•Soil aggregate stability is an important ecosystem property that is altered by anthropogenic disturbance. Yet, the generalization of these alterations and the identification of the main contributors are limited by the absence of cross-site comparisons and the application of inconsistent methodologies across regions.
•We assessed aggregate stability in paired remnant and post-disturbance grasslands across California, shortgrass and tallgrass prairies, and in manipulative experiments of plant composition and soil microbial inoculation.
•Grasslands recovering from anthropogenic disturbance consistently had lower aggregate stability than remnants. Across all grasslands, non-native plant diversity was significantly associated with reduced soil aggregate stability. A negative effect of non-native plants on aggregate stability was also observed in a mesocosm experiment comparing native and non-native plants from California grasslands. Moreover, an inoculation study demonstrated that the degradation of the microbial community also contributes to the decline in soil aggregate stability in disturbed grasslands.
•Anthropogenic disturbance consistently reduced water-stable aggregates. The stability of aggregates was reduced by non-native plants and the degradation of the native soil microbial community. This latter effect might contribute to the sustained decline in aggregate stability following anthropogenic disturbance. Further exploration is advocated to understand the generality of these potential mechanisms.
Soil is a physically, chemically and biologically active material that is shaped by past geologic, climatic, vegetative and anthropogenic factors. The interactions of these variables influence important ecosystem properties, such as soil aggregate stability, which is the ability of soil aggregates to resist disintegration occurring as a result of a disruptive force (Tisdall & Oades, 1982). Studies have revealed that the enhancement of soil aggregate stability is strongly related to increased carbon storage (John et al., 2005), organic matter stabilization (Six et al., 1998; Balabane & Plante, 2004), water-holding capacity (Shukla et al., 2003) and resistance to erosion (Teixeira & Misra, 1997; Barthès & Roose, 2002). Therefore, soil aggregate stability is crucial for the maintenance of ecosystem functions, such as land stabilization, soil productivity and agricultural sustainability.
The ecological relevance of aggregate stability has motivated increased research as scientists seek to understand the drivers that affect soil aggregation and stabilization. The hierarchical theory of soil aggregation proposes that primary particles (sand, silt, clay) are bound together into micro-aggregates (< 250 μm) by organo–mineral bonds, and that the micro-aggregates are held together by roots and fungal hyphae to form macro-aggregates (> 250 μm) (Tisdall & Oades, 1982; Oades & Waters, 1991; Bossuyt et al., 2001; Six et al., 2004). Recent findings have reported organic matter (Lu et al., 1998; Chenu et al., 2000) and clay mineralogy (Reichert et al., 2009) to be the main abiotic binding agents in aggregate formation and stabilization, and soil microbes have been reported as key biotic aggregating agents, particularly in micro-aggregation (Tisdall, 1994). For instance, recent studies have revealed differences in the distribution of bacterial species in different aggregate size classes (Mummey et al., 2006); however, the role of these bacteria in aggregation is uncertain. Fungi can also be prominent contributors to soil stabilization, as fungal biomass has been shown to have a stronger correlation than total microbial biomass with stable aggregates (Cosentino et al., 2006). One group of root-associated fungi, arbuscular mycorrhizal fungi (AMF), has a particularly important role in macro-aggregate formation and long-term stabilization (Miller & Jastrow, 1990), both through direct binding of particles with hyphae and through the production of glomalin, a glycoprotein that has a capacity to bind soil particles together and is positively related to soil aggregate stability (Wright & Upadhyaya, 1998; Rillig & Steinberg, 2002; Rillig et al., 2002; Wilson et al., 2009). Plant roots promote aggregation by producing substances that directly stabilize soil particles, as well as by favoring microbial activity in the rhizosphere which, in turn, affects soil structure (Jastrow et al., 1998; Chaudhary et al., 2009) The degree to which plant species vary in their contribution to aggregate stabilization requires further study (Moreno-Espíndola et al., 2007). Recent attempts to test the interactive biotic and abiotic contributions to soil stability using structural equation modeling have found that biological crusts and AMF are the largest contributors to soil stability in semiarid shrublands (Chaudhary et al., 2009), but that abiotic factors, specifically soil texture, are more determinant across a range of soil types in mesic systems (Barto et al., 2010).
Although physical and biological factors contribute to the formation and stabilization of aggregates in soil, anthropogenic disturbance may alter this process. The decline in stability of aggregates has been repeatedly demonstrated to be related to land use and soil management practices in a variety of ecosystems. Specifically, tillage can degrade aggregate stability (Six et al., 1999; Wright et al., 1999; Norton et al., 2006; Shukla et al., 2006; Pikul et al., 2007) and soil aggregate chemical composition, as organic carbon, total nitrogen and organic matter are also reduced in the aggregate fractions (Kasper et al., 2009; Pikul et al., 2009; Fernández et al., 2010). Consequently, tillage can also result in a reduction in carbon stocks (Barreto et al., 2009) and particulate organic matter (Beare et al., 1994; Pikul et al., 2007). Grazing has also resulted in the degradation of surface soil structure and reduced soil water infiltration rates of woodlands in Australia and aggregate stability in China (Yates et al., 2000; Li et al., 2007). However, grazing has had no detrimental effect in the semiarid areas of Argentina (Quiroga et al., 2009). In addition, it has been reported that increased land use intensity (mowing, grazing and fertilization) has resulted in increased soil aggregate stabilization in managed grasslands across a range of soil types in Germany (Barto et al., 2010). The findings reported so far correspond to studies of particular sites. Furthermore, there are great variations in the methods used to assess soil aggregate stability. Therefore, the generalization of the effects of anthropogenic disturbance on soil aggregate stability is limited by the absence of cross-site comparisons and the application of inconsistent methodologies across regions.
In this study, we tested whether the effects of anthropogenic disturbance on soil aggregate stability can be generalized across grasslands of North America by surveying old agricultural fields (sites with a history of tillage and/or grazing) and remnant grasslands (sites with no history of tillage and/or grazing). We also tested the correlation between aggregate stability and each site’s vegetation composition and soil physical–chemical properties. In two independent experiments, we tested to what extent the changes in soil aggregate stability could be mediated by a variation in plant and soil microbial communities. By identifying possible factors contributing to the degradation of soil aggregate stabilization, this work suggests conservation practices that could potentially promote soil stabilization.
Materials and Methods
Field sampling: patterns of soil aggregate stability in grasslands
Our experimental design consisted of paired plots of remnant prairie (R) and post-disturbance old agricultural fields (O). The goal of this paired experimental design was to compare the effects of disturbance on fields with a history of agricultural land use with remnant fields, which we presume to have had similar soil properties as O had before disturbance. Some of the eight sites selected for this study were located within Long Term Ecological Research and Nature Preserves in North America (Table 1). The sites represented tallgrass and shortgrass prairies as well as California grasslands. The historical records of land use of the sites were used to identify remnant and old agricultural field sites. The land use history accounts for two types of disturbance: post-tillage and grazing. At each site, we randomly sampled three to five plots. This hierarchical sampling (Table 1) allowed us to test for both consistent effects of disturbance and grassland types over the variation among sites within grassland types (see the Statistical analysis section).
Table 1. Grassland type, site identification code, number of plots per site, site location, soil textural class, disturbance history and non-native plants of all remnant (R) and old agricultural (O) fields included in the study
Using a clean trowel, we collected an estimated volume of 500 cm3 of intact top soil (0–20 cm) from each plot in the site. Soil collection occurred under grassland canopies after the vegetation cover had been removed. The soil samples were not bulked. Sampling was carried out between May 2007 and May 2008. In order to preserve relatively undisturbed samples, the whole soil sample, that is the intact soil sample collected from the field, was air dried over 7 d at room temperature and stored in plastic containers in a cold room, maintained at 4°C, until laboratory analysis was performed.
Following Kemper & Rosenau (1986), we took a 500-cm3 sample of air-dried soil and separated the sample into six aggregate groups (< 0.25, 0.25–0.5, 0.5–1, 1–2, 2–4 and > 4 mm) using a rotary sieve for a 2-min cycle. Each aggregate group was weighed to obtain the dry aggregate distribution. The mean weight diameter (MWD) was calculated as the sum of the mass fractions remaining on each sieve after sieving, multiplied by the mean aperture of the adjacent sieves.
For the estimation of the proportion of water-stable aggregates (WSAs), three replicates, which are referred to as technical replicates, of c. 7 g of dry aggregates from groups 0.5–1 and 1–2 mm were analyzed from each soil sample. This method targets these specific soil size fractions to enable a comparison between soils with different initial structures (Le Bissonnais, 1996). In addition, this method is often utilized in studies considering the effect of mechanical stress, such as tillage, on soil profiles (Díaz-Zorita et al., 2002). Roots, seeds and rocks were handpicked from each fraction, and the remaining sample was placed on a 250-μm sieve. The three technical replicate sieves were placed on equipment designed for wet sieving, where the sieves were pre-wetted by capillary action for 10 min. Then, the sieves holding the soil were agitated in distilled water for 20 min at 35 cycles min–1 (Kemper & Rosenau, 1986). The fraction of the aggregates that were broken down by water and passed through the sieves was considered to be the water-unstable fraction. The material that remained on the sieves was considered to be WSA and sand particles. To separate WSA from sand, the sieves were passed through a 1 M NaOH solution for 10 min. Then, the sand retained in each sieve was rinsed with distilled water and the WSA–NaOH solution was collected in stainless steel evaporation pans. The pans and sieves were oven dried at 110°C and weighed. The weights were corrected for the weight of the sodium solution fraction that remained in the samples. The water-stable fraction was calculated as the mass of aggregates that remained after wet sieving as a percentage of the initial mass of soil. The measures of WSA from the three technical replicates were averaged before statistical analysis.
Additional portions of the field-collected soil samples were used for soil chemical composition analysis by the National Research Center for Coal and Energy (Morgantown, WV, USA). The soil textural class description was obtained from the Web Soil Survey Database (Soil Survey Staff, Natural Resources Conservation Service, USDA).
Vegetation composition was assessed with three to five 1-m2 plots per site using the modified point intercept method (Barbour et al., 1999). Within each plot, a 0.75-m flag was extended into the vegetation at 14-cm intervals. Each plant that touched the flag pin was identified, and the number of times that an individual plant touched the flag was also recorded at each interval. Multiple species may have touched the flag pin at any given interval. Plant species occurrence and abundance data from each plot, obtained from the point intercept method, were used to calculate species density, species richness, evenness and Shannon’s index of diversity (Hayek & Buzas, 1997). These measures were calculated independently for native and exotic species in the plots, as our previous observations have suggested that belowground communities may differ in their response to native or exotic plant species (Vogelsang & Bever, 2009). Sites located in Indiana, Illinois, Minnesota and Kansas were sampled once at the peak of the growing season (July) and once in the spring (May–June) in order to capture the full diversity of the flora. Sites located in Colorado, New Mexico and California were sampled in October.
We set up two independent glasshouse experiments to test to what extent the changes in soil aggregate stability could be mediated by the variation in plant and soil microbial communities. In the first experiment, we tested the effects of plants on soil aggregate stability by constructing mesocosms with native or naturalized non-native plant communities that grow on typical local soil and soil microbial communities for one growing season. In the second experiment, we tested the effect of the soil microbial communities on soil aggregate stability by constructing mesocosms with native plants growing on common background soil that received soil microbial inoculation from remnant or disturbed fields.
Experiment 1: non-native plant effects on soil aggregate stability We constructed an experiment using mesocosms of typical California coastal grasslands. The experiment consisted of either all native species or naturalized non-native species. The two communities varied in species richness along a gradient of one, two, four, seven or eight species (see details of experimental design in Supporting Information Table S2). Each community type (native or non-native) contained monocultures of eight species from their respective species pools, with each monoculture replicated four times and all species being equally represented within each richness treatment.
We filled each 40-l pot to 75% capacity with a 1 : 1 homogeneous mixture of washed sand and clay loam soil from Irvine, CA, USA. We imposed an 8-wk watering and weeding regime on the pots to reduce the seed bank. As the background soil was derived from an area dominated by non-native plant species, we inoculated each pot with 225 ml of whole soil cultured from soil originating in an upland coastal grassland wildlife sanctuary (Starr Ranch, Trabuco Canyon, CA, USA) and augmented with fresh field soil to ensure that microbes from both native and non-native plants would initially be present. The soil cultures were started by mixing fresh field soil into sterile soil, which is sand background soil, and then growing a mixture of native and non-native plants as hosts for a growing season. The Starr Ranch site was characterized by native and non-native species, similar in composition to the Santa Rosa Plateau. We capped the inoculum with a 4-cm layer of sterile soil and sand, and then planted eight native or non-native seedlings into each pot, depending on the community type being modeled. Pots were hand watered as necessary to supplement natural precipitation, and each plant/microbial community was grown for 8 months, allowing sufficient time for fungal colonization, but before the roots became pot bound. Aboveground biomass was harvested, dried and weighed by species, and samples of the pot soil were removed to test soil aggregate stability and for use as inoculum to test for mycorrhizal density (as reported in Vogelsang & Bever, 2009).
Air-dried aggregates were processed similarly to field sampling, with the following modifications. Approximately 8 g of aggregates in the 1–2-mm range were loaded onto 500-μm mesh sieves and vapor wetted in a vaporizing chamber for 45 min before wet sieving. Vapor wetting improved the sensitivity of our assays on these soils.
Experiment 2: microbial mediation of differences in soil aggregate stability with site history In order to test the possibility that soil microbial composition mediates the effects of land use history on soil aggregate stability, we used whole soil from all tallgrass sites (Table 1) as inoculum in a test of the aggregate stability potential of soil microbial communities. Forty pots (four tallgrass paired sites × two disturbance levels (remnant vs disturbed) × five plots) were set up as follows: 5-l pots were filled with 3 l of a double autoclaved mixture of 1 : 1 v/v of low nutrient loamy-sand soil and glasshouse sand. On top of this, 200 cm3 of field-collected soil samples (4.76% of total soil volume) were placed and covered with another liter of double autoclaved mixture. Single individuals of pre-germinated seedlings of four native prairie plant species (Andropogon gerardii Vitman, Coreopsis palmata Nutt., Amorpha canescens Pursh and Lespedeza capitata Michx) were evenly spaced and planted in the pots. Plants were grown in a glasshouse at Indiana University, Bloomington, IN, USA for 2 yr, allowing the re-establishment of the soil microbial community. The pots were watered by hand as needed. After 2 yr, the pots were allowed to air dry for 4 wk at room temperature and all above-ground vegetation was removed before the remaining content of each pot was chopped, mixed and stored at 4°C. Three subsamples (50 ml) were taken from each pot to separately assess soil aggregate stability according to the method of Kemper & Rosenau (1986) described above, and these ‘technical replicates’ were averaged before analysis.
Before analysis of the field experiment, the measures of WSA data were arcsine-square root transformed to improve the homogeneity of variance. Statistical differences between the means of WSAs in remnant and old agricultural fields were determined using a mixed model analysis of variance (ANOVA) employing PROC MIXED in SAS (Cary, NC, USA). PROC MIXED is sufficiently robust in unbalanced designs such as ours with variable numbers of sites within grassland types and numbers of plots within sites. In the mixed model ANOVA, the grassland and anthropogenic disturbance types were the fixed effects and the sites within grassland types (tallgrass prairie, shortgrass prairie or California grassland) and plots within these sites were the random effects. Identifying the sites within grassland types as a random effect provides a conservative test of the effects of disturbance and grassland types, as these effects are tested over variation across sites within grassland types. We first tested for the overall effect of anthropogenic disturbance, and then tested the effect of the type of disturbance (tillage or heavy grazing) in a second analysis. In this second analysis, we used a priori orthogonal contrasts to test for differences between grazing and tillage. Tests of grassland and anthropogenic disturbance with the mixed model ANOVA are conservative, and the effects of disturbance are much more significant in a fixed effect model. By identifying the sites as random effects, however, we tested for robust patterns that should hold for other grassland sites in North America.
In order to test the effect of plant composition (plant density, native and non-native Shannon’s index of diversity) and soil physical characteristics (soil texture and chemistry) as potential causal mechanisms for the variation in aggregate stability, we included them as covariates in the mixed model analysis of covariance. Covariates that mediate the effect of disturbance on aggregate stability will reduce the sums of squared deviations explained by disturbance in the ANOVA (Sokal & Rohlf, 1995). To test for significance of the observed reductions, we construct F-tests on the difference in the sums of squared deviations explained by disturbance with and without the covariate. In all analyses, P < 0.05 was regarded as statistically significant.
For Experiment 1, which tests for differences in aggregate stability caused by plant community composition, we analyzed the percentage of stable aggregates (arc sine square root transformed) using a two-way analysis of covariance to assess the main effects of species richness and community type (native or non-native) using proc GLM in SAS. We included the total plant productivity from the conditioning period as a covariate because of the large growth differences observed within and among the experimental array. The experimental test of microbial mediation of differences in aggregate stability with site history was analyzed in the same way as the aggregate stability of the field soils.
The proportion of WSAs was sensitive to grassland type and disturbance. The proportion of WSAs varied significantly with grassland type (F2,5 = 5.8, P =0.05). California grasslands exhibited the highest proportion of stable aggregates, followed by tallgrass prairie and shortgrass prairie (Fig. 1). Disturbance reduced significantly the proportion of WSAs within grasslands (F1,5 = 14.2, P =0.01, Table 2). Aggregates from disturbed sites were less stable on exposure to water than those from remnant sites (Fig. 2). Sites with a history of tillage had significantly fewer WSAs than sites with a history of grazing (F1,10 = 5.5, P =0.04, Fig. 3).
Table 2. Analysis of water-stable aggregates (WSAs) and non-native plant diversity
Type 3 tests of fixed effects
Number degrees of freedom
Denominator degrees of freedom
P > F
We first present the mixed model analysis of variance, which shows a strong effect difference in WSA with disturbance. We then include potential environmental measures that may mediate this shift with disturbance, and find that the non-native diversity is a significant covariate and that the inclusion of non-native diversity as a covariate explains 60% of the variation in aggregate stability previously explained by disturbance.
Grassland × disturbance
Variance parameter estimates
P > Z
Grassland type × site
Grassland × site × disturbance
Type 3 tests of fixed effects
Number degrees of freedom
Denominator degrees of freedom
P > F
Grassland × disturbance
Variance parameter estimates
P > Z
Grassland type × site
Grassland × site × disturbance
Difference in variance explained by disturbance with covariates
The proportion of WSAs was also affected by soil physical properties and non-native plant diversity. The proportion of silt in the soil was a significant positive covariate with WSAs (F1,56 = 4.46, P =0.04). The inclusion of silt as a covariate did not modify the significance of the grassland type or disturbance. Other soil characteristics, such as nitrogen, phosphorus and other mineral contents, were not significant covariates. Further details on the physical–chemical soil properties and non-native plant diversity are provided in Table S1.
Shannon’s index of native plant diversity and both native and non-native plant abundance were not significant covariates of WSAs. Shannon’s index of non-native plant diversity, however, was a significant negative predictor of the water stability of soil aggregates (F1,56 = 6.1, P =0.02, Fig. 4). The interaction of the water stability of soil aggregates with disturbance was not significant (F1,56 = 0.23, P =0.64), indicating that the relationship was consistently negative in both remnant and disturbed grasslands (Fig. 4). When including non-native diversity as a covariate, the significance of grassland type was not affected, but disturbance was no longer significant. Including non-native diversity as a covariate explained 60% of the sums of squared deviations originally explained by disturbance (F1,5 = 11.76, P =0.02, Table 2); suggesting that non-native diversity contributes to the effect of disturbance on soil aggregate stability. Shannon’s index of non-native plant diversity was significantly higher in disturbed sites (F1,5 = 42.28, P =0.001). No differences were detected in aggregate MWD within the field samples.
In Experiment 1, soil aggregate stability was higher in pots containing native plants than in pots containing non-native plants (F1,153 = 8.4, P =0.004, Fig. 5). There were no significant effects of species richness. Plant productivity from the first year covaried positively (F1,153 = 4.3, P =0.04) with increasing aggregate stability, as expected given the physical function of plant roots in soil aggregate formation and stabilization.
In Experiment 2, soil aggregate stability was sensitive to land use history across tallgrass prairie sites. After inoculation with whole remnant soil, soil aggregate stability was significantly higher than that produced by inoculation with whole post-agricultural soil (F1,3 = 9.9, P =0.05, Fig. 6). Aggregate stability did not vary significantly among soils from the four tallgrass sites.
The effect of anthropogenic disturbance on WSAs across grasslands
We found that the proportion of WSAs varied across grasslands (California, shortgrass and tallgrass), suggesting that soil aggregate stability is a unique characteristic of ecosystems. The proportion of silt was a positive predictor of aggregate stability; however, it did not explain the variation across grassland type, suggesting that the variation in aggregate stability across grassland type may be independent of soil physical properties.
Although the soil textural class varied, the reduction in WSAs caused by disturbance was consistent across grasslands. We found that aggregates from remnant soils were more stable than those in soils with a history of anthropogenic disturbance. Moreover, the effect of soil disturbance was still evident 10–20 yr after the cessation of tillage; suggesting a long-term effect of disturbance and that the recovery of soil aggregate stability can be a slow process. Our results are consistent with previous observations of long-term effects of anthropogenic disturbance on the biochemical properties of soils (Compton et al., 1998; Jangid et al., 2010). We observed tillage to have a larger effect than grazing on aggregate stability. This result is consistent with previous work in mesic and arid grasslands (Eynard et al., 2004; Li et al., 2007), suggesting that it is a general result.
Non-native plants and WSAs
We provided two lines of evidence supporting a negative contribution of non-native species on aggregate stability. First, our statistical analysis of field samples suggested that non-native plant diversity explains part of the negative effect of disturbance on soil aggregate stability. Specifically, we found that Shannon’s index of non-native plant diversity was significantly higher in disturbed grassland sites, consistent with other findings on the influence of land-use history on plant composition (McIntyre & Lavorel, 1994; Kulmatiski et al., 2006; McDonald et al., 2009; Mosher et al., 2009). We also found that non-native plant diversity was significantly negatively correlated with soil aggregate stability (Fig. 4). This correlation could be spurious, as disturbance probably has both direct negative effects on aggregate stability and positive effects on non-native diversity. Alternatively, non-native diversity could be causally related to the reduced stability of aggregates, with non-native diversity mediating the negative effect of disturbance in two ways: non-native diversity could reduce directly stable aggregates or non-native diversity could slow the rate of recovery of aggregate stability after disturbance. In support of these causal hypotheses, the inclusion of non-native diversity as a covariate explained 60% of the variation in aggregate stability previously explained by disturbance (Table 2). Moreover, the negative relationship between aggregate stability and non-native diversity was consistent among remnant sites and among anthropogenically disturbed sites (Fig. 4). Although these field patterns are consistent with a causal role of non-native plant species in reduced aggregate stability, they are also vulnerable to spurious interpretations, and additional experiments are required for confirmation of causation.
Second, we found direct evidence for the negative effect of non-native plants on soil aggregate stability in a manipulative experiment (Experiment 1). In this study, 1 yr of association with non-native plant species typical of southern California grasslands resulted in significantly lower aggregate stability than soil associated with native species. This negative effect of non-native plants was consistent in low and high richness treatments, in contrast with the pattern observed in the field samples. We note, however, that this study differed from the field patterns in that native and non-native species were kept in separate treatments in the experiment, whereas they were mixed in many field plots. In this experiment, the physical soil matrix and the soil microbial communities were common to all pots prior to the conditioning period; thus, the changes observed after one growing season were caused by differences in native vs non-native plant communities. These mesocosms with non-native plant communities showed reductions in AMF density (Vogelsang & Bever, 2009), which may have contributed to the reduction in aggregate stability observed. Similar reductions in the density of AMF infection have been observed in non-native relative to native grasses in Texas, suggesting a potential difference in the functional aspects of the soil fungal community and plant relationship (Kivlin & Hawkes, 2011). Alternatively, altered composition of AMF in response to non-native plants has also been observed (Bever, 2002; Mummey et al., 2005; Hawkes et al., 2006, Batten et al., 2006), and these changes could also mediate the reduced aggregate stability in this study, as AMF is an important contributor of soil aggregate stabilization.
Although we observed consistent negative effects of non-native plant species on soil aggregate stability across our field sampling and mesocosm experiment, previous studies of non-native plants on soil aggregate stability have not been consistent (Rillig et al., 2002; Lutgen & Rillig, 2004; Batten et al., 2005). Studies of an invasive grass (Bromus hordeaceus) and an invasive forb (Centaurea maculosa) found negative effects on soil aggregate stability within California (Eviner & Stuart Chapin, 2002; Lutgen & Rillig, 2004; respectively). However, Batten et al. (2005) found no consistent difference across native and invasive forb species in the Northern California Coast Range. It is likely that the effect of non-native species on soil aggregate stability varies with the strength of microbial associations and root architecture. For instance, communities including B. hordeaceus showed a decreased density of mycorrhizal fungi (Vogelsang & Bever, 2009). This alteration probably results in a decrease in soil aggregate stability as mycorrhizal fungi, through physical binding and enzymatic activity, play a prominent role in soil aggregation (Rillig & Mummey, 2006). Given the consistency of our analysis over a great geographical breadth of grasslands and an experimental manipulation in the California grassland, our results suggest that non-native plant species may, on average, be poorer at aggregating soil than are native grassland species of North America. This interpretation is consistent with recent evidence that non-native plant species in North American grasslands are, on average, less dependent on mycorrhizal fungi (Pringle et al., 2009). However, further work is required to test the generality of this pattern.
Soil microbial community and soil aggregate stability
The present study provides experimental evidence suggesting that microbial activity is an important contributor to changes in soil aggregate stability. In Experiment 2, we observed greater aggregation of sterilized background soil inoculated with a small amount of fresh soil from remnant than from disturbed fields. These differences in aggregate stability are unlikely to be a direct consequence of the physical addition of the field soil. As the original remnant field soils averaged a 13% greater proportion of WSAs than soils from disturbed sites, and the experiment was inoculated by the mixing of < 5% field soil by volume, the original difference in aggregate stability could account for < 1% of the difference between the remnant and disturbed inoculation experiment (i.e. 0.05 × 0.13 = 0.0065). Instead, we observed 9.7% greater stable aggregates in pots inoculated with soil communities from the remnant sites than from the disturbed sites. This result suggests that the soil microbial function remains degraded long after anthropogenic disturbance ceases, consistent with measures of microbial composition (Buckley & Schmidt, 2003; Jangid et al., 2010).
We note that, in order to control for the effects of plant species, we used the same suite of native prairie plant species across both remnant and anthropogenically disturbed soils. Given the likelihood of ongoing local coevolution occurring between populations of plants and mycorrhizal fungi, as recently reviewed by Hoeksema (2010) and Johnson et al. (2010), there may be a concern that the higher aggregation observed with the native plant–soil microbe combination reflects local coadaptation, such that the soil communities associated with anthropogenically disturbed soils may aggregate best in association with the weedy plants that dominate following disturbance. We do not see this as likely; however, as the pairing of weedy plants with the soil communities that follow disturbance did not generate high levels of aggregation in the field, as shown in Fig. 4.
Although we have not identified the component of the soil community that caused this difference in aggregate stability, we have observed differences in composition of the mycorrhizal community in these soils (W. Kaonongbua & J. D. Bever, unpublished), and mycorrhizal fungal species have been shown to vary in their ability to aggregate soil in laboratory studies (Bedini et al., 2009). Recent findings report less fungal and more Gram-positive bacterial biomass in disturbed fields (Drenovsky et al., 2010). In particular, tillage resulted in the alteration of the soil microbial community structure (Helgason et al., 2010), and in the reduction of the proportion of culturable heterotrophic bacterial communities able to produce soil bindings (Caesar-TonThat et al., 2010). Further research separating the effects of AMF and other microbial guilds within soil communities could provide important information on which community members are the strongest drivers of soil aggregate development and stability.
We found that anthropogenic disturbance generally has negative effects on soil stabilization, and that these effects can persist for many years following the cessation of disturbance. Given the importance of soil stability to overall ecosystem health, its restoration should be a priority for ecosystem management. Based on our data, we can suggest that the restoration of the native plant community and the repopulation of native soil microbes will help to improve soil aggregate stability in post-disturbance communities. Although restoration practitioners have developed effective approaches for the re-establishment of native plant communities (Middleton et al., 2010), the potential additional benefits on soil stabilization remain to be tested. Although a few studies have shown the potential for the re-establishment of native soil communities to improve native plant restoration (Bever, 2003; Rowe et al., 2009; Middleton & Bever, 2012), more work needs to be performed to confirm the utility of native microbe inoculation in the restoration of soil aggregate stability.
Anthropogenic disturbance was consistently associated with reduced soil aggregate stability across North American grasslands. Tillage had a greater negative effect than grazing on soil aggregate stability, and its effects can last for at least 20 yr. In addition, we found evidence that the stability of soil aggregates was weakened by the increase in non-native plants and the alteration of the native soil microbial community that followed disturbance. Overall, this work illustrates that non-native plants and soil microbial communities are potential contributors to the consistent reduction in soil aggregate stability, which is an important measure of ecosystem function. Although further work is required to test the generality of the role played by plant and soil microbial composition in soil stabilization, focus on the restoration of these factors may enhance soil stability, a key to soil conservation and erosion prevention.
Gratitude is expressed to the Fulbright/LASPAU (Academic and Professional Programs for the Americas) Faculty Development Program for providing a scholarship to J.D. This research was supported by a grant from the Inter-American Institute for Global Change Research (IAI) (grant CRN2014) and National Science Foundation grants GEO-04523250, DEB-0616891 and DEB-0919434.