3　Results

3.1　Basic properties of soils

Table 2 shows the physical and chemical properties of the soil samples.The sand contents of all the soils from the two parent materials were the lowest(＜20%)compared with the silt and clay contents.The texture of the soils ranged from silt loam to clay,and the clay content was higher in soils derived from Quaternary red clay than that from the soils derived from Shale with the same land use.All soils were strongly acidic with p H ranging from 4.28 to 5.56.Soil organic matter content in most soil samples was less than 20g·kg-1 except Prunus×cistena upland from Quaternary red clay.Cation exchange capacity values of the studied soils ranged from 11.48 cmol·kg-1 to 19.38 cmol·kg-1,showing a minimal difference among them.The contents of SOM from the Prunus×cistena upland(Qp and Sp)surface soils were the highest,followed by tea gardens and orchardswith the lowest content appearing in the cropland.The low organic matter of the cropland might be associated with frequent soil tillage for a prolonged period(Zhang and Xu,2005).The relatively lowcation exchange capacity of the soils could be attributed to the leaching process following high intensity rainfall in the region.

The dithionite-and oxalate-extracted Fe and Al are shown in Table 2.Dithionite Fe and Al in all soils were higher than were the oxalate extracted forms.Parent materials apparently had a strong influence on the contents of the dithionite extractable Fe and Al in soils:dithionite extractable Fe and Al ranged from 17.00～23.07g·kg-1 and 2.16～2.65g·kg-1 from Quaternary red clay whereas they were from 8.75～14.96g·kg-1 and 0.89～1.83g·kg-1 from Shale.The difference of the oxalate extractable Fe and Al in all soils was not obvious.Oxalate extractable Fe and Al ranged from 1.38～1.95g·kg-1 and 1.04～2.08g·kg-1 from Quaternary red clay,and they were from 1.39～2.07g·kg-1 and 0.94～1.42g·kg-1 from Shale,respectively.

3.2　The size and stability of macro-and micro-aggregate

Table 3 shows the percentage of＞5mm aggregates was the highest except for the Sp(10～20cm)sample,and the percentage of 2～1mm aggregates was the lowest except for the Qc(0～10cm)and Sp(10～20cm)samples.Little difference in percentages was found among macroaggregates of the three sizes(5～2mm,1～0.5mm and 0.5～0.25mm)for most soils.These results differed from those of Horn(1990),who concluded that the large aggregates(＞3mm)were broken and small aggregates(0.25～3mm)were significantly increased.

The percentages of all five size fractions of the macro-aggregates from Quaternary red clay were greater than those from Shale,suggesting that the macro-aggregates from Shale suffered more damage after wet sieving.Regardless of the parent materials,the percentages of＞5mm aggregates from the orchard samples(Qo and So)were the highest whereas the percentages from the cropland(Qc and Sc)were the lowest among different land uses.

The results of the aggregate stability from the wet sieving method were slightly different due to the choice of the stability indicators(Fig.2).The value of the ADR was inversely proportional to that of the WSA,and the higher the ADR value,the worse the aggregate stability.For the soils in this study,the values of WSA,MWD and ADR ranged from 8.34%to 73.80%,0.08mm to 2.42mm,and 23.30%to 91.49%,respectively.For similar land uses of the two parent materials,the water stability of the aggregates from Quaternary red clay was stronger than that from the Shale(p＜0.05).In this study,the stability among surface and subsurface soil aggregates showed no significant variation.

The particle size distribution of the micro-aggregates is shown in Fig.3.Because of the different soil parent materials,the particle size distribution of the micro-aggregates was significantly different.The percentage of 0.25～2mm aggregates from Quaternary red clay(51.85%to 65.18%)was obviously higher than that from Shale(9.35%to 38.00%).The percentage of 0.05～0.25mm particles from Quaternary red clay was the highest(ranged from 18.35%to 28.20%),followed by 0.05～0.02mm(5.02%to 12.52%)and＜0.002 mm(3.82%to 5.82%).However,the percentage of 0.02～0.002mm aggregates from Shale was the highest(20.00%to 38.14%),followed by 0.25～0.05mm(14.00%to 30.45%)and＜0.002mm(4.92%to 10.48%).Regardless of the soil parent material,the difference of the aggregates size distribution among land use types or soil depth was insignificant.

Macro-aggregates were composed of micro-aggregates,and different size aggregates had different aggregation and stability mechanisms(Oades and Waters,1991).The micro-aggregate stability indices of the soils are presented in Table 4.The clay dispersion ratio(CDR)was used to estimate the instability of the micro-aggregates whereas aggregated silt and clay(ASC)followed a contrary tendency to that of CDR.The results of CDR and ASC analyses showed that the micro-aggregate stability of the two parent materials was significantly different under the same land use types or soil layers.The values of CDR and ASC from Quaternary red clay ranged from 0.08 to 0.15 and 70.44 to 78.84%,respectively;by contrast,the values from Shale ranged from 0.16 to 0.38 and 19.90%to 58.74%,respectively.Among the land use types,the micro-aggregate stability was significantly different,but they exhibited no trend under the two parent materials or soil layers.

3.3　Macro-and micro-aggregate stability versus related soil properties

The correlation coefficients of the individual soil properties are listed in Table 5.A preliminary analysis indicated high correlation among most soil properties.The results found that the basic soil properties(silt,clay and CEC)were well correlated with different forms of iron and aluminum(Fed,Ald,and Alo).The SOM was well correlated with CEC.However,the Feo seemed to be more independent than the other soil properties.The silt was found to be negatively correlated with the other soil properties(clay,CEC,Fed,Ald,and Alo).Furthermore,the silt and clay were more sensitive to Fed and Ald than to Alo whereas the CEC was more sensitive to Fed and Ald.Among the different forms of iron and aluminum,Fed was more sensitive to Ald than to Alo.

Table 6 summarizes the five PLSR models constructed separately for macro-and micro-aggregate stability(WSA＞0.25,MWD,ADR,CDI,and ASC).The macro-aggregate characteristic(WSA＞0.25,MWD,and ADR)models extracted one PLSR component that was relevant to 8 predictor variables(related soil properties).The first component accounted for 80.65%,69.17%,and 82.19%of the variance in the dataset with WSA＞0.25,MWD,and ADR,respectively.The addition component cumulatively accounted for 84.87%,80.41%and 85.93%of the total variance in the WSA＞0.25,MWD,and ADR values,respectively.However,the cumulativevalues decreased from0.75,0.60 and 0.77 to 0.72,0.55,and 0.75,respectively(Table 6),which suggested that it did not substantially improve the description of the contributions to the variance.The relative importance of the related soil properties could be obtained by exploring their VIP values and their regression coefficients(RCs)(Table 7).In the case of macro-aggregate stability,the highest VIP value was obtained for Ald from the three indicators,followed by clay,Fed,silt,and Alo for the WAS＞0.25;clay,silt,Alo,and Fed for the MWD;and clay,silt,Fed,and Alo for the ADR.Macro-aggregate stability appeared to increase with increasing Ald,clay,Fed,and Alo(due to the positive regression coefficients for the WSA＞0.25 and MWD,and the negative regression coefficients for the ADR);silt contributed to lower stability.All of the considered soil properties were related to macroaggregate stability to some extent,and some of themhad VIP＞1.Others with VIP values below 1(CEC,SOM,and Feo)were of minor importance for macro-aggregate stability.

The first component accounted for 83.91%of the variance in the dataset with CDI of the microaggregate stability.The addition component cumulatively accounted for 88.36%of the total variance,and the cumulativevalues increased from 0.83 to 0.84.However,the best PLSR prediction was also described by the lowest RMSEE.The addition component did not produce a significant change inand RMSEE.For the CDI model,the RMSEE minimum value was achieved with one component,accounting for 83.91%of the CDI.For the ASC model,the RMSEE minimumvalue was achieved with two components.The addition of the second component significantly improved the model(Table 6).In the micro-aggregate stability models,the highest VIP value was observed for the clay content,followed by the silt,Fed,Ald,and Alo for the CDI,and silt,Fed,Alo,and Ald for the ASC.Other properties were of minor importance for micro-aggregate stability(VIP＜1).

Using PLSR modeling,we explored the main soil properties that affect macro-and micro-aggregate stability.The results showed that Ald was superior in binding the macro-aggregates and clay was superior in binding the micro aggregates,followed by Fed,Alo,CEC and SOM;Feo was the weakest contributor to the formation of macro-or micro-aggregates.