Statistical Analysis of Major Element Patterns in Basalts

Abstract

Discriminant analysis enables basalts from different tectonic settings to be identified on the basis of their complete major element patterns. By using discriminant functions it is possible to represent visually the separation of six tectonically-defined magma types in three dimensions instead of the original eight. Analysis of variance confirms that this separation is highly significant. Numerical classification is possible by applying Bayes' decision rule, using either normal distribution functions or empirical discriminant functions. Tests on known samples showed that ocean-floor basalts, low-potassium tholeiites (from island arcs), calc-alkali basalts and shoshonites (from volcanic arcs) and within-plate basalts could be correctly classified about 90 per cent of the time by all these methods. A subdivision of within plate basalts into ocean island and continental basalts could not however be achieved chemically with any great success. Classification of weathered and metamorphosed ocean-floor basalts showed that alteration can greatly reduce the success rate of the classification. Application of the numerical and visual methods to lavas of 'unknown' tectonic settings (Archaean greenstones and high-potash basalts) mostly gave geologically consistent results.