AGGREGATE DISTRIBUTION, STABILITY AND RELEASE OF WATER DISPERSIBLE CLAY FOR TWO SUBTROPICAL OXISOLS

The behavior of a soil regarding the dispersion and aggregation of its particles is very important for the development of environmental and agricultural soil functions. This study was conducted to determine how aggregate distribution and stability are impacted by land uses and how the release of Water Dispersible Clay (WDC) relates to disaggregation in Oxisols from subtropical Brazil. Samples from two Oxisols, collected at three depths from sites under no-tillage (NT), conventional tillage (CT) and native vegetation (NV) land uses were shaken in 250 mL plastic bottles for intervals up to 27 hours. The mass of aggregates was measured in five size classes ranging from 53 to 2000 m. Most aggregates larger than 500 m disappeared during the first 7.5 hours of shaking, concurrent with an increase in WDC release and without change in soil suspension pH and electrical conductivity, without increase in smaller aggregates. Therefore, there is no aggregate hierarchy in these soils and the release of WDC was caused by breaking aggregates within the 500 to 2000 m range. Land uses affect mass of aggregates in each size class, but the aggregate stability depends on its size, not land use.


INTRODUCTION
Because soil aggregates are dynamic and respond rapidly to environmental changes, interest on them as soil quality indicators is growing (Caravaca et al., 2004;Boix-Fayos et al., 2001).In addition, the capability of soil particles to protect organic matter from oxidation allows to approach soils as a carbon sink to the atmosphere (Torn et al., 1997).Land use change caused by tree harvesting (Yanai et al., 2003) and/or change from forest to agriculture has an extensive impact on aggregation (Carpenedo & Mielniczuk, 1990;Perin et al., 2003) and carbon dynamics in tropical, acid soils (Leite et al., 2004;Zinn et al., 2005).The lack of aggregate hierarchy (Oades & Waters, 1991) and great amount of iron and aluminum oxides Sci.Agric.(Piracicaba, Braz.), v.64, n.1, p.36-43, January/February 2007 make stability and size of aggregates from weathered Oxisols require specific studies.In Brazil, such soils tend to have very stable aggregates less than 2 mm in diameter.As a result, clayey Oxisols behave like medium-textured soils allowing agricultural activities such as tillage or harvest to occur soon after raining (Buol & Eswaran, 2000).The potential to release water dispersible clay (WDC) due to disaggregation is not well known for these soils (Azevedo & Bonumá, 2004) and such knowledge may help to better manage these soils during changes in land use, and is important for evaluating the environmental mobility of herbicides, pesticides, and other xenobiotic compounds used in agriculture (Seta & Karathanasis, 1996;Bertsch & Seaman, 1999).
The objective of this study was to evaluate the long-term impact of management practices on the distribution and stability of aggregates of less than 2 mm diameter and the release of WDC during disaggregation for two Oxisols from Southern Brazil.

MATERIAL AND METHODS
The two chosen sites were located on the Sul Riograndense Plateau, a geomorphic region developed on the lava flows of the Serra Geral formation (Figure 1).The soil at the Santo Ângelo site was a Typic Haplorthox developed from basaltic rocks, occupies 7.26% of the Rio Grande do Sul State and occurs at elevations between 200 and 400 m above sea level (a.s.l.).The climate is Cfa in the Köppen system, with mean annual temperature of 19.5°C and mean precipitation of 1,850 mm yr -1 (BRASIL, 1973).Samples from conventional (CT) and no-tillage (NT) systems treatments were collected in Santo Ângelo, Rio Grande do Sul State (28 o 16' S, 54 o 13' W, approx.280 m a.s.l.).The tillage experiment was established in 1979 in a field that had been under a wheat-soybean rotation since 1964 (Dalla Rosa, 1981).Samples from an undisturbed, forested soil were collected from the nearest protected area of original forest vegetation (28°12' S, 54°13' W) about 15 kilometers from the experimental site.The soil at the Passo Fundo site was a Typic Haplohumox developed from a mixture of basalt and sandstone and occurs at elevations between 460 and 700 m a.s.l..The climate is Cfa1, with mean annual temperature 18°C and mean precipitation of 1,750 mm yr -1 .Samples under conventional and notillage systems were collected in Passo Fundo, Rio Grande do Sul State (28°14' S and 52°24' W) in an experiment established in 1983 (Kochham & Denardim, 1997).Soils under native vegetation were sampled in a forested reserve on the experimental station within 0.5 km of the experimental site.
In each of both sites, three samples of treatment NT and CT were randomly collected in the experimental field, and the three samples from native vegetation (NV) were also randomly collected in forests.Care was taken to keep similar distances among collecting points in the experimental field and in the forests (5 to 10 m).In this study, NT, CT and NV were refered to as "land uses".Samples from 0 to 5 and from 10 to 15 cm depth were collected from small pits about 0.3 0.3 0.20 m depth and samples from 40-60 cm depth were collected using a bucket auger.Airdried samples were gently crushed and passed through a 2 mm sieve (fine earth fraction).The three field replications of each land use from each site were combined, thoroughly mixed, and then stored in sealed plastic bags.Therefore, lab analyses were performed on one composite sample from each of the three depths, three land uses and two soils, summing up eighteen samples.
Organic carbon was determined by digestion in K 2 Cr 2 O 7 and titration with Fe(NH 4 ) 2 (SO 4 ) 2 .6H 2 O and particle size distribution was determined by the pipette method after dispersion with 6% NaOH (EMBRAPA, 1997).Both analyses were performed twice for each composite sample.
Both WDC and aggregate stability of the composite samples were made three times (three runs) by the standard method for WDC determination (USDA, 1996).Shortly, five 10 g aliquots of fine earth were weighed and placed into 250 mL plastic bottles.The bottles were filled with 175 mL of deionized (DI) water and shaken (120 excursions per minute, 4 cm horizontal displacement) for 0, 3.75, 7.5, 15 and 27 hours (one aliquot for each time period).The suspensions were then poured through a nest of five sieves of 1000 m, 500 m, 250 m, 106 m and 53 m.Disaggregated clay and silt were gently washed from the soil on the sieves with DI water and collected in 1 L cylinders for WDC determination.The soil material retained on the sieves was dried at 110° for 24 hours and weighed.Soil material that passed through the nest of sieves was collected in one liter cylinders and re-suspended for WDC measurement by the pipet method (USDA, 1996).Slacking was assumed to be negligible, since samples submitted only to rapid wetting were almost entirely aggregated (0 h shaking time on Figures 2 and 3, sum of aggregates).
The mass of aggregates in each size class, A i , was calculated by: where B i is the dry mass of soil material in size class i, C i is the dry mass of sand in size class i and D is the initial oven-dry mass of soil.C i was the average of three previous determinations of sand content in each composite sample.Sand was subtracted from the numerator of equation [1] to avoid counting individual sand grains as aggregates.For brevity, we use the term "aggregates" to refer to Ai, , and "sample" for the composite sample, in the discussion that follows.
The distribution of aggregates was analyzed as a split plot design with three treatments (NT, CT, NV) with three blocks (each one of the three runs) replicated over time.Data for each soil type, depth and size class were analyzed separately.Shaking time was considered the whole unit and land use the sub unit of the split plot experiment.For the percent soil mass data, a square root transformation was performed prior to analysis of variance in order to achieve homogeneity of error variance.Error (a) was pooled with error (b) because it was not significant (P = 0.25) in the majority of cases.Time and land use versus time effects were partitioned into orthogonal polynomial contrasts.Regressions on the means of the dependent variables as a function of land use and time was followed by the analysis of variance (ANOVA) with the regression model determined by the significant treatment effects and contrasts.Comparison among regressions models of disaggregation from each land use (along shaking times) through ANOVA is shown in Table 2.

RESULTS AND DISCUSSION
Because sand content was subtracted from the mass of soil material retained in each size class (equation [1]), the sandier Typic Haplohumox (Table 1) had fewer soil aggregates in each size class than the Typic Haplorthox.
Prior to shaking (0 h shaking time; Figures 2  and 3), from 73 to 91% of the Typic Haplorthox and from 54 to 69% of the Typic Haplohumox occurred in aggregates.With only two exceptions, regressed disaggregation models were different (P = 0.01) among the three land uses for the 0-5 and 5-10 cm sampling depths, but land use had less effect on disagregation models for the 40-60 cm depth (Table 2).Some common behavior could be observed in both soils (Figures 2 and 3): there were very few 106-  53 m and 250-105 m aggregates in soil under native vegetation at 0-5 cm and 10-15 cm depths; aggregates larger than 500 m in diameter dominated both at 0-5 cm and 10-15 cm depths, but at 40-60 cm depth, aggregates 2,000 -1,000 m in diameter were the least abundant; and although the stacking order of curves in a single graph changed depending on size class and depth, it was noted that curve shapes were, in general, similar for each size class.
There was a small release of WDC upon initial wetting (0h shaking) in all samples, supporting the assumption that slacking was not a significant disaggregation process under the experimental conditions.
The largest aggregates (2000-1000 m and 1000-500 m) broke down quickly within the first 7.5 hours of shaking, but the mass of smaller aggregates did not increase concurrently, which showed that such aggregates were mainly broken into primary soil particles and not to small aggregates (Figures 2 and 3).Therefore these soils did not have aggregate hierarchy (Oades & Waters, 1991).Since WDC increased as aggregates larger than 500 m disaggregated, without change in pH suspensions, disaggregation appears to be the main mechanism of WDC production.
The lack of aggregate hierarchy allows the description of disaggregation by a first order process model (Beare & Bruce, 1993;Parkin & Robinson, 1992;Olson, 1963): where A i,t is the mass of aggregates in size class i at time t, A i,0 is the mass of aggregates at time 0, and k is a curvature parameter.This model does not account for additions of aggregates to sieve i, produced by the disaggregation of aggregates in the sieve i+1 above it, and so can only be used in soils with no aggregate hi-erarchy.The curvature parameter (k) was assumed as disaggregation rate index, and the greater its absolute value, the smaller the aggregate stability.No recognizable pattern was found when organizing the k values according to land use.However, when the range of variantion of k was plotted according to aggregate size class, a reasonably clear trend appeared (Figures 4 and  5), indicating that k values were clustered around successively larger values as size class increases.This suggests that land use had more impact on the amount of soil material in each size class (aggregate distribution, Figures 2 and 3) than on aggregate stability.
The effect of land use on the k values can be inferred from the dispersion of values at each depth and size class (points along each line on Figures 4 and 5).Although not dominant, the effect of land use is greatest (larger range) in size classes larger than 500 m and in surface horizons (Figures 4 and  5).This is in agreement with the aggregation model suggested by Oades & Waters (1991), in which large aggregates are more dependent of fungal hyphae and fine roots, and therefore on land use, while the small aggregates depends more on soil colloidal properties and chemistry.
Considering that the release of WDC was closely related to the disaggregation of aggregates larger than 500 m, that the conventional and no-tillage systems promoted a decrease in the amount of large aggregates, and that there is no aggregate hierarchy, a considerable amount of clay can potentially be lost during the change from forest to agricultural land use.In all cases, maintenance of large aggregates is, according to these results, essential to avoid an increase in WDC.Adding to this, complexes of clay minerals and organic colloids, which increase the potencial for dispersion, should be greater in the surface horizons (Tombácz et al., 2004).

Figure 1 -
Figure 1 -Location of sampling sites.

Table 2 -
Analysis of variance among the regression models for disaggregation curves from the three land uses.(P values).