engleski

Visualisation and interpretation of the mineralogical data from the boreholes in Quaternary deposits of the southern part of the Pannonian Basin System, by using multivariate statistics combined with volume interpolation

The investigated area is situated along the Sava River and in between the Požeška and Dilj Mts. in the north and the Motajica Mt. in the south. This study utilises 18 boreholes as observations, drilled in the area of 470 km2 of Quaternary deposits. Sampling of the boreholes was performed irregularly to a maximum depth of 10.25 m. Modal analyses were made on thin sections of the samples, prepared from heavy and light mineral fractions. The final data set included proportions of 11 variables (rock fragments, quartz, feldspar, tourmaline, staurolite, garnet, pyroxene, amphibole, epidotezoisite, chlorite, opaque minerals) in 46 samples. The purpose of this study was to visually present the distribution of mineral phases on 3D maps of the investigated deposits with the aim to build a meaningful 3D subsurface model, by using open source software. The results from modal analyses were expressed in percentages, thus representing compositional data (CoDa ; AITCHISON, 1986). CoDa are constrained, i.e. they are always positive and can vary from 0 to some constant value (e.g. if data are expressed in percentages, they can vary from 0 to 100 % and 100 is the constant). Such scale is relative, which means that sample compositions are related in a way that an increment (or decrement) of one variable must result in a decrement (or increment) of at least one other variable. Standard statistical methods are based on the Euclidean geometry where scale is absolute and values can vary freely, from –∞ to +∞. Here, isometric log- ratio transformation (EGOZCUE et al., 2003) was applied to transform the data to the Euclidean space, so standard statistical methods can be applied. A first step in the workflow was done in R 3.0.1 software (2013) and consisted of the determination of mineral assemblages, together with the estimation of the compositional variability by the means of multivariate statistical analysis, i.e. the principal component analysis (PCA) for CoDa (FILZMOSER et al., 2009). PCA loadings were used to describe mineral assemblages and PCA scores were used for an interpolation in the next step of the workflow. The interpolation was performed in GRASS GIS 6.4.4 software (2012), using regularised spline with tension (RST) algorithm (MITAŠOVA & MITAŠ, 1993 ; HOFIERKA & MITAŠOVA, 1993). Tension and smoothing parameters were set by minimising the cross- validation (CV) error along with the generalisation of the phenomena, because small number of samples could give the misleading CV error. The final step in the workflow was the visualisation of the results which was done in ParaView 4.3.1 (2015) and MeshLab (2012) software. The first two principal components (PCs) explain approximately 53 % of the total data variability. PC1 distinguished between two mineral assemblages: the more stable garnet, staurolite and opaque minerals which had positive loadings and the less stable rock fragments, feldspar and chlorite which had negative loadings. Such mineral assemblages very likely represent more or less mature sediment. Visualization of the positive PC1 scores (Fig. 1a) possibly revealed the older, Orljava River structure lying beneath the Sava River alluvial plain (negative PC1 scores ; Figure 1b). PC2 distinguished between garnet, tourmaline, epidotezoisite and staurolite which had positive loadings and amphibole and quartz which had negative loadings. Such mineral assemblages could be a signature of different metamorphic source rocks (e.g. gneisses against amphibolites). The negative PC2 scores (Figure 1d) are tied up with the Orljava River, carrying sediment from the Papuk and Psunj Mts., in addition to the Požeška and Dilj Mts.. Positive PC2 scores (Figure 1c) fill the rest of the volume and probably represent mixed sediment from the Požeška and Dilj Mts. and the Sava River. This study revealed specific mineral assemblages and sedimentary bodies in 3D space based on the Quaternary sediment modal composition. To confirm or reject interpretation of the results, more analytical methods are required, but the proposed method presumably enlightened the origin of analysed Quaternary sediments and indicated its application in e.g. 3D geological mapping, natural resources or hazards management.

Quaternary deposits ; compositional data ; multivariate statistics ; volume interpolation ; 3D model

nije evidentirano