How to determine the optimal balance for geochemical pattern recognition and anomaly mapping based on compositional balance analysis?

Balance analysis of two groups of parts within a whole has become an important method for compositional data analysis. A compositional balance is a particular orthonormal coordinate that is depicted by the log-ratio between two groups of components. Two available approaches to compositional balance analysis (CoBA) can be adopted to generate targeted balances for geochemical pattern analysis and anomaly identification, so-called data-driven CoBA and knowledge-driven CoBA. For the data-driven CoBA, the balance is produced strictly by the rules of sequential binary partition (SBP), while for the knowledge-driven CoBA, the first group within a balance is composed of the interesting parts of the whole and the second group is defined by the remaining parts of the whole. Commonly, it is difficult to conceptualize balances, particularly for high-dimensional data, because it will produce a large number of orthonormal bases or balances based on CoBA. For a certain geochemical pattern, it might be represented by multiple compositional balances generated by data-driven and knowledge-driven CoBA. Thus, how to determine an optimal balance for geochemical pattern analysis and anomaly identification needs to be further explored. In the present study, this question was thoroughly investigated based on a case study from the Chinese Western Tianshan (CWT) region. Fourteen compositional balances and three principal factors associated with different geochemical patterns including gold and copper mineralization, and particular lithological units were selected for comparative studies to illustrate how to determine the optimal balances from the perspective of CoBA and multivariate statistical analysis.Thematic collection: This article is part of the Applications of Innovations in Geochemical Data Analysis collection available at: https://www.lyellcollection.org/cc/applications-of-innovations-in-geochemical-data-analysisSupplementary material:https://doi.org/10.6084/m9.figshare.c.6083724