Under natural conditions, from the upper limit of the Bw horizon to the surface (Fig. 1 and Table 2), the reference soil (CER) presented a characteristic succession of layers, with a strong fine to very-fine granular structure (1), two intermediate layers in which a weak to moderate medium subangular blocky structure developed in the granular matrix (2 and 3), and the upper layers showed typical subangular blocky (4) and granular structures produced by faunal activity (5). We compared the structures found between 0.05 and 0.50 m depth, where the macromorphological transformations are the most significant (Fregonezi et al., 2001) (Fig. 1, Tables 2 and 3). We found that the soil of the old CB pasture, compared with that under native vegetation (CER) had undergone major modifications down to a depth of 0.5 m, resulting in a general loss of visible porosity and a change in the type of subangular blocky (4) and granular structures (5) to coarse to fine angular blocky and coarse subangular blocky with cracks (7), and a platy structure in the upper layers (8) (Fig. 1, Table 2). The transformation of this old CB pasture into new pastures (RB or RBS), observed after 4 years of development, resulted in recovery of a fine porosity, with the formation of new small and medium-sized rounded aggregates included in a subangular blocky structure, and a fine to medium subangular blocky to fine to coarse granular structure in the upper layer (structures 4 and 10, Table 2).

The volumetric water contents of the structural volumes are shown in Table 4. The relationship between tension applied and pore class (given in equivalent pore radius) was as follows: −1 and −3 kPa = 150–47 μm; −3 and −10 kPa = 47–15 μm; −10 and −33 kPa = 15–4.7 μm ; −33 and −1500 kPa = 4.7–0.094 μm. In the 0.5–0.35 m layer and for the pore class 47–150 μm (−1 to −3 kPa), structural volumes 3 of CER and 4 of RB showed the highest values; for the 15–47 μm pores (−3 to −10 kPa), the highest values were in structural volumes 4 of RB and 7 of CB; for the 4.7–15 μm pores (−10 to −33 kPa) structural volumes 7 of CB and 3 of CER showed significant variation; finally, for the finest pores, structural volumes 3 and 4 of CER and 7 of CB were notable but not significantly different.

In the 0.35–0.50 m layer, the values obtained in structures 9 of RB were significantly the highest in the porosity range −1 to −3 kPa (47–150 μm), whereas there were no particularly significant differences in the other pore classes.

In the micro-aggregated structures, structure 1 (CER, CB, RB and RBS), the values were equivalent in all pore range.

The largest biomasses, observed in the monoliths sampling, were founded in the CER, followed by those in the RB and RBS sites (Table 5). Termites and ants were present everywhere, whereas earthworms, which were virtually absent in the CER and CB (where only a few individuals were observed) were well represented in the RB and RBS plots. Results showed that earthworm density was particularly high in the 0.1–0.2 m layer in the RBS plot. The PCA correlation matrix showed a negative correlation within the D_b values and termite biomass values (Table 6). The PCA produced an optimized factorial model with three factors accounting for 79.8% of the total variance (Fig. 2). Factor 1 explained 41.0% of the variance and was negatively correlated with soil bulk density (−0.80), aggregate bulk density (−0.60) and positively with Isoptera biomass and density (0.95; 0.65) and biomass and density of Formicidae and Coleoptera. Thus, factor 1 represented the effects of the soil Isoptera, Formicidae and Coleoptera on bulk density, since 58% of the data on factor 1 were contributed by their biomass. Factor 1 clearly differentiated the contribution of insects (termites, ants, coleoptera) from that of earthworms. Factor 2, accounting for 20.3% of the variance, was positively correlated with earthworm biomass (0.61) and aggregate bulk density (D_b ag); and negatively with bulk density of structural features (−0.68). The D_b ag is accounted for 45.5% in factor 2. The faunal survey in RBS, 0.1–0.2 layer, shows high earthworm density, in contrast to the RB 0–0.1 layer in which insects were dominant. Factor 2 differentiated CER, in terms of the contribution of termites and ants, from the pasture upper layers.

Results showed a fairly good textural homogeneity at each depth among the ferralsols studied; values for particle density and soil colour confirmed this homogeneity (Tables 1 and 3). Textural variations were caused by material brought back up by termites and ants and subsequent mixing, which we were able to observe in the field. Brazilian ferralsols are normally known for featuring weakly developed fine granular macrostructure, which expresses a stable microstructure with aggregates of 80 to 350 μm in size (Furian et al., 1999; Lima and Anderson, 1997; Neufeldt et al., 1999). A more detailed analysis was performed to separate this microstructure into individual structural units (Balbino et al., 2001), and to quantify them (Volland-Tuduri et al., 2005). The structures of the CER soil were similar to those previously observed (Balbino et al., 2002a; Reatto et al., 2007; Volland-Tuduri et al., 2005).

Some studies show that when ferralsols are transformed into pasture or intensively farmed, the structure is altered and bulk density values increase (Curmi et al., 1993; Kondo and Dias Junior, 1999; Stone and Silveira, 1978; Tavares Filho and Tessier, 1998). This was not observed in our study, bulk density values not showing any significant variation at the soil horizon scale for the different types of usage (Table 1). However, if we examine the bulk density values of the structures (D_bvol), results showed an increase in D_bvol from CER to CB, and a drop from CB to the two recent pastures, RB and RBS (Table 3).

The changes that occur can be defined by calculating pore volumes. The total pore volume of the structures (V_pt) was calculated based on particle density (D_p) and bulk density (D_b) (Table 3). It has been suggested that there is a close linear relationship between the pore volume due to the assemblage of clay particles and the clay content in ferralsols (Balbino et al., 2002b). Using this relationship (Table 3), we estimated that the mean pore volume associated with the clay of the soil studied is 0.185 ± 0.013 cm³ g⁻¹. This calculation shows that 28 and 32% of the total soil porosity (V_pt) were due to the organisation of the clay particles. At the soil surface, the changes that had taken place in the old pasture (CB) compared with the soil under natural conditions (CER) are related to the development of physical barriers to infiltration, with laminar structures at the surface (Fig. 1, Table 2) and a noticeable reduction of pore volumes above 47 μm down to a depth of 0.35 m (Table 4). This result confirms earlier observations (Balbino et al., 2001; Carvalho et al., 1998; Grimaldi et al., 1993). It is notable that the transformation of this ancient pasture to RBS and RB showed a similar change in structures 4, similar to the structure 4 described for the CER (Fig. 1, Tables 2, 3 and 4). In the 0.35–0.50 m layers, the renewal of CB into RB significantly changed the porosity > 47 μm and contributed to a total porosity greater than RBS (Table 4). In the micro-aggregate layers, we found a wide variation of the increase in the porosity for the pore classes > 15 μm. This is explained by the presence of small aggregates whose porosity was comparable in all the plots (Table 3), but the total porosity of the whole horizon was highest in CER (0.693 dm³ kg⁻¹), as shown by the bulk density values (Table 3). Much of the intra-aggregate and inter-clod porosity is due to vigorous rooting (Table 2). In the RBS, there was a clear reduction in the internal porosity of the macro-aggregates. This was one of the characteristics of these coarse aggregates, previously described as compact to very compact and whose bulk density values (D_bag) were higher than those of the structural volume from which they came (columns D_bvol and D_bag, Table 3). Consequently, the water supply from these soils is provided by a coarse network of inter-clod voids and aggregates characteristic of an assemblage that is piling up.

Coarse aggregates, described as compact to very compact (Tables 2 and 4), are structures resulting from the activities of two species of earthworm observed which have strongly colonised these plots (Fig. 2, Table 5). This result is in line with previous work (Blanchart et al., 1997, 1999) describing compacting worms. The colonisation was compatible with the higher carbon contents observed in the surface horizon of RB and RBS in comparison to the other sites (Table 1).

Usually, cultivating the soil leads to the disappearance of macropores formed by biological activity (Borges et al., 1999). In ferralsols on basalt, development of the structure into micro-aggregates at the surface and subsurface following farming practices has been demonstrated (Tavares Filho, 1995). If macropore shrinkage can be attributed to the introduction of cultivation, our study clearly shows that the drop in porosity is largely offset by rearrangement into elementary structures, the macro-aggregates observed in structures 4 and 10 (Table 2). What therefore characterised these ferralsols was this juxtaposition of a microstructure with coarse aggregates. The mass proportion of micro-aggregates (< 0.84 mm, Table 3) varied with bulk density (Φ_<0.84 = 1.97 [1/D_b] − 1.52) (Volland-Tuduri et al., 2005). The mass proportion of micro-aggregates, denoted 1 (Fig. 1, Table 2), estimated using this equation, indicated the absence of micro-aggregate structures in the macro-aggregates of RBS and CB according to the D_bag values, and their significant presence in the structures of CER and in the deep layers of CER, CB and RB (Φ_ag in Table 3). Their absence from RBS was due to earthworms that sorted and totally reorganised the microparticles as they passed through the intestine and were compacted during excretion (Barois et al., 1993; Chapuis-Lardy et al., 1998). The insects, termites and ants tended to create porosity in transporting the soil and tunnelling, thereby helping to reduce soil bulk density, as illustrated by the PCA (Fig. 2). In contrast, endogeic earthworms promoted the formation of large stable macro-aggregates (Lavelle and Spain, 2001) that eventually led us observe a close relationship between the bulk density of the soil layers and the structure of the earthworm community (Rossi, 2003). This was observed in PCA and the correlation matrix where the bulk density values of soil layers and aggregates were negatively correlated with termite biomass and density (Table 6).