| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a C.C. Cerri, Centro de Energia Nuclear na Agricultura, Universidade de São Paulo, CP.96. Piracicaba, Brazil
b Agriculture and the Environment Division, IACR-Rothamsted, Harpenden, Hertfordshire AL5 2JQ, UK
c Institut de Recherche pour le Développement, France
* Corresponding author (cepcerri{at}cena.usp.br).
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
Abbreviations: BIO, microbial biomass • CEC, effective cation exchange capacity • DPM, decomposable plant material • HUM, humified organic matter • INPE, Instituto Nacional de Pesquisas Espaciais • IOM, inert organic matter • RothC, the Rothamsted C turnover model • RPM, resistant plant material • RMSE, root mean square error • M, mean difference between measurements and simulation
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
Over the last 25 yr more than 70 million ha of native vegetation in Brazil have been replaced by pastures for beef production. The substitution of native vegetation on such a large scale with African grasses (often from the genus Brachiaria) is likely to have an impact on nutrients and organic matter composition, as well as a regional impact on hydrology and water quality. It would be expected to affect CO2 respiration and C sequestration in soils. In turn, these effects on C dynamics at large landscape scales would be important in global C budget and climate change.
One striking example of such changes is the replacement of forest by pasture in the Amazon Basin (Moraes et al., 1996; Piccolo et al., 1996; Steudler et al., 1996; Fearnside and Barbosa, 1998; Cerri et al., 1999). The Amazon Basin covers an area of some 7 million km2, and the central part is almost entirely located within Brazilian territory (Pires and Prance, 1986). This region has the highest rates of deforestation in the world (Skole and Tucker, 1993; INPE, 1998). Between 15000 and 29000 km2 of forest were cleared every year over the past two decades, and the total area deforested now exceeds 500000 km2 (INPE, 1998). Cattle pasture dominates this once-forested land in most of the basin (Pires and Prance, 1986; Skole and Tucker, 1993). Fearnside and Barbosa (1998) estimated that 75% of the deforested land had been managed as pasture at one stage or another.
A decline in soil C stocks is almost universally observed when tropical forest is cleared and cropped (Batjes and Sombroek, 1997; Shang and Tiessen, 1997; Lal, 1998; Bruce et al., 1999). In contrast, pasture grasses have the potential to introduce large amounts of organic matter to the soil (Fisher et al., 1994; Boddey et al., 1996; Rezende et al., 1999). Increased soil C concentrations in surface horizons are a common consequence of pasture formation from cleared moist tropical forest in the Amazon Basin (Bonde et al., 1992; Moraes, 1995; Trumbore et al., 1995; Moraes et al., 1996; Neill et al., 1997; Bernoux et al., 1998a). Fearnside and Barbosa (1998) reviewed soil C changes brought about by conversion of the Brazilian Amazon forest to pasture and found that pasture soils can be a net sink or a net source of C, depending on management.
Our goal in this paper was to model long-term impacts of converting forest to pasture on C stocks, using data collected from the Nova Vida ranch in the Western Brazilian Amazon basin. Some of these data are new and some come from earlier work.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
The Nova Vida Ranch covers an area of 22000 ha, and is a mixture of native forest and pastures of different ages. Different areas of the ranch were converted from forest to pasture in 1989, 1987, 1983, 1979, 1972, 1951, and 1911. Dates of conversion were established from Nova Vida records and from satellite images (Moraes et al., 1996). All the pastures studied in this work are within 5 km of each other and the study area has a similar topography (Moraes et al., 1996; Steudler et al., 1996).
According to Neill et al. (1997) this chronosequence represents one of the longest sequences in the Amazon region. The pastures were converted directly from forest without intermediate use for annual crops. This makes them particularly valuable for evaluating the effects of continuous pasture, without the confounding factor of those brief cropping phases that are common in the Amazon and which complicate many pastures studies.
Soils are classified as Podzólicos Vermelho-Amarelo (Red Yellow Podzolic) in the Brazilian classification scheme and as Ultisols (kandiuldults) in the U.S. soil taxonomy (Moraes et al., 1995). This soil type covers about 35% of the Brazilian Amazon Basin (Bernoux et al., 1998a; Bernoux et al. 1998b; Cerri et al., 1999). According to Moraes et al. (1996), the Red Yellow Podzolic occurs on both the top and the upper slope of the low hills that characterize this region. Below the leaf litter, the surface horizon is normally a very thin (10 cm) layer of bleached sand. The A horizon is a weakly structured sand clay loam, about 10 cm thick. The AB horizon extends to 25 cm. The B1 horizon is brown or yellowish red, gradually changing with depth to a sandy clay, with a weakly developed blocky structure. The underlying B2 horizon is yellowish red to red, soft, porous, sandy clay in texture with a massive structure. The lower part of this horizon may contain up to 50% gravel and stones (from 2–20 cm in diameter) composed of flat ferruginous sandstone, subrounded ferruginous rock, and angular and subangular quartz. The average depth of this gravel and stone layer is 150 cm, with an observed range of 50 to 250 cm. In the underlying BC layer, no gravel is encountered but some fragments of weathered rock can be observed. The clay minerals consist of kaolinite and small amounts of gibbsite in both B and C horizons.
The native forest vegetation of the ranch is classified as 
open humid tropical forest,
with large numbers of palms. The most common palms are Orbignya barbosiana, Oenocarpus spp., Jessenia bataua, Euterpe precatoria, and Maximiliana regia (Pires and Prance, 1986). Selective logging removed three or four economically valuable trees per hectare in the forest sites between 1987 and 1990 (Piccolo et al., 1996). The pasture sites were developed by a slash and burn technique used to clear the original forest and then establish the grass species (Graça et al., 1999). This was done by cutting brush in March, followed by tree harvest in June and July. The remaining trees and bush were burned at the beginning of the next rainy season in September or early October, followed by seeding of pasture grasses. All pastures were created in a similar manner. The pastures cleared in 1989, 1987, 1972, and 1911 are dominated by brachiarão (Brachiaria brizantha) and the pastures cleared in 1983, 1979, and 1951 are dominated by colonião (Panicum maximum). Mechanized agricultural practices or chemical fertilizers were not used on any of the pastures (Steudler et al., 1996).
Soil Chemical, Physical, and Biological Properties
Most of the soil data used in this paper was taken from Feigl et al. (1995), Moraes (1995), Moraes et al. (1996), Neill et al. (1996), Piccolo et al. (1996), and Fernandes et al. (2001) (Table 1). Soil data from the 1999 sampling has not been previously published (Table 2). Data was combined from pasture sites sampled in 1991, 1992, 1993, 1994, and 1999, which provided a pasture chronosequence of 0 (forest), 2, 3, 4, 5, 6, 9, 10, 12, 13, 14, 16, 19, 20, 21, 27, 41, 42, 81, 82, or 88 yr.
|
|
2 kg) were taken by knife from the 0- to 5-, 5- to 10-, 10- to 20-, and 20- to 30-cm layers down the side of each pit. Samples were air-dried and sieved (2 mm) to remove stones and root fragments before analysis. Soil chemical and physical properties for the samples collected in 1999 were analyzed in the same laboratory, using the same techniques that were used to analyze the samples collected in 1991, 1992, 1993, and 1994. This procedure was adopted so that the analytical techniques were consistent over the whole time span, using the same quality controls. Soil bulk density, clay content, pH, effective cation-exchange capacity (CEC), base saturation, soil BIO C, C and N concentrations were measured as described by Anderson and Ingram (1989), Embrapa (1979), Piccolo et al. (1996), Moraes et al. (1996), Feigl et al. (1995), and Fernandes et al. (2001).
Bulk density was measured in the field with volumetric steel rings. Particle-size fractions were determined by hydrometer after destruction of organic matter with H2O2, followed by dispersion in a mixer with sodium hexametaphosphate (2 g in 250 mL water). Soil pH was measured in water (2.5:1) on air-dried soil. Total C and N were measured by dry combustion on a LECO CN elemental analyzer (furnace at 1350°C in pure oxygen). Microbial biomass C, measured by Feigl et al. (1995) and Fernandes et al. (2001), was determined using the chloroform fumigation-extraction technique (Brookes et al., 1985; Vance et al., 1987). Soil samples (corresponding to 25 g of dry weight) were fumigated with CHCl3 for 24 h at 25°C (Jenkinson and Powlson, 1976). After removal of the CHCl3, the C was extracted from fumigated and unfumigated samples with 0.5 M K2SO4 for 1 h on an end-over-end shaker (soil/solution 1:4). Organic C in the filtered extracts was determined by the acid dichromate oxidation method. Microbial biomass C flush (difference between extractable C from fumigated and unfumigated samples) was converted to soil BIO C using a KEC factor of 0.30 (Feigl et al., 1995).
Soil Carbon Stocks Adjusted for Bulk Density
Measurement of soil bulk density is essential to calculate soil C stocks (Bernoux et al., 1998b). Several studies have reported increases in soil bulk density in tropical pastures relative to the original forest (Bonde et al., 1992; Trumbore et al., 1995; Moraes et al., 1996; Neill et al., 1997; Bernoux et al., 1998b; Fearnside and Barbosa, 1998). We used the methodology described in Moraes et al. (1996), to correct soil C stocks to an equivalent depth basis, that is, the depth of pasture soil that contains the same mass of soil as the corresponding layer of the original forest. Therefore, we corrected soil C stocks to 30 cm based on sampling of a soil mass in the pastures that was equal to the mass to 30-cm depth in the native forest. This resulted in calculating C stocks based on a depth of slightly <30 cm when bulk density increased under pasture and slightly >30 cm when the bulk density decreased.
The Rothamsted Carbon Model
We used the Rothamsted Carbon Model (RothC-26.3) to simulate changes in soil C in the chronosequence. The RothC-26.3 (described in detail by Jenkinson et al., 1992; Coleman and Jenkinson, 1996 and Coleman et al., 1997) model predicts organic C turnover in non-waterlogged topsoils according to soil type, temperature, moisture content, and plant cover. It uses a monthly time step to calculate total C, BIO C, and 
14C on a years-to-centuries timescale. In this model, soil organic C is split into four active fractions and one small inert organic matter (IOM) fraction. The four active fractions are decomposable plant material (DPM), resistant plant material (RPM), BIO, and humified organic matter (HUM). Each fraction decomposes by a first-order process with its own characteristic rate. The IOM fraction is resistant to decomposition.
RothC is solely concerned with soil processes and does not contain a submodel for plant production as does the CENTURY model (Parton et al., 1987). The RothC model's main advantage is that it runs on data that are readily available (Smith et al., 1997).
We used the RothC under the assumptions that (i) all sites were the same before forest clearance and, (ii) that subsequent pasture developments proceeded similarly in the different sites. It is now not possible to check these assumptions because the soils were not sampled when pastures were created, nor was plant productivity measured. However, there were differences in clay content among sites that would not have changed. On the other hand pasture management was similar for all pastures (e.g., establishment, animal stocking rate, grass species), which means disturbance and C inputs were likely similar across the chronosequences.
Statistical Analysis
We used the following tests to compare the difference between measured and simulated data from the Nova Vida Ranch chronosequence: sample correlation coefficient (r), root mean square error (RMSE), and mean difference between measurements and simulation (M). Details of these tests can be found in Smith et al. (1996) and Smith et al. (1997).
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
|
Carbon/N ratios were generally higher in the surface soils of the pastures than in the native forest (Table 3). This behavior is consistent with the relatively greater accumulation of C compared with N in the older pastures (Moraes, 1995; Neill et al., 1996).
Forest clearance caused soil pH in the top 10 cm to increase by about two units within 5 yr after which it slowly decreased (Table 3). Clay content increased with soil depth and was different between sites, but these are not related to pasture age (Moraes et al., 1996; Bernoux et al., 1998a). The bulk density of forest soil in the surface 0- to 5-cm soil layer ranged from 1.14 to 1.24 Mg m-3 compared with pasture which was higher and ranged from 1.25 to 1.48 Mg m-3 in the 0- to 5-cm layer. Averaging the data in Tables 2, showed that 56% of the total soil C was in the top 10 cm.
Carbon stocks in the top 10 cm tended to increase with pasture age (Table 4). All seven pastures created at different times had higher C stocks than the forest. The smallest gains were in the youngest pasture and the highest gain, about 11 Mg ha-1, in the oldest.
|
The CEC varied from 3.79 to 8.41 cmol(+) kg-1 (Table 4). The CEC increased in young pastures (3 and 5 yr old), but thereafter declined to levels markedly below that of the native forest soil. Base saturation increased in young pastures and decreased somewhat in the older pastures, driven by declines in exchangeable base cations and increases in exchangeable Al3+. According to Neill et al. (1997), the short-term increases in soil pH and in CEC at Nova Vida arise because the forest was cleared by burning.
Modeling the Soil Carbon Dynamics
Modeling Soil Organic Carbon Before Forest Clearing
Before fitting the model to the data from the seven pasture sites in the chronosequence, it was necessary to run RothC to generate the C content in the soil at the starting point, taken to be 34.4 Mg C ha-1 for all seven preclearing forest sites. At that time the organic C content of the forest soil was assumed to be in steady state.
To do this, RothC was run for 10000 yr with a DPM/RPM ratio value of 0.25, the value specified by Jenkinson et al. (1992) for deciduous or tropical woodland. Inert organic matter was calculated from the equation of Falloon et al. (1998), since the radiocarbon age of the soil had not been measured on any of the soils in Table 3. From the measured C content of the 0- to 30-cm layer (34.4 Mg C ha-1), the Falloon et al. (1998) equation gave a value of 2.75 Mg C ha-1 as IOM. In the 0- to 10-cm layer the measured C was 15.7 Mg C ha-1 (Table 4), giving an IOM of 1.13 Mg C ha-1. Although we have no radiocarbon measurements to set IOM, work by Gomes (1995) and Pessenda et al. (1998) indicates that the Falloon et al. (1998) equation is not seriously in error when applied to these samples. From the radiocarbon measurements made by Gomes (1995) on soil sampled in 1994 from a forested area of the Nova Vida ranch, we used RothC to calculate that 6.6% of the organic C in the 0- to 30-cm layer was in IOM. The corresponding value calculated from the Falloon et al. (1998) equation was about 8%. The weather inputs were taken from the meteorological station at Nova Vida Ranch. Except when otherwise noted, the clay content of the soil was taken as the mean (weighted averages) of all the data, that is, 234 g kg-1 for the 0- to 10-cm layer and 286 for the 0- to 30-cm layer.
The annual input of C (from plant debris, roots, etc.) needed to generate 34.4 Mg C ha-1 at the starting point was calculated to be 4.52 Mg C ha-1 yr-1 for the 0- to 30-cm layer (monthly input 0.38 Mg C ha-1). In the same manner, to obtain 15.7 Mg C ha-1 in the 0- to 10-cm layer we needed an input of 2.04 Mg C ha-1 yr-1 (0.17 Mg C ha-1 mo-1).
Predicting Soil Organic Carbon After Forest Clearing
According to Boddey et al. (1996), pasture needs some time to become fully established after forest clearance. To allow for this, we arbitrarily set the plant input to zero in the first year of pasture creation: thereafter annual input was taken as constant, even though it probably took 2 or 3 yr to achieve this constant value. The input was applied uniformly throughout the year, even though measurements of aboveground production in Brachiaria pastures in Brazil and Colombia show production to be markedly greater in the wet season than in the dry (Boddey et al., 1996; Rezende et al., 1999). Jenkinson et al. (1992) specified two DPM/RPM ratios for grassland: 0.67 for unimproved grassland and scrub; and 1.44 for improved grassland. We decided that the pasture sites at Nova Vida are more similar to improved than unimproved grassland and used a 1.44 DPM/RPM ratio. The 0- to 30- and 0- to 10-cm layers were modeled separately: the weather inputs were used for the forest phase. The clay content used for modeling was 286 g kg-1 soil in the 0- to 30-cm layer and 234 g kg-1 soil for the 0- to 10-cm layer, which was the same for forest.
The annual input was iteratively adjusted to give the least squares best fit between model and data. The best fit for the 0- to 30-cm layer is shown in Fig. 1 for an input of 8.28 Mg C ha-1 yr-1. Figure 2 shows the corresponding best fit for the 0- to 10-cm layer; the input is 4.68 Mg C ha-1 yr-1. Visual evaluation shows tolerable agreement between modeled and measured data, especially in older pastures. The model predicts an initial decline in soil C stock in the first years, following conversion from forest to pasture, and then a steady increase during pasture establishment. The qualitative visual examination of the simulations (Fig. 1 and 2) is supported by the statistical analysis. Statistical tests between modeled and measured data in Fig. 1 gave a correlation coefficient (r) of 0.73, RMSE (showing total error) of 4.93% and a M (showing bias) of 0.67 Mg C ha-1. The corresponding values for Fig. 2 were 0.64, 3.69m, and 0.98, respectively (Table 5).
|
|
|
There are no measurements where total annual C input was reduced to compare with our modeled value of 8.28 Mg C ha-1 yr-1 for the 0- to 30-cm layer of soil, although there are some measurements of aboveground production that are broadly compatible with our calculations. Boddey et al. (1996) measured a mean herbage production (aboveground) of 22.1 Mg DM ha-1 yr-1 in a range of Brachiaria pastures—equivalent to 8.84 Mg C ha-1 yr-1. Rezende et al. (1999) developed a value of 20.6 Mg DM ha-1 yr-1 (8.24 Mg C ha-1 yr-1) for Brachiaria pastures with a stocking rate of 2 animals ha-1. Herbivores will consume some of this aboveground production, of which part will be returned in feces, making annual return to the soil lower than aboveground production by an amount dependent on intensity of grazing. Belowground production will contribute directly to the annual input.
The most likely explanation for the variation in the stock of soil C along the chronosequence (Fig. 1 and 2) is variability in the input of C from the pastures, particularly in the early years. Another possible explanation is that there were pre-existing differences in the total amount of soil C between sites at the time of clearing (Neill et al., 1997). Again this would be most important in the early years after pasture establishment.
Table 2 suggests that a Panicum dominated pasture (that started in 1983) accumulated less C in the top 10 cm of soil than pastures dominated by Brachiaria (those started in 1987, 1972, and 1911). However, the unknown spatial variability of soil properties among pastures mean that these differences must be treated with caution.
Analysis of RothC Assumptions
Figure 3
shows how three of the assumptions made in modeling the data in Fig. 1 affect the fit. The first assumption is related to the DPM/RPM ratio. The value specified in RothC for improved grassland (1.44) gives the fit shown in Fig. 1 with an input of 8.28 Mg C ha-1 yr-1. Using a DPM/RPM ratio of 0.67 (as specified for unimproved grassland and scrub) and the same annual input, increases the modeled C content of the soil 100 yr after conversion to pasture by 6.2% (Fig. 3).
|
The third assumption is the use of average clay content for all the pastures, regardless of the measured content as given in Table 3. Figure 3 illustrates the worst case—the pasture created in 1911, with clay content of 153 g kg-1 soil, as opposed to the average for all the soils of 286 g kg-1 soil. One hundred years after conversion to pasture the modeled C content, using 153 g clay kg-1 soil, of this site is only 7.4% lower than when modeled with the average clay content (Fig. 3). Smith et al. (2000) examined some of the data in Table 3 and came to a similar conclusion about the effects of clay. It is therefore unlikely that differences in clay content are in themselves sufficient to explain the discrepancies between predicted and measured in Fig. 1 and 2.
Modeling Microbial Biomass Carbon
Microbial biomass C measurements provide an independent test of model performance. Fig. 4
shows predicted and measured values for soil BIO C in the 0- to 10-cm layer of the Nova Vida chronosequence, using the same average clay content (234 g kg-1) for all the pastures. The annual input was also set to be the same as in Fig. 2 (4.68 Mg C ha-1 yr-1). Bearing in mind that no model adjustments were made to generate the values in Fig. 4 and that there were some large errors in measuring BIO; the modeled quantities of soil BIO C fit the measured data reasonably well. Table 5 gives the results of statistical tests for the accuracy of the BIO C simulation, and the tests are similar to those already used for total C simulations. The correlation coefficient is less than that for the corresponding total C simulation, presumably as a result of larger errors in some of the biomass C measurements. A M value of 0.03 Mg C ha-1 and RMSE value of 0.15% shows no consistent error in the fitting, and a negligible mean error between simulated and measured data points.
|
|
PERSPECTIVES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
Our conclusions on the turnover of organic C in Amazonian pastures are based on data from a single, well-managed ranch. Furthermore, more work is needed on other ranches to determine if the Nova Vida pastures are representative of other pastures in the Amazon region. It is commonly considered that cattle ranching in the Amazon region can never be a profit-making venture, as long as the only revenue is from the sale of cattle. But if monetary incentives or other subsides become available for C sequestration and land rehabilitation, this might provide additional motivation to improve management of areas already converted to pasture.
Models that consider soil, pasture, and animal variables will be essential if we are to understand the impact of management on pasture sustainability and calculate the financial implications of any specified system of management. Models for the turnover of organic C in soil, such as RothC, can contribute to our understanding of long-term ecosystem sustainability.
|
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
Received for publication March 15, 2003.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONPERSPECTIVESREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. deB. Richter Jr., M. Hofmockel, M. A. Callaham Jr., D. S. Powlson, and P. Smith Long-Term Soil Experiments: Keys to Managing Earth's Rapidly Changing Ecosystems Soil Sci. Soc. Am. J., March 12, 2007; 71(2): 266 - 279. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
) and forest (*). Bars indicate standard errors.
) and forest (*).
) and forest (*).