer. Record unit conversions, reprojection, resampling, compositing, censoring, imputation, clipping and any category mapping.
If several source products are aligned onto one computational grid, that common grid is an analytical convenience—not proof that the source observations share the same support or accuracy.
4. Carry QA/QC and detection context with the values.#
Geochemical values need analytical method, units, detection limits, below-detection handling, standards, blanks, duplicates and relevant batch context. Drilling needs hole identities, collar and survey controls, interval rules, recovery context and a traceable connection between geology and assays.
Geophysics and remote-sensing products need survey or product identity, acquisition and processing history, native resolution and known artefacts. A clean image without this context can be visually persuasive and analytically ambiguous.
5. Make absence and uneven coverage visible.#
A missing value is not automatically background. Sampling density, access, cover, tenure history, survey design and previous exploration focus can shape where evidence exists. Supply coverage masks, detection support and known access or surface context where relevant.
These layers can remain outside the geological vote while still being displayed as warnings or alternative explanations.
6. Include the current interpretation—but label it.#
Existing targets, geological models and expert interpretations are valuable inputs to a challenge review. They should be identified as interpretations and kept separate from the observations used to support them. This makes it possible to test for confirmation bias rather than silently encoding the preferred answer.
A practical handover checklist#
Use this as your own preparation record. Selections last while this page is open; they do not validate project readiness.
What this preparation does not guarantee#
Good preparation cannot create evidence that was never collected or remove every historical bias. It does allow the review to state what the available data can support, which gaps matter and whether further analysis is responsible.
See the LodeCraft engagement for MindSync’s data-suitability gate and end-to-end deliverables.