Abstract
This study presents a multivariate framework for geochemical data processing and anomaly detection to support mineral exploration in the Hattfjelldal area, Norway. The workflow integrates data levelling, multivariate analysis, and spatial evaluation to improve the detection and interpretation of geochemical anomalies associated with volcanogenic massive sulfide (VMS) mineralization. Soil geochemical and magnetic susceptibility data were log-transformed and subsequently levelled using Z-score normalization by soil type and lithology. Both linear (principal component analysis, PCA) and non-linear algorithms (hierarchical clustering, isolation forest, and angle-based outlier detection) were applied to construct anomaly detection vectors. Hierarchical clustering proved particularly effective in defining element assemblages that refine anomaly detection, including associations of Type 1 (Ag, Mo, S, Sb, Bi, Pb); Type 2 A (Fe, Zn, Co, Mn) and Type 2B (Fe, Zn, Co, Mn, As, Cu). These groupings provide a robust geochemical and geological context within established VMS zoning models.
Magnetic susceptibility, although less reliable as a stand-alone exploration vector, enhances interpretation when integrated with geochemical anomalies. Fractal analysis applied to both, geochemical vectors and magnetic susceptibility data effectively distinguished background from anomalous values, delineating areas of potential economic interest. Spatial Feature Embeddings (SFE), derived from clustering radiometric, topographic, and spectral datasets, further improved the spatial characterization of anomalies. When combined with airborne magnetics, SFE enabled the refinement and prioritization of specific targets within broad anomaly zones.
Overall, this framework demonstrates the value of integrating statistical, geochemical, and geophysical methods within their spatial context, providing a transferable approach for exploration programs in Arctic environments.