GEOECOLOGICAL DRIVERS OF CERRADO HETEROGENEITY AND 13C NATURAL ABUNDANCE IN OXISOLS AFTER LAND-USE CHANGE(1)

The 13C natural abundance technique was applied to study C dynamics after land-use change from native savanna to Brachiaria, Pinus, and Eucalyptus in differently textured Cerrado Oxisols. But due to differences in the δ13C signatures of subsoils under native savanna and under introduced species, C substitution could only be calculated based on results of cultivated soils nearby. It was estimated that after 20 years, Pinus C had replaced only 5 % of the native C in the 0–1.2 m layer, in which substitution was restricted to the top 0.4 m. Conversely, after 12 years, Brachiaria had replaced 21 % of Cerrado C to a depth of 1.2 m, where substitution decreased only slightly throughout the entire profile. The high δ13C values in the subsoils of the cultivated sites led to the hypothesis that the natural vegetation there had been grassland rather than Cerrado sensu stricto, in spite of the comparable soil and site characteristics and the proximity of the studied sites. The hypothesis was tested using aerial photographs of 1964, which showed that the cultivated sites were located on a desiccated runoff head. The vegetation shift to a grass-dominated savanna formation might therefore have occurred in response to waterlogging and reduced soil aeration. A simple model was developed thereof, which ascribes the different Cerrado formations mainly to the plant-available water content and soil aeration. Soil fertility is considered of minor significance only, since at the studied native savanna sites tree density was independent of soil texture or nutrient status.


INTRODUCTION
The Cerrado region covers roughly two million square kilometers of Central Brazil, i.e. 20 % of the national territory, and is characterized by a seasonal climate, low-fertility soils, and a savanna-like vegetation (Adámoli et al., 1986). During the past four decades, the Cerrado region underwent severe changes as gradually more areas were cleared for cultivation. Today pastures cover around 400,000 km 2 , and approximately 150,000 km 2 are cultivated with annual crops and tree plantations (Resck et al., 2000).
After land-use change, C derived from the introduced species (predominantly Brachiaria pastures, maize and soybean as annual crops, and Pinus or Eucalyptus in tree plantations) gradually replaces the original C. By the 13 C natural abundance technique (Balesdent & Mariotti, 1996) this alteration can be calculated based on the physiologically different CO 2 assimilation pathways between C-3 and C-4 plants. C-3 plants have δ 13 C signatures ranging from -32 ‰ to -22 ‰ with an average of -27 ‰, and C-4 plants have values ranging from -16 to -9 ‰, with a mean of -13 ‰ (Balesdent & Mariotti, 1996).
Since Cerrado sensu stricto (s.s.) comprises both C-3 trees and shrubs and C-4 grasses, its δ 13 C signature is between that of pure C-3 and C-4 vegetations. Roscoe et al. (2001) and Wilcke & Lilienfein (2004) used this feature to quantify the replacement of C derived from Cerrado vegetation after introduction of Brachiaria and Pinus on clayey Oxisols. Neufeldt (1998) applied the 13 C technique to differently textured Cerrado Oxisols, but was not able to calculate C replacement due to surprisingly high subsoil δ 13 C values on the sites with the introduced crops.
Neufeldt (1998) hypothesized that slightly higher groundwater levels at the catchment scale had induced a higher natural C-4 grass abundance in the past. Accordingly, catchment scale variations of the moisture balance could explain the occurrence of Cerrado savannas of varied densities on the nearly flat terrain (< 2 % inclination). Similarly, Ker & Resende (1996) observed that the Cerrado on hilltops was tree-dominated and gradually changed into grassdominated Cerrado formations towards the effluents. In studies of forest-savanna boundaries within the Cerrado region, Emmerich (1989) found that the soil moisture balance was the main determinant factor of vegetation formation, so that mesophytic forests developed where the moisture balance was more stable, whereas soil fertility had no marked influence.
However, this hypothesis disagrees with the common belief that, within the Cerrado biome, there is a relation between, the shift from grass to tree dominated Cerrado formations (3) and soil fertility (Lopes & Cox, 1977;Haridasan, 2000), water availability (Alvim, 1996), or soil acidity and aluminum toxicity (Goodland & Pollard, 1973;Goodland & Ferri, 1979). Furthermore, frequent burning strongly affects height and density of the Cerrado vegetation (Eiten & Sambuichi, 1996;Roscoe et al., 2000), but fires are nowadays seen as a predominantly anthropogenic factor of Cerrado degradation (Alvim, 1996). These contrasting hypotheses on the determining factors of Cerrado formations call for more in-depth research.
For this study, δ 13 C of soils under Cerrado, Brachiaria pastures, and Pinus and Eucalyptus plantations were analyzed and aerial photographs were assessed to test the hypothesis of waterlogging as a key factor of the varied catchment-scale Cerrado formations.

Site location and history
The study area lies between 19.10 ° to 19.20 ° S and 48.12 ° to 48.18 °W, at a distance of approximately 25 km SSE from Uberlândia, Minas Gerais, Brazil. The mean annual temperature is 22 °C and average precipitation 1.650 mm, 90 % of which falls between October and April. Coarse-loamy, mixed, isohyperthermic Typic Haplustox and very fine, allitic, isohyperthermic Anionic Acrustox (Soil Survey Staff, 1997) were chosen for a regionally representative sampling . According to the Brazilian taxonomy, the soils were classified as Latossolos vermelho-amarelos álicos A moderado textura média / muito argilosa fase cerrado (Camargo et al., 1986). On the clayey soil, a degraded Brachiaria decumbens pasture, a Pinus caribaea plantation and a Cerrado s.s. savanna were chosen, while on the coarse-loamy soil a degraded Brachiaria decumbens pasture, a Eucalyptus citriodora plantation and a Cerradão were selected. The sites were chosen for their longstanding management history and for being close to each other (< 2 km). The savanna sites differed in tree density, because no comparable sites could be identified within an acceptable distance. Detailed management histories as well as a discussion on the chemical and physical properties of the soils under study are provided in . Table 1 gives an overview of site locations, management records, and soil properties.

Sampling and sample preparation
Soil samples were collected in March 1995, at the end of the rainy season, from each of the six sites. At five randomly selected points per site, samples were taken with an Edelman soil auger from the 0-0.1, 0.1-0.2, 0.2-0.4, 0.4-0.6, 0.6-0.8, 0.8-1, and 1-1.2 m layers and pooled. The samples were dried in an airforced oven at 40 °C and passed though a 2 mm sieve to remove roots. For the chemical analyses, subsamples from each depth were ground for greater homogeneity. Litter samples were collected using a 0.25 × 0.25 m 2 frame. Thirty samples per treatment were randomly collected, pooled, and dried in an airforced oven at 65 °C for at least 48 h. One litter subsample per treatment was finely ground for further analysis.

Analytical methods
The analytical methods for the determination of soil properties are described in . The signature of stable carbon isotopes was obtained after burning samples of 20-100 mg in the presence of CuO under a pure oxygen stream at 900 °C in a dry combustion furnace. In the resulting CO 2, 12 C and 13 C were determined with a mass spectrometer (Europa Scientific Roboprep Tracermass) and corrected for 17 O influence (Craig, 1957). All analyses were performed in triplicate and presented as mean ± standard error. The δ 13 C value was calculated according to the international V-PDB standard and expressed in tenths of a percent (‰), according to the equation: (1)
In the clayey soil under Cerrado s.s., δ 13 C was -20.0 ‰ in the 0-0.2 m layer and increased slightly with depth, reaching -18.7 ‰ at 0.8 m. Below that, the signal remained almost constant. Conversely, in the coarse-loamy soil under Cerradão, δ 13 C in the 0-0.1 m layer was -23.1 ‰, which reflects the higher tree component, and asymptotically approached -17.3 ‰ in the 1.0-1.2 m layer. Between 0.6 and 1.0 m the signals were erratic, which might reflect past vegetation shifts or distinct rooting distributions. According to Wedin et al. (1995), the frequently observed δ 13 C increase at greater depths is related to Despite lower δ 13 C values of the litter, the soil signatures under Pinus and Eucalyptus clearly suggested enrichment of δ 13 C throughout the profiles, in comparison to the respective Cerrado soils. There was a curvilinear signal increase in the upper 0.6 m ranging from -18.9 ‰ to -15.9 ‰ in the clayey soil and from -20.8 ‰ to -14.0 ‰ in the coarse-loamy soil, followed by a continuous decrease to -17.4 ‰ and -15.8 ‰ in the clayey and coarse-loamy soil, respectively. The signal range was therefore larger in the coarse-loamy soil.
The δ 13 C signals in clayey soils under Brachiaria were constant at around -15.2 ‰ to a depth of 0.6 m and subsequently decreased to -17.1 ‰ in the deeper subsoil. Conversely, in the coarse-loamy soil under Brachiaria, signals increased stepwise from -16.8 ‰ in the 0-0.1 m layer to -14.9 ‰ in the 1.0-1.2 m layer. Figure 2 shows a section of an aerial photograph of 1964. Very clayey soils occur on the unconsolidated fine-textured Tertiary sediments of the tableland, to the right of the FEPASA railroad tracks at around 950 m altitude. Left hand of the tracks, the clay content of the soils continuously decreases as the sandstones of the Marília formation (Eocretaceous) are exposed (Nishyama, 1989), reaching an altitude of approximately 900 m at the BR 050 interstate highway (then under construction). Hence, the sites with clayey soils were located near the railroad tracks, whereas the sites with coarse-loamy soils were close to the road (site locations encircled).

Interpretation of aerial photographs
Savanna dominates the natural vegetation between the effluents and is generally denser on hilltops ( Figure 2). The other natural vegetation forms are closely associated to the drainage system. Mata de galeria (gallery forest) grows where inclination gives the runoff a downhill direction. Veredas, which are palm groves characterized by the Burití palm (Mauritia flexuosa), are found in shallow valleys with diffuse runoff and high groundwater levels throughout the whole year. Murundus are convex microstructures (probably termite mounds) of up to 1.5 m height and 20 m diameter within flat depressions or runoff heads with very low inclination (Ribeiro & Walter, 1998). The vegetation of the runoff heads is predominantly grassy with few or no woody components. In the study area, Murundus only occur on the fine-textured sediments of the tableland (Schneider, 1996). Figure 3 shows these structures more clearly, together with former runoff heads, later dried out, so that Pinus was planted on them in 1979.

Replacement of Cerrado C after land-use change
Since the cultivated soils and the adjacent Cerrado sites did not show similar subsoil δ 13 C signatures, the Cerrado soils being depleted in δ 13 C compared to the soil under introduced species, and hence lacking a common baseline, a direct calculation of C substitution after land-use change was not possible (Figure 1). On the clayey soils this was probably caused by a higher C-4 grass component in the past due to their proximity to an ancient runoff head (Figure 3). On the coarseloamy soils, the Cerrado site seems to have been exceptionally dense, suggesting that the grass component had again been higher on the Eucalyptus and Brachiaria sites prior to cultivation (Figure 2).
An estimation of the C proportion derived from Pinus and Brachiaria in the clayey soils however seems possible, based on results of Wilcke & Lilienfein (2004) of comparable soils nearby. For the loamy soils, the calculation of C replacement by introduced species was not undertaken, since neither reference studies were available to compare the measured values with nor a similar subsoil δ 13 C distribution between Cerradão, Eucalyptus, and Brachiaria that could serve as baseline. Wilcke & Lilienfein (2004) showed that 20 yr of Pinus caribaea plantation on a clayey soil only led to C replacement in the topsoil (0-0.3 m), whereas C in the subsoil (0.3-2 m) remained unaffected by the landuse change. Similar results were obtained by Trouve et al. (1994) for 30 -year-old Pinus caribaea stands in Congo. Wilcke & Lilienfein (2004) further reported that Brachiaria decumbens -derived C replaced 5-31 % of the original in the 0-2 m layer after 12 yr of cultivation. Roscoe et al. (2001) described a similar C substitution throughout the whole soil profile after 23 yr under Brachiaria spp. pasture in Sete Lagoas, MG, Brazil. Hence, C substitution rates for Pinus and for Brachiaria were comparable at different locations in Brazil and elsewhere, and it therefore seems valid to assume a similar response to the introduction of Pinus and Brachiaria on the study sites as well.
C substitution after vegetation change to Pinus was calculated using the measured δ 13 C values, and assuming that no C replacement occurred below 0.4 m under the studied Pinus plantation. It was estimated that C derived from Pinus replaced 27 % of the original C in the 0-0.1 m layer, 10 % in the 0.1.0.2 m layer, and 1 % in the 0.2-0.4 m layer. In the total layer (0-1.2 m), only 5 % C were Pinus-derived after 20 yr. The estimated values strongly agree with the results of Wilcke & Lilienfein (2004) and Trouve et al. (1994), and probably reflect the fact that the Pinus litter accumulates on the soil surface, forming a thick organic layer (5-20 cm) that is only slowly incorporated into the soil (Neufeldt et al., 2002), and a comparatively low rooting intensity in the subsoil.
C substitution by Brachiaria on the clayey soil was calculated by relating the δ 13 C pasture values to those of the corresponding layer of the Pinus site, because both appeared to have had a similar grassland vegetation in the past due to the adjacent runoff head (Figure 1). C substitution under Brachiaria was estimated to be nearly constant at about 24 % to 0.6 m depth, and then decreased gradually to 6 % in the 1.0-1.2 m layer. In the total layer (0-1.2 m), 21 % C were pasture-derived. Both the amounts of C replacement and their distribution in the profile are in close agreement with the results of Wilcke & Lilienfein (2004) and Roscoe et al. (2001). The high C substitution is a response to the profound and dense pasture rooting system and the relatively low recalcitrance of the soil organic matter despite high clay contents (Roscoe et al., 2001;Neufeldt et al., 2002). Summing up, the underlying assumptions allowed the calculation of C replacements, which agree very well with published results. This suggests that the initial hypothesis, according to which the vegetation on the Brachiaria and Pinus sites had been dominated by C-4 grasses in the past due to waterlogging, appears to be correct. Aerial photographs (Figures 1 and 2), which were taken into consideration thereupon, reinforced the premise. However, analyses of stable isotopes and soil moisture contents along natural Cerrado vegetation gradients are required to verify the hypothesis.

Cerrado formations in response to geoecological drivers
Aerial photographs (Figure 2) and 13 C natural abundance ( Figure 1) indicated a strong heterogeneity of the studied savanna formations and suggest that high groundwater levels limit Cerrado tree growth in favor of grasses. This is confirmed by Eiten (1972), according to whom Cerrado trees are sensitive to waterlogging. Since the lateral extent of the groundwater influence is related to inclination, the affected zone may be very large on flat topographies. On the other hand, the coarse-loamy soils are clearly less fertile than clayey soils as indicated by their substantially lower cation exchange capacities (CEC) and C concentrations (Table 1). This does not seem to affect Cerrado physiognomy, as dense Cerrado s.s. and Cerradão frequently occur on the coarse-loamy soils derived from the Marília formation sandstones (Figure 1). Hence, on a catchment scale of several square kilometers, soil fertility does not seem to play such a prominent role for Cerrado density as frequently proposed (Goodland & Pollard, 1973;Lopes & Cox, 1977;Goodland & Ferri, 1979;Adámoli et al., 1986;Haridasan, 2000). This is consistent with Emmerich (1989) who showed that mainly the soil moisture balance determined the vegetation form along savanna-forest boundaries within the Cerrado region, whereas soil fertility played only a minor role.
Recently, Ruggiero et al. (2002) studied the soilvegetation relationships in Campo cerrado, Cerrado s.s., and Cerradão, but found no similar trends in the Cerrado formations in terms of nutrient status, base saturation, Al saturation, and CEC. However, in contrast to Emmerich (1989), they could clearly distinguish the soil fertility of the Cerrado formations from the adjacent semideciduous forest. The forest plots presented higher clay contents and CEC and the Cerrado higher Al saturation in surface soils.
Based on these observations, a simple model is proposed in an attempt to integrate the distribution of Cerrado formations. The model centers around the plant-available water content and soil aeration, in view of the plants' water stress during the dry season (Alvim, 1996) and the sensitivity of the Cerrado trees to waterlogging (Eiten, 1972), respectively. On the other hand, soil fertility is given a minor role only. The model is restricted to the catchment scale (up to several km 2 ) and explicitly does not question the superordinate relevance of climate for the occurrence of the phytomorphological domains on the Brazilian territory. The Cerrado biome prevails where the amount and temporal distribution of precipitation is intermediate between the moist Amazonian and Atlantic forest biomes on the one hand and the dry Caatinga and Chaco biomes on the other (Ab 'Sáber, 1971;Adámoli et al., 1986). The effects of fire on the Cerrado features are not discussed in this context either, since burning is seen as a predominantly anthropogenic cause of Cerrado degradation (Alvim, 1996;Eiten & Sambuichi, 1996). Table 2 shows the different vegetation formations that occur naturally within the study region vs. different geoecological drivers. With exception of the Vereda, which is characterized by the Burití palm tree, the vegetation formations are in the order of increasing canopy height and tree density (Goodland & Ferri, 1979). While the Vereda is flooded for most of the year, the grasses on the Murundu and Campo limpo or Campo sujo may temporarily suffer from drought during the dry season due to the restricted rooting depth, which limits the plant-available water content. While in the study area rooting depth is restricted by a high groundwater level in the vicinity of the effluents, solid bedrock may limit root penetration in other regions. According to Eiten (1972), Campo sujo and Campo limpo are frequently found on very shallow soils (e.g. steep slopes or rock outcrops) with good or poor drainage. Goodland & Ferri (1979) suggested that the few dwarf trees on the Campo sujo might reflect faults within a laterite crust or the bedrock. As rooting depth and thus the plant-available water content increases, the grass-  (Eiten, 1972). Under otherwise comparable conditions, soil fertility should also play a role for the distribution of the vegetation forms and certainly has an influence on the phytosociological composition of the Cerrado (Goodland & Pollard, 1973;Alvim, 1996). In conclusion, tree density rises with increasing plant-available water content provided that good internal drainage conditions a well aerated solum which does not impose physical resistance (e.g. waterlogging, laterite, bedrock) to deep rooting. The Mata de galeria is restricted to the effluents with directed runoff and occurs on deep, well-drained, and possibly more fertile soils. Plants of the Mata de galeria, which floristically belong to the mesophytic Atlantic forest biome (Eiten, 1972), have access to water from the capillary fringe of the stream and are therefore less exposed to drought during the dry season. Apparently, Cerrado vegetation cannot compete with the gallery forest under these circumstances. Similarly, Emmerich (1989) reported that this feature can be observed on the intra-montane planes west of Brasília (Cerrado biome), where additional water (and possibly nutrients) from the surrounding slopes allows mesophytic forest to thrive. The centers of these planes, where the groundwater level reaches its minimum are, however, dominated by dense Cerrado.
Recapitulating, the presence of different Cerrado formations can mainly be ascribed to good internal drainage and the absence of waterlogging or other physical barriers to rooting. The soil nutrient status or Al saturation, frequently cited to explain the occurrence of different Cerrado formations, are seen as less important.