SSSAJ Grow Your Career with SSSA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by da Silva Fraga, V.
Right arrow Articles by Salcedo, I. H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by da Silva Fraga, V.
Right arrow Articles by Salcedo, I. H.
Agricola
Right arrow Articles by da Silva Fraga, V.
Right arrow Articles by Salcedo, I. H.
Related Collections
Right arrow Soil Erosion
Right arrow Isotopes
Right arrow Soil Organic Matter
Published in Soil Sci. Soc. Am. J. 68:215-224 (2004).
© 2004 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA

DIVISION S-6—SOIL & WATER MANAGEMENT & CONSERVATION

Declines of Organic Nutrient Pools in Tropical Semi-Arid Soils under Subsistence Farming

Vânia da Silva Fragaa and Ignacio H. Salcedo*,b

a Centro de Ciências Agrárias/Universidade Federal da Paraíba, 58597-000 Areia, PB, Brazil
b Dep. Energia Nuclear/Universidade Federal de Pernambuco, 50740-540 Recife, PE, Brazil

* Corresponding author (salcedo{at}ufpe.br).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Soils under subsistence farming in semi-arid northeastern Brazil are showing progressive declines in soil fertility. In this study, we assessed the effects of subsistence farming on losses of organic nutrient pools in 10 independent sites that were distributed in five districts of two states. Each site had adjacent areas under dry forest and subsistence farming. Areas under each land use were separated in two groups that had different degrees of land degradation. Group assignment was based on land use history and visual observations of vegetation-soil degradation and was further evaluated using 137Cs activity to estimate erosion. Four combinations of land use degradation were defined: undisturbed (UDF) and disturbed (DDF) dry forest, and preserved (PC) and degraded (DC) cultivated land. Soils were sampled in the 0- to 7.5- and 7.5- to 15-cm layers. The 137Cs concentrations declined in the order UDF > DDF {approx} PC > DC. Soil C and N concentrations in DC (8.9 and 0.94 g kg–1 soil, respectively) were half of those in UDF (17.8 and 1.51 g kg–1 soil, respectively). The organic P (Po)/inorganic P (Pi) ratio was 1.47 in UDF and decreased to 0.82 in PC and DC. Carbon losses in DC were attributed to erosion (43%) and to mineralization (57%) processes. Differences between UDF and other land uses were greater in the 0 to 7.5 cm than in the 0- to 15-cm layer. Carbon and N losses in low-P status soils, in addition to limited water availability and unsuitable land management techniques, are likely to restrict the recovery of degraded soils by traditional bush fallow techniques.

Abbreviations: Alf, Haplustalf • DC, degraded cultivated land • DDF, disturbed dry forest • L-Alf, Lithic Haplustalfs • L-Ent, Lithic Usthortents • PC, preserved cultivated land • Pi, inorganic P • Po, organic P • Pt, total P • SOM, soil organic matter • UDF, undisturbed dry forest • Ult, Haplustult


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
THE SEMI-ARID REGION in NE Brazil extends for almost one million km2 and has a predominant native vegetation of deciduous thorn forest or thorn bush savannah, locally known as caatinga (Sampaio, 1995). Subsistence agriculture occupies large areas within the region, creating mosaics of <1 to 2 ha of land showing various stages of caatinga regrowth (bush fallow), subsistence crops, or overgrazed grasslands. Soils were mostly derived from igneous rocks and for this reason the vegetation variability is also underlain by long-range (Sampaio et al., 1995) and short-range (Tiessen and Santos, 1989) soil variability. Studies of the influence of land use on soil organic matter (SOM) are rare in this region, where there is no tradition for running the type of replicated long-term trials that are typically used to assess these effects.

Due to increasing population pressure, crops are planted before soil fertility has recovered through long bush fallows (Tiessen et al., 1992), resulting in further depletion of soil organic nutrient pools (Salcedo et al., 1997). Productivity under subsistence agriculture relies strongly on organically held nutrients (Tiessen et al., 2002), because nutrient inputs from external sources are normally absent. Six years of cultivation of a Paleustox reduced soil organic C, N, and P concentrations 30%, resulting in loss of production and land abandonment (Tiessen et al., 1992). Due to the evenness of the site, those authors attributed these SOM losses to mineralization processes (Tiessen et al., 1992). However, in regions with uneven relief (37% of the semi-arid has slopes between 4 to 12% and 20% greater than 12%, Brazil, 1983), it is reasonable to expect that organic matter will also be lost by erosion, particularly in the upper- and mid-slope positions (Silva et al., 1986a; Albuquerque et al., 2001) that farmers normally occupy with subsistence agriculture.

The effect of cultivation has been frequently assessed by comparing cultivated areas with contiguous undisturbed areas (Detwiller, 1986; Davidson and Ackerman, 1993). However, in most cases, inferences about significant differences and processes involved are limited by the absence of treatment replication and randomization. Furthermore, this methodology presumes that areas with undisturbed native vegetation (Tiessen et al., 1992) have not undergone erosion (Ellert and Gregorich, 1996). Both assumptions are quite restrictive for the semi-arid region of northeastern Brazil because the mosaic created by subsistence agriculture rarely contains areas with truly undisturbed caatinga vegetation in a pristine condition. Thus, although it is possible to find adjacent areas of cultivated and uncultivated land, uncultivated areas do not necessarily represent a condition of undisturbed vegetation or of absence of erosion. This means that differences in organic matter content in paired comparisons would be also largely influenced by the specific condition of each uncultivated area, limiting the extrapolation of results.

To overcome the limitations mentioned above, treatment replications were obtained from 10 independent paired situations. Adjacent dry forest and cultivated areas were separated into two groups each, yielding four groups with increasing disturbance intensity. Group assignment was based on land use history and in situ observations and further verified with results from 137Cs analysis. Soil classification or slope was not considered as criteria for group assignment. While a number of studies on the effects of land use practices have been completed, few were replicated at the regional level and none integrated 137Cs and nutrient measurements to quantify the processes of nutrient depletion. We postulated that the gradient of increasing vegetation-soil degradation would show corresponding declines in the organic nutrient pools as a result of erosion and mineralization processes and that these declines would be greater than the within group variability caused by different soils and/or slopes.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Description of Study Sites
We selected 10 study areas distributed across five districts of two states in northeastern Brazil, Pernambuco and Paraiba. Study areas where several kilometers apart from each other, with the exception of two areas that were <1 km apart (Table 1). Average annual rainfall in this region varies between 500 and 1000 mm. The highest precipitation level is usually considered as the boundary for the semiarid region (Sampaio, 1995). However, rainfall in the region is so erratic in space and time (Reddy, 1983) that annual averages are a poor climatic parameter (Sampaio, 1995). Average temperatures are 23 to 27°C, with <5°C monthly and 5 to 10°C daily variations. Seven to 11 mo per year have a negative water balance (Brazil, 1985). For an overview of soils and vegetation in this region see Sampaio (1995). The soils at the study sites included six Lithic Haplustalfs (L-Alf) (Lithic Cromic Luvisol), one Haplustalf (Alf) (Orthic Cromic Luvisol), two Lithic Usthortents (L-Ent) (Lithic Eutrophic Neosol), and one Haplustult (Ult) (Red-yellow Eutrophic Argisol) (Table 1); abbreviations for each soil were indicated between brackets as well as the corresponding names by the Brazilian Soil Classification System (Brazil, 1999). The sampling extended from March 1998 until June 1998, under a severe dry spell that lasted 2 yr.


View this table:
[in this window]
[in a new window]
  		Table 1. Locations of the study sites, soil types, management histories, and vegetation cover.

 
Management Histories and Grouping of Sites
The overriding criterion for site selection was the presence of two adjacent areas in the same slope position, one with dry forest or bush fallow vegetation, and the other with subsistence agriculture or pasture. At each site we set a sampling transect in the backslope that maintained a constant slope position within both areas. We assumed that soil samples taken along this transect would represent similar soil and landscape characteristics, so differences in organic nutrient pools between neighboring areas would be mainly due to differences in vegetation cover and land use histories (Ellert and Gregorich, 1996). Interviews with landowners provided land use information of the areas (Table 1). The site-selection criterion resulted in our sampling of a variety of soils, slopes, and management histories. In situ observations showed that it was possible to differentiate two vegetation and/or soil degradation levels within each land use group (cultivated–dry forest), which were consistent with management histories. This yielded four groups of increasing land degradation, with each group including a mixture of soil types and slope gradients. The first group, under dry forest, included the first six sites listed in Table 1, two with undisturbed caatinga and four with bush-fallows of 20 yr or longer. Only one of them (L-Ent-2), with a bush fallow of 35 yr, had signals (stumps) of very limited logging. This group was categorized as UDF. The second group (the last four sites in Table 1) included sites with DDF; the group included three selectively logged areas and one area with a short 6-yr bush fallow. The logging of the caatinga vegetation had been relatively intense and the soil surface showed signs of erosion.

All farmed areas had signs of soil erosion, except for the Haplustalf at Custódia-PE. However, erosion was less marked in the Lithic Haplustalf (L-Alf-6), with prior cultivation in contour lines, and in the Lithic Usthortents (L-Ent-1 and 2), which had pebbles and cobbles on the soil surface. Farmers indicated that this type of soil surface helped control soil erosion. Thus, these four areas were grouped under the PC category. The remaining cultivated areas showed signs of more intense erosion; the worst case was the Ult, cultivated to corn (Zea mays L.) with lines in the down-slope direction, which had lost most of the Ap horizon. Although the intensity of soil degradation was not the same at all six sites, as a group they represented the worst situation in terms of soil degradation and were included in the DC category.

Soil Sampling and Analysis
Soil samples at each site were taken approximately in the middle of the backslope. The sampling transect in each area started 20 to 30 m away from the boundary line separating the adjacent areas, and had five sampling points at regular intervals and same slope position. Sampling in the dry forest areas was not stratified by vegetation cover (under tree vs. inter-space). Available areas for sampling were not generally wider than 150 to 200 m, so sampling intervals varied between 20 to 30 m among sites, depending on size. A pit was opened at each site near the boundary line to classify the soil. Single soil samples from two fixed-layer depths (0 to 7.5 cm and 7.5 to 15 cm) were taken at each sampling point. Most soils were extremely dry and hard at the time of sampling and, with two exceptions, had pebbles at the surface and inside the profile. For this reason we used a metal chisel 30 cm long, 6 cm wide, and 0.8 cm thick with a sharpened end, to cut a soil block 25 cm long, 12 cm wide, and 7.5 cm thick from each layer. This resulted in a total of 200 single samples (10 sites x two areas per site x two depths x five sampling points) that were processed and analyzed separately. Results from the first sample in each area (nearest to the boundary line) were disregarded afterwards, because at some sites they had intermediate organic matter contents with respect to the remaining samples in each area, suggesting a ‘boundary effect’. Thus, the total number of samples was reduced to 160.

After breaking down the soil blocks and separating roots and large organic debris, the total weight of each sample was recorded. Subsamples of soil were collected and oven dried at 105°C for 24 h to determine gravimetric moisture content; the remaining soil was air dried and ground manually with mortar and pestle, to pass through a 2-mm sieve. The weight of pebbles and of <2 mm soil were also recorded. Due to the difficulty of keeping the sampling volume constant at every sampling point, the volume of each sample had to be individually estimated. The volume of pebbles was calculated from its mass and a particle density of 2.65 Mg m–3. The volume of <2 mm soil was calculated from its mass and from its bulk density, determined by adding three consecutive <2-mm soil subsamples ({approx}35 mL) into a 100-mL graduated cylinder, tapping the cylinder 10 times on a 5-mm rubber foam from a height of approximately 10 cm after each addition (Brazil, 1997), and then recording the volume and mass of dispersed soil. Finally, the bulk density of the soil (without pebbles) was calculated dividing the mass of <2-mm soil by the total sample volume (volume of pebbles plus <2-mm soil). This methodology for bulk density determination was kept unchanged for the few samples without pebbles, to keep results comparable within the sample set. Bulk density results were only used for the calculation of 137Cs stocks, from which estimates of soil erosion were derived. Thus, 137Cs stocks and erosion results may have been underestimated (5–10%) because bulk density of dispersed soil is normally smaller than that of undisturbed samples. Particle size of the fine earth fraction was determined by the hydrometer method (Gee and Bauder, 1986). Only the mass of pebbles was expressed on a whole-soil basis.

Fine earth samples were analyzed for extractable P with Mehlich-1 (0.05M HCl + 0.0125M H2SO4) and extractable N with 1M KCl. Phosphate was determined colorimetrically (Murphy and Riley, 1962), and NO3 and NH4 with autoanalyzer (USEPA, 1971). Exchangeable cations (Ca, Mg, K and Na) were extracted with 1M NH4Cl and determined with atomic absorption spectrophotomer. Subsamples ground to pass a 0.015-mm sieve were analyzed for total organic C by wet oxidation-diffusion (Snyder and Trofymow, 1984); total N by Kjeldahl (Bremner and Mulvaney, 1982) and organic N by subtraction of mineral N from total N; total Po by ignition and acid extraction (Walker and Adams, 1958); and total P (Pt) by digestion with a mixture of sulfuric acid and hydrogen peroxide (Thomas et al., 1967).

Cesium-137 Analysis and Estimation of Soil Erosion
Cesium-137 activity was determined by high efficiency {gamma} spectrometry using a multichannel analyzer with a lithium-drifted germanium detector and a counting time of 80 ks (Bajracharya et al., 1998). Counting efficiency was determined mixing an aqueous solution with known 137Cs activity to a previously counted sample, which was recounted following air drying and thorough homogenization. The data were expressed as becquerels per kilogram (Bq kg–1) and were not corrected for radioactive decay, since the samples were taken during a short time interval compared with the half-life of 137Cs.

Because of the homogeneous distribution of 137Cs fallout at the regional level and high retention by clays, changes in the soil concentration of 137Cs among sites in a landscape have been interpreted as movement of soil material (e.g., review by Ritchie and Mc Henry, 1990). Several empirical relationships and models of varying complexity relating 137Cs loss from the soil to the rate of soil erosion have been used (Ritchie et al., 1971, Kachanoski and de Jong, 1984; Yang et al., 1998). About half of the total 137Cs fallout until 1982 occurred in 1963 (Ritchie and McHenry, 1990). Estimates of soil erosion based on the 137Cs data were calculated assuming a linear loss rate in the interval 1963 through 1998 (35 yr) and a constant 137Cs distribution within the 0- to 7.5-cm layer. Soil bulk densities of the 0- to 7.5-cm layer were used to convert 137Cs concentrations into stocks using the following equation:

[1]
where 137Cs stock (Bq m2) is the total element content in the 0- to 7.5-cm layer, {rho} is the dry soil bulk density (Mg m–3), z (m) is the soil layer thickness, 137Cs activity (Bq kg–1) is the element concentration and 1000 is a unit conversion factor. The mass of soil loss (Mg ha–1) was calculated with the expression:

[2]
where (i) represents the land use group of interest.

Statistical Analysis
Areas within each land use were divided in two groups of different land degradation based on qualitative information (land use histories and in situ observations). Mean 137Cs activities (n = 4) of areas belonging to each combination of land use degradation were averaged and confidence intervals (90%) for the group means calculated using the Student-t distribution for small samples. These confidence intervals did not overlap within each land use, which was interpreted as additional evidence of the existence of two-land degradation levels in both, dry forest and farmed areas. Differences in organic nutrient pools among the four land use degradation groups were assessed by analysis of variance, using the GLM procedure for unbalanced design with respect to number of replicates, and the Tukey's HSD test at P < 0.10 for the comparison of means (SAS Institute, 1985). Because land use groups included soils that were quite different in some of their properties, comparisons of other selected physical and chemical properties between dry forest and cultivated areas were done using the t-test for paired samples (P < 0.10) (Sokhal and Rolf, 1995). These properties probably also affected organic pools, meaning that our analysis are more conservative than the P < 0.10 cutoff implies.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Cesium-137 Activity and Estimates of Soil Losses
Cesium-137 activities were grouped by land use-degradation categories, but means were also shown for each site because we could not find published regional data about this nuclide (Table 2). Activities in samples from the 0- to 7.5-cm layer ranged from 0.499 to 1.37 Bq kg–1 of soil and were between 20 and 50% of the activities found in soils of southeastern Brazil (Schuch et al., 1994; Andrello et al., 1997). In comparison with published results for the northern hemisphere, results in Table 2 were one order of magnitude lower, for example, 30 Bq kg–1 in the 0- to 13-cm layer of a reference area in Ohio (Bajracharya et al., 1998). This is because 137Cs fallout was smaller in the southern than in the northern hemisphere (Ritchie and McHenry, 1990). We did not find measurable activities in samples from the 7.5 to 15 cm. This likely resulted from the steep negative distribution gradient of 37Cs with soil depth (Wallbrink et al., 1999) combined with the smaller fallout mentioned before. In addition, shallow manual tillage ({approx}10 cm) and climate and landscape conditions that do not favor water infiltration probably contributed also to this absence of measurable activity. In relation to the effect of relief on erosion, we expected increasing differences in 137Cs activities between adjacent areas as slopes became steeper, which did not occur (Table 2). Soil management and degree of surface coverage (Table 1) seemed to be more relevant than slope gradient in determining erosion intensity. Corn planted down slope for several years in Ult (Table 1), with 9% slope, resulted in an erosion comparable with L-Alf-4 and L-Alf-5, which had steeper slopes and were also cultivated, but not down slope. The soil at L-Alf-6, with 40% slope but planted in contour lines (Table 1), was less eroded (higher 137Cs activity) than the first three soils in Table 2, that had smaller slope gradients. Soils at L-Ent-1 and L-Ent-2, with slopes of 32 and 44%, respectively, had pebbles and cobbles on the soil surface (Table 1) and, although cultivated for several years, were less eroded than sites with less steep slopes (Table 2).


View this table:
[in this window]
[in a new window]
  		Table 2. Means (±SE) of 137Cs activity in cultivated areas adjacent with areas under dry forest, divided in two groups according to land use history and in situ observations (0- to 7.5-cm soil layer, n = 4).{dagger}

 
The mean 137Cs concentration under UDF was the highest and it was used as a baseline activity for a condition of unknown, but probably limited erosion (Table 3). The choice of using a group mean as baseline 137Cs activity, as opposed to a single site mean (Table 2), seemed a conservative approach, because the former incorporated some degree of regional variability (Table 3). The stock of 118 Bq m–2 calculated for UDF (Table 3) for an approximate latitude of 8° S (Table 1), coincided with the pattern of decreasing 137Cs stocks with decreasing latitude within Brazil (Schuch et al., 1994). These authors reported 137Cs stocks of 329, 159, 150, and 107 Bq m–2 for latitudes of 28, 26, 22, and 13° S, respectively.


View this table:
[in this window]
[in a new window]
  		Table 3. Mean concentrations (±90% confidence intervals) and stocks of 137Cs as a function of land use intensity (0- to 7.5-cm soil layer) and soil losses calculated in relation to undisturbed dry forest.

 
Cesium-137 concentration under DDF was approximately 30% lower than under UDF (Table 3). Farmers usually do not consider selective logging as detrimental to the ecosystem, but this difference suggests it can contribute to soil erosion, probably due to the increase of patches with uncovered soil surface. However, the periods of bush-fallows were relatively short (Table 1) in comparison with the time of peak 137Cs fallout, which occurred in 1963 (Ritchie and McHenry, 1990). Thus, part of the 137Cs loss may have occurred during a previous cropping phase, before the present bush-fallow was initiated.

The extent of 137Cs loss from the PC group was similar in intensity to the DDF group but was much smaller than in DC (Table 3). The smaller 137Cs loss from this subgroup of cultivated sites substantiates the farmer's idea that pebbles and cobbles help prevent or reduce soil erosion. Planting in contour lines and avoiding overgrazing of pastures, also represented in PC, are well-known practices to reduce soil erosion.

The highest 137Cs losses were observed in areas with overgrazed grasslands or cultivated down-slope to annual crops, grouped under DC (Tables 1 and 3). Soil losses in this region are caused by frequent rainstorm events (annual occurrence of short duration rains with 80 mm h–1, and a 10% probability of 100 mm of precipitation in 24 h) (Tiessen and Santos, 1989) that occur during the rainy season. Unprotected soil surface due to excessive logging, slash-burn, overgrazing, or recent land abandonment to bush fallow, in addition to the uneven relief, aggravates the degree of soil erosion, (Margolis et al., 1985; Silva et al., 1986b, 1989; Melo Filho and Silva, 1993; Albuquerque et al., 2001).

Annual rates of soil erosion are calculated using empirical equations or more elaborate models, comparing 137Cs activities in areas of uncultivated or reference soils with areas under a given land use (e.g., pastures or crops) (Ritchie et al., 1971; Kachanoski and de Jong, 1984; Yang et al., 1998). During the 35-yr interval considered in the present work, most sites had frequent shifts in land use (Table 1) that certainly modified the rates of erosion. Moreover, the amount and erosivity of rainfalls are highly variable in time: an 8-yr monitoring in plots 25 m long by 5 m wide, with 6% slope and bare soil, recorded an accumulated sediment loss of 438 Mg ha–1. Almost 40% of this amount was lost in a single year and additional 50% in three different years (about a third in each) (Albuquerque et al., 2001). Due to this large variability in annual losses, estimates at DDF, PC, and DC were reported as integrated losses for the 35-yr period (Table 3).

The greatest soil loss occurred under DC and reached an estimated mass of 500 Mg ha–1, which is equivalent to the redistribution of 3.5 cm of soil (Table 3). As a comparison, Albuquerque et al. (2001) quantified an accumulated 8-yr loss of 227 Mg ha–1, in this case using large plots (0.5–1 ha) cleared of caatinga vegetation; thus, it seems reasonable to expect that in 35 yr (time period for results in Table 3) losses would be at least twice as large.

Land Use Effects on Organic Nutrient Pools
Farming produced significant losses in C concentrations (Table 4). In comparison with UDF (0–7.5 cm layer) declines were close to 28% in DDF and PC, but reached 50% under DC that was the most degraded situation. Losses of about 30% of C after cultivation of native vegetation areas have been often reported (Detwiller, 1986; Tiessen et al., 1992; Davidson and Ackerman, 1993) but losses of half of the initial organic C are more unlikely (Woods and Schuman, 1988; Hajabbasi et al., 1997). Because a relatively shallow sampling depth of 0 to 7.5 cm would tend to enhance changes in C concentrations, we also reported these changes for the 0- to15-cm layer (Table 4) by pooling C concentrations and bulk densities of individual samples from the 0- to 7.5- and the 7.5- to 15-cm layers (n = 160). Carbon concentrations became smaller overall when the soil layer thickness increased to 15 cm; soils under dry forest were relatively more affected because C decreased more with depth in those sites than in the cultivated areas. In spite of this, the effect of land use degradation remained significant in the 0- to 15-cm layer (P < 0.10), with declines of 21% for DDF and PC, and 43% for DC, in comparison with UDF (Table 4).


View this table:
[in this window]
[in a new window]
  		Table 4. Effects of land use intensity and sampling depth on total organic C and N concentrations (±SE) and C/N ratios.

 
Effects of land use degradation on total organic N followed the same trends as C (Table 4), except that N concentration between PC and DC did not differ significantly (P < 0.10). Carbon/N ratios decreased significantly (P < 0.10) when undisturbed dry forestland was degraded by farming (Table 4); this was probably due to the combined effects of change in the quality of litter inputs (Anderson et al., 1989) and modifications in the degree of SOM protection by soil aggregates (Oades, 1989; Cambardella and Elliott, 1994). Although specific data for caatinga are not available, litterfall from deciduous forest is richer in lignin and polyphenols than substrates from grasses and crop residues (Loranger et al., 2002; Kögel-Knabner, 2002). Inputs of organic residues that are relatively more resistant to decomposition and the absence of cultivation are consistent with the wider C/N ratio in UDF. Conversely, easily decomposable substrates combined with cultivation-erosion processes, which expose aggregate-protected organic matter, would enhance the action of microorganisms and yield a narrower C/N ratio under DC.

Concentrations of Mehlich-1 extractable P were similar and in the deficiency range in all land use types, thus illustrating the absence of fertilizer usage in farmed areas (Table 5). The effect of land use type on the Po pool was reported as a ratio (Table 5) to remove the effect of varying Pt concentrations on Po. Organic P was 1.5 times greater than Pi in the 0- to 7.5-cm layer of UDF, became equal to Pi in DDF and decreased even more, although not significantly, in the farmed areas. Modifications in the relative amounts of Po and Pi as a result of cultivation have been observed in soils from temperate (Stewart and Tiessen, 1987) and tropical (Tiessen et al., 1992) regions. Tiessen et al. (1992) also determined that following a long bush-fallow, the ratio returned to a value that was similar with the uncultivated site.


View this table:
[in this window]
[in a new window]
  		Table 5. Effects of land use intensity and sampling depth on Mehlich-1 extractable P and Po/(Pt – Po) ratio (±SE).{dagger}

 
Modifications in the Po/Pi ratio among land use groups were no longer significant (P < 0.10) when the 0- to 15-cm layer was considered (Table 5). This occurred because, under UDF, Po was more stratified in depth than Pt (the average Po/Pt ratio was 0.52 in the 0- to 7.5-cm layer and 0.44 in the 7.5- to 15-cm layer) and consequently mixing with the 7.5- to 15-cm layer decreased the Po/Pi ratio from 1.47 to 1.11. Thus, as was found for C and N, a thinner layer of superficial soil was more adequate to study inorganic and organic P dynamics in these soils.

Differences in soil types within groups did not mask the effect of land use intensity on organic nutrient pools, but were sufficient to limit comparisons of other soil physical and chemical properties. Changes in some of these properties became more evident when the soils were grouped by soil classification criteria and land uses were compared using t-tests for paired samples (Table 6). Although data was shown for all soil classes, the analysis was limited to the Entisols and Lithic Alfisols to avoid comparisons based on pseudo-replicates (single site comparisons) (Fraga, 2002).


View this table:
[in this window]
[in a new window]
  		Table 6. Selected physical and chemical properties of the samples, grouped according to soil classification and sampling depth.

 
All the soils showed increases in bulk densities due to farming (Table 6). These increases were mostly limited to the 0- to 7.5-cm layer, since bulk densities did not increase any further when the 0- to 15-cm layer was considered. Erosion induced an increase in the average sand content of the Lithic Alfisols as a group, but this increase was particularly large only in L-Alf-3 and L-Alf-5 (not shown). The increase in sand content explains the corresponding decrease in effective cation-exchange capacity (ECEC). Soil organic matter losses (more than 50% in L-Alf-3 and L-Alf-5, not shown) probably also contributed to the decrease in ECEC. The influence of SOM in the ECEC is also illustrated by results of the Ultisol: the sand content in the strongly eroded farmed area was smaller because samples included part of the B horizon, but despite the increase in the fine fraction the ECEC still declined.

Estimates of Carbon Losses through Erosion and Mineralization
Soil losses reported in Table 3 were used to estimate how much C was redistributed in the landscape through erosion (Gregorich et al., 1998) or was effectively lost as CO2–C by mineralization (Ritchie and Rasmussen, 2000). Using the comparison between DC and UDF as an example, we first calculated how much C would be lost with the loss of 3.5 cm of topsoil from UDF sites (Table 3). A sample taken from the 0- to 7.5-cm layer after this topsoil loss, would include 4 cm of the original layer plus 3.5 cm of the layer underneath and its C concentration would be a weighted contribution of both thicknesses and their corresponding C concentrations (assuming constant C concentration in each layer). Carbon concentrations in the 7.5- to 15-cm layer, although not explicitly shown in Table 4, can be calculated from results in this table (e.g., for UDF it is 9.6 mg C kg–1 soil), since changes in bulk density between soil layers were very small (Table 6). The resulting C concentration for the above example was estimated as [17.8 (4/7.5) + 9.6 (3.5/7.5)] = 13.97 g C kg–1 soil. This result is 3.83 g C kg–1 soil lower than under UDF and represents the decline associated with the topsoil loss. The difference between the estimated 13.97 g C kg–1 soil and the experimentally determined 8.90 g C kg–1 soil under DC, represents an additional decrease of 5.07 g C kg–1 soil, associated with organic matter mineralization. Thus, 43% of the total C decrease was due to soil redistribution and 57% was to mineralization processes.

The limitation of these estimates lies in the assumption that C concentrations in soils under UDF and DC were similar before the land was cultivated. Since land use groups combined different soil types, this assumption would probably not be valid. To overcome this limitation, we repeated the above calculations for UDF sites (first six sites in Table 1) in single pair comparisons with three degraded (L-Alf-4, L-Alf-5, and Ult) and three preserved farmlands (L-Alf-6, L-Ent-2, and Alf). Due to the proximity of the paired areas, it seemed reasonable to assume similar C concentrations before the land was cultivated. Differences in 137Cs concentrations (converted to 137Cs stocks) between paired areas at each site (Table 2) were used to calculate erosional topsoil losses. Based on these losses we estimated how much of the differences in soil C between paired areas could be attributed to erosion or to SOM mineralization, as exemplified above. Soil C losses were more intense in DC (40–57%) than in PC areas (<30%), in paired comparisons with UDF areas (Fig. 1) . In two of the DC areas (L-Alf-4 and L-Alf-5, Fig. 1) erosion was relatively more important than SOM mineralization in explaining C losses. Conversely, SOM mineralization was relatively more important than erosion in two areas of the PC-UDF comparison (L-Alf-6 and L-Ent-2, Fig. 1).



View larger version (32K):
[in this window]
[in a new window]
  		Fig. 1. Relative contribution of mineralization and erosion processes to soil C losses, when comparing areas of undisturbed dry forest (UDF) adjacent with areas of degraded-cultivated (DC), or preserved-cultivated (PC) farmlands.

 
Interrelationships between Soil Carbon and Phosphorus
The mean C/Po ratio of all the samples was 125:1 and was quite variable (n = 160, coefficient of variation [CV] = 47%). This variability prevented the analysis of trends among land use categories, as was done for the C/N ratio (Table 4). Larger variability in C/Po than in C/N ratios (CV = 20%, not shown) has been reported (Duxbury et al., 1989; Stevenson, 1994).

Carbon, Po, and Pt were significantly related when only samples under dry forest were considered and, based on the magnitude of R2, nonlinear functions fitted the data better than linear ones in both layers (Fig. 2 and 3) . Seven samples had Pt concentration of less than 0.3 g P kg–1 soil (Fig. 3), which agrees with reviews that report a generalized low P status in semiarid soils from northeastern Brazil (Sampaio et al., 1995), although exceptions have also been observed (Agbenin and Tiessen, 1994). The highly variable P concentrations are illustrated by the L-Alf soils, three of which are in the upper end and three in the lower end of the range of Po and Pt concentrations (Fig. 2 and 3).



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 2. Relationship between organic P and C concentrations in areas with dry forest vegetation (0- to 7.5- and 7.5- to 15-cm soil layers). Values are means for each site (n = 4) and bars are standard errors. (L-Alf = Lithic Haplustalf, L-Ent = Lithic Ustorthent, Alf = Haplustalf, Ult = Haplustult).

 


View larger version (23K):
[in this window]
[in a new window]
  		Fig. 3. Relationship between C and total P contents in areas with dry forest vegetation (0- to 7.5- and 7.5- to 15-cm soil layers). Values are means for each site (n = 4) and bars are standard errors. (L-Alf = Lithic Haplustalf, L-Ent = Lithic Ustorthent, Alf = Haplustalf, Ult = Haplustult).

 
The curve for the 0- to 7.5-cm layer in Fig. 2 indicates that large C concentrations (22 g C kg–1 soil) correspond to relatively narrow C/Po ratios (96:1) while at low C concentrations (10 g C kg–1 soil) the C/Po ratio widens (167:1). Thus, SOM becomes relatively more impoverished in Po as the soil C content decreases. This relative impoverishment suggests that Pt in the lower concentration range, by limiting available P, is controlling Po and C accumulation (Fig. 3), as has been noticed previously (Tate and Salcedo, 1988). As Pt increased C increases tended to level off (Fig. 3), which is in contrast with the linear relationship found by other authors (Walker and Adams, 1958; Brossard and Laurent, 1992). The range of Pt concentrations in Brossard and Laurent (1992) was very similar to that shown in Fig. 3, but the range of total C concentrations was almost twice as large. Since data in Brossard and Laurent (1992) were from the humid tropics, it can be hypothesized that increases in C tended to level off as Pt became less limiting because another factor, probably water, began controlling C accumulation.

Relationships summarized in Fig. 2 and 3 became evident only when data from cultivated areas were excluded from the graphs, because the Po/C and C/Pt ratios in cultivated areas varied notably. Greater soil erosion likely modified C concentrations to a greater extent than the concentrations of Pt or Po because those variables were less stratified with depth (Fig. 2 and 3). The fact that a single non-linear function described Po and Pt concentrations in samples from both depths and land uses seems to support this hypothesis (Fig. 4) . Soil C stratification was also likely responsible for depth based differences observed in data summarized in Fig. 2 and 3.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 4. Relationship between organic P (Po) and total P (Pt) contents in cultivated and dry forest areas (0- to 7.5- and 7.5- to 15-cm soil layers). Values are means for each area (n = 4).

 
The non-linear trend between Po and Pt (Fig. 4) is in contrast with the linear relationship found by Walker and Adams (1958). The type of curvature in Fig. 4 is the opposite of what we anticipated, as we thought the relationship between Po and Pt (Fig. 4) would follow that of C and Pt (Fig. 3). Since Po is part of the organic matter pool, we expected the mechanisms of control exerted by Pt to be similar for Po and C. Although no experimental evidence is available, one possible explanation for the trend in Fig. 4 is that as Pt becomes less limiting, caatinga litterfall becomes relatively more enriched in P. At the same time, since vegetation growth remains restricted by low water availability, litterfall production remains small and limits C accumulation in soil. This variation in litter composition due to soil nutrient availability has been reported for a dystrophic rainforest (Tiessen et al., 1994).


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Areas under dry forest or subsistence farming (ten of each) were each divided into two different levels of soil and/or vegetation disturbance to yield four groups. Group assignment was initially based on in situ observations and management histories, and was later corroborated with 137Cs activity measurements to assess soil erosion. Grouping allowed a general overview of the intensity of land degradation caused by subsistence farming. Losses of C and N, and enhanced mineralization of organic P were observed under both main land uses. Logging in areas under dry forest was detrimental to SOM conservation and equivalent in its effect to well-managed pastures, cultivation in contour lines, or cultivation of soils with surface coverage of pebbles and cobbles. The different intensities in SOM losses were not related to slope. Erosion processes were relatively more important than SOM mineralization in driving losses of soil C and N from DC farmlands, which contained approximately 50% less C than adjacent UDF areas. The opposite occurred in PC farmlands, where SOM mineralization mostly drove losses that resulted in 20% less soil C and N than in adjacent UDF areas. The largest losses in SOM seem incompatible with a system that relies solely in bush fallow to recover the fertility of degraded areas. The combination of severe degradation, low P status soils and water limitations should impose severe restrictions to the rate of recovery of degraded farmlands or could even impair its recovery.


   ACKNOWLEDGMENTS
 
Financial support from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq-Brazil) and the InterAmerican Institute for Global Change Research (IAI-USA) is sincerely acknowledged. The authors also thank Dr. Ivandro de F. Silva for help in locating sites, Dr. Romilton Amaral for help with 137Cs analysis, and technicians at the Radioagronomy laboratory (DEN/UFPE) for their technical support. Comments provided by Dr. Michelle Wander, Dr. Holm Tiessen, and three anonymous reviewers helped to improve earlier versions of this manuscript.

Received for publication October 28, 2002.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by da Silva Fraga, V.
Right arrow Articles by Salcedo, I. H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by da Silva Fraga, V.
Right arrow Articles by Salcedo, I. H.
Agricola
Right arrow Articles by da Silva Fraga, V.
Right arrow Articles by Salcedo, I. H.
Related Collections
Right arrow Soil Erosion
Right arrow Isotopes
Right arrow Soil Organic Matter

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