Hydrogeochemical Vectoring to Concealed Porphyry Copper Systems Using Unsupervised Machine Learning: A Case Study from Central British Columbia, Canada

Boago Felicity Kukudi

Environments with thick cover present a challenge to surficial exploration efforts the Canadian
Cordillera has extensive Quaternary glacial till and sediment cover, making it difficult to find
hidden ore bodies. An alternative is using hydrogeochemistry as an exploration tool, to detect
anomalies associated with mineralization. This study demonstrates the hydrogeochemical
vectoring to porphyry copper deposit using unsupervised machine learning and a spatial workflow
to explore beneath the cover of British Columbia, Canada. A framework of 3501 stream and
groundwater data points was compiled, cleaned for string-censored limits of detection (LOD) and
evaluated across 12 physicochemical variables simultaneously. An unsupervised K-Means
algorithm successfully separated the groundwater into two clusters: "Fresh Baseline Water"
(n=3314) and "Ore-Interacting Anomaly" water (n=187). The univariate evaluation achieved a
clean separation of 90.67% dissolved Sulfate, confirming sulfate as an effective exploration vector;
the multivariate pipeline resolved the 9.33% overlap zone between background and mineralized
samples. A custom trail gradient index (Cu * Mo)/ (Zn) was deployed was deployed to segment
the anomalies into three distinct trails: Proximal Hypogene Core (n=75) with mean Mo=16.81ppb,
Zn=205.75ppb, a Secondary Reducing Zone (n=21) exhibiting peak copper precipitation; (mean
Cu: 1.26ppb) and a Distal Exploration Halo (n=90) with mean Zn=51.04ppb and elevated
Electrical Conductivity mean= 580.44μS/cm. QGIS mapping show these clusters align with the
major regional fault networks, providing a non-invasive exploration tool.