From licence polygon to ranked drill targets — in nine days.
This walkthrough uses a fully synthetic Cu-Co exploration brief over a fictional 184 km² licence. Every figure on this page was regenerated client-side from the same deterministic library that powers the homepage demos. No client data appears here.
Define the AOI
The client supplied a single licence polygon. We buffered it by 500 m to capture edge effects and intersected with the Sentinel-2 tile grid — eleven cloud-filtered L2A scenes from the previous nine months passed acceptance.
Compute spectral indices
Iron-oxide and Al-OH (clay) indices were the leading hypotheses given the regional analogues. Both rasters below are computed from the same synthetic tile via the band ratios shown.
Add structure and drainage
Sobel-filtered hillshade plus a Hough transform extracts dominant lineaments. Drainage is derived from a 30 m DEM. Both are routinely the strongest predictor of where mineralised systems daylight.
Stack, train, rank
32 co-registered feature channels were fed to an XGBoost ranker, calibrated on regional analogue licences (leave-one-out). The composite below is the final prospectivity surface; pixels exceeding the 60th percentile trigger a candidate target.
Outcome
Of the eight ranked targets, three sit in Rank A. The client trimmed a planned 28-hole programme to a phased 12-hole campaign on the A and top-B targets — projected drill saving in the synthetic exercise: ~USD 410k before any discovery upside.