Principal component analysis of textural characteristics of fluvio-lacustrine sandstones and controlling factors of sandstone textures

Abstract

Textures are important features of sandstones; however, their controlling factors are not fully understood. We present a detailed textural analysis of fluvio-lacustrine sandstones and discuss the influences of provenance and depositional environments on sandstone textures. The upper Permian – lowermost Triassic Wutonggou sandstones in the Bogda Mountains, NW China, are the focus of this study. Sandstone thin-sections were studied by point counting and their textures were analysed using statistical and principal component analysis. Fluvial lithic, fluvial feldspathic, deltaic lithic, deltaic feldspathic, littoral lithic and littoral feldspathic sandstone were classified and compared. These comparisons indicate that lithic and feldspathic sandstones from the same depositional settings have significant differences in graphic mean, graphic standard deviation and roundness; in contrast, sandstones from different depositional settings but with similar compositions have limited differences in textures. Moreover, three principal components (PCs) are recognized to explain 75% of the total variance, of which the first principal component (PC1) can explain 44%. In bivariate plots of the PCs, sandstones can be distinguished by composition where lithic and feldspathic sandstones are placed in different fields of the plots along the axis of PC1. However, sandstones from different depositional settings overlap and show no clear division. These results indicate that provenance, mainly the source lithology, is the most significant controlling factor on sandstone texture, whereas the depositional environment has limited influence. This study improves our understanding of textural characteristics of fluvio-lacustrine sandstones and their controlling factors, and shows the potentiality of principal component analysis in sandstone studies.

Automated Classification of Estuarine Sub‐Depositional Environment Using Sediment Texture

“…Early approaches employed statistical measurements of particle size distribution of sand to discriminate between beach, dune, river, and aeolian environments (Biederman, 1962;Friedman, 1961;Mason & Folk, 1958;Moiola & Spencer, 1979;Moiola et al, 1974;Sevon, 1966;Vincent, 1998). More advanced approaches employ Analysis of Variance (ANOVA), bivariate or multivariate discrimination, such as principal component analysis (PCA), with increased success over simple statistical measurements (Flood et al, 2015;Purkait & Das Majumdar, 2014;Simon et al, 2021;Zheng & Wu, 2021;Zubillaga & Edwards, 2005). Simon et al (2021) proved a strong statistical link between surface sub-depositional environment and sediment texture in the Ravenglass Estuary, UK, and devised a simple machine learning classification scheme for the classification of sub-depositional environment from sediment texture data.…”