Artículos de revistas
Stepwise Regression For Recognition Of Geochemical Anomalies: Case Study In Takab Area, Nw Iran
Registro en:
Journal Of Geochemical Exploration . Elsevier Science Bv, v. 168, p. 150 - 162, 2016.
0375-6742
1879-1689
WOS:000381648200012
10.1016/j.gexplo.2016.07.003
Autor
Nazarpour
Ahad; Paydar
Ghodratolah Rostami; Carranza
Emmanuel John M.
Institución
Resumen
Stream sediment geochemical data represent compositional materials derived from various sources, including single or multiple lithologic units, soil types, rocks types, etc. In order to delineate geochemical anomalies, stream sediment geochemical data are usually subjected to suitable multivariate analysis, and not simply using univariate threshold values because these are not reliable for delineation of geochemical anomalies in areas with complex geological units. Relationships among multiple major/trace elements and rock types are more important than single major/trace elements for delineation of geochemical anomalies. In this study we present an approach based on robust stepwise multiple regression using values major oxides (SiO2, Al2O3, Fe2O3, MnO, and MgO) in stream sediments to predict elemental content related to rock types and to recognize geochemical anomalies. The major/trace element data were subjected to isometric logratio transformation to address the compositional data closure problem. For further examination of the stepwise regression method, its performance was compared to robust principal components analysis (RPCA), median + 2MAD and concentration-area (C-A) fractal methods. The results show that multi-element anomalies obtained by the stepwise regression method, compared to those obtained by the other methods, have stronger spatial association with the known deposits, such as Chichaklo and Ay-Ghale-Si in the Takab 1:25,000 scale geological map (NW) Iran, and the anomalies have stronger spatial correlation with structural features and prospects, and thus can be used as guides to new exploration targets. (C) 2016 Elsevier B.V. All rights reserved. 168 150 162