2　Results and Discussion

2.1　Aggregate Breakdown Induced by Rainfall

The MWDrain curves of eight soil samples exposed to rainfall are shown in Fig.1.The MWDrain decreased exponentially as the amount of rainfall increased.The changes in MWDrain could be grouped into two stages.The MWDrain decreased sharply at the beginning of rainfall and subsequently approached a constant value asymptotically.Shale aggregates broke down more rapidly than quaternary soil aggregates.

The breakdown of shale aggregates mainly occurred at 0 to 0.5mm cumulative rainfall and approached a constant value at approximately 2mm cumulative rainfall.In contrast,quaternary aggregate breakdown mainly occurred at 0 to 1mm and approached a constant value at 24mm cumulative rainfall asymptotically.The MWDrain of shale was lower than that of quaternary soil when the samples were subjected to the same cumulative rainfall.Evident differences in MWDrain were observed among four quaternary aggregates.In contrast,shale aggregates differed slightly.

Three mechanisms that contributed to aggregate breakdown under rainfall are(i)slaking or breakdown caused by the compression of entrapped air during wetting;(ii)breakdown by differential swelling;and(iii)breakdown by raindrop impact(Le Bissonnais,1996).In our previous study(Shi et al.,2010;Yang et al.,2012),slaking was the main mechanisms of Ultisol aggregate breakdown at 60mm h-1 rainfall.In addition,the breakdown of all aggregates mainly occurred during the first 1mm of rainfall(Fig.1).Concaret(1967)and Legout et al.(2005a)believed that slaking is present and dominant during the first millimeter rainfall of heavy rainfall.Thus,slaking is the main mechanism of aggregate breakdown,followed by differential swelling and mechanical breakdown.Zhang and Horn(2001)and Li et al.(2005)compared the aggregate water stability of Ultisols by using the Le Bissonnais(1996)methods and demonstrated the importance of slaking during the aggregate breakdown process.

2.2　Aggregate Water Stability and Relations to Aggregate Breakdown

Aggregate water stability is presented in Table 3.The MWDsieving followed this trend:Qpc＞Qt＞Qs＞Qc＞Spc=Sp＞Sc＞St.Quaternary aggregates were more stable in water than shale aggregates.The MWDsieving of quaternary aggregates ranged from 0.87 to 2.16mm;the MWDsieving of shale aggregates ranged from 0.31 to 0.67mm.Quaternary aggregates exhibited a higher number of fragments of greater than 1mm in size where shale aggregates exhibited a higher number of fragments of less than 0.5mm in size.Approximately 50%of the shale fragments were less than 0.1mm in size.Shale aggregates showed no evident differences because of weaker stability than quaternary aggregates.

Wet sieving method simulates the process of aggregate breakdown under rainfall and measures the breakdown of aggregates on rapid wetting(slaking)and the subsequent breakdown of aggregates caused by mechanical sieving action.Therefore,wet sieving may be the best index of aggregate breakdown under rainfall and soil erosion.In this study,MWDsieving correlated strongly(R2＞0.8)with MWDrain under different cumulative rainfall(Fig.2).However,compared with the test of tensile strength,the aggregate water stability test is more complicated and time consuming.

2.3　Aggregate Tensile Strength and Relations to Aggregate Breakdown

The mean values of tensile strength decreased significantly in the following order:Qpc＞Qt=Qs＞Qc＞Spc＞St=Sp＞Sc.The TS of quaternary soils was higher than that of shale.The TS of shale soils ranged from 181 to 276kPa;for quaternary soils,the TS ranged from 338 to 735kPa.

At the same amount of rainfall,MWDrain of the eight aggregates was closely correlated with TS.High correlation coefficients were found between TS and MWDrain at different cumulative rainfall(Fig.3).Both aggregate breakdown and TS depend on aggregate microstructure.

According to the theory of brittle fracture addressed by Freudenthal(1968)and Braunack et al.(1979),flaws of various magnitudes are distributed throughout the aggregate,such as cracks or microscopic pores,material packing heterogeneities,and macroscopic holes(Perfect and Kay,1991).These flaws are the〝failure zones〞of the solid and determine the distribution and strength of aggregate.Under critical stress,aggregates possibly break along these flaws.The TSis the critical stress required to cause soil aggregates to fail in tension.Therefore,aggregates with low TS contain greater flaws.

As discussed previously,slaking is the main mechanism of aggregate breakdown under rainfall.Slaking is caused by the compression of air trapped inside aggregates during wetting.Compression effect depends on the volume of air trapped inside the aggregates(Le Bissonnais,1996).Flaws comprise the air space that exists inside the aggregate,indicating that a lower TS corresponds to a higher aggregate porosity(Macks et al.,1996)and a lower aggregate stability under slaking.The TSis closely correlated with fragment size distribution after aggregate breakdown by slaking.

2.4　Aggregate Penetration Resistance and Relations to Aggregate Breakdown

Aggregate penetration resistance showed less evident differences with different soil aggregates compared with TS(Table 4).The APRdry was approximately 10 to 18 times greater than TS and nine to 30 times greater than that of APRwet.The APRdry of quaternary soils was greater than that of shale.The APRdry was ranked in the following order:Qpc＞Qt=Qs＞Sp＞St=Sc=Qc=Spc.The APRwet ranked in the following order:Sp＞Spc=Qs＞Qpc＞St＞Sc=Qt＞Qc.The large variation in APRdry showed less evident differences among different soil types compared with APRwet.

Previous studies have shown a relationship between soil penetration resistance and soil erosion(Al-Durrah and Bradford,1982;Callebaut et al.,1985;Bradford et al.,1992;Maestre et al.,2002;Stavi et al.,2008).In most of these studies,penetration resistance was determined on saturated soil.In the current study,penetration resistance was measured both on dry and saturated aggregates.However,under the same cumulative rainfall,only APRdry was closely correlated with TS and MWDrain(Fig.4).No linear relation between MWDrain and APRwet was observed(R2＜0.04).

Information on the penetration resistance of dry aggregate is insufficient because this factor is more difficult to determine than that of wet aggregates.Penetration resistance is determined using wet soil samples,which allows deformation of the sample without cracking under penetration stress.Farrell and Greacen(1966)indicated that penetration resistance covers two components:the pressure required to expand a cavity in the soil and the frictional resistance to the probe.The APRdry was tested using the same needle penetrometer as the one used for wet aggregates.Tensile failure occurred during the penetration process because of low water content.Therefore,it is a process for evaluating APRdry similar to that for TSin principle but cannot reflect the soil frictional resistance and the ability to resist deformation.A good correlation between APRdry and TS was observed(r=0.906,P＜0.01).Thus,APRdry also showed a good relationship with MWDrain.

No significant linear relationship(R2＜0.04)was observed between MWDrain and APRwet,which was different from the results of previous studies on soil penetration resistance and soil erosion(Al-Durrah and Bradford,1982;Callebaut et al.,1985;Bradford et al.,1992;Maestre et al.,2002;Stavi et al.,2008).These erosion tests were conducted on wet soil,and raindrop impact was the main mechanism of aggregate breakdown.In contrast,the current study was conducted using dry aggregates,and slaking was the main mechanism of aggregate breakdown.Scale differences between bulk soil and aggregate may be another reason that contributed to this contradiction.

2.5　Aggregate Stability and Splash Detachment

The regression relationships between aggregate stability indexes and splash are shown in Fig.5.The relationship between different aggregate stability parameters and splash detachment proportion was similar to the relationship between the different aggregate stability parameters and MWDrain.No significant linear relationship was observed between APRwet and splash detachment(R2=0.014),whereas MWDsieving(R2=0.758)and TS still showed a better relationship with splash(R2=0.865)than APRdry(R2=0.563).The MWDsieving consistently showed the best relationship with MWDrain.at different cumulative rainfall(R2=0.857,0.853,0.851,and 0.888,respectively).In contrast,TS showed the best regression relationship with splash detachment proportion(R2=0.865).The MWDsieving could be more efficiently used to predict fragment size distribution,whereas TS could be more efficiently used to predict splash detachment.However,the difference in efficiency between MWDsieving and TS to estimate fragment size distribution and splash is small.

The good relationship between TSand fragments size distribution and splash helps predict soil resistance to erosion by TS.Tensile strength can be determined with simple and rapid tests using aggregates of very different sizes compared with wet sieving.Tensile strength is a sensitive indicator of soil structure(Dexter and Kroesbergen,1985;Watts and Dexter,1998;Mosaddeghi et al.,2006).Thus,TS could serve as a good index to predict soil resistance to erosion.

2.6　Linking Soil Properties and Aggregate Stability

The basic soil properties related to aggregate stability are presented in Table 5.All of the materials are strongly acidic and are poor in soil organic matter(SOM).The materials contain a relatively small amount of total exchangeable bases,which is typical of Ultisols from southeastern China(Yan et al.,2008).A large amount of clay and silt is observed.Soil texture varies from silt loam to clay.These properties vary with the soil parent materials and the land use history.The clay composition in soils derived from quaternary red clay is greater than those derived from shale,whereas shale soils contain larger amount of silt.Quaternary soils contain more Fed,Ald,Feo,and Alo than shale.Soils underused for long-term cultivation contained less organic matter because of a low organic matter supply or frequent cultivation,which promotes the mineralization of SOM(Zhang and Xu,2005).

A correlation analysis for MWD,TS,APRdry,MWDrain,and splash showed that Fe oxide and clay are the main properties contributing to aggregate stability.However,APRwet did not show any correlation with soil properties.The Feo was more strongly correlated with these aggregate parameters than the other soil properties.Soil organic matter and Al oxide were only correlated with MWDrain or APRdry,whereas CEC and the contents of sand and silt did not correlate well with any of the properties.

The contribution of organic matter to aggregate stability is recognized.Organic matter is the main stabilizing substance in aggregates.In our study,the effect of SOM on aggregate stability was not evident.This contradiction may be caused by the high composition of clay and oxides in the soil samples,indicating that the stability mechanism of Ultisol aggregates is more complicated than that of other soil types.Chappell et al.(1999)indicated that organic carbon in one of the most important governing factors,accounting for 56%of the variance in aggregate stability.Huang et al.(2010)mentioned that the percentages of greater than 0.25mm water-stable aggregates correlated well with the organic matter contents of eroded Ultisols.However,Zhang and Horn(2001)and Li et al.(2005)found that clay and oxides are the main contributors to aggregate water stability,whereas little or no relationship was observed between MWD and organic matter content.Amézketa(1999)considered that the aggregate stability of Ultisols is determined not only by the organic matter content but also by the contens of clay and oxides.The effect of organic matter on aggregate stability depends on its composition and the relative contribution of other aggregate-stabilizing substances,including clay and oxides(Goldberg et al.,1990).In addition,SOC varied from 15.34 to 21.16 k·kg-1 in this study.This small range of variation may be another contributing factor to the low correlation between SOM and aggregate stability.