Partition of subvolcanic rocks on the geodynamic classes using classification equations based on petrophysical data
O. Shabatura and N. Kostenko
Event name: 15th EAGE International Conference on Geoinformatics - Theoretical and Applied Aspects
Session: 3. Geological Modeling
Publication date: 10 May 2016
Info: Extended abstract, PDF ( 457.74Kb )
Price: € 20
A new approach to the classification of geodynamic volcanic formations based on the involvement of a wide range of petrophysical characteristics of rocks (34 volcanic centres of the Great Caucases and the Precaucases). The geodynamic classes of geological objects are obtained with Pierce geochemical discrimination use as a "training sample". For their help it have been created a matrix of probabilities each geological object related to geodynamic classes. Next step of analysis is a creation of a petrophysical patterns for each geological object and the mixing of the petrophysical characteristics in feature hyperfield. Linear discriminant functions received by minimizing the classification error and they include a number of standardized coefficients for variables petrophysical characteristics (density, magnetic, thermal, capacitive). These linear fuctions shared a hyperspace of features on regions that correspond to the desired classes (geodynamic groups of geological structures). An accuracy of class assignment to syn-collisional rifting achieve 85%, subduction-collision - 68%. Unclassified objects that do not fall in any class is likely to correspond complex and transitional geodynamic regimes that experienced geological objects in its history. Thus, the approach may be useful for tasks of geological reconstruction, correlation and comparison, searching of the analogues in other regions, to complement the existing (formational, geochemical) methods of converged geodynamics.