Fig 1.
Image processing workflow for this study depicted for a small two-dimensional subset (a). Noise is removed with a non-local means filter (b). Image segmentation is performed in several steps. First, gray values are tentatively segmented into pores (black), aggregates (blue) and garnet (red) via simple thresholding (c). Then particles are detected with a Laplacian of Gaussian Filter (LoG) (c) and subsequent hysteresis thresholding of the LoG Image. Note that the edges of large grains are masked out during particle detection (not shown). For the final segmentation (e) partial volume voxels tentatively assigned to the aggregate class are set to unassigned (white) with a morphological opening of the aggregate class by a small structuring element (dSE = 5 voxels). The tentative garnet class is set to unassigned (white) and overwritten by the thresholded LoG image (red). Pores are further differentiated with respect to whether they are fully enclosed in soil aggregates (yellow) or not (black) (f). These images are subjected to different types of analysis (5.-7.).
Fig 2.
3D rendering of a sample at a bulk density of (a) ρ = 1.1 g/cm3, (b) ρ = 1.3 g/cm3 and (c) ρ = 1.5 g/cm3.
The green circles highlight aggregates which are strongly (#1) or weakly (#2) covered with particles or pores in which detached particles gather (#3). The yellow circles highlight that photon absorption in garnet (#1) is higher than in iron-free minerals like quartz (#2). The red circles highlight in the incorporation of particles into the soil matrix in the course of compaction.
Fig 3.
Pore space analysis for soil at three bulk density levels (ρ = 1.1, 1.3, 1.5 g/cm3) with five replicates: (a) porosity profiles, (b) cumulative pore size distribution, (c) mean pore size and Γ connectivity as a function of pore diameter.
Because of the depth gradient in porosity results are shown separately for the top and bottom of the samples.
Fig 4.
(a) Spatial distribution of big garnet grains and small garnet particles at a bulk density of 1.1 (green) and 1.3 g/cm3 (red). (b) Elastic image registration leads to a very good spatial alignment between the deformed and the original image. (c) The resulting displacement field shows a gradient in vertical displacement. (d) Compaction to 1.5 g/cm3 increases this gradient even further.
Fig 5.
(a) Frequency distribution of contact distances between particles and pores with average and standard deviation of five replicates. The inset shows the contact distance distribution between bulk soil and pores with is generally much larger. (b) The mean contact distance of particles and bulk soil is shown for each replicate and further separated into top and bottom part of the sample. The mean contact distance for particles scales linearly with the mean contact distance for bulk soil during compaction.
Fig 6.
The mean contact distance between bulk soil and pores scales exponentially with decreasing porosity during soil compaction.
This exponential trend follows from the non-linear increase of contact distances as pores start to vanish completely at higher bulk density. The scatter with subgroups of similar bulk density is caused by different degree of intra-aggregate porosity mostly due to crack-formation. This variability in intra-aggregate porosity causes a linear scaling relationship between contact distances and porosity. All fitted curve converge to a similar contact distance at vanishing porosity which is mainly determined by the average size of aggregates.
Fig 7.
Conceptual scheme for the quantification of soil structure turnover rates.
(a) A two-dimensional packing of aggregates is covered with garnet particles. (b) The aggregates are first compressed by soil compaction. Subsequently, soil structure turnover is initiated through new root channel formation, micro-crack formation and the partial refilling of pores with earthworm casts, where (c) and (d) represent two consecutive moments in time. (e) The distribution of contact distances between pores and particles or between pores and bulk soil (inset). (f) The mean contact distance of particles is initially much smaller than the mean contact distance for bulk soil. Soil compaction does not lead to a trajectory towards the 1:1 line (randomized position of particles), whereas structure turnover does.