Sanja Lozić

 

PRIMJENA MULTIVARIJANTNIH STATISTIÄŒKIH METODA U GEOMORFOLOGIJI

 

DOI: 10.35666/28310438.2008.2.120

UDC: 911.5/.9[551.4:311](497.5)

 

Abstract: This work deals with multivariate statistic approach to the relief classification and typology in respect of slope processes, on the example of Zrinska and Trgovska mountain. Methods used in this research are: multiple linear correlation and regression, cluster analysis, discriminant and canonical analysis. Multiple linear correlation and regression are useful as criterion for establishing the relationship between chosen variables but also for determination the hierarchy of influence of independent variables on the dependent one.Cluster analysis represents a statistical technique of classification individual patterns of terrain in aspect of their similarity or dissimilarity according to their characteristics. The main purpose is determination of more relatively homogenous cluster groups of terrain patterns which represents particular relief types. By discriminant analysis, it is possible to establish the complete discriminant influence of variables on system in whole and slo the level of influence of each variable on the differentation of cluster groups (relief types). After this procedure, the extracting of discriminant functions need to be done by method of canonical correlation, for the purpose of gaining orthogonal linear combinations of predictor variables (discriminant functions). By analysis of loading of each discriminant function with variables it is possible to define each function according to domination of certain geomorphological processes. This method also enables constructing the models of relationships between cluster groups and discriminant functions, which can be established as a base for typology and evaluation.

Keywords: multivariate statistical methods, Zrinska and Trgovska mountain, slope processes, relief types.

Full text pdf