It has been suggested that some management practices and farming systems that promote C sequestration may exacerbate the risk of groundwater pollution due to fast preferential transport in soil macropores. However, soil organic C (SOC) may also impact the soil pore structure at scales smaller than the macropore scale, where complexes of SOC and clay form microaggregates that may increase pore volumes in the micrometer size range. These effects of SOC per se on pore network architecture, water flow, and solute transport have hardly been investigated. Therefore, to investigate this question, we measured tracer transport through soil cores sampled along a transect on a field under grass–clover ley with a natural gradient in SOC content. The strength of preferential transport was characterized at two flow rates (2 and 5 mm h−1) and related to the volume, size distribution, heterogeneity, and connectivity of pore networks quantified by X-ray tomography. The results showed that soils with a larger SOC content had larger volumes of pores in the smallest imaged size range (200–600 μm) that were also more uniformly distributed. These effects of SOC on the imaged pore networks were only apparent up to a threshold value of the ratio between clay and SOC of 10:1, which is assumed to correspond with the amount of SOC needed for C saturation of the clay fraction. The increased flow capacity of these smaller macropores in soil columns with larger SOC contents prevented flow from being activated in larger pores, which significantly reduced the strength of preferential transport.
Soil organic C is intimately associated with the structure of the soil, which may be defined as the three-dimensional arrangement of the soil solids and pore space (i.e., the volume, size distribution, and connectivity of pores). Soil structure is critical for sustainable crop production and food security because it enhances infiltration and drainage of water, reduces surface runoff and erosion, and improves soil aeration, plant root growth, and nutrient uptake (Lal, 2004a, 2004b; Johnston et al., 2009; Powlson et al., 2011). Indeed, it has been suggested that the global value of soil structure for crop production could be on the order of US$300 billion annually (Clothier et al., 2008). Historically, arable cropping has depleted soils of organic C, which has contributed significantly to CO2 emissions (Lal, 2004a, 2004b). Thus, changing land use and soil and crop management practices to reverse this process by sequestering organic C in degraded agricultural soils has the potential to both contribute to mitigating climate change and restore the productive capacity of degraded agricultural land (Lal, 2004b). However, SOC sequestration may also involve adverse effects and trade-offs that are not yet fully understood. One concern that has been expressed (e.g., Sojka and Upchurch, 1999; Bates et al., 2008) is that some management practices (e.g., fertilization, organic amendments) and farming systems (e.g., rotational leys, cover crops, conservation tillage) that promote SOC sequestration will enhance soil macroporosity (Jarvis, 2007) and may therefore exacerbate the risk of groundwater pollution due to preferential flow.
However, the effects of SOC per se on solute transport in arable soil have hardly been investigated and are not well understood. The results from two earlier studies that attempted to address this question were inconclusive (Vendelboe et al., 2013; Soares et al., 2015), probably because the range of variation in SOC content (17–22 g kg−1) was too small. Other studies performed at the field and catchment scales for a limited number of soil columns have suggested that preferential transport may be weaker in conventionally tilled arable topsoils with larger SOC content (Jarvis et al., 2007; Ghafoor et al., 2013; Paradelo et al., 2016). Furthermore, a recent global meta-analysis of the effects of soil properties and site attributes on tracer transport also identified SOC as a potentially important factor reducing preferential flow in soil, although a correlation between SOC and clay content made unequivocal interpretation of the data difficult (Koestel and Jorda, 2014). One likely explanation of these results is that SOC and pore space architecture are linked at multiple scales, not only at the scale of millimeter-sized macropores but also at smaller scales where SOC sequestration may increase the volume of pores in the micrometer size range (Peth et al., 2008; Schlüter et al., 2011). In this respect, complexes of SOC and clay are thought to form the fundamental building blocks of soil microaggregation that provides long-term physical protection of organic matter against decomposition (Hassink, 1997; Six et al., 2002, 2004) and improves structural stability in tilled soils (Watts and Dexter, 1997; Dexter et al., 2008).
Extensive, well-connected pore networks in the micrometer size range would tend to promote more uniform transport by increasing the near-saturated hydraulic conductivity, thereby reducing the risk of “triggering” and sustaining fast preferential transport in the larger macropores (Jarvis et al., 2012; Larsbo et al., 2014; Katuwal et al., 2015). The objective of this study was to test this hypothesis by quantifying the effects of SOC on soil structure and non-reactive solute transport for a clay soil. To achieve this objective, we conducted non-reactive tracer transport experiments under steady water flow through soil cores sampled in a field with a large natural gradient in SOC content. The strength of preferential transport, characterized by the normalized 5% arrival time of the applied solute, was measured at two flow rates (2 and 5 mm h−1) and related to the volume, size distribution, heterogeneity, and connectivity of macropore networks quantified by X-ray tomography on the same samples.
Materials and Methods
Field Site and Sampling
The site is located at Måtteby (59°43′59″ N, 16°55′7″ E), approximately 50 km west of Uppsala in eastern Sweden. The USDA texture class of the soil is clay (63.7–70.5% clay, 28.9–34.9% silt, 0.6–1.6% sand). This site was chosen because preliminary measurements indicated a large gradient in SOC content in the direction of a slight slope (<1°) at the site, with the largest SOC contents at the valley bottom, close to a drainage ditch. The SOC gradient is probably a legacy of decreasing water table depths toward the valley bottom. The field was under grass and clover ley (50% timothy [ Phleum pratense L.], 25% red clover [Trifolium pratense L.], 20% meadow fescue [Festuca pratensis Huds.], and 5% white clover [Trifolium repens L.]) at the time of sampling and had not been plowed since autumn 2012.
On 20 May 2015, 20 soil columns were taken in polyvinyl chloride (PVC) cylinders (height 20 cm, inner diameter 12.7 cm) along a transect, starting close to the drainage ditch and finishing 100 m farther upslope. The soil samples were taken from the topsoil using a tractor-mounted hydraulic press. The soil surface was left undisturbed. Measurements made on samples extracted from each column (see below) showed that the organic C content decreased from 151 to 42.4 g kg−1 along the transect in the upslope direction, while clay contents (expressed as a fraction of the mineral soil mass) increased slightly, from 63.7 to 70.5% (Fig. 1). Because of the considerably larger relative variation in organic C content compared with clay content, we assume that any differences in pore network characteristics and transport can be attributed mainly to differences in SOC content.
The GE Phoenix v|tome|x m X-ray scanner installed at the Department of Soil and Environment at the Swedish University of Agricultural Sciences was used to scan the columns. It has a 240-kV X-ray tube, a tungsten target, and a GE 16″ flat panel detector. We collected 2000 radiographs per column with a discretization of 2024 by 2024 pixels. The X-ray scans were performed at a voltage of 150 kV and a current of 300 μA. The exposure time for each radiograph was 333 or 500 μs, varying among columns depending on the soil’s density. A 3-mm-thick Cu filter was attached to the detector to reduce beam hardening artifacts. The radiographs were inverted to a three-dimensional image using the GE image reconstruction software datos|x. The resulting spatial resolution of the reconstructed three-dimensional images was 100 μm in all directions. X-ray tomography imaging was performed in June 2015, prior to the solute transport experiments.
Image processing and analysis were performed using the Fiji distribution (Schindelin et al., 2012) of the software ImageJ (Abramoff et al., 2004) including the plugin BoneJ (Doube et al., 2010) and R (R Core Team, 2016). First, the resolution of each image was reduced by a factor of two in all directions to limit the computation time during the following image processing steps. As a result, each image voxel had an edge length of 200 μm. A three-dimensional median filter with a radius of one voxel was then applied to reduce noise. Illumination differences between and within scans in the vertical direction were corrected using the gray values in the PVC cylinder and the air inside the column as references. There were no differences in illumination in the radial direction. A single gray value threshold was then applied to all images. This approach resulted in a satisfactory separation between the pore space and the soil solids. All visible pores are referred to as macropores based on the definitions given by Jarvis (2007).
Pore Space Characteristics
The total imaged porosity and volumes of pores in three size (“thickness”) classes (0.2–0.6 , 0.6–1.2, and >1.2 mm) were calculated using BoneJ (Doube et al., 2010). The pore thickness is defined for each pore voxel as the diameter of the largest sphere that fits into the pore and contains the voxel. The Particle Analyzer in BoneJ was used to label all individual pore clusters in an image. The files containing the labeled pore clusters were imported into R to calculate the connected porosity, which is defined as the porosity connected to both the top and the bottom of the sample. The mass fractal dimension, a measure of the heterogeneity of the spatial distribution of imaged porosity (i.e., how well the macropore network “fills up” the sample volume), was calculated with the box-counting algorithm in ImageJ. Pore networks were analyzed both for the whole sample volume and for cylindrical subvolumes (8 cm high, 10-cm diameter) located centrally within the soil columns 3 cm below the average depth of the soil surface. The subvolumes, which were used for analysis of the effects of SOC on pore network measures, were assumed to be free from artifacts created at sampling close to the column walls. We used pore network measures calculated for the whole sample volume in the analysis of effects of differences in structure on preferential transport.
Solute Transport Experiments
The columns were set up in a laboratory irrigation chamber allowing free drainage at the base. Tracer breakthrough experiments were performed at two different irrigation rates (2 and 5 mm h−1). In Sweden, a rainfall intensity of 5 mm h−1 has a return period of 0.4, 4, and 30 yr for durations of 2, 6, and 12 h, respectively (Hernebring, 2006). In each experiment, the columns were first irrigated for 5 d with artificial rainwater prior to the first solute application to achieve stable flow rates and background electrical conductivities in the effluent. A 2-mL pulse of KBr solution (250 mg Br mL−1 water) was applied during approximately 30 s at the surface of each column using a pipette. Application within 10 mm of the column walls was avoided to limit transport through any artificial pores created at sampling.
Bromide breakthrough curves were derived from electrical conductivity measured in the effluent at 5-min intervals, following the method described by Larsbo et al. (2014). The soil columns were weighed directly after the irrigation nozzles had been turned off to enable calculation of the degree of saturation for the whole column and in the imaged pore space during the transport experiments at both flow rates from measured water contents and porosities (see below). Ponding occurred on one column during irrigation at 2 mm h−1 and seven columns at 5 mm h−1. These were excluded from the subsequent analysis to ensure that flow rates were identical through all columns.
Preferential transport is associated with an early arrival and pronounced “tailing” in tracer breakthrough curves. In this study, we used the normalized 5% arrival time, t0.05, as a measure of preferential transport (Knudby and Carrera, 2005; Koestel et al., 2011). This measure, which reflects the tendency for early arrival of the solutes, is defined as the time between solute application and the arrival of the first 5% of the applied solute at the bottom of the column normalized to the first temporal moment of the breakthrough curve. Small values of t0.05 thus indicate a high degree of preferential transport.
After completion of the breakthrough experiments, the soil samples were split into three parts: the uppermost 3 cm, 3–11 cm (corresponding to the location of the cylindrical subvolume chosen for structure characterization), and the remaining soil at the bottom of the columns. The top and bottom parts were dried at 105°C and weighed. The central part was split into two approximately equal-sized parts. One half was dried at 20°C and then analyzed for particle densities, SOC content by dry combustion on a TruMac CN (LECO Corp.), and particle size distribution by the pipette method. The other half was dried at 105°C and weighed. The average degree of saturation in the imaged pores and total pore space during the solute transport experiments was calculated from the weights of these samples, the volumes of the samples, and the imaged pore volumes determined by X-ray, particle densities (assuming that the value measured at 3–11 cm was representative for the whole column), and the weights of the soil columns at the end of the solute transport experiments.
Spearman rank correlation coefficients, ρ, were calculated between imaged pore space characteristics, soil properties, and t0.05. The statistical significance of the correlations was calculated from exact permutation distributions in R (R Core Team, 2016). Results were considered significant for p values <0.05.
Results and Discussion
The column breakthrough experiments showed that soils with larger SOC contents were associated with weaker preferential transport. The 5% arrival time of the tracer, t0.05, was positively correlated with the SOC content at both flow rates (Fig. 2; ρ = 0.90, p < 0.001 and ρ = 0.54, p = 0.047, respectively). A threshold behavior can be discerned at the smaller of the two flow rates (2 mm h−1), with t0.05 values increasing at low SOC contents, reaching an average value of about 0.4 at SOC contents larger than approximately 60 to 70 g kg−1. A t0.05 value of approximately 0.4 is within the range of values found for artificial homogeneous porous media as well as sieved and repacked soils without any macrostructure (Koestel et al., 2012), which suggests convective–dispersive transport in the high-SOC columns at 2 mm h−1. Preferential transport was stronger at the 5 mm h−1 flow rate. One soil column in particular showed much stronger preferential transport (t0.05 ≈ 0.1) at this flow rate despite a high SOC content (≈140 g kg−1, Fig. 2), which suggests that flow in large biopores was triggered in this column at the higher irrigation rate. The t0.05 was negatively correlated with the degree of saturation of the total pore space at the 2 mm h−1 flow rate (ρ = −0.52, p = 0.021), which confirms that preferential solute transport was increasingly triggered as larger pores became active. Finally, Fig. 2 also shows that many of the columns were far from fully saturated during the irrigation experiments.
Figure 3 shows examples of the imaged pore networks for columns of low, intermediate, and high SOC contents. The porosity in the sample with a low SOC content was dominated by a few large macropores, while the pore networks in the samples representing soils of intermediate and high SOC contents were characterized by a dense, homogeneously distributed network of smaller pores in addition to the larger macropores (Fig. 3). Similar effects of SOC on the soil pore space architecture have been documented in two other recent X-ray tomography studies (Peth et al., 2008; Schlüter et al., 2011). These differences in the soil pore space architecture are illustrated quantitatively in Fig. 4b and 4c. Columns with high SOC contents were associated with large porosities in the smallest imaged size class (0.2–0.6 mm in diameter) (ρ = 0.59, p = 0.0067), whereas no significant correlation was found between SOC and either the total imaged porosity (Fig. 4a) or the porosity in the two largest pore classes (Fig. 4b). The calculated fractal dimensions showed that the pore space in the samples with smaller SOC content was less uniformly distributed (Fig. 4c; ρ = 0.50, p = 0.026). The fraction of the imaged porosity that was connected to both the top and the bottom of the analyzed subvolume ranged from zero for the two columns with the smallest SOC contents to an average value of 0.56 for the five columns with the largest SOC contents (Fig. 4a). The SOC/clay ratio has been suggested to play an important role in soil structure formation. This is because a fraction of the clay is assumed to be uncomplexed below a threshold value of this ratio when the soil becomes “saturated” with C (Hassink, 1997; Dexter et al., 2008). Based on an analysis of Polish and French agricultural soils, Dexter et al. (2008) proposed that this threshold should be around 0.1. Noting that clay contents were between 63.7 and 70.5% (Fig. 1), Fig. 4b and 4c suggest a similar threshold value above which the effects of SOC on the volumes of pores <1.2 mm in diameter and the fractal dimension become negligible.
All pore network measures for the whole volume were significantly correlated with the same measures for the subvolume (not shown). However, imaged porosities were, on average, about twice as large for the whole sample. The reason for these differences is that the whole volume contained artificial pores close to the column walls created at sampling. For the subvolumes, three columns (all with SOC < 60 g kg−1) did not have a continuous path connecting the top of the subvolume to the bottom. For the whole volume, such a path existed for all columns. Figures 5a and 5b show that t0.05 was positively correlated with the volume of imaged pores <0.6 mm in size (ρ = 0.75, p < 0.001 and ρ = 0.85, p < 0.001 for the 2 and 5 mm h−1 irrigation rates, respectively) and with pores in the size class of 0.6 to 1.2 mm, although less strongly in this case (ρ = 0.66, p = 0.0029 and ρ = 0.78, p = 0.0026). In contrast, the volume of macropores >1.2 mm in size was not correlated with t0.05 (Fig. 5c). The positive correlation found between SOC and t0.05 (Fig. 2) would probably be weaker for flow rates >5 mm h−1 because more of these larger macropores would be active in the transport.
The results from our study show that high-SOC soils were characterized by pore networks in the micrometer size range (200–1200 μm) that had a sufficiently large hydraulic conductivity to prevent the occurrence of fast preferential transport in the larger macropores at the applied flow rates. These effects of SOC on pore network characteristics and preferential flow were most noticeable at the smallest flow rate of 2 mm h−1 and in columns with smaller SOC contents. No apparent effects were found above a threshold SOC content equivalent to C “saturation,” assuming that saturation occurs at a ratio of organic C to clay content of 0.1. With this definition, about 30% of Swedish arable topsoils have a deficit of SOC, which suggests that there may be considerable scope in Sweden for improving soil and water quality through policies that promote C sequestration. However, we reported data from only one field site, and further work would be needed to establish the general validity of our results for other soil types.
We thank the land owner, Dr. Martin Larsson, for letting us take samples from the field at Måtteby. The work was funded by the Swedish Research Council Formas.