2　Materials and methods

2.1　Soils

Soil samples were collected from Xianning City,Hubei Province in the south-central part of China(Fig.1).The sampling sites were in the subtropical zone,which has an annual rainfall of 1572mm and annual average temperatures of 16.8℃.The elevation of this area is 32～52m above sea level.These areas are hilly with different degrees of erosion.Soil erosion in some areas of the ultisol region reaches 7000t·km-2 per year(He and Sun,2008).Eight soil samples were collected from the surface layer(0～10cm)and eight from a subsurface layer(10～20cm).Each sampled area was about 0.1 to 0.3ha.This number of samples is representative for the study area.The eight sample sites were geographically representative for the soils in Hubei Province.They covered the most soil parent materials in subtropical China.Considering the soil parent materials and different land use,we selected the eight soil samples derived from Quaternary red clay and Shale from Xianning City.General characteristics of these sampling sites are shown in Table 1.

Table 1　Basic description of selected soils

2.2　Laboratory methods

The wet sieving method of Yoder(1936)was modified to measure macro-aggregate stability.Air-dried soil samples were sieved by hand using a column of five sieves at sizes of 5.00,2.00,1.00,0.50 and 0.25mm.The percentages of different aggregates in the bulk soil as determined with the dry-sieving method were calculated and recorded.Based on these percentages,soil samples of different aggregate sizes were prepared.Approximately 50g of composed soil samples was placed on the first sieve of a sequential nest and gently moistened to avoid sudden rupture of the aggregates.The sample was then sieved in distilled water at 30 oscillations per minute(along 4cm amplitude)for 30min.The resistant soil materials on each sieve were transferred into clean beakers.These soil materials were then oven-dried gently at 40℃for 48h and weighed.Macro-aggregate stability was expressed by the content of water-stable aggregates that were greater than 0.25mm(WSA＞0.25),mean weight diameter(MWD),and the aggregate deterioration rate(ADR)by wet sieving.Equations used in this research include:

Fig.1　Location of the study area and sampling sites:(a)location of Xianning County in Hubei Province,(b)location of Heshengqiao town in Xianning County

where xi is the mean diameter of each size class(＜0.05mm;0.05～0.10mm;0.10.0～0.25mm;0.25～0.50mm;0.50～1.00mm;1.00～2.00mm;＞2.00mm),and wi is the proportion of each size class with respect to the total sample.The upper and lower limits of mean intersieve aperture are 3.5 and 0.05 mm,respectively.

where Cd(%)is the content of dry aggregates,of which the size is greater than 0.25mm,and Cw(%)is the content of water-stable aggregate,of which the size is greater than 0.25mm.

Particle size distribution was measured by the hydrometer method(Gee and Bauder,1986).The micro-aggregate distribution was determined as particle size distribution analysis above except that there was no chemical dispersant(sodium hydroxide)applied(only mechanical agitationwith an endover-end shaker).Clay dispersion ratio(CDR)and aggregated silt and clay(ASC)were used to express the microaggregate stability as described by Dong et al.(1983)and Igwe et al.(1999).They were obtained using the following equations:

where Wc(%)and Ws(%)are the content of clay(＜0.002mm)and silt(0.05～0.002mm)from the micro-aggregate stability determination,Cc(%)and Cs(%)are the content of clay(＜0.002mm)and silt(0.05～0.002mm)from the particle size distribution measured by the hydrometer method,respectively.The more stable the micro-aggregates,the lower the CDR value or the higher the value of the ASC(%).

The basic chemical properties of the soils were determined using the routine methods.Soil p H was measured with a p H meter in a 1.0∶2.5 soil:water mixture.Soil organic matter(SOM)was determined using the K2 Cr2 O7 wet oxidation method(Jackson,1958).Cation-exchange capacity(CEC)was determined using the ammonium acetate method buffered at p H 7(Sumner and Miller,1996).Free Fe and Al oxides(Fed and Ald)were extracted using dithionite-citrate-bicarbonate(DCB)solution(Mehra and Jackson,1960).The extraction of amorphous Fe and Al oxides(Feo and Alo)was carried out using oxalic acid ammonium oxalate at p H 3.0(Mckeague and Day,1966).Pyrophosphate Fe and Al(Fep and Alp)were extracted using sodium pyrophosphate at p H 10(Bascomb,1968).The extracted Fe and Al were diluted and determined by an inductively coupled plasma spectrometer(ICP)(VISTAMPX,Varian,Inc.,USA).

2.3　Data analysis

Statistical analyses were performed using Excel and SPSS 11.0 software(Statistical Package for the Social Sciences,2001).Significant differences among the soil samples for macro-aggregate stability were determined using the LSD(Least Significant Difference)procedure for a multiple range test at the 0.05 significance level.

Partial least squares regression(PLSR)was used to determine the relationships among related soil properties and macro-and micro-aggregate stability.In this study,the relationships among the predictors(related soil properties)and the response function(the macro-and micro-aggregate stability)can be inferred from the variable importance for the projection(VIP)and regression coefficients(RCs)of individual predictors in the most explanatory components.The VIP was used to identify the importance of a predictor for both the independent and dependent variables.Terms with large VIP values are the most relevant for explaining the dependent variable.In general,an independent-variable VIP value greater than 1 significantly explains the dependent variable,while a value lower than 0.5 indicates that the independent variable does not significantly explain the dependent variable.The regression coefficients of the PLSR models were used to show the direction of the relationship between the related soil properties and macro-and micro-aggregate stability.Thus,it is possible to determine which related properties most strongly interact with macro-and micro-aggregate stability.

The analyses were performed using the PLSR procedure implemented in SIMCA-P(Umetrics AB.Sweden).The cross-validation was enforced by using the criterion to determine among the number of significant PLSR components.The fraction of the total variation of the dependent variables that could be predicted by a component(Q2)and the cumulative Q2 over all the selected PLSR componentsin SIMCA were computed according to the following formula:

where PRESS is the prediction error sum-of-squares,SS is the residual sum-of-squares,and m is the number of PLSR components.A model is thought to have a good predictive ability whenis greater than 0.5.In addition,the root mean square error(RMSEE)also provides useful information for calibrating and developing the regression model.