RUSSIAN JOURNAL OF EARTH SCIENCES VOL. 10, ES4001, doi:10.2205/2007ES000218, 2008

5. Algorithm of Geodynamical Classification

[45]  Data preparation for cluster classification includes the formation of spatially identical matrixes for all the parameters used (see Section 3) and their values standardization required for this algorithm accounted for calculation of distances (required uniform parameter dimension). Then the tabulated data is loaded into program environment.

[46]  The next step should be the classification test by minor N values. Here algorithm should step by step accomplish the classification of space analyzed into clusters, geologically valid. Starting with N =2 algorithm divides the area analyzed into oceanic and continental (shelf areas). At the next step ( N =3) the oceanic area is divided into basins and most elevated parts of MAR. During the following steps (up to N =5) successive isolation of MAR zone, including flanks and division into "cold'' and "hot'' parts takes place.

[47]  Starting from step N > 5 trivial solutions are followed by situations not visually discernible. For example, differentiation of basins, MAR flanks and continent-ocean transition zones appears. At steps from N =8 to N =10 flank MAR zones obliquely oriented and locally deeply incised into basins along with isolation of MAR zone north of Iceland and pseudosymmetric superimposed effects start to appear. Final stable differentiation of MAR zone as well as most of basins and continental margins into clusters with physically clearly specification takes place on steps N =11 to N =13. Steps from N =14 to N =15 show final extraction of nontrivial clusters superimposed on main oceanic structural elements. The parameters of these elements differ by a value above parameter scatter within isolated zones belonging to one of the clusters.

2007ES000218-fig11
Figure 11
[48]  From the steps N > 15 a sudden "scattering'' of the largest cluster occurs. It "scatters'' to small cell groups chaotically distributed in the space of deep ocean basins. Difference between these groups is comparable to mean dispersion of parameters in standardized space values. At steps from 16 up to 100 avalanche increase in cluster number for all areas of the Atlantic whose profiles are concentrated in zero variation area and do not bounce significantly from zero like at step N =15 (see Figure 11) is observed. It means that the physically valid limit for cluster classification with the available data set has been attained. Further N increase with its asymptotic approximation to the number of objects (or to infinity, depending on degree of detail of area analyzed) won't result in solution of classification problem.


RJES

Citation: Sokolov, S. Yu., N. S. Sokolov, and L. V. Dmitriev (2008), Geodynamic zonation of the Atlantic Ocean lithosphere: Application of cluster analysis procedure and zoning inferred from geophysical data, Russ. J. Earth Sci., 10, ES4001, doi:10.2205/2007ES000218.

Copyright 2008 by the Russian Journal of Earth Sciences

Powered by TeXWeb (Win32, v.2.0).