Comparing 2D and 4D K-Means for Enhancing Interpretation of FDEM Data in Archaeological Prospection

This study evaluates the efficiency of incorporating spatial information into the clustering of Frequency Domain ElectroMagnetic (FDEM) data for archaeological interpretation. Building on previous work that applied K-Means clustering to two-dimensional space defined by the in-phase and quadrature components of the electromagnetic field, this research tests a four-dimensional clustering approach by adding geographic coordinates to the properties set. The goal is to assess whether including spatial information improves the identification of areas of archaeological interest. FDEM data from the Torre Galli site in Calabria, Italy, were analyzed using both 2D (EM components only) and 4D (EM components plus spatial coordinates) clustering, with cluster validity assessed through Silhouette analysis. Results indicate that the 2D clustering approach yields more distinct and archaeologically meaningful clusters, showing better alignment with excavation-validated features compared to 4D approach. The inclusion of spatial coordinates in the clustering process seems distorting the cluster geometry. This reduces the discrimination power of electromagnetic features, leading to a less clear differentiation between natural background and archaeological target areas including roads, walls and metal artifacts. These findings suggest that, for FDEM data interpretation in archaeology, focusing on the intrinsic electromagnetic properties without spatial embedding enhances the accuracy and efficiency of anomaly detection.