Zonificación geoquímica diferencial de dominios asociados a Tierras Raras Ligeras, empleando Lantano (La) y elementos geoquímicos proxy (Th, U, P), mediante Kmeans adaptativo (Mini-Batch) y kriging.

Abstract

Abstract: This research develops a methodology for the geochemical zoning of domains associated with Light Rare Earth Elements, integrating multivariate analysis and geostatistical modeling. These domains were delineated through the joint analysis of Lanthanum as a direct tracer, and Thorium, Uranium, and Phosphorus as proxy elements. A cleaned database of 4,000 geochemical samples with quality control was processed, applying Mini-Batch K-means for unsupervised clustering and Ordinary Kriging for spatial interpolation. Exploratory analysis results revealed log-normal distributions and a strong lithological control, where felsic volcanic units concentrated the highest Lanthanum enrichment factors. The algorithm classified the area into four optimal geochemical domains, optimizing the balance between statistical cohesion and geological coherence. The interpolation maps revealed discrete, high-intensity anomalies, validating the spatial co-occurrence of the proxy elements. It is concluded that the combination of Mini-Batch K-means and Kriging constitutes an efficient, reproducible, and objective tool for identifying Light Rare Earth Elements exploration targets, effectively overcoming the limitations of traditional methods based on absolute concentration thresholds.