CLASIFICACIÓN DE LOS SUELOS CON FINES PERICIALES MEDIANTE MICROSCOPÍA ELECTRÓNICA DE BARRIDO CON ANALIZADOR DE RAYOS X Y EL EMPLEO DE HERRAMIENTAS QUIMIOMÉTRICAS PARA EL PROCESAMIENTO MULTIVARIANTE DE LOS DATOS

Abstract

Soil forensic evidence samples are very difficult to process, due to the greater number of general and individuals characteristics presents at the same time and the low discriminative information that the surface layer has , and this part is the one that is collected in crime scene or in shoes and other objects
belonging to a suspect. The main purpose of this paper is demonstrating the feasibility and benefits of the Scanning Electron Microscopy with X-ray Analyzer coupled (SEM-EDS) for the analysis and chemical characterization of samples of soils of the municipality of Guanabacoa in Havana City, and the
construction of an automatic classification model for soil samples discrimination, in correlation with the present-day genetic classification existing of this place, in order to predict the origin and soil type in unknown samples related with a case using Chemometrics tools for the multivariate processing of the
data. The data from the SEM-EDS analysis was submitted to an exploratory analysis using the Hierarchical Clusters Analysis, demonstrating the feasibility and differentiation of 6 types of soils of the 9 presents. Elements Mg, Si, K, Ca, Ti, Fe and Al were the most significant elements in the discrimination.
From these results a model for the automatic classification for 6 types of soil samples was constructed using a Support Vector Machines classifier. The model was validated with external samples not present in the training set) with 97 % of efficacy.
Keywords: soils, scanning electron microscopy, chemometrics, HCA.