Interactions among soil texture, pore structure, and labile carbon influence soil carbon gains

Perennial vegetation with high plant diversity, e.g., restored prairie, is known for stimulation of soil carbon (C) gains, due in part to enhanced formation of pore structure beneficial for long-term C storage. However, the prevalence of this phenomenon across soils of different types remains poorly understood. The aim of the study was to assess the associations between pore structure, soil C, and their differences in monoculture switchgrass and polyculture restored prairie vegetation across a wide range of soils dominating the Upper Midwest of the USA. Six experimental sites were sampled, representing three soil types with texture ranging from sandy to silt loams. The two vegetation systems studied at each site were (i) monoculture switchgrass ( Panicum virgatum L.), and (ii) polyculture restored prairie, also containing switchgrass as one of its species. X-ray computed micro-tomography ( µ CT) was employed to analyze soil pore structure. Structural equation modeling and multiple path analyses were used to assess direct and indirect effects of soil texture and pore characteristics on microbial biomass C (MBC), particulate organic matter (POM), dissolved organic C (DOC), short-term respiration (CO 2 ), and, ultimately, soil organic C (SOC). Across studied sites, prairie increased fractions of medium (50 – 150 µ m Ø ) pores by 11 – 45 %, SOC by 3 – 69 %, and MBC by 18 – 59 % (except for one site). The greater were the prairie-induced increases in the medium pore volumes, the greater were the prairie-induced SOC gains. Greater C losses via CO 2 and DOC contributed to slower C accumulation in the prairie soil. We surmise that the interactive feedback loop relating medium pores and soil C acts across a wide range of soil textures and is an important mechanism through which perennial vegetation with high plant diversity, such as restored prairie, promotes rapid SOC gains.


Introduction
Conversion from fossil fuels to bioenergy is a part of the global endeavors of combatting climate change (IPCC, 2014).Considering concerns of food security and land availability, cultivating perennial bioenergy plants on lands unsuitable for food production is a promising option for cellulosic bioenergy feedstocks production (Sang and Zhu, 2011;Mehmood et al., 2017).Ethanol yields from a number of perennial bioenergy plants are comparable to those of annual crops (Varvel et al., 2008;David and Ragauskas, 2010).Moreover, perennial vegetation can reduce carbon (C) losses due to soil disturbance by tillage, often necessary when growing annual crops, and can lead to soil C gains (Lal, 2001;Schulte et al., 2017).Benefits of the bioenergy cropping can be further amplified if the crops can capture atmospheric carbon dioxide and store it as soil organic matter resulting in soil C sequestration (Robertson et al., 2008(Robertson et al., , 2011)).
The fundamental principles of soil C sequestration involve promoting C inputs into the soil and stabilizing added C against immediate losses (Six and Jastrow, 2002), where physical and chemical protections of C within the soil matrix drive soil C stabilization (Clough and Skjemstad, 2000;Six et al., 2002).Plant roots and residues are major sources of soil C, while elevated soil biological activity stimulates C accumulation via the enhanced production and processing of the added C (Lange et al., 2015;Chen et al., 2018).Whether soil C is decomposed by microorganisms and their enzymes or remains protected from them depends on its physical accessibility enabled by presence and properties of soil pores (Strong et al., 2004;Kravchenko et al., 2019a and2021).
Cropping systems influence soil pore characteristics (Helliwell et al., 2019;Lucas et al., 2022b) due in part to differences in root architectures and biomass, quantities and qualities of C inputs, and rhizosphere microbial community composition (Pagliai and De Nobili, 1993;Sprunger et al., 2017;Lucas et al., 2022a).Of particular relevance for microbial activity and, thus, for processing and protection of the newly added C are the pores in the tens to hundreds micrometer size range, which we will refer to here as medium pores.Such pores are often associated with greater mineralization of newly added C (Strong et al., 2004;Quigley et al., 2018), higher microbial activity (Wright et al., 1995;Kravchenko et al., 2019a;Liang et al., 2019), and faster and greater microbial turnover (Ruamps et al., 2011;Kravchenko et al., 2021).The chemical composition of dissolved organic matter residing in pores of different size ranges also differs, leading to differential decomposition rates (Bailey et al., 2017).Previous work by our team (Kravchenko et al., 2019b) demonstrated that bioenergy cropping systems with high plant diversity developed greater volumes of medium pores than low diversity systems.This finding is important for elucidating the drivers through which diverse assemblages of perennial bioenergy plants can generate significant soil C gains (Sanford, 2014;Sprunger and Robertson, 2018), thus for developing strategies to maximize the climate mitigation benefits of biofuel production.However, Kravchenko et al. (2019b) study was based on only a single soil (sandy loam Alfisol), raising questions regarding universality of the observed phenomenon.
Pore formation and characteristics also depend on inherent soil physical properties, especially soil texture.Generally, the volume and size of pores increases with increasing size of soil particles, as larger particles are likely to give rise to larger pores in-between them (Nimmo, 2013;Ding et al., 2016;Fan et al., 2021).Thus, the formation of medium pores in soils of contrasting textures may differ in their contribution to the decomposition of soil organic matter and plant residues and to the stabilization of the processed C. Indeed, greater decomposition of soilincorporated root and leaf residues was observed when the residues were surrounded by soil with a greater abundance of > 30 μm Ø pores compared to that of the soil with prevalence of smaller pores (Negassa et al., 2015;Kim et al., 2020).Also, greater amount of the decomposed substrates was found to be occluded and subsequently protected in finetextured soils than that in coarse-textured soils (Kölbl and Kögel-Knabner, 2004;Schweizer et al., 2019;Haddix et al., 2020).Furthermore, Shen (1999) found that finer textured soils had a higher capacity for sorption of dissolved organic matter, likely caused by slow water movement and greater microbial processing of the dissolved organic matter (Don and Schulze, 2008;Kaiser and Kalbitz, 2012).Thus, we surmise that soil texture can modify the contribution of plant community composition to formation and size-distribution of soil pores and modulate the importance of pore structure for soil C processing.Modifications in pore structure may subsequently lead to changes in labile organic matter and then influence soil organic C (SOC) contents.
The first objective of this study was to compare pore size distributions (PSD), labile C characteristics, and organic C contents of the soils from two bioenergy cropping systems, a monoculture switchgrass and a polyculture restored prairie which includes switchgrass as one of the species after their multiple years of implementation at six experimental sites with disparate soil characteristics.We hypothesized that increases in volumes of medium pores and SOC in prairie vegetation, previously observed in the sandy loam Alfisol soil (Kravchenko et al., 2019b), would also be present across a range of soil types and textures.
The second objective was to explore the joint influences of the cropping systems and pore characteristics on microbial biomass C (MBC) and other forms of labile C (i.e., particulate organic matter (POM), dissolved organic C (DOC), and short-term CO 2 respiration (CO 2 )) along with the potential contributions of such influences to the long-term soil C accumulation.We hypothesize that greater formation of medium pores under prairie vegetation will contribute to greater MBC and other labile C forms and will lead to increases in SOC contents.

Experimental design and soil sample collections
The six experimental sites from Great Lakes Bioenergy Center used for this experiment were established in 2013 at Lux Arbor (LA), Lake City (LC), and Escanaba (ES) locations in Michigan and Hancock (HA), Rhinelender (RH), and Oregon (OR) locations in Wisconsin (Table 1) (Kasmerchak and Schaetzl, 2018).In three of the sites, namely LA, ES and OR, the soils belonged to Alifisol, in two, i.e., LC and RH, to Spodosols, and in HA to Entisol types, and were regarded as overall low fertility soils unsuitable for food production.
Two bioenergy cropping systems selected for the study were monoculture switchgrass (Panicum virgatum L.) and polyculture restored prairie.The restored prairie consisted of 18 plant species of grasses (including switchgrass), forbs, and legumes.At each site the experiment was set up as a randomized complete block design with four (LA, ES, RH, and OR) or three (LC and HA) replicated blocks.The two cropping systems were randomly assigned to the plots within each block.Both cropping systems shared the same managing and harvesting practices since the establishment of the six sites, that is, both were not fertilized and not tilled.Soil from the first five sites was sampled in November of 2019, and samples from OR site were collected a year later.
Two types of soil samples were collected from each replicated plot.First, three intact soil cores (5 cm in height and 5 cm in diameter) were taken from 5 to 10 cm depth to be used for X-ray computed microtomography (µCT) scanning.This specific depth was chosen because it represents the portion of the soil profile most significantly affected by the root systems.The loose soil surrounding the cores was collected for subsequent measurements of soil texture and C characteristics (MBC, POM, DOC, CO 2 , and SOC contents).All collected samples were stored at 4 • C until the measurements and scanning.

Soil measurements
The following chemical properties were measured in soil passes through 2-mm sieve at the MSU Soil & Plant Nutrient Laboratory (East Lansing, Michigan, USA): soil pH was measured in a 1:1 soil: water slurry; available phosphorus (Bray-P) was determined by Bray-Kurtz P1 (weak acid) test; concentrations of potassium, calcium, magnesium for cation exchange capacity (CEC), and metals including: zinc (Zn), manganese (Mn), iron (Fe), and copper (Cu) were measured using inductively coupled plasma (ICP) spectrophotometers after extracting these cations from soil samples (Warncke et al., 2009).Soil texture was determined using the hydrometer method (Gee and Or, 2002).
Chloroform fumigation-incubation method was used to determine soil MBC (Paul et al., 1999).Two sets of 10 g soil samples were prepared, and water was added to the samples with a pipette to achieve 50 % of their water holding capacity.Then, the samples were preincubated for days.After the preincubation, one set of the samples was fumigated for 24 h by ethanol-free chloroform vapor, while the other set of the samples remained unfumigated.After that, both sets of soils were subjected to a 10-day incubation at 20 • C in the dark.Emitted CO 2 was measured using Infrared Photoacoustic Spectroscopy (1412 Photoacoustic multi-gas monitors, INNOVA Air Tech Instruments, Ballerup, Denmark) in the gas circulation mode.Differences between quantities of CO 2 emitted from fumigated and non-fumigated samples were used to calculate MBC.The quantities of CO 2 emitted from unfumigated samples were reported as measures of the short-term respiration.This respiration indicates the amount of decomposed C substrates from soils during the incubation period (Adviento-Borbe et al., 2006;Haney et al., 2008).
Soil POM contents were determined using physical fractionation (53-2000 μm) method (Cambardella and Elliott, 1992).Specifically, Rhizosphere soil was used for the DOC measurements.Plants from both systems were carefully uprooted and manually shaken, and the soil adhering to roots, regarded as rhizosphere soil, was collected.Six grams of the rhizosphere soil was extracted by 30 ml of 0.5 M K 2 SO 4 by shaking at 200 rpm for 1 h, and the extracts were filtered with Whatman grade 202 filter paper.Concentration of DOC was determined using a vario TOC cube (Elementar Americas Inc., New York, USA).Means of 3-year DOC concentrations were used as one of the labile C characteristics in this study, besides MBC, POM, and CO 2 .

Soil core scanning and image analysis
Microscale quantification of soil pore size distributions was obtained via X-ray µCT, a tool that can provide visualization of soil in its intact state (Udawatta et al., 2008;Vogel et al., 2010).Prior to X-ray μCT scanning, all soil cores were brought to the matric potential of 28 kPa, which ensured that pores larger than 5 μm Ø were air-filled and thus easily detectable on the μCT images.For that, the cores were first saturated for 24 h in a water filled sand bath, and then transferred to a pressure chamber (5 bar pressure plate extractor, Soilmoisture Equipment Corp., Goleta, CA, USA) and kept there for 2 days at 28 kPa.Bringing all cores to the same matric potential enabled comparisons of detectable via X-ray μCT pore volumes among experimental sites and plant systems.
The cores were scanned using X-ray μCT instrument (North Star Imaging, X3000, Rogers, USA) at the Department of Horticulture facility, Michigan State University.The scanning resolution and projected energy level were 18.2 μm and 75 kV with 450 μA, respectively.The high resolution was achieved using the Subpi-mode of the scanner, combining four individual scanning's.Scanned images from 3014 projections were reconstructed by efX software (North Star, Rogers, USA).
Image analyses for size distributions of pore volumes were performed using ImageJ-Fiji software (Schindelin et al., 2012) and SimpleITK in Python (Beare et al., 2018).The following image pre-processing steps were conducted to remove artifacts and noises.First, the images were cropped into squares (2.7 cm in length, 2.7 cm in width, and 4.1 cm height corresponded to 1500 × 1500 × 2240 pixels) from the center to exclude sampling artifacts near the soil core walls.Then, 'Remove Background' tool in Xlib/Beat plugin of ImageJ software was used to remove shadowing effects from the images, and removal of ring artifacts was conducted on the image polar domain using a stripe filter of the Xlib/Beat plugin.Finally, 2-D non-local mean filter (σ = 0.1) was applied to reduce the noise (Darbon et al., 2008;Buades et al., 2011).
The pre-processed grayscale images were segmented into pore and solid binary images.Mean threshold values were obtained by averaging the thresholds derived from 8 segmentation methods (Otsu, Kittler, Triangle, Huang, ISO, Li, Renyi, and Moments) using SimpleITK (Lucas et al., 2022a).The rationale for averaging thresholds is to mitigate biases of the individual methods, thus enhancing accuracy in pore threshold calculation (Schlüter et al., 2014).Pore size distributions of 3D stacked images were determined by the 'Local Thickness' approach in ImageJ, based on the maximum inscribed sphere method (Hildebrand and Rüesgsegger, 1997;Vogel et al., 2010).The volumes of pores of different size classes were expressed as fractions of the total soil volume (mm 3 /mm 3 ).
Pores larger than 500 μm Ø were not included in further analyses as they were rarely related to soil C cycling (Nunan et al., 2003;Franklin et al., 2021).In accordance with the reported size ranges, we decided to define 50-150 μm Ø as the range of medium pores and consider this range as a potential determinant of labile C characteristics and consequent C accumulation of soil over the studied sites.

Statistical analysis
The effects of the two plant systems at the studied experimental sites were assessed using SAS 9.4 (SAS Institute Inc., NC, USA) procedures of PROC MIXED and PROC GLIMMIX.Since we did not expect that 6-7 years of prairie and switchgrass vegetation growth will influence soil texture, the statistical models for sand, silt, and clay contents included only the experimental sites as the fixed effect.The statistical models for soil C characteristics (MBC, POM, DOC, CO 2 , and SOC), volumes of medium pores, and other soil chemical properties consisted of the fixed effects of plant systems, experimental sites, and their interaction.All models included the random effects of experimental blocks nested within the sites and effects of soil cores nested within plots, plant systems, and sites.The assumptions of normality and variance homogeneity were assessed using normal probability plots, plots of residuals vs. predicted values, and Levene's tests for equal variances.
Linear relationships among the texture variables, e.g., sand content, medium pore volumes, and C characteristics were assessed using Pearson's correlation coefficients.For each of the studied soil properties, except texture, we also calculated the differences between the switchgrass and prairie systems (Δ) within individual experimental blocks of each site.Working with Δ enabled us to focus on the changes generated by vegetation differences, while reducing the influence of the inherent variability among the experimental sites.

Structural equation modeling and path analysis
Complex causal relationships through which soil texture, pores, microbial biomass, and labile C may contribute to SOC required that, in order to address the study's hypotheses, we employed path analysis and structural equation modeling (SEM) (Pérès et al., 2013;Eisenhauer et al., 2015;Zhao et al., 2019;Liao et al., 2022).Path analysis can reveal the causal relationships among a set of observed variables (Grace, 2006;Lange et al., 2015), and SEM uses latent variables, which are hypothetical constructs encompassing comprehensive effects of more than one observed variable on other variables (Grace, 2006;Eisenhauer et al., 2015).The analyses were performed using PROC CALIS procedure of SAS 9.4.The conceptual structure of the explored models is shown on Fig. S1.Individual percentage of sand and clay contents were used to construct an exogenous latent variable, "soil texture", which directly affects SOC contents, fraction of medium pores, and "labile C" (arrow 1, 2, and 3 on Fig. S1).Measured MBC, POM, DOC, and CO 2 constructed an endogenous variable, "labile C", affected by the "soil texture" and the fraction of medium pores (arrow 3 and 4 on Fig. S1).The "labile C" and the fraction of medium pores are also declared to be exogenous variables, with both direct and indirect effects on SOC contents (arrow 4, 5, and 6 on Fig. S1).The multi-relationships postulated in our model were developed based on and are supported by the literatures listed in Table 2.The models for prairie and switchgrass systems were fitted separately.
It should be noted that soil pores and SOC are closely intertwined.SOC accumulation and pore formation are linked in a feedback cycle (Marinari et al., 2000): an increase in one accelerates the increases in the other, and simultaneous examination of both sides of the cycle is not possible.Since our data set includes 6 experimental sites with very different soil textures but rather comparable SOC levels, and texture is the major primary driver for pore formation, in this study we decided to focus on the pores → SOC side of the cycle and to explore the contribution of pores as drivers of SOC accumulation.
An additional model was constructed to examine the effect of texture on the differences (Δs) between prairie and switchgrass systems in terms of the difference in the labile C characteristics (ΔMBC, ΔPOM, ΔDOC, ΔCO 2 ), fractions of medium pores (ΔMedium), and SOC contents (ΔSOC) that developed during the 6-7 years of system implementations.The values of Δs were calculated as described in the previous section, and the model for Δs used the same conceptual structure of the original model (Fig. S1).Means of sand and clay contents from both plant systems were used as an exogenous latent variable (arrow 1, 2, and 3 on Fig. S1), and Δs of the "labile C" components and the pore fractions were defined as endogenous variables under the "soil texture" (arrow 2 and 3 on Fig. S1).Δs of "labile C" components and the pore fractions were also used as exogenous variables for ΔSOC (arrow 5 and 6 on Fig. S1).
Multiple path analyses were performed to explore the effect of individual components within the "labile C" on SOC contents and hypothesized causal relationships of it with other observed variables.For that we used the same SEM concepts (Table 2) to contract four path analysis models using percentage of sand contents (arrow 1, 2, and 3 on Fig. S1).We choose to work with sand as opposed to silt content in these path analyses because the relationship between pore volumes and silt contents is known to be less significant compared to that with sand or clay contents (Ding et al., 2016).Because of a very narrow range of clay contents observed in the six studied sites (Table 1), path analyses with clay content were also found to be less informative than those with sand.Each of the four models included one of the observed variables from the "labile C" as endogenous variable being affected by sand contents and fractions of medium pores (arrow 3 and 4 on Fig. S1).Each component of "labile C" was also used as exogenous variable influencing SOC contents (arrow 6 on Fig. S1).Means of sand content were used to construct another four models including difference of each observed variable within "labile C" between the two plant systems (ΔMBC, ΔPOM, ΔDOC, ΔCO 2 ).The list of the path analysis models was attached on Table S1, and the hypothesized effects of the individual component of "labile C" were described by additional statements of Table S2.
We followed the two-index presentation strategy for model evaluation suggested by Hu and Bentler (1999).That is, a chi-square test (χ 2 ) and goodness of fit index (GFI) were used to determine model fitness and adequacy.The indices measure the degree to which the model accounts for the variance and covariance among the observed variables (Bentler, 1990).The models with χ 2 test p values > 0.05 and GFI > 0.90 are regarded as acceptable (Hu and Bentler, 1999;Eisenhauer et al., 2015).
Since the observed variables differed in their units and scales, the standardized coefficients were computed using standardized data through PROC CALIS procedure.The computed coefficients were used to compare the relative impact of the initially incommensurable variables and to indicate the strength of the paths (Kwan and Chan, 2011).

Basic soil properties
Soils of the six sites ranged in texture from silt loam to loamy sand.HA and LC sites were loamy sands with more than 80 % sand content, while silt loam of OR site contained less than 10 % of sand (Table 1).The prairie and switchgrass soils within each site did not differ from each other in terms of the soil chemical properties (Table 3).The only exception was a tendency for the higher Mn content in prairie soils, which was statistically significant in LC site.

SOC and labile C characteristics
The interaction between experimental sites and plant systems in the ANOVA tests were not statistically significant for the studied soil C variables, except for POM (Fig. 1).Across all studied sites, SOC and MBC were higher in prairie than in switchgrass system (p < 0.05) (Fig. 1a and  1b).SOC was higher than that of switchgrass in all sites (Fig. 1a), and MBC was higher than that of switchgrass in all sites, but LC (Fig. 1b).POM was higher in prairie than in switchgrass in four of the sites (Fig. 1c).There were no significant differences between the two systems in terms of either DOC or short-term respiration (Fig. 1d and 1e).

Volume of medium pores and its relationship to SOC
The pore-size distributions differed among the studied six sites and between two plant systems (Fig. S2).Since the pores within 50-150 μm Ø size range were hypothesized to be of the greatest importance for decomposition and protection of the newly added C, the fractions of medium pores were extracted from the entire data set and analyzed separately (Fig. 2).Prairie system had greater fractions of medium pores than switchgrass across all sites (p = 0.005) (Fig. 2).Examples of distributed pores within this size range in the two systems of Oregon and Lake City sites, having the least and most sand contents (Table 1), are provided in Fig. 3.
The fractions of medium pores were not correlated to SOC contents in either prairie (p = 0.19) or switchgrass (p = 0.41) soils (Fig. 4a and  4b).But the differences in SOC contents between prairie and switchgrass soils, ΔSOC, were positively correlated to the respective differences in medium pore fractions, ΔMedium, when assessed across all studied sites (Fig. 4c).

Structural equation modeling
The two plant systems showed similar trends in terms of their paths of direct and indirect effects of soil texture on SOC contents.The texture affected the fraction of medium pores and labile C in both systems (Fig. 5a), and labile C affected SOC contents (Fig. 5a).However, the medium pore fractions did not influence either labile C or SOC contents in either of the systems.
The texture influenced the differences between prairie and switchgrass in terms of the fractions of medium pores, which then affected the differences in labile C, ΔLabile C (Fig. 5b).These prairie-switchgrass differences in the fractions of medium pores along with ΔLabile C, mediated by texture, influenced the prairie-switchgrass differences in SOC contents.Prairie system had greater medium pore fractions and SOC contents compared to switchgrass (Fig. 1a and 2).Thus, the magnitude of the increases due to prairie in medium pore fractions directly and indirectly affected the magnitude of the increases in SOC contents (Fig. 5b).

Multiple path analyses
Sand contents had a direct negative effect on MBC and direct positive effects on DOC and CO 2 (Fig. S3a, S3c, and S3d).There was no direct effect of sand contents on POM in both plant systems, while the fraction of medium pores was the main driver of the positive indirect effect of sand contents on POM (Fig. S3b).For DOC and CO 2 that indirect effect was negative (Fig. S3c and S3d).Soil MBC and POM had direct positive effects on SOC contents (Fig. S3a and S3b), whereas DOC and CO 2 of the soil had no direct effect on SOC contents in both plant systems (Fig. S3c  and S3d).As expected, the effects of clay contents had opposite signs of those of sand content (Fig. S4).
The fraction of medium pores was higher in prairie compared to switchgrass system (Fig. 2), and the increases in fractions of such pores due to prairie system (ΔMedium) were positively influenced by soil sand content (Fig. 6).Sand content negatively, while clay content positively (Fig. S5), influenced the magnitude of increases in MBC due to prairie system (ΔMBC), but it did not affect either ΔDOC or ΔPOM (Fig. 6a, 6c,  and 6d).The ΔMedium led to greater ΔMBC and ΔSOC (Fig. 6a).While increase in MBC due to prairie (ΔMBC) positively influenced ΔSOC (Fig. 6a), the contributions of ΔDOC and ΔCO 2 were negative (Fig. 6c  and 6d).

Discussion
After polyculture restored prairie and monoculture switchgrass systems were in place for several (6-7) years their soils diverged in terms of Note: Bolded values of chemical properties within each site indicate that differences between prairie and switchgrass soil were statistically significant at the p < 0.05 level.
the volumes of medium (50-150 μm Ø size range) pores, as well as in terms of the soil MBC and SOC contents.The results supported our hypotheses that prolonged prairie vegetation leads to greater formation of medium pores which further stimulate soil C gains across a range of soil textures and types.When the effects of sand content and changes in the volume of medium pores were accounted for, the increases in MBC due to prairie vegetation led to increases in SOC contents over those of monoculture switchgrass, while the increases in DOC contents and shortterm respiration led to SOC decreases.

Monoculture switchgrass system is slower in soil C accrual than restored prairie
The finding that long-term implementation of polyculture prairie system increased SOC contents as compared to monoculture switchgrass (Fig. 1a) is in agreement with other published studies, which consistently observed greater soil C in polyculture system than in monoculture.High plant diversity positively affected soil C accumulation in bioenergy cropping systems of the US Midwest (Fornara and Tilman, 2008;Sanford, 2014;Sprunger and Robertson, 2018), and increasing the number of plant species directly promoted soil C gains and soil microbial biomass (Prommer et al., 2020).Increases in soil C contents were positively correlated with the plant species richness in grasslands of UK and Central Europe (De Deyn et al., 2011;Lange et al., 2015).
Plant systems with highly diverse perennial vegetation stimulate soil C accrual via several mechanisms.Among them are greater inputs of active C into rhizosphere, faster rates of microbial growth and turnover (Lange et al., 2015;Eisenhauer et al., 2017;Sprunger and Robertson, 2018), and greater nitrogen use efficiency in cases of joint presence of C4 grasses and legumes (Lange et al., 2015).Greater microbial activity is another recognized driver of soil C gains in diverse plant communities vs. monocultures (Lange et al., 2015), but this does not seem to be the case for monoculture vs. polyculture switchgrass system (Jesus et al., 2016;Zhang et al., 2017).
Greater above-and belowground productivity is yet another driver of soil C gains suggested by large scale meta-analysis (Chen et al., 2018) and experimental work (Furey and Tilman, 2021).However, switchgrass monoculture cropping system appears to be an exception, because its root biomass and aboveground productivity are massive, yet the soil C  gains are very slow to non-existent (McLaughlin and Kszos, 2005;Chimento et al., 2016;Sprunger et al., 2020).Indeed, when working with soils of LA experimental site of this study, Gelfand et al. ( 2020) reported lower root biomass in the prairie than in the switchgrass system, while harvested aboveground biomass was more than twice that in the switchgrass as in the prairie.Significantly greater aboveground  as (c) differences between soils of prairie and switchgrass systems in terms of of medium pores (Δ Medium) plotted versus respective differences in SOC contents.The differences developed after 6-7 years of system implementation and were calculated by subtracting the pore fractions and SOC contents of switchgrass system from those of the prairie system.Shown are observations from the six studied sites (dots), the linear regression fitted to the data (red), and the r 2 value for the fitted regression model (in (a) and (b) p > 0.05, in (c) p < 0.001).productivity of switchgrass was observed compared to that of prairie in all other experimental sites reported in our study, i.e., OR, ES, RH, HA, and LC (Li et al., 2022), as well as in yet another low fertility soil of the U.S. Midwest (Cooney et al., 2023).Another experiment in the direct vicinity of our LA site (Lei et al., 2021) reported no statistically significant differences in root biomass between the restored prairie and the monoculture switchgrass system.Thus, contrary to the expectations, our results and published studies suggest that the productivity does not always serve as a significant determinant of soil C accumulation.
Faster C accrual in the prairie than in the switchgrass soil was related to greater formation of medium pores (Fig. 2.).Changes in the pore systems, e.g., volumes of medium pores, likely responded to the differences in the root systems of the plant communities (Pagliai and De Nobili, 1993;Lucas et al., 2022a).We surmise that the volume of fine roots is the key contributor to the formation of medium pores (Bodner et al., 2014;Koebernick et al., 2017).Prairie vegetation has greater densities of fine (<2mm Ø) roots as compared to monoculture switchgrass (Sprunger et al., 2017), and even fine roots of the prairie tend to be thinner than those of the switchgrass.For example, only 35-71 % of switchgrass fine roots had their diameters within < 0.5 mm size range, while a number of other native grasses had more than 80 % of their fine roots in the < 0.5 mm Ø size class (De Graaff et al., 2013;Liu et al., 2022).

Contributions of soil texture and medium pores to SOC gains under prairie vegetation
The advantages of the prairie over monoculture switchgrass in terms of soil C accrual appeared to be much more pronounced in the coarsethan in the fine-textured soils of this study.For example, the SOC content of the prairie system was ~ 3 % higher than that of the switchgrass in the silt loam soil (OR), while it was 15-69 % higher in the coarser textured soils (Fig. 1a).Consistent with our findings, Juyal et al. (2021) reported only minor differences in soil C between prairie and monoculture switchgrass systems in fine-textured soils of topographic depressions, while substantially higher C in prairie than in switchgrass in the coarse-textured soils of uphill areas.Kasanke et al. (2021) reported that soil C contents under switchgrass vegetation even decreased after 6 years of growth on sandy soils.
According to our hypotheses (Table 2), the influence of the soil texture on the magnitude of increases in SOC contents under prairie vegetation was manifested through the texture's contribution to formation of soil pores, to MBC, and to C protection on mineral surfaces.Soil texture drives the size distribution of soil pores (Nimmo, 2013;Ding et al., 2016;Fan et al., 2021).Higher sand content was associated with greater presence of the medium pores in the studied soils, in both plant systems (Fig. S3).Medium sized pores are easily accessible to fine roots and root hairs (Koebernick et al., 2017;Lucas et al., 2022b).Thus, it can be surmised that prolific fine roots of prairie vegetation readily explored the existing medium pores in the sandier soils of this study, while still contributing to the formation of new ones.Bulkier switchgrass roots might have been at a disadvantage since they did not have as much access to the already existing medium size pore space.
Pores determine access to the soil organic matter by microorganisms and thus its processing and protection (Strong et al., 2004;Negassa et al., 2015).Medium pores are known to function as optimal microbial habitats, supporting high microbial activity (Wright et al., 1995;Nunan et al., 2003;Xia et al., 2022).They ensure high oxygen and water availability, yet do not limit the accessibility of microorganisms to organic matter substrates (Rawlins et al., 2016;Keiluweit et al., 2018).They were also reported as the primary locations of rhizodeposition inputs (Quigley and Kravchenko, 2022), thus providing C and nutrients for the resident microorganisms.A diversity of C inputs from polyculture prairie vegetation probably stimulated development of high species diversity in the microbial communities.The resultant greater microbial activity in such pores could lead to faster processing of new C additions and potentially to generation of more microbial decomposition products and necromass, which then can be protected within the soil matrix (Strong et al., 2004;Quigley et al., 2018;Guidi et al., 2021).
Our study does not allow us to directly examine the utilization of the medium pores by the roots of the two studied systems.However, it is reasonable to assume that if the medium pores were less used by the switchgrass with its thicker roots, they did not receive as much new C inputs and were not as attractive to microorganisms as they were in the prairie systems.It is important to emphasize that it is not just the volume of the medium pores per se that stimulates SOC gains, but the volume of such pores that are supplied with new C. Indeed, the observed SOC contents were not related to the fractions of medium pores (Fig. 4a and  4b).Yet, the differences between the two systems, i.e., ΔMedium and ΔSOC, were positively associated with each other (Fig. 4c), suggesting that formation of such pores, which in prairie system was likely accompanied by C inputs, went hand-in-hand with soil C gains.
This explanation is further supported by our MBC results.Consistent with other reports (Kaiser et al., 1992;Franzluebbers et al., 1996), MBC decreased with increasing sand contents in this study, thus was negatively correlated with the volume fractions of the medium pores (Fig. S6).However, the increase in the medium pores due to prairie vegetation, ΔMedium, promoted MBC increases, ΔMBC (Fig. 6a).Thus, extra formation of such pores via prairie vegetation, likely accompanied by the rhizodeposits and other C inputs into them, stimulated  2) and literatures (Miltner et al., 2012;Oduor et al., 2018;Prommer et al., 2020), the ΔMBC positively influenced ΔSOC, likely through greater quantities of living microorganisms and their necromass accumulation.
MBC, POM, DOC, and short-term respiration are often found to be positively correlated to each other and to the SOC contents (Franzluebbers, 1999;Dou et al., 2008;Oduor et al., 2018).Such correlations reflect joint common effects of soil texture, management practices or land use changes on soil C inputs and protection, which simultaneously drive root growth and subsequent POM inputs (Rasse et al., 2005;Ontl et al., 2015), abundance of microorganisms (Anderson and Domsch, 1989;Jinbo et al., 2006), and subsequent DOC and CO 2 production in a course of microbial activity (Jinbo et al., 2006;Mavi et al., 2012;Woloszczyk et al., 2020).Similarly, when examined across the range of the studied soils, path analysis of this study demonstrated that greater MBC, POM, and DOC is related to an overall greater SOC (Fig. S3).
Yet, analysis of the differences between the two systems enabled unraveling some of the co-variations between these labile C characteristics and offered us an opportunity to assess the mechanisms of their individual impacts on SOC gains.Prairie-induced increases in DOC and short-term respiration negatively influenced ΔSOC (Fig. 6c) -the result consistent with the expectation that greater C losses as either CO 2 or DOC slow down SOC gains.Greater DOC suggests that there is more organic C available for immediate microbial decomposition or possible loses with outflowing water into deeper soil layers (Kalbitz and Kaiser, 2008;Andrews et al., 2011), and coarse-textured soils tended to retain less DOC than the fine-textured ones (Shen, 1999;Filep et al., 2022).C mineralization from newly added C substrates can be more rapid in coarse-textured soils compared to that in fine-textured soils (Franzluebbers, 1999).
Even though other studies, e.g., Fukumasu et al. (2022) found positive associations between volumes of 30-100 μm Ø pores and POM contents, suggestive that the abundance of such pores was likely to be associated with higher growth of fine roots, the POM effect was not significant in our study.That likely was due to substantial variability of our POM data (Fig. 1).

Conclusions
Multiple years of prairie vegetation led to greater volume fractions of medium pores compared to monoculture switchgrass across several soils of the U.S. Upper Midwest (Alfisol, Spodosol, and Entisol).The magnitude of the formation of such pores tended to be greater in coarse-than in fine-textured soils.Stimulation of such pore formation by prairie system led to greater microbial biomass, which, in turn, led to greater SOC contents compared to monoculture switchgrass system.The more prairie vegetation promoted development of such pores, the higher were its SOC contents as compared to the monoculture switchgrass.On the contrary, potentially higher C losses via CO 2 respiration and DOC in the prairie system contributed to slower soil C accumulation.The study provides an evidence that the interactive feedback loop connecting soil physical characteristics of texture and pore structure with microbial activity, and C accumulation acts across a wide range of soils and is an important mechanism of C gains in polyculture prairie vegetation.

Fig. 1 .
Fig. 1.Soil C characteristics of the prairie and switchgrass bioenergy systems in the six studied experimental sites (a: SOC, b: MBC, c: POM, d: DOC, and e: 10-day CO 2 respiration) (OR: Oregon; LA: Lux Arbor; ES: Escanaba; RH: Rhinelender; HA: Hancock; LC: Lake City).Error bars represent standard deviation, and ANOVA test results are shown in the top-right corner of each figure.Symbol ** on Fig. 1c denotes statistically significant differences in POM between the two systems in each studied site at the p < 0.05 level (POM was the only variable where the Plant*Site interaction was statistically significant).OR, LA, ES, and RH sites: n = 4 and HA and LC sites: n = 3.

Fig. 2 .
Fig. 2. Fractions of medium (50-150 µm Ø) pores in intact soil cores from prairie and switchgrass systems of the six studied sites (OR: Oregon; LA: Lux Arbor; ES: Escanaba; RH: Rhinelender; HA: Hancock; LC: Lake City).Means are represented by white circles and medians by horizontal black lines.ANOVA results are shown in the top-right corner (based on 12 scanned soil cores per site from OR, LA, ES, and RH and on 9 cores per site from HA and LC).

Fig. 4 .
Fig. 4. Relationships between fractions of medium (50-150 μm Ø) pores and soil organic C (SOC) contents (a) in prairie system and (b) in switchgrass system as well

Fig. 5 .
Fig. 5. Structural equation models examining the hypothesized effects of soil texture, represented by sand and clay contents on (a) soil organic C (SOC) and (b) on the difference in SOC between soils under either prairie or switchgrass vegetation.We hypothesize that the texture influences the overall formation of medium (50-150 μm Ø) pores (a), as well as their enhanced formation under prairie vegetation (b).Directly and indirectly (through medium pores) it affects soil labile C, while all of them influence the SOC contents (a) as well as SOC increases due to diverse vegetation of the prairie system (b).Observed and latent variables are given with rectangular and oval boxes, respectively.Numbers represent standardized path coefficients, and bold arrows (solid for positive and dashed for negative) represent statistically significant effects (*, **, and *** denote statistical significance at p < 0.05, < 0.01, and < 0.001 levels, respectively).On (a), orange and blue mark the results from switchgrass and prairie systems, respectively.χ 2 and GFI values are shown under the corresponding models.

Table 1
Soil taxonomy, geographical locations, and texture of the six studied sites.
J.H.Lee et al.

Table 2
Theoretical supports for the hypothesized paths in the structural equation modeling (Arrows in the conceptual model of Fig. S1 indicate hypothesized mechanisms of the individually numbered paths).

Table 3
Soil chemical properties of the prairie and switchgrass bioenergy systems of the six studied sites.