GEMOC ARC National Key CentrePLUTONIC SUITE RECOGNITION IN THE BEGA AND MORUYA BATHOLITHS USING A MULTIVARIATE EXTENSION OF ENTROPY ANALYSIS OF THE CHEMICAL DATA BASE.
R. H. Flood1,2, J. Forrest1,and B. W. Chappell2,3
1. School of Earth Sciences, Macquarie University
2. Department of Geology, ANU
3. GEMOC National Key Centre
One of the important aspects of granite suites is that they represent a range of rock types that are genetically connected rather than a group of near identical rocks. This makes the problem of computer-based suite recognition much more difficult than might be encountered subdividing a single group of rocks like basalts. For a statistical package to work it has to be able to ignore the very different bulk compositions of intermediate and felsic granitoids and "see" the subtle similarities that might indicate the relatedness of a particular group of diorites, granodiorites and adamellites. Although it may be useful to examine how the chemical data base can be subdivided statistically, there is no obvious reason why suites need to be able to be recognised/defined in this way and it is probable that as granitoid suites are groups of plutons, the chemical variation within a particular pluton may provide the most important control on suite definition.
In the Bega Batholith of the Lachlan Fold Belt the large chemical data base has allowed statistical methods to be used to test the distinctiveness of the component granitoid suites. Although this statistical approach has shown that cluster analysis using a restricted group of elements does distinguish the suites ( Whitten and Chappell, 1984, Whitten et al. 1987), the use of the full complement of elements has been less successful .
A multivariate extension of entropy analysis used by Forrest and Clark (1989) in a sedimentology application has the advantage over some other statistic packages, of being able to treat the totality of the chemical data without imposing any of the customary limitations on the form of the data such as normality of distribution. The advantages of this method is discussed by Forrest and Clark (1989) and are reviewed in detail by Johnson and Semple (1983). In summary the method groups analyses with similar major and trace element profiles, where the profile is the exact shape of each distribution across all elements. Furthermore, unlike other grouping techniques, it minimises the amount of within-group variance by testing all possible groupings of observations; each level of grouping is entirely independent of any other.
To date our preliminary study has shown that subdividing the chemical data base into groups using this approach indicates that:
(b) The geographic distribution of most of the groups are north-south elongate; the well documented east to west variation is being recognised but these groups do not match exactly the recognised suites.
(c) One group that consists of samples from the southern part of the Bega Batholith has a subcircular outline.
(d) The removal of the twenty most silica-rich samples
(initially expected to make the suite recognition clearer by removal of
samples that may so strongly reflect the minimum melt composition and/or
have unusual trace element values as a result of strong fractionation)
was found to make little difference to the groupings.
REFERENCES
Forrest, J, and Clark, N R, 1989. Characterizing grain size distributions: evaluation of a new approach using a multivariate extension of entropy analysis. Sedimentology, 36, 711-722.
Johnson, R J, and Semple, R K, 1983. Classification using information statistics. Geobooks, Norwich.
Whitten, E H T, and Chappell, B W, 1984. Suites within a granitoid batholith: a quantitative justification based on the Lachlan Fold Belt, SE Australia. International Geological Congress, 27th, Moscow, Abstract vol. iv, p 489.
Whitten, E T H, Bornhorst, T J, Li, G, Hicks, D L, and Beckworth, J P, 1987. Suites, subdivision of batholiths and igneous-rock classification: geological and mathematical conceptualization. American Journal of Science, 287, 332-352.
