Variation of Soil Aggregation along the Weathering Gradient: Comparison of Grain Size Distribution under Different Disruptive Forces

The formation and stabilization of soil aggregates play a key role in soil functions. To date, few studies have been performed on the variation of soil aggregation with increasing soil weathering degree. Here, soil aggregation and its influencing factors along the weathering gradient were investigated. Six typical zonal soils (derived from similar parent materials) were sampled from temperate to tropical regions. Grain size distribution (GSD) in aggregate fragmentation with increasing disruptive forces (air-dried, water dispersion and chemical dispersion) was determined by laser diffraction particle size analyzer. Different forms of sesquioxides were determined by selective chemical extraction and their contributions to soil aggregation were identified by multiple stepwise regression analysis. The high variability of sesquioxides in different forms appeared with increasing free oxide content (Fed and Ald) from the temperate to tropical soils. The transformation of GSD peak to small size varied with increasing disruptive forces (p<0.05). Although in different weathering degrees, zonal soils showed a similar fragmentation process. Aggregate water stability generally increased with increasing soil weathering (p<0.01), with higher stability in eluvium (A) horizon than in illuvium (B) horizon (p<0.01). Crystalline oxides and amorphous iron oxides (Feo), especially (Fed-Feo) contributed to the formation of air-dried macroaggregates and their stability against slaking (R2 = 55%, p<0.01), while fine particles (<50μm) and Feo (excluding the complex form Fep) played a positive role in the formation of water stable aggregates (R2 = 93%, p<0.01). Additionally, water stable aggregates (including stability, size distribution and specific surface area) were closely related with pH, organic matter, cation exchange capacity (CEC), bulk density (BD), and free oxides (including various forms) (p<0.05). The overall results indicate that soil aggregation conforms to aggregate hierarchy theory to some extent along the weathering gradient and different forms of sesquioxides perform their specific roles in the formation and stabilization of different size aggregates.


Introduction
techniques. The main objectives of this study were: (i) to compare grain size distributions under different treatments and further check the soil aggregation process; (ii) to identify the major factors that affect soil aggregation along the weathering gradient. These results will supplement the theoretic aggregation mechanism of soils in subtropical-tropical regions. All the abbreviations used in this paper are summarized in Table 1.

Materials and Methods
We declare that all soil samplings were performed under the permission of the owners of the farmlands and the field studies did not involve endangered or protected species.

Study site and soil sampling
For a better understanding of soil aggregation, five typical zonal soils derived from quaternary clay and one temperate soil were selected in a similar soil texture. These soils from central and south China covered three monsoon climatic regions (temperate, subtropical and tropical monsoon climate). The mean annual temperature and precipitation ranged from 14 to 20°C and from 640 to 1778 mm respectively, indicating an increasing trend of water and heat conditions contributing to soil weathering. The sampling sites were separately located in Zhengzhou City (ZZ) in Henan province, Xiangyang City (XN), Jingshan County (JS), Xianning City (XN) in Hubei province, Changsha City (CS) in Hunan province, Shaoguan City (SG) in Guangdong province, China. The soils ZZ derived from alluvium were classified into Cambosols, XY, JS and XN into Argosols, CS into Ferrosols, and SG into Ferralosols according to Chinese Soil Taxonomy [22], which were Cambisols (ZZ), Luvisols (XY and JS), Alisols (XN and CS) and Acrisols (SG) according to the WRB (2014). All sampling sites were located on gentle slopes or plain with the slope gradient smaller than 5%, and were cultivated by human beings with intact genetic soil profile and slight erosion. Detailed information of soil sampling sites including locations, climates, farming systems, topographies, soil groups and soil profiles is listed in Table 2 and Fig 1 (S1 Fig).
At each site, a soil profile pit was dug, and soil samples were taken from two distinct profile horizons, namely eluvium (A) and illuvium (B), from July to September in 2012. Undisturbed soil cores (100 cm 3 ) were taken for the analysis of bulk density. As previously reported, the structure of aggregates smaller than 3 mm was not influenced by cropping history or other external forces, but by the nature and proportions of soil constituents [23]. Therefore, in this study we used soils smaller than 2 mm as experimental objects. After collection, the field-moist disturbed soil samples were gently broken up manually so that the large clods broke along natural fissures, and then fully air-dried and ground through 2 mm sieve after the removal of large roots and other fresh organic materials.

Soil analyses
Soil basic properties were determined using standard analytical methods: bulk density (BD) by the core method; soil pH in a 1:2.5 soil/water mass ratio by the pH meter; soil organic matter (SOM) by acid potassium-dichromate (K 2 Cr 2 O 7 , c = 0.4M) digestion; cation exchange capacity (CEC) by ammonium acetate exchange method. [24] Free Fe and Al oxides (Fe d and Al d ) were extracted by dithionite-citrate-bicarbonate (DCB) [25]; amorphous Fe and Al oxides (Fe o and Al o ) by acid ammonium oxalate [26]; complex Fe and Al oxides (Fe p and Al p ) by sodium pyrophosphate [27]. The Fe and Al extracted by the above-mentioned procedures were determined by ICP-OES (VISTA-MPX, Varian, America) after dilution. All analyses were run in triplicate and averaged for statistical analysis (Table 3). In addition, we calculated the contents of crystalline oxides (Fe d -Fe o and Al d -Al o ) and the amorphous oxides excluding complex oxides (Fe o -Fe p and Al o -Al p ) by the difference between free and amorphous oxides, and between amorphous and complex oxides, respectively [28]. These indices would be selected for further multiple stepwise regression analysis.

Laser diffraction analyses
Selected soils (<2mm) were pretreated using the following three treatments: (i) air-dried (AD); (ii) water dispersion (WD), 1g of each samples was immersed in water for 24 h and then shaken for 2 h; (iii) chemical dispersion (CD), 0.3~0.5g of each sample was fully-dispersed in dispersant (0.5mol L -1 sodium hydroxide or sodium oxalate). Following each treatment, grain (including aggregates and particles) size distribution was analyzed using the Mastersizer3000 equipment, which measures the volume content (%) in 100 bin distribution ranging from 0.01 to 3000μm. In the three treatments, AD sample was measured with laser diffraction technique in dry mode (the particle absorption index was 0.3, particle refractive index was 1.60 and dispersant refractive index was 1) and the other two treatments (WD and CD) in wet mode (the particle absorption index was 0.1, particle refractive index was 1.60 and dispersant refractive index was 1.33) [19,29]. A background measurement was performed firstly to subtract the ambient light signal from the total scattered light received from the sample, followed by five consecutive analyses of soil sample per lens. The data of five measurements were averaged to obtain relative volume data. In this study, it is assumed that the volume contents of grain sizes measured by the dry and wet mode in laser diffraction technique are comparable to the previous studies [19,20]. Besides, we also obtained other parameters, such as specific surface area (SSA) and characteristic grain size (d 10 , d 30 and d 60 ) [30]. d 10 , d 30 and d 60 denote constrained grain size, median grain size and effective grain size respectively, which referred to the diameters corresponding to the soil cumulative volumes of 10, 30 and 60%. They could be used to describe soil grain size distributions. The indices, uniformity coefficient (C u ) and curvature coefficient (C c ), were calculated based on these three characteristic grain sizes to describe the uniformity of soil grain size distribution [30]: The larger C u value indicates the broader range of GSDs and higher unevenness of grains. C c reflects the entire morphology of grain size cumulative curve, especially the distribution of the grains ranging from d 10 to d 60 .
Mean volume diameter (MVD, mm) was calculated as follows: Where v i (%) is the volume percentage of the grains in the bin i with the average diameter r i (mm). MVD AD , MVD WD and MVD CD denote the mean volume diameters in AD, WD and CD treatments. To evaluate the fragmentation degree of air-dried aggregates after being immersed in water and the aggregation degree of water stable aggregates from particles, detachability index (DI) and aggregation index (AI) were calculated by the following formulas [31]:

Data analysis
To investigate the similarity of soil aggregation among these zonal soils in different weathering degrees, cluster analysis was conducted on the volume differences (ΔV, %) of different size grains between the treatments of WD and AD, and between CD and WD, respectively. The Euclidean square distance was adopted and the soil samples were classified into three clusters. The differences among the selected soils were statistically analyzed by three-way ANOVA (p<0.05). Normality tests were carried out for all variables using the Shapiro-Wilk method. The variables not conforming to normal distribution were transformed by natural logarithmic treatment. Pearson's correlations were made among variables at the level of p<0.05, 0.01 and 0.001. Multiple stepwise regressions were performed between soil aggregate stability (MVD WD , SSA WD , C u , C c , DI and AI) and basic properties (pH, BD, SOM, CEC, SSA CD , >50μm, 2~50μm, <2μm, MVD CD , different forms of sesquioxides including Fe d -Fe o and Al d -Al o , Fe o -Fe p and Al o -Al p , Fe p and Al p ) to determine the variables accounting for the majority of soil aggregation indices. Besides, the variance inflation factor (VIF) was adopted to evaluate the collinearity between the explanatory variables. The VIF <5 indicated the weak collinearity. All data analyses were performed using the SPSS16.0 [32].

Physicochemical properties
The basic physicochemical properties of the studied soils are listed in Table 3. The selected soils generally exhibited alkaline to strongly acidic from the central to south China, and the pH of B horizon was higher than that of A horizon. Soil organic matter content was lower in B horizon than in A horizon, and the same trend was observed in CEC except for ZZ. Bulk density showed the least variation with CV of 9% among all the tested properties. The ranges of Fe d and Al d were 7.02~150.35g kg -1 and 1.04~21.21 g kg -1 separately with the CVs of 95% and 82%. The contents of free oxides increased significantly from central to south China and Fe d was remarkably higher than Al d , indicating the increase of the soil weathering degree from ZZ to SG [21]. Amorphous oxides (Fe o and Al o ) varied in a narrow range with the CVs of 41% and 33%. Complex oxides also exhibited high variations, especially Fe p with a CV as high as 96%; besides, the content of Fe p was much lower than that of Al p .  (Table 4).

Grain size distribution
When subjected to slaking in water, the studied soils displayed significant aggregate disruption (Fig 3). The peaks of GSDs in WD treatment shifted to smaller sizes to some extent relative to AD treatment. Soils XY, XN and JS-B mostly exhibited distributions of double peaks and most of water-stable aggregates were mainly concentrated in the range of 20~600 μm. Except that in ZZ-A and ZZ-B, the C u above 10 in the other soils indicated the broad range and unevenness of water-stable aggregate distribution (Table 4).
Large amounts of macroaggregates (>700 μm) broke down into smaller aggregates and particles (>10 μm), with the temperate ZZ soils being around 60 μm, and less macroaggregates   (>500 μm) of XN-A, CS-A and SG-B were fragmented into less aggregates or particles of a similar size (>10 μm) (Fig 3). The difference of the aggregate fragmentation between the second cluster and the other two clusters was in the amount of broken macroaggregates and the homogeneous distribution of the yielded materials (>2 μm). Compared with WD treatment, all the curves of particle size distributions (PSDs) in CD treatments shifted to smaller sizes in a certain degree (Fig 2). The peaks of PSDs for ZZ soils were situated at 40 μm, and those for other soils were located in the range of 5 to 20 μm. C u (7.6~11.3) and C c (0.8~1.3) of PSDs indicate a similar narrow range of particle sizes among these zonal soils. Based on the similarity in the fragmentation process, these zonal soils were grouped into three clusters as well (Fig 4). In the first cluster, the water stable aggregates of XY-A, XY-B, JS-B and XN-B showed a similar fragmentation process to that of ZZ-A and ZZ-B, which indirectly reflects their similarity in aggregations of primary particles to form water stable aggregates although in different weathering degree. In the second cluster, there were more large size water stable aggregates (>50 μm) (JS-A, XN-A, CS-B and SG-A) than in the first cluster, suggesting more fine particles (0.4~50 μm) were correspondingly released in the former cluster, and this phenomenon was more remarkable for soils in the third cluster.

Soil aggregate stability
For air-dried soils, mean volume diameters (MVDs) varied from 0.72 to 1.61 mm with a CV of 22% and the soil surface specific area (SSA) ranged from 9 to 164 m 2 kg -1 . The data of MVD AD and SSA AD provided the initial reference in GSD of air dried soils, which would facilitate the investigation of the disintegration of aggregates in WD and CD treatments.  Table 5). The MVD and SSA values in CD treatment were much smaller and larger than those in WD treatment, respectively. In the WD treatment, the MVD was overall smaller in B horizon than in A horizon (p<0.01) except for XY and XN, and the interaction of soil × horizon influenced the MVD (p<0.01) significantly. The values of MVD WD generally increased with increasing soil weathering (Fig 5a). SSA WD was significantly (p<0.05) affected by soil types, and this parameter was the largest for soil JS, but no significant differences were observed in the other soils, indicating a similar release rate of the fine grains in the disintegration process of air-dried aggregates in WD treatment (Fig 5b).
In CD treatment, there was an extremely significant difference in MVD between soil types (p<0.01). MVDs of soil particles were the largest in ZZ soils, and the smallest in XY (Fig 5c). SSA CD was significantly influenced by soil types, horizon and their interaction (p<0.01); SSA CD was larger in A horizon than in B horizon. Contrary to MVD, SSA CD in ZZ soils was the smallest and the other soils displayed no significant difference in SSA (Fig 5d).
Detachability index (DI) decreased from 95% (ZZ-B) to 67% (CS-A and SG-B) (Fig 6). The larger value of DI indicates the lower soil water stability, which represents weaker resistance to water erosive force. Aggregation index (AI) ranged from 31 to 89% indicating remarkable differences in aggregation of soil particles, and the aggregation was found to be highest and lowest in CS-A and ZZ-B, respectively.

Relationships between soil aggregation and physicochemical properties
Correlation analysis (Table 6) showed that MVD WD had a significant positive correlation with SOM and different forms of sesquioxides (r!0.57, p<0.05), but a negative one with pH (r = -0.60, p<0.05). Among these forms of sesquioxides, ln(Fe o -Fe p ) had the most significant relationship with MVD WD (p<0.001). SSA WD was significantly (r = 0.65 and 0.69, p<0.05) related  Table) entered the models with low VIPs (<2) (S2 Table) at significant level of p<0.05. The ln(Fe o -Fe p ) accounted for the majority of variations in MVD WD (R 2 = 77%) and BD explained 43% of the variance in SSA WD . The characteristics of water stable aggregates distribution (C u and C c ) were related to ln(Al d -Al o ) and BD, respectively. Besides, DI was negatively related to (Fe d -Fe o ) exponentially. AI was positively related with ln(Fe o -Fe p ),but negatively with MVD CD , both of which explained 54 and 31% of its variance, respectively, followed by silt (2~50 μm) (8%) and BD (4%). It can be concluded that free oxide especially Fe o -Fe p likely has the most significant influence on aggregate stability along the weathering gradient and that soil particle compositions (especially fine particles) and bulk density varied in their influence on the formation and stability of soil aggregates.

Discussion
A high variability was observed in the iron and aluminum oxides, indicating that the tested soils varied in different weathering degrees in the subtropical-tropical regions. Aggregate water stability generally increased along the weathering gradient with the maximum value in CS-A. Meanwhile, the water stable aggregates size distribution gradually transformed to the large size (>500 μm). Under increasing disruptive forces, the peak of grain size distribution gradually changed towards small size to a different degree, indicating the stepwise fragmentation process, which is similar to the findings of Oades and Water [3]. When subjected to different disruptive forces, the soils in different weathering degrees generally displayed a similar aggregation process (Figs 3 and 4), further indicating that soil aggregation of these zonal soils along the weathering gradient conforms to aggregate hierarchy theory to some extent. Asano and Wagai proposed a conceptual model of aggregate hierarchy in Andisol at micron scales (<53 μm) [5]. More recent studies have shown the hierarchical structure in Oxisols analogous to the temperate soils [33,34]. The main differences in aggregate breakdown among these soils lay in the amount of water stable aggregates. Miller and Schaetzl suggested that the clay-silt break be set at 6 μm in the grain measurement by laser diffraction [35]. Therefore, macroaggregates were assembled from sand-and silt-size grains (>20 μm) while water stable aggregates consisted of the particles of <40 μm.  In this paper, crystalline and amorphous iron, complex aluminum oxides and fine particles (<50μm) jointly contributed to aggregate water stability against slaking while pH and coarse particles (>50μm) played the inverse role. Imhoff et al. showed that clay and silt fractions (<50 μm) acted as cementing agent in water stable aggregate formation [36]. Lu et al. found that Fe d and clay greatly influenced porosity in Ultisols [37]. The effect of crystalline oxides on soil porosity has been verified in another paper (Wu et al., unpublished). The aforementioned results suggest that crystalline oxides especially Fe d -Fe o , influenced aggregate formation and stability most likely through porosity. According to the degree of aggregate breakdown subjected to slaking, the binding effects of amorphous oxides and clay on the stability of macroaggregates (250~3000 μm) were relatively limited compared to crystalline oxides. Pronk et al. reported that the crystalline iron oxides are present as primary particles [38]. Peng et al. showed that organic matter played a primary role in aggregate stability (250-2000 μm) [11]. However, in the present study, we failed to observe the significant effect of SOM on aggregate stability, probably due to its small quantity in soils (below 23 g kg -1 ) [8].
When compared to the air dried macroaggregates, a higher similarity was observed in the formation of water stable aggregates, with both free sesquioxides and particle composition playing an important role in aggregation (Fig 5). Peng et al. showed that Fe/Al oxides seemed to be the major agents of <250 μm aggregates [11]. Huang et al. found that water stable macroaggregates (>250 μm) were correlated with the organic matter contents of eroded Ultisols [39]. In this study, water stable aggregates (SSA and MVD) and size distribution (C u , C c ) were significantly related to pH, SOM, CEC, BD and free oxides, with ln(Fe o -Fe p ), ln(Al d -Al o ) and BD most closely linked to water stable aggregates along the weathering gradient. Duiker et al. reported that Fe o and SOM were well correlated with aggregate stability [9].
It should be noteworthy that free oxides especially Fe o -Fe p and Al o -Al p and fine particles (<50 μm) were the dominant agents in aggregation along the weathering gradient. Likewise, the relative low content of organic matter (below 23 g kg -1 ) resulted in its less significant effects than sesquioxides and clay content on the formation of water stable aggregates [8]. The large surface area of Fe and Al oxides might facilitate reactions with clay particles through Coulombic forces [40]. Amorphous oxides exhibited the stronger cementing role in the aggregate formation than crystalline oxides. Previous studies showed that aggregate stability increased with an increase in clay content due to its cementing effect especially for the inactive clay mineralogy such as kaolinite [36,[41][42][43]. Overall, the ln(Fe d -Fe o ), MVD CD and ln(Fe o -Fe p ) could be well used to evaluate and predict soil aggregation. As soil texture and sesquioxides are not easily managed at a small scale, organic matter could enhance soil aggregation, especially macroaggregate formation and stability. Further works are required to elucidate the mechanistic interaction between soil aggregation and these active agents.

Conclusion
This study about the aggregation of zonal soils and its relationships with physicochemical properties from the temperate to topical regions indicates that, under increasing disruptive forces, soil aggregates showed a stepwise fragmentation and the peak of grain size distribution grain transformed to small size gradually. For zonal soils in different weathering degrees, the formation process of soils aggregates conformed to the aggregate hierarchy theory to some extent. Aggregate water stability generally increased with increasing soil weathering degree. Free oxides, organic matter, bulk density and particles composition exert their specific roles in different stages of soil aggregation. Crystalline (Fe d -Fe o ) and amorphous iron oxides (Fe o -Fe p ) in combination with fine particles (<50 μm) remarkably contributed to the formation and stability of air dried aggregates and water stable aggregates along the weathering gradient, respectively. However, the experiment evidence about the effects of sesquioxides and fine particles on soil aggregation remains to be explored in future studies.