Comparison of digital image analysis methods for morphometric characterization of soil aggregates in thin sections

The purpose of this study was to investigate the applicability of semiautomatic segmentation methods for obtaining and evaluating morphometric parameters of soil aggregates in artificially prepared loose samples in soil thin sections. The object of the research is typical arable Chernozem. The aggregates were separated by wet sieving method from loose sample of upper 10 cm of the plowing horizon after erosion by a model shallow water flow on a large erosion tray. The aggregates, loosely scattered on the glass and fixed with polyester resin, were used to produce the thin sections. Images of the thin sections were taken under a polarizing microscope and then were processed using two methods compared: Adobe Photoshop + CTan and Thixomet Pro. Data on morphometric parameters of aggregates were obtained: the shape factor, the degree of roundness and the coefficient of aggregate surface roughness. The convergence of the results obtained using Photoshop + CTan by three researchers was evaluated by comparing samples using the Student's test and the Mann-Whitney test. The convergence of the averaged results obtained using Photoshop + CTan and the results obtained using Thixomet Pro was evaluated using the Mann - Whitney test. No significant differences were found between the parameters of the same aggregates obtained using a combination of Adobe Photoshop and CTan programs by different researchers. No significant differences were found between the parameters of the same aggregates obtained by the compared methods. So, one can conclude that the reliability of determining the morphometric parameters of soil aggregates using Thixomet Pro is comparable to the reliability of results when working with images of sectionsin CTan after binarization in Adobe Photoshop. The method of obtaining data on morphometric parameters of soil aggregates using Thixomet Pro completely eliminates the possibility of subjective error, shows a high degree of automation, reproducibility and reliability of the results obtained, and is faster.