Análisis de las relaciones geométricas entre la clasificación de alteraciones y las concentraciones de metales (K, Na, S, Fe) usando modelos de regresión Logística Múltiple Condicionada

Abstract

Abstract: The characterization of hydrothermal alterations is a key tool for the exploration of mineralizing systems, as their geochemical and spatial patterns reflect the circulation of hydrothermal fluids. Although elements such as K, Na, S, and Fe have been used as indicators of these processes, their relationships with alteration types are not always uniform or linear, due to the lithological and structural heterogeneity of the rock mass. This study aimed to estimate the probability of the presence of advanced argillic, argillic, and silicic alteration types in an area of ​​Ecuador (location withheld), using a Conditioned Multiple Logistic Regression (CMLR) model. The model incorporated K, Na, S, and Fe as explanatory variables conditioned by lithology and elevation based on a dataset of 2,147 samples. The model was statistically significant (pseudo R² = 0.26) and demonstrated good discriminative capacity (AUC = 0.81), albeit with moderate accuracy (63.8%; Kappa = 0.44). Na exhibited a pattern most consistent with classical genetic models, whereas K displayed non-monotonic behavior. It is concluded that CMLR serves as a statistically valid tool for the probabilistic estimation and mapping of hydrothermal alteration in support of exploration efforts.