engleski

Assessing natural and anthropogenic impact on soil chemistry using principal component analysis for compositional data

For the purpose of this study, two surface soil samples (0-5 and 5-10 cm) were collected on each of the 36 regular grid locations around the fertiliser plant in Kutina, Croatia. Samples were analysed on chemical composition and statistically processed using principal component analysis. Concentrations of chemical elements are a typical example of compositional data. Such data form simplex space which has different properties than Euclidean space. Accordingly, data have to be properly transformed so that standard statistical methods can be safely applied. In this paper, efficiency of log and isometric log- ratio transformation in principal component analysis was tested. It was shown how compositional data methodology can improve the results of analysis and provide new possibilities in interpretation and determination of the associations of chemical elements and observations.

compositional data ; principal component analysis ; soil ; geochemistry ; fertilizer plant Petrokemija

nije evidentirano