Alkane Resources | Costerfield Gold Mine
Using geochemical vectoring to identify proximity to mineralisation at Costerfield
Alkane Resources, operator of the Costerfield gold–antimony mine in Victoria, needed a reliable method to use routine drill-hole geochemistry to prioritise exploration drilling.
Datarock developed a geochemical vectoring workflow that analyses multivariate assay data to predict whether samples fall within 15 m or 30 m of ore.
Challenge
- Incomplete element coverage made data suitability for modelling uncertain.
- No clear geochemical controls; teams relied on assumption over evidence.
- No scalable way to flag near-ore samples without subjective interpretation.
Solution
- Assay data cleaned, standardised, and filtered for spatial coverage.
- Geochemical ratios engineered as inputs: Sb/As, Mn/Zn, Fe/Mg, As/Fe, Ba/K.
- Two gradient-boosted models predict proximity within 15 m and 30 m of ore.
- Validated via leave-one-hole-out testing with a 40 m exclusion buffer.
Result
- 15 m model: 69% accuracy. 30 m model: 60% accuracy.
- Arsenic drives the near-ore signal; barium defines the 15–30 m halo.
- Spatial outputs flag where predictions diverge from geological interpretation.
- Alkane can screen new assay batches and identify targets earlier.
“Multi-element geochemistry contains strong signals about proximity to mineralization, but those signals can be difficult to interpret consistently across large datasets. This approach helped translate complex assay data into a clear and practical tool for identifying samples that may be close to ore.”
Braden VerityGeology Manager, Alkane Resources
