1 Introduction

The arrangement of the particles and aggregates of soil matrix plays an important role in the architecture of soil pores. Differences in the characteristics of soil particles such as their size, shape, and orientation help to create a complex arrangement of pore types, which has great influence in many processes that happen into the soil matrix [1]. These processes can involve additions, losses, transformations, and translocations [2]. All of them have in common dynamic events like, for instance, mineral material movement into the soil due to flooding (addition), water movement from the plants to atmosphere through their roots and leaves or from the soil directly (losses), and clay and organic matter movement when water percolates downward into the soil profile (translocation). These three processes described are dependent on transformations like, for example, mineral conversion into clays, which affects the soil structure over time [3].

Many factors can affect the soil pore system with direct impacts in the dynamic processes taking place in it. Climate, management systems, wetting and drying cycles, biological activity, and mechanical procedures have important influence in the soil structure over short and long periods of time [4,5,6,7]. The combination of these events changes the complexity of the soil pores affecting their spatial and size distributions, shape, connectivity, tortuosity, etc. [8,9,10,11]. These changes can modify the soil permeability influencing the movement of water and air through the soil profile, affecting the capacity of plant roots to develop and penetrate the soil, the water infiltration and distribution, and the amount of water and nutrients available to plants [12,13,14].

Due to the importance of the complexity of the soil physical system to many of the events occurring in this porous material, soil scientists have been investigating in the last decades the internal structure of soils through two-(2D) and three-dimensional (3D) image analysis techniques. Optical and electron microscopy, nuclear magnetic resonance, and tomography are examples of methods employed to extract information from images by means of digital image processing associated with computational mathematical algorithms [15,16,17,18]. Nowadays, tomography is one of the techniques that has been widely utilized for detailed 3D analyses of the soil structure [19].

Tomography is a non-destructive and non-invasive technique that allows imaging different types of materials such as rocks, ceramics, soils, metals, and polymers, among others. It is considered an important tool in material sciences for non-destructive tests and evaluation of internal structures, metrology, and quantitative analysis of morphological and geometrical properties of porous materials [20]. New advances in the spot sizes of X-ray sources and detector systems have opened the possibility of obtaining 3D images with high spatial resolutions from micrometers to nanometers [21,22,23]. It is important to point out that 3D tomographies can be obtained using not only X-rays but also neutrons, positrons, gamma rays, electrons, seismic waves, etc. However, in this paper, we will focus on the use of X-ray computed tomography (XCT) to inspect the internal structure of porous materials such as soil.

X-ray computed tomography is an interesting technique because it allows scanning samples up to several centimeters in size, in most of the commercial benchtop scanners, or up to a few millimeters in synchrotron (SR) light sources [21]. A trade-off between commercial XCT systems and SR-based XCT (SRXCT) is spatial resolution vs. sample size, due to much smaller X-ray source sizes that can be achieved in SRs but limited associated field of view due to an often designed parallel X-ray beam. For instance, Ferreira et al. [24, 25] presented a complementary study using undisturbed soil blocks of 80 × 80 × 80 mm with a voxel size of 60 µm scanned in an industrial XCT system and soil aggregates of up to 4 mm with voxel size of 1.64 µm scanned by SRXCT, respectively. In XCT, the composition of the material scanned permits building a data matrix of the radiation interaction with their internal structures [26, 27]. These data matrix are then converted to images allowing to investigate 3D morphological properties of the soil from the micropore to macropore scales, offering the opportunity of the analysis of the soil pores based on their functions [28].

The characterization of the spatial distribution of the soil pores and their morphological and geometrical features through XCT has opened the possibility to understand many of the dynamic processes that take place in the soil [29]. As said before, these processes involve mainly the fluid flow within the soil, the plant root system development, and hydro-mechanical behavior of the soil. However, many of the studies dealing with XCT combined with image analysis have been restricted to a single scanning of the samples and subsequent characterization of their morphological characteristics or computational simulation of dynamic events through the soil [30,31,32]. On the other hand, 4D tomography (3D + time) seems to be an interesting alternative to perform time-lapse (or time-resolved) experiments. This technique permits not only to characterize the materials pore system but also to study dynamic processes occurring in it [33, 34].

Some studies have focused on changes in the soil structure over time such as those caused by soil use, compaction, wetting and drying cycles using either single scanning of replicate samples subjected to different treatments for an average characterization [6, 35], or repeated scanning of the same samples reinstalled in the field over the course of years [28]. In 4D XCT investigations, usually the same samples are scanned many times after being submitted to contrasting treatments like wetting and drying cycles, shear strength tests, compression, etc. This type of approach can create many problems such as the constant manipulation of the samples to transport them to the tomographic system, the need for special calibrations, scanner voltage and electron flux adjustments, and the need for guaranteeing the same experimental conditions before each scan, to cite some of them. In addition, with the advent of third- and fourth-generation SRs (the latter presenting the highest photon flux among all currently available systems), the temporal resolution of 4D XCT experiments has been increased to the order of seconds. This has raised the necessity to develop instrumentation to perform 4D XCT in situ (mounted directly at the SRXCT beamline or XCT equipment).

Dynamic XCT opens the possibility to evaluate the role of pore continuity and tortuosity on soil water movement, distribution, and retention, to investigate crack formation in clayey soils, to study the soil response to compression and triaxial loads, and to evaluate the development of roots over time [36,37,38,39]. It can also be utilized to monitor the evolution of hydraulic properties in soils subject to compaction, to observe the influence of the connectivity of the pores in the transport of soil gases, and to map soil deformation around plant roots, among many other types of studies [36,37,38,39]. As we can see, many studies can be proposed using 4D XCT. A literature search shows that it is almost easy to find papers dealing with time-lapse electrical resistivity tomography, which is based on measurements of soil resistivity. This technique has been used in studies of piping caused by seepage and soil erosion, rainfall simulation and water infiltration experiments in soils, temporal variation of hydrologic connectivity and soil structure dynamics, changes in pressure head due to spray irrigation, preferential infiltration, and so on [40,41,42,43,44]. However, the use of 4D XCT to image soil processes is still scarce, especially considering real soils as most of the available literature deals with artificial porous media. It is important to highlight, as described before, that modelling and understanding many processes in the soil require dynamic analyses, indicating the need for the development of new studies focusing on 4D XCT.

Having in mind the importance of the 4D XCT for the analysis of dynamic processes in the soil, this paper describes the current state of the art presenting researches related with two hot topics in soil science: (i) fluid flow and water distribution in soil and (ii) soil-root interactions. In addition, some future perspectives that are hoped to substantially improve the experimental conditions to develop innovative 4D XCT investigations in Brazil will be presented aiming to bring to the readers an outline of near-future possibilities.

2 Materials and Methods

The search for literature in this paper was carried out through the following search bases: Scielo (Scientific Electronic Library Online), Science Direct, Web of Science (Institute of Scientific Information), and Google Scholar. The utilized search terms were “soil AND 4D AND X-ray tomography,” “soil AND time resolved AND X-ray tomography,” “soil AND time lapse AND X-ray tomography,” “soil AND 4D AND synchrotron tomography,” “soil AND time resolved AND synchrotron tomography,” and “soil AND time lapse AND synchrotron tomography.”

The relevant literature selected for this paper back 18 years comprising studies dealing with 4D or 3D analysis combined with modeling and simulations of processes. Only papers published in journals were selected for this study. All the papers were peer-reviewed, and full texts are available through the different databases searched. After a careful analysis of the documents, 125 publications were considered to be more aligned with the current paper (i.e., presenting important relation between 4D XCT and soil science). Three important subject branches were identified in this search: 50 documents were related to fluid flow and water content distribution in soil (or any other porous media with some relevance to soil), 20 documents about rhizosphere processes, and 15 documents related to soil mechanical properties. The other 40 documents comprised studies dealing with XCT to evaluate geometrical and morphological properties of the soil, computer simulation of processes, and image analysis. From the 125 identified publications, 95 were included in this overview paper.

The two first topics (fluid flow and water content distribution and rhizosphere processes) were chosen to elaborate an overview of examples of dynamic processes that have been investigated up to date. This choice relies on the fact that the papers found for the third topic (mechanical properties) had less relation with agricultural implications (i.e., more focused on engineering issues). However, it is important to highlight that this paper seeks to describe some of the studies carried out on 4D XCT, but it does not intend to present a comprehensive review which is beyond the scope of this contribution.

The new Brazilian synchrotron light source (Sirius), which is one of the three fourth-generation SR sources in the world, will count on a world-leading micro and nano SRXCT beamline (Mogno: https://www.lnls.cnpem.br/facilities/mogno-en/), which is expected to be available for users in the near future (2022/2023). This beamline will present very relevant capabilities to circumvent current limitations already recognized in 4D XCT studies, which will be briefly described in the section of future perspectives. Moreover, in the overview of selected dynamic studies in both proposed topics, some limitations and research gaps referred by the authors were identified and will be used to exemplify potential solutions at Mogno, aiming to help readers setting a more practical meaning of the new experimental capabilities that will be available in Brazil.

3 Overview of 4D XCT in Soil Studies

In the following sections, a limited presentation of some applications of XCT combining dynamic studies is covered. The selection of papers aims to show to the readers the main capabilities of 4D XCT for the study of the dynamic processes taking place in the soil and other porous media.

3.1 Fluid Flow and Water Distribution

It is known that XCT is a powerful tool that allows the acquisition of 3D high-resolution images on multiple scales (from nanometers to millimeters). After image reconstruction, processing steps, and segmentation, XCT provides many types of information about the architecture of pores, roots, soil particles, and organic matter distribution in the soil [6, 25, 45,46,47]. Some of the parameters that can be measured through XCT are volume fractions, size and shape distributions, anisotropy, intrinsic permeability, porosity, connectivity, tortuosity, orientation, fractal characteristics, among several others [48,49,50,51]. The combination of morphological and geometrical parameters related to the pores has profound effect in the water and matter dynamics in the soil [8, 14, 19, 52, 53].

The analysis of pore-scale multiphase flow experiments has become nowadays possible with the new advances in XCT scanners and image analysis softwares dedicated exclusively for the analysis of materials. One of the first studies applying XCT to minute pore-scale multiphase flow has been presented by Wildenschild et al. [54]. Small cylindrical pressure cells packed with sandy material were utilized in the measurements. The authors concluded that XCT has great potential as an analytical tool for multiphase flow studies, and they also observed relations between the characteristics of the pores, like size distribution and continuity and their relation with the drainage taking place. The movement of water was also investigated by Carminati et al. [55] through experiments using SRXCT. The use of SRXCT refers to the high brilliance and coherence of the X-ray source and the possibility of the acquisition of micrometer and sub-micrometer 3D images within a very short time [56]. The authors found that micro-heterogeneities present in the structure of the soil control the overall hydraulic behavior showing the influence of the gravity and capillary forces during water drainage. An example of an XCT experimental setup for the study of dynamic processes is presented in Fig. 1 [14].

Fig. 1
figure 1

Schematic drawing of a custom-built confining-pressure flow cell (left) used for fluid flow experiments and the X-ray computed tomography (XCT) scanner (right) built to perform dynamic studies of fluid movement through porous systems. In this experimental XCT scanner setup, the gantry rotates around the sample, which remains static [14]. All images are reproduced with permission

Full size image

Jelinkova et al. [57] examined the effects of air entrapment on water flow, combining magnetic resonance and XCT images. Tomographic images were used to assess possible changes of the soil structure caused by water and solute movement. The authors noticed that the main pathways responsible for conducting the dominant portion of water are strongly influenced by trapped air bubbles. Jung et al. [58] investigated, through time-resolved 2D X-ray radiography, the dynamic movement of the water front in sand samples of contrasting particle diameter ranges. The authors found that the initial water content strongly influences the water transport and infiltration patterns (from bulb to trapezoidal shape). However, when the porous media is saturated, the preferential flow disappears with the water being very uniformly distributed in all directions, as expected [3]. As an example, Fig. 2 represents 3D images showing time sequences of a drainage event in a porous system [14]. These images highlight the variation between the non-wetting and the wetting phases. Time-resolved 3D analyses of water preferential flow in an undisturbed silty-clay soil were proposed by Sammartino et al. [59]. Images with resolution of 370 μm were acquired every 3 and 5 min (3D image), depending on the flow regime (transient and stationary), at a helical medical XCT scanner. The authors showed that even in intense rainfall events, the macropores remained mostly unsaturated during water infiltration. They also indicated that their study presented some drawbacks associated with the measurements discrete in time (every 3 min) and in space (spatial resolution) for the analysis of the water movement and retention.

Fig. 2
figure 2

Four-dimensional X-ray computed tomography (XCT) images showing a time sequence of a pore filling event (circled region in the images) in a fluid flow experiment [14]. In red is represented the non-wetting phase, while in blue is the wetting phase. The upper and bottom images represent two viewing angles. The time intervals (t) showed indicate the starting time of the XCT acquisition (12 s for each acquisition). All images are reproduced with permission

Full size image

Koestel and Larsbo [60] visualized the transport of potassium iodide in a small undisturbed soil column under steady-state hydraulic conditions. 4D XCT was utilized by the authors to measure the dilution index, which is related to preferential flow. The soil columns were scanned in different periods of time (5, 10, 30 min, and 1 h). Some of the results observed by them indicated that most of the macropore system was non-conducting, which was consistent with the presence of isolated macropores that did not contribute to the tracer conduction through the soil column. The speed of the solute front in the soil matrix downward was about 1 mm h−1, as visualized in the 3D images. Figure 3 shows examples of reconstructed 3D images presenting solute infiltration in the macropore system of limestone [14]. The role of macropores and intragranular pores on solute transport is visualized in the 3D images. Sammartino et al. [52] also presented an interesting study to monitor water infiltration and preferential flow in a structured soil. The authors showed that the geometric properties of the active macropore network play an important role in the infiltration of the water. Peng et al. [61] monitored and characterized the water flow in 3D in sand and porous clay sphere arrangements. A voxel size of 6.3 × 10−3 cm in all three dimensions was obtained. The temporal resolution was around 25 min per scan. Depending on the size of the pores, the authors observed that the water infiltration into the clay spheres was negligible, and it only occurred in the sand matrix during the time range of infiltration. The authors also visualized that the waterfront had a heterogeneous shape, consistent with the heterogeneity of the porous media.

Fig. 3
figure 3

Image intensity histograms of a limestone pore space voxels during brine injection is presented in the graph [14]. The highest intensities indicate the high CsCl concentrations. The 3D X-ray computed tomography (XCT) images present the variation of the CsCl concentration in the macropores. A region corresponding to an intragranular pore in which the flow is stagnant is indicated as a dotted circle in the images. All images are reproduced with permission

Full size image

Rad et al. [62] investigated the salt precipitation evolution under evaporation. The main idea was to study the role of particle and pore sizes on salt precipitation dynamics and patterns. Quartz sand samples differing in particle size distribution were selected for the study. The authors noticed that the existence of preferential evaporation sites on the surface of the samples affects the patterns and dynamics of salt precipitation [63]. The formation of thick and discrete salt crusts in the sample columns was associated with the presence of fine pores at the surface. Biological soil crusts formed in the topsoil of arid areas were investigated by Couradeau et al. [64]. SRXCT was employed to scan the samples sequentially for 3 h every 10 to 40 min as the soil sample dried. The authors demonstrated the influence of the crusts in the water distribution through tomographic images of high resolution. The analysis of water loss through evaporation was also made by Yang et al. [65] working with a limestone. Through SR radiation experiments, water content changes were observed only in time-differential phase contrast images, demonstrating the potential of XCT to visualize water distribution during infiltration. Another interesting paper dealing with the influence of particle size distribution on solute transport in porous media during evaporation at the pore- and macro-scales using dynamic XCT was presented by Shokri-Kuehni et al. [66].

Sub-second analysis of dynamic pore micro-scale fluid transport processes in 4D was made by Dobson et al. [67] using SRXCT. Sandstone gravels with different size fractions were scanned. The authors presented in their study an interesting analysis of fluid dynamics showing the importance of inter-grain and intra-grain pores in fluid percolation. Liu et al. [68] studied natural sediments (sand and silty-clay) from impounded rivers to investigate methane bubble growth and movement through time-lapse XCT. Their results demonstrated the importance of pore structure, macroporosity, and pore size distribution for ebullition dynamics in sediments. Measurements of solute transport through an unsaturated porous medium (glass bead packing) with voxel size of 3.25 μm and time resolution of 6 s (3 s for imaging and 3 s for sample rotation back to the original position) were made by Hasan et al. [69]. One of the main conclusions of these authors is that the saturation topology has an important role in defining the solute transport in unsaturated porous media.

Mascini et al. [70] estimated the importance of pore-scale contact angles in fluid movement through 4D XCT. Sintered glass bead pack and limestone porous media were analyzed. The authors demonstrated that tomographic images can be utilized to analyze contact angles and capillary pressures, which exert strong influences on fluid displacement [71, 72]. Recently, Amato et al. [73] studied the importance of freezing processes in remoulded clay samples. Consolidated cylindrical samples (70-mm diameter and 70-mm height) were scanned through XCT obtaining a time series of 3D images of the freezing process. The authors showed that capillary pressures arise due to water–ice interfacial tensions and the small pore sizes cracking and displacing the clay structure, resulting in visible segregated ice structures. Petroselli et al. [74] investigated the phosphorus transport in soil combining microdialysis and XCT. The authors demonstrated the importance of the soil particle size and soil porosity on the phosphorus dispersion through time-resolved techniques, which can offer new insights about the phosphorus dynamics aiming to improve its use efficiency.

3.2 Soil-Root Interactions

A decade ago in relation to the present work, Mooney et al. [75] proposed an important review on XCT application in studies of root visualization over the previous 30 years (~ between 1981 and 2011). These authors emphasized the importance of XCT to monitor root development temporally and spatially in undisturbed environments. In their review, a majority of studies based on qualitative observations rather than quantitative analyses was recognized, indicating the need to circumvent limitations existent at that time. For instance, the discrimination between roots and water in soil at the segmentation step (usually performed by thresholding) was pointed as a major impediment due to their similar X-ray absorption properties [27]. The absorption contrast between the growth medium solids, air/water filled pores, roots, and organic matter vary with soil type/moisture, proximity of roots to organic matter or air/water filled pores, and root water status [75]. In turn, the limitations on image segmentation strongly affect quantitative interpretations of rhizosphere dynamic processes (e.g., root growth) imaged in time.

An alternative to image low-density tissues that present low X-ray absorption is to exploit the partial coherence of X-rays generated at SRs. In this case, the imaging relies on the refraction of X-rays and is referred as phase contrast imaging [76]. The experimental setups for X-ray absorption and phase contrast imaging are similar, but the later requires larger sample to detector distance. Small increases in sample to detector distance will typically just enhance interfaces between features of interest. However, as the phase signal increases with larger sample to detector distance, a condition that is favorable to achieve in SRs, the collected image becomes increasingly blurred due to interference fringe patterns and requires proper phase retrieval algorithms prior to image reconstruction. Karunakaran et al. [76] emphasized that phase contrast imaging is a powerful technique to perform real-time investigations of water-refilling process in xylem vessels of plants. These authors pointed out the need of further investigations on coupling increased photon flux from new beamlines and higher energies (> 40 keV) for above and below ground phase contrast imaging.

A known limitation for in vivo XCT experiments is also the dose of radiation to which the living plant is exposed, especially in repeated scanning as is the case of 4D investigations. In this context, Jenneson et al. [77] proposed an XCT system dedicated to image soil/plant root systems with reduced radiation dose. With such XCT system, the authors reported the capability of generating multiple time-lapse 3D images, in approximately 30 min each and a voxel size of 100 µm, of a cylindrical soil sample (25 mm diameter × 25 mm height) where a maize root would grow. The dose reported was 0.1 Gy per XCT scan, considered below the level of significant cellular damage. A routine was then employed to measure the center of mass, cross-sectional area, lengths, and volumes of each root over time. By following this procedure, Jenneson et al. [77] successfully imaged 11 mm and then 24 mm of the primary root growth in 1 and 2 days, respectively. More recently, Blaser et al. [78] evaluated the influence of cumulative X-ray dose on two different plant root species: Vicia faba and Horderum vulgare. The cultivars were grown in columns of 70 mm diameter × 250 mm height filled with a silty clay loam soil. X-ray computed tomography scanning was performed over the first 17 days after planting in two modes (n = 5 each, i.e., 5 replicates): frequent scanning (every 2 days) and moderate scanning (every 4 days). Cumulative doses were estimated as 7.8 Gy and 4.2 Gy, respectively, using the Rad Pro Calculator software (http://www.radprocalculator.com/RadProDownloads.aspx). The authors observed that the distinct cultivars responded differently to the same X-ray dose: Vicia faba was affected significantly, especially at frequent scanning, while Horderum vulgare showed no influence. Although the Rad Pro Calculator has been often reported for X-ray dose estimative in rhizosphere studies [79,80,81], some limitations of this tool have been discussed in Lippold et al. [82] (e.g., underestimation of measured dose).

The low phosphorus mobility in soils is well known due to its strong sorption on mineral surfaces, which limits the plant access to this important nutrient. Ahmed et al. [83] applied 4D XCT to investigate root development over 14 weeks after planting around discrete granules of two contrasting phosphorus fertilizers. Plastic (polyvinyl chloride) columns (110 mm diameter × 500 mm height) were filled with phosphorus deficient sandy loam soil, and wheat seeds were planted incorporating three treatments: triple superphosphate only, struvite only, and a 50:50 mixture of both (n = 3 for each treatment). The columns were imaged with a laboratory XCT system with ~ 56 min per column and voxel size of 60 µm. The authors highlighted the same aforementioned challenges related with root segmentation, leading to the choice of complete manual segmentation. Results from Ahmed et al. [83] showed that the growth of wheat root system occurred in three stages: (i) slow rate of new root production and extension of existing roots, (ii) roots get close to the nutrient granule sources, rapidly increasing the rate of root length, and (iii) plants enter grain-filling stage slightly reducing the rate of root growth (Fig. 4a). Interestingly, the authors observed that the struvite granules had very slow dissolution rate (interpreted from image-based volume analysis) when roots were absent within 5 mm from the granules. An acceleration of dissolution happened when roots entered a 10 × 10 × 10 mm3 region around the granules, slowing down again when root tips grew away from the granules. One of the hypotheses suggested by these authors to explain their observations was the direct interaction between fine roots, root hairs, and fungal hyphae with granules, influencing the phosphorus release, which however could not be tested due to limitation in spatial resolution.

Fig. 4
figure 4

(a) Change in root length over time in a soil column from the 50:50 triple superphosphate: struvite treatment along with examples of X-ray computed tomography (XCT) image cross sections (red and yellow spots represent the location of roots at the end of week 8 and 10, respectively) and 3D renderings of the segmented root system at weeks 8 and 14 [83]. (b) Schematics of experimental design showing maize supplied with superphosphate in the form of localized granules (ML) and uniform powder (MU) [80]. (c) XCT derived root length over time. (d) 3D rendering of temporal development (DAS = days after sowing) of maize root systems extracted from XCT images for localized and uniform phosphorous supply [80]. All images are reproduced with permission

Full size image

Gao et al. [80] contributed with new insights through XCT on the 4D root response to localized phosphorus supply. In this study, polycarbonate columns (70 mm inner diameter × 250 mm height) were filled with a silty clay loam soil. Two phosphorus supply patterns have been tested: uniform triple superphosphate powder and localized triple superphosphate granules (Fig. 4b), and two plant species differing in root architecture (maize and faba bean) were considered. The columns were imaged at increasing number of days after sowing with an industrial XCT system (~ 8 min 20 s per scan, voxel size 40 µm, and estimated dose at the pot wall 0.45 Gy per scan). Gao et al. [80] also reported manual segmentation of the root networks for faba bean, but a new procedure name “Rootine” was successfully applied for automate root segmentation of maize. The authors found that the phosphorus supply pattern significantly affected maize root growth, but not faba bean (Fig. 4c, d). An analysis along the sample depth profiles showed that for maize, root length increased significantly when crossing the layer where the localized phosphorus granules were placed. The authors associated these differences with the following reasons: (i) the finer root system of maize favors the increasing root length towards the triple superphosphate granule for phosphorus acquisition, (ii) in low phosphorus condition, maize invests more biomass in root growth, while faba bean acidifies the rhizosphere to increase phosphorus availability, (iii) faba bean has larger seed phosphorus reserves and exploits it for a larger fraction of its phosphorus requirement than maize, and (iv) the dependence of the two species on root hairs and mycorrizal symbioses of phosphorus uptake is different.

The availability and concentration of nitrogen in soil is also very relevant to root system architectures. Blaser et al. [79] investigated temporal and spatial dynamics of root responses in situ to different nitrogen forms: commercial urea fertilizer granules with and without a nitrification inhibitor (U + NI and U, respectively). Acrylic columns (70 mm diameter × 250 mm height) were filled with silty clay loam soil treated with U and U + NI where seeds of two contrasting root system architectures (Vicia faba and Horderum vulgare) were planted. The root growth dynamics of Vicia faba was characterized by XCT at 8, 12, and 16 days after planting, with ~ 8.5 min per scan, voxel size of 40 µm, and estimated dose of 2.3 Gy h−1 [79]. A semi-automated region growing method was used to segment roots of Vicia faba, and a new and less time-consuming type of quantitative measurement (based on relative frequencies of soil-root distance) was proposed to distinguish between tap root, first-, and second-order lateral roots [79]. For comparison, columns with Horderum vulgare went through destructive harvesting, root washing, and analysis with WinRHIZO because the XCT image resolution was not sufficient to capture the typical finer roots of this specie [79]. The only root response found by the authors that was in line with previous studies was a higher initiation of Horderum vulgare lateral roots for U + NI. However, Blaser et al. [79] emphasized that most of the preexistent investigations on root responses to nitrogen supply had been performed in artificial growth media (e.g., agar, quartz, etc.) instead of real soils, which significantly decreases the natural heterogeneity of the rhizosphere.

Another important aspect that has been studied is associated with the inhibition of root growth by soil mechanical impedance. Keyes et al. [5] developed for the first time an in vivo experiment simultaneously focusing on the 3D dynamics of soil deformation and root tip. Maize seeds were planted in polypropylene centrifuge tubes (30 mm diameter × 115 mm height) filled with a sandy loam soil (one control and one to be XCT scanned). XCT scans were preformed hourly over a growth period of 20 h, lasting ~ 7 min each, and resulting in a voxel size of 30 µm. From the root growth perspective, the root extension rate decreased gradually over time, while the angular change rate peaked at 8 h, then returning to a baseline (Fig. 5a). Digital volume correlation was applied to map full-field soil displacement around the growing plant root (Fig. 5b). From these analyses, the authors identified an inverse relationship between soil displacement and root tip extension rate, evidencing that the resistance to root extension is reduced when it grows into macropores in comparison to when it grows displacing material to form a cavity [5].

Fig. 5
figure 5

(a) The growth direction of the root over the period ∆t = tn − tn−1 is represented by the vector \(\underset{{v}_{n}}{\to }\), and its modulus defines the root extension rate, with θ denoting the change in heading between successive growth vectors [5]. (b) Displacement fields over time generated using digital volume correlation analysis with the displayed vector length (L) scaled from the actual displacement magnitude (D, associated with the color map) by a factor k, for visualization purposes [5]. All images are reproduced with permission

Full size image

Later on, Keyes et al. [84] used 4D SRXCT to reveal plant/root interactions at a magnification an order higher than that of Keyes et al. [5]. Syringe barrels of 1 ml were used as columns to prepare different soil conditions (wetting and compaction) and grow different maize cultivars (decapped and intact) (6 treatments in total) [84]. Synchrotron-based X-ray computed tomography scans were performed in monochromatic condition (19 keV), in ~ 60 s each, every 6 min over 48 min (8 scans per sample), with voxel size of 1.6 µm (n = 3 for each soil/cultivar combination, totalizing 144 3D images). Their results showed reconfiguration of primary mineral grains, textural phase, pore gas, and pore fluid as influenced by root elongation. For the first time, the authors applied a discrete grain analysis to extract primary mineral grain geometry from real soil images, demonstrating good agreement between discrete grain tracking and digital volume correlation. Such analyses allowed to conclude that grater soil deformation per unit of root growth occurred for decapped roots than intact ones, which indicates that the presence of root cap affects the ability to penetrate high strength (drier condition) soils [84].

Schlüter et al. [85] introduced an important root distance model (mixed triangular-gamma model) to describe frequency distributions of Euclidean distances from soil to root, at several growth stages of Vicia faba. Their study was based on the same XCT data/experimental setup from Blaser et al. [78] (described before). The authors demonstrated a good agreement between metrics derived from analyses based on the root perspective (root network traits) and soil perspective (root distance traits) [85]. Several reasons were pointed as advantages of quantifying root growth from the soil instead of the root perspective, but also some limitations. Therefore, the authors argued that both perspectives should be taken into account for robust and complementary investigations [85]. In addition to understanding root growth patterns for agricultural purposes, the root and soil perspectives of study are highly important to other subjects, such as the mechanical root reinforcement of soil to prevent slope instability [86]. Bull et al. [86] established a methodology for in situ XCT experiments to quantify 3D responses of soil mechanics and soil/root interactions to a direct shear load. The authors designed a test rig that would fit within the XCT scanner and accommodate a willow rooted soil specimen, consisting of two 250-mm-length sections of a cylindrical tube with a plane in between where controlled direct shear could be applied. For data analysis, they applied digital volume correlation to generate full-field 3D displacement and strain information. Their proposed methodology enabled the measurement of relative soil and root displacements and deformation during shear, which are relevant to understand the root behaviors in this process (e.g., bending, stretching, slippage at the soil/root interface, etc.).

The readers are encouraged to consult other studies that were found in the original search but have not been covered in this brief overview [87,88,89,90,91,92,93,94,95].

4 Future Perspectives

As stated in the introduction section of this survey, fourth-generation SRs are characterized by delivering a high photon flux. In addition, the X-ray sources of different techniques made available in these facilities are considerably reduced. Therefore, the substantial gains in both photon flux and source size lead to increased temporal and spatial SRXCT resolutions, respectively. At the Mogno beamline (micro and nano SRXCT beamline at the Sirius SR in Brazil), the temporal resolution will reach 1 full tomography s−1 and the spatial resolution will vary from 55 µm up to 120 nm (associated with field of views varying from 80 × 80 × 80 mm3 to 150 × 150 × 150 µm3, respectively). Here, it is important to highlight that 120 nm is the maximum true expected geometrical spatial resolution (the correspondent effective voxel sizes, usually reported in XCT related studies, is 70 nm). Additionally, the maximum sample diameter (80 mm) can be achieved depending on the level of X-ray attenuation through the sample. For a typical soil composition with bulk density of 1.3 g cm−3 and 50% porosity, we expect that 40-mm-diameter samples will be nicely imaged at 67.5 keV (the height can go up to 80 mm independent of the composition, as long as stability during sample rotation is ensured).

The flexible range in spatial resolution is not very common in SR facilities worldwide as the majority of them are designed with parallel X-ray beams. Mogno, however, will count on a cone beam geometry, giving flexibility to field of view and respective spatial resolution. Such flexibility allows what is known as zoom tomography, meaning that a same sample can be imaged in multiscale, just by repositioning the specimen across the ~ 27-m-long rail over which the cone beam spreads through. In terms of energy, this beamline will work in three quasi-monochromatic (ΔE/E approx. 10−2) tender (22 and 39 keV) and hard (67.5 keV) X-rays. Yet, a major advance that will be found in this beamline is the phase contrast imaging regime (with high phase signal) which comes along with high spatial resolutions achieved in large sample to detector distances over the X-ray cone beam. Mogno will count on a direct detection system in order to ensure fast data collection, but traditional indirect detection systems will also be available (for more details, please consult the official website in https://www.lnls.cnpem.br/facilities/mogno-en/).

As one can rapidly imagine, all the mentioned capabilities of Mogno beamline are in line with the natural evolution of the studies described in the proposed overview on fluid flow and water distribution in soil and on soil-root interactions. Therefore, next we present a discussion on limitations and research gaps, with examples identified in some of the described studies that we foresee as a potential to be overcome by applying Mogno’s new capabilities: increased spatial and temporal resolutions, multiscale analysis, high energy, and phase contrast imaging.

The fluid flow is complex and can greatly vary in time and location, which means the need for 4D XCT systems to rapidly acquire images allowing the study of such process [61]. Limitations in time acquisition will not allow following accurately the position of an infiltrating fluid and its drainage front [52, 59]. When the time and spatial resolutions are not adequate uncertainties on the water front position occur [61]. Limited X-ray flux always hampers 4D analysis of dynamic pore-scale processes [14]. This means that higher intensity of the radiation is necessary to overcome this type of limitation such as those obtained in SR facilities (e.g. Mogno beamline).

Limitations in the spatial and temporal resolutions under 4D XCT can affect the quality of the images generating artifacts depending on the scan system employed [59, 72]. Some of the artifacts generated are graining, which results from the low signal to noise ratio. As many images are obtained in 4D experiments, to follow the dynamic processes under study, precautions should be taken to avoid the X-ray tube overheating (benchtop systems), which sometimes affect the obtention of adequate signal to noise ratios. On the other hand, high signal to noise ratio is much more feasible in SRXCT, especially in those of fourth-generation SR source. Motion artifacts can sometimes also happen during fluid flow analyses due to the processes taken place in the soil. These motion artifacts can arise both from low temporal resolution to capture the event of interest and from lack of a proper sample environment. At Mogno and other SRXCT beamlines around the world, the development of in situ experimental setups (usually called sample environments) has become highly important. Such setups are custom-made for the considered XCT system, taking all specifications into account to minimize any undesirable external motion of the sample undergoing a dynamic process (e.g., fluid flow injection, root growing, localized compression, etc.). A few sample environments are already currently available at the Mogno beamline, but future users are encouraged to develop new ones that will best work to perform different in situ soil fluid flow experiments.

Sometimes, the spatial resolution needed to follow a specific event in porous media is lower than the system detection threshold [59]. This limitation is mainly related with the size of the samples analyzed having in mind the representativeness of the fluid flow experiments through the soil [50]. For fluid flow experiments, the sample size is often not representative and the volumes accessed through image analysis are not suitable for representative measurements of fluid movement in pore scale analysis conducted on multiphase images [72]. In the same way, a relevant limitation pointed by Gao et al. [80] was that the trade-off between image resolution and pot size (the better the image resolution, the smaller the sample diameter) led them to choose a coarse voxel size (40 µm) to fit relatively large pots in their study (70 mm inner diameter × 250 mm height).

A multiscale approach using zoom tomography would be very helpful to solve the reported trade-off between image resolution and sample diameter, both in fluid flow and soil-root interaction contexts. For instance, considering that a sample of ~ 40 mm diameter was scanned at Mogno with a voxel size similar to that used in Gao et al. [80], the same study could be developed with the extra benefit that regions of interest in the rhizosphere (e.g., the interface between the phosphorus granule and roots) could be selected for a zoom tomography. In this way, the sample would be dislocated to a position further from the detector (closer to the X-ray source) where the field of view would be decreased along with the effective voxel size (increase of image resolution). Note that, according to the authors, some of the pointed reasons that could be underlying the root responses to different phosphorus supply between two different plant species were not tested, for instance, the hypothesis that the species depend differently on root hairs and mycorrizal symbioses of phosphorus uptake. Indeed, the 40-µm voxel used in their study would not allow such investigation, but a multiscale analysis would enable new insights in this direction. The aforementioned representativeness issue in fluid flow analyses would also be greatly benefited by the multiscale approach.

In the study performed by Keyes et al. [84], where SRXCT was used, the trade-off between sample diameter and spatial resolution was also a limitation. In their case, a narrow column (less than 10 mm) was used to achieve the desired small voxel size (1.6 µm). However, the authors expressed a concern regarding the possible constraint exerted by the column walls on the soil displacement and root elongation. Additionally, they recognized the merit of a “complementary approach using SRXCT to probe the micro-scale fundamentals of root-soil interactions, and coarser resolution XCT to study deformation at whole and root system scales.” We, in turn, advocate that it would be even more interesting to develop a full in situ multiscale study by exploiting both 4D and zoom tomography.

Many of the studies covered in the overview section reported limitations regarding the image segmentation of phases that present similar X-ray absorption properties (e.g., air, water, and roots; also organic matter, air, water, and solid phase). As pointed by Karunakaran et al. [76], the phase contrast imaging can be a good alternative to enhance interfaces between low density materials, facilitating post image processing and segmentation. They chose a sample to detector distance that was only sufficient to enhance edges, limiting the phase signal due to lack of a proper phase retrieval algorithm that would apply to their data at that time. This is indeed one of the challenges nowadays in the realm of XCT. At the Sirius SR, the Scientific Computing Group is leading this effort with the aim to provide robust and reliable phase retrieval solutions to support Mogno and other beamlines. The phase contrast imaging coupled with 4D XCT at Mogno is expected to provide unprecedent paths in soil-related investigations. Additionally, Karunakaran et al. [76] pointed out the need of further investigations on coupling increased photon flux from new beamlines and higher energies (> 40 keV) for aboveground and belowground phase contrast imaging. In fact, this is something that will demand characterization at the Mogno beamline, especially considering that higher energies cause less radiation damage in biological samples and that soil-related investigations will take great advantage of increased sample diameter, which is maximum at the higher energy at Mogno (67.5 keV).

In conclusion, future developments in X-ray sources and detection systems can help to increase temporal and spatial imaging resolutions [14, 19]. The development of new XCT reconstruction algorithms and 4D image analysis software aids in the improvement of image quality and data processing. Therefore, new advances in computing performance are another requirement for working with large amount of data such as those obtained in time-series datasets [14]. It is of crucial importance to develop efficient, precise, and flexible protocols and strategies for image segmentation that will allow data analysis in a timely manner. In this context, the Sirius Scientific Computing Group has also developed an internal powerful software based on machine and deep learning principles, which has already demonstrated promising results for soil image segmentation.