Leveraging spatial anomaly detection for mineral exploration

Mineral deposits in exploration geochemistry are often identified by elevated concentrations of specific elements, resulting in an elemental composition that differs from that of nearby samples. Local anomaly detection techniques are particularly well-suited for identifying these contrasts by focusing on spatially varying compositions. Unlike traditional anomaly detection methods, which often neglect spatial context, these approaches combine multivariate analysis with spatial considerations. A cutting-edge local outlier detection method, which utilizes covariance matrices that are locally and robustly estimated, is introduced, and its application to geochemical soil data is demonstrated for mineral exploration, while accounting for the compositional nature of soil samples.