The characterization of the shape of mineral grains in thin sections of soils by quantimet and besi

Abstract

Silt-sized mineral grains were analysed for their shapes or forms in thin sections of soils using backscattered electron scanning images (BESI) and an image analyzer (Quantimet 720). This work was done with an image-editor and required manual editing of the shape of mineral grains on the display screen of the Quantimet. Shape analysis of coarse and fine silt grains could thus be performed at magnifications of x 240 and x 480. The shape of the mineral grains could be characterized by five parameters: area ( A), perimeter ( Pe), largest and smallest Feret's diameters ( Fmax and Fmin), lobation ratio ( Lr) and indentation ratio ( Ir). A relation was found, however, between the APe 2 ratio and the Lr and Ir ratios. Consequently, shape analysis of mineral grains can be done by measuring three of the five parameters, i.e. A, Pe and Fmax/ Fmin. These parameters, however, are still not be able to reproduce the shape of individual mineral grains and images of them therefore remain necessary. Several of the 54 mineral grains were found to have different shapes but the same A Pe 2 ratio. An explanation for this phenomenon is given by using a theoretical model in which two-dimensional forms change their shapes from a circle ( A Pe 2 = 79 ) to a line ( A Pe 2 = 1 ). Form separation of mineral grains with the same ( A Pe 2 ) ratio can be done by using the F max F min ratio. The measurement of the shape of the mineral grains, the theoretical model and observation of the micrographs of the individual mineral grains, made it possible to draw a diagram with shape classes which can be used for shape classification purposes. Knowledge gathered in soil micromorphology on shape analysis of pores and soil aggregates has also been used to obtain boundaries of shape classes in the diagram. The present study is intended both as a proposal as to how two-dimensional shape classification can possibly be done, and as a work-model. The boundaries of the shape classes can easily be adjusted if data from forthcoming studies require it.