Soil structural transformation under multiple freeze–thaw cycles with comprehensive morphological and topological analysis: The hypothesis of periodicity

Abstract

Freezing–thawing cycles significantly affect the dynamics of soil pore space structure, both directly and through indirect processes. Previous studies have reported varying observations and interpretations regarding the qualitative and quantitative changes in soil structure induced by these cycles. In this study, we conducted freeze–thaw experiments on three different disturbed soil samples. The samples were imaged with an X-ray scanner in their original state and after 1, 5, 10, and 20 freeze–thaw cycles. To process the obtained data, we employed two novel but essential methods: (1) a custom-built image registration technique to establish identical 3D regions of interest within the scans, and (2) a set of soil structural descriptors with high information content, incorporating both morphological and topological information — correlation functions, pore-network statistics, Euler numbers, and connectivity. Registration and careful consideration of individual grains within the soil structure enabled robust segmentation of grayscale images into solids and pores. Unlike most previous studies, we did not observe a steady or nearly monotonic change in structural metrics. Instead, we detected a type of chaotic behavior of these metrics between the freeze–thaw cycles. Using vector descriptors, we demonstrated that the experimental data can be interpreted as hypothetical oscillatory changes within soil structure. This finding leads us to hypothesize that disturbed soils — and possibly natural soils after multiple cycles — undergo periodic structural dynamics. The novel idea behind this hypothesis is simple: as at some point the same temperature impact will not produce the same effect, the structure disturbance will stagnate. From the original state, soils exhibited the strongest structural degradation due to freeze–thaw cycles, whereas in subsequent cycles the dynamics involved both dispersing and aggregating processes, as observed in the X-ray tomography images. We conclude by discussing the necessary future research to confirm or refute this hypothesis and emphasize why soil structure in such experiments should be described using a novel class of vector metrics with high information content, which also subsume most classical soil structural metrics.