SPATIAL VARIABILITY OF SOIL PHOSPHORUS OF A LOW PRODUCTIVITY Brachiaria brizantha PASTURE

Studies on soil phosphorus (P) of low productivity cultivated pastures in Cerrado (Brazilian Savanna) areas and surveys on other possible problems related to P are scarce. The spatial variability of soil phosphorus content of a Rhodic Ferralsol was studied in a low productivity pasture of Brachiaria brizantha (BB) grown for 10 years, without fertilizer application, in an experimental area at Planaltina (GO), Brazil. Soil samplings were performed on a regular grid of 10 by 10 meters, with 98 sampling points before (between tussocks and under tussocks) and after the establishment of the experiment (after fertilizing). On the same grid, forage plants were collected and separated into fractions for N and P content analyses. Soil available phosphate was determined by the resin method (Pr) and complemented by the P isotopic exchange kinetics analysis. Descriptive statistical and geostatistical analyses were utilized to describe the spatial variability. The Pr content on soil samples under tussocks presented mean and median values 45% larger than in soil samples taken between tussocks. The higher variation is probably related to the greater concentration of BB roots, soil organic matter content and soil P recycled through the plants tussocks. The spatial variability of Pr in this soil was high especially after fertilizer application. This variable did not present spatial dependence for the regular 10 m sampling. The generated knowledge on P variability of soils under low productivity cultivated pastures revealed problems related to the sampling methodology traditionally utilized and to P application.


INTRODUCTION
In the Cerrado Region (Brazil) cultivated pastures (i.e.Brachiaria spp.) represent 49.5 million ha (Sano et al., 2000) and approximately 80% of which present low productivity (Kluthcouski et al., 1999).Inadequate pasture management, overgrazing, and the lack of maintenance fertilization are pointed as causes for decline yield.
Scientia Agricola, v.60, n.3, p.559-564, Jul./Sept.2003 The reduction in areas of natural vegetation destined to opening and establishing pastures and, conversely, the growing pressure for the conservation of non-renewable natural resources and the development of technologies for better exploration of areas previously grown as pastures, have contributed to an increased interest on cultivated pastures.Priority should be given to guidelines for management that would allow the reclamation of areas in early stages of decline or already degraded, instead of simply encouraging expansion of the agro-silvo-pastoral front according to the occupation models that have occurred so far (Lopes, 1997).
The reclamation of cultivated pastures presumes a knowledge on soil fertility fluctuations through soil analyses.The greatest interest usually refers to P content, since it is one of the most limiting nutrients for the production of forage plants.In addition, P inputs have high costs in Brazil.Even though it is generally thought that the P variability is low and that it is not necessary to gather information on it to begin a pasture reclamation strategy.The variability of P contents in pastures depends on the management.Replacing natural vegetation with the forage plant, soil tillage, liming and fertilizing with P at pasture establishment are some management practices that influence soil phosphorus variability.In addition, the irregular return of manure and urine (spatial and temporal) to the pasture (Petersen et al., 1956a, b), combined with reduced soil tillage and cultivation as compared to soils that are annually tilled, create a potential for considerable development of soil heterogeneity with respect to nutrient concentration (Fisher et al., 1998).
The heterogeneous supply of nutrients in the soil results in plant growth at various scales.Within the space of a defined root system, the proliferation of roots and the increase in the rate of P absorption occurs as a response to sites that have been enriched by the addition of P (Jackson et al., 1990).At field scale, soil heterogeneity contributes to species coexistence in pastures, through a variety of different niches where plant species compete individually with greater success (Reynolds et al., 1997).Also at field scale, soil heterogeneity per se is considered to favor plant communities locally when compared with isolated plants (Birch & Hutchings, 1994).In a superior scale of individual root systems, the exchangeable P in the soil is related to the distribution of many plants (Tyler, 1996).
A knowledge of phosphorus spatial variability in soils under pasture is essential to establish adequate reclamation strategies.Some authors have reported extremely low phosphorus levels in soils under low productivity pastures (Oberson et al., 1999).However, those results refer to composite samples collected in the pastures, not to systematic surveys devised to check for phosphorus variability and other soil physical and chemical attributes.
A literature review on this subject reveals that field data on phosphorus in soils under degraded or low productivity pastures in the Cerrado Region are scarce.This work is, therefore, an attempt to evaluate the spatial variability of soil phosphorus in a Brachiaria brizantha cv.Marandu (BB) low productivity pasture, identifying the distribution of soil phosphorus contents and their possible relations with the Brachiaria brizantha.

MATERIAL AND METHODS
The present work was carried out on a 29.11 ha experimental area at Planaltina-GO, Brazil (15°27'10''S and 47°36'51''W), altitude 826 m.The Brachiaria brizantha cv.Marandu (BB) grass, introduced in 1990 after the native vegetation was cleared.The BB carrying capacity during the rainy season was lower than 1 animal unit (AU) ha -1 (AU = 450 kg live weight).The little live weight gain in the rainy season and the live weight loss in the dry season characterized the low yield capacity per animal and per hectare of this pasture.The area was selected for the installation of an experiment to study pasture reclamation under grazing.Dolomitic lime 3 t ha -1 have been applied at soil surface.The mean results of 10 soil samples collected from the 0-10 cm layer on a Rhodic Ferralsol were as follows: clay = 800 g kg -1 ; pH H 2 O = 5.9; P (Mehlich-1) = 0.3 mg L -1 ; organic matter = 28 g kg -1 ; cation exchange capacity = 99 mmol c kg -1 , and base saturation = 50%.
Planimetric and altimetric surveys and spatial location of trees, remainder pile and fences preceded the subdivision of the area into six plots of 4.9 ha each for the installation of the following treatments, in February 1999: i) control; ii) 40 kg ha -1 P as Gafsa reactive phosphate (GRP) (T2); iii) ditto as T2 with the introduction of Stylosanthes guianensis cv.Mineirão (SG) (T3); iv) 80 kg ha -1 P as GRP (T4); v) 40 kg ha -1 P as triple superphosphate (TSP) (T5); and vi) ditto as T5 with the introduction of SG (T6).The areas were subsequently divided into four paddocks for animal rotation (P1, P2, P3, P4 -Figure 1a).Phosphates were applied to the surface and incorporated to ± 10 cm with heavy harrow.
Single soil samples were collected with an auger from September to October 1998 before installation of the experiment.Collections were made on 54 points at depths 0-10, 10-20 and 20-30 cm, equidistant 40 m, in the five rows of the treatment divisors (Figure 1a), W-E direction, to evaluate P variability in the total area of the experiment.A probability drawing was performed on the selected treatments (T6) and paddock two (P2).This area was located 70 m in the S-N direction, and 140 m in the W-E direction and marked with stakes on a 10 by 10 m grid design, with 98 sampling points, a total area of 9,800 m 2 (Figure 1b).The ascending direction of the slope runs in the direction of the coordinate y from 0 to Scientia Agricola, v.60, n.3, p.559-564, Jul./Sept.2003 130 m, with ± 5% slope.In the marked row there is a remainder pile that was built by piling up vegetation stubble when the area was deforested and then burned.At the 98 points, single soil samples were collected from the surface layer (0-10 cm) between tussocks, and under tussocks.At the same location, a 0.5 m 2 area of BB was collected by cutting at ground level and separated botanically and into fractions of BB as green matter (GM), dry matter (DM) and litter (L) on a dry weight basis; P content was determined in samples consisting of each row of the y coordinate, except for remainder pile samples.In September 1999, after the installation of the experiment, soil samples were collected at the same previously sampled points, in addition to another 24 points sampled on a 2.5 x 2.5 m grid marked with small crosses (Figure 1b) to evaluate the effect of TSP application on soil P variability.In addition, 30 simple random soil samples under adjacent native vegetation (cerrado) were collected to constitute 3 composite samples.
Available phosphorus (P r ) was colorimetrically determined by the resin method (Raij et al., 1986).Total P contents in the plant were determined by colorimetry after wet digestion with perchloric acid and peroxide (Adler & Wilcox, 1985).
The isotopic exchange kinetics (IEK) analyses were performed in the soil solution system after the samples had reached equilibrium in a 1:10 soil-solution ratio (Salcedo et al., 1991;Fardeau et al., 1991;Frossard et al., 1994).The system consisted of 10 g soil, sifted through a 2 mm mesh sieve, and 99 mL deionized water in a 250 mL capacity plastic container.Samples were stirred for 16 hours and then placed over a magnetic stirrer, and 1 mL of a solution containing carrier-free 32 P radioisotope, with an activity of approximately 1 Megabecquerel mL -1 , was added at time zero (t=0).After t 1 = 1, t 2 = 3, t 3 =10 and t 4 = 15 min, an aliquot of ± 8 mL of the suspension was drawn with a plastic syringe, which was immediately connected to a pressure filter holder with a 0.05 µm membrane filter.The suspension was filtered in ± 10 seconds and a 0.25 µL aliquot was removed with a micro pipet; the volume was completed to 20 mL for later determination of the 32 P counts by Cerenkov effect in a liquid scintillation counter.At the end of the process, 60 mL syringes were utilized to obtain a larger amount of solution to determine the concentration of phosphate (Cp) present in the soil solution by colorimetry through the malachite green method (Rao et al., 1997).The concentration of phosphate ions in the filtered solution was determined with 3 x 2 cm membranetype anionic resin pieces (BDH15 1TD) as recommended by Salcedo et al. (1991).The obtained soil solution phosphate concentration values of a standard sample had a 1.23% coefficient of variation (CV; n=3).
Two statistical analyses procedures were utilized.The first consisted of a general descriptive analysis of the results and correlations between soil P and plant variables, performed with the Statistica for Windows Statsoft (1996) software package.The second corresponded to a spatial continuity study of the variables under consideration.This study was preceded by an exploratory analysis of data including a median refinement (Hamlett et al., 1986) to remove the large scale variation effect (trend) of variable P in the experimental area.Spatial dependence was studied by semivariograms obtained from the collected samples and analyzed with the Variowin 2.2 (Pannatier, 1996) software package.

RESULTS AND DISCUSSION
The P r contents of the treatment divider were between 1 and 9 mg kg -1 , minimum and maximum values, respectively (Figure 2).The widest amplitude was found for the 0-10 cm depth.P r contents decreased with depth, as expected; the decline, however, was not abrupt.The spatial dependence study revealed that spatial continuity did not exist for the 54 samples from the 40 m spaced divisors.
The construction of P r semivariograms of soil samples between tussocks and after fertilizing (10 m grid), indicated no space variation structure.Spatial independence was also verified when the same data were analyzed and samples located on the remainder pile line (Figure 1b); values on average higher than in other points,  were disregarded.However, for samples collected under tussocks, a decreasing tendency was observed for P values in the ascending direction of the slope (Figure 3a), which was also reflected in the semivariogram that corresponds to the original data (Figure 3b), a tendency of indefinite growth being observed, indicating that an intrinsic stationarity, a model assumption in any geostatistical study, was not met.
The semivariogram of refinement residues (Figure 3b) for P r under tussocks indicates that this variable does not present a spatial dependence structure (pure nugget effect).The growth tendency of P r contents in the descend-ing direction of the slope verified in the original data was efficiently removed through median refinement.Overall, no spatial dependence exists for the sampled distance (10 x 10 m systematic grid) of variables P r in the soil between tussocks, under tussocks and after fertilizing.As performed, sampling was not sufficient to detect spatial dependence.
Distinct situations were observed in the comparison of P r between tussocks and under tussocks (Figure 4): Samples between tussocks had smaller means and medians than those under tussocks, approximately 45% lower.Even though the minimum values are close, the maximum was approximately fourfold greater under tussocks (Figure 4), demonstrating greater variation of the data (amplitude, variance and standard deviation), at least twice as high as the values for samples between tussocks.This greater variation in P contents CV between tussocks = 29% and under tussocks = 62%) is probably related to the greater concentration of the root system and to P cycling via litter in these samples.The observation of soil profiles in the experimental area had a poorer distribution of roots between tussocks, even when soil conditions under tussocks presented physical properties that would limit root system development.With regard to litter, it is worth noting that at the time of sampling the soil was bare between tussocks and the litter was under the tussocks.
The construction of a correlation matrix to evaluate the relationship between soil P r and plant P did not indicate correlation of P r between tussocks and P in the GM, DM and L (Figure 5a).However, linear correlations were found between P r under tussocks and plant P (Figure 5b) (P = 0.0006).These results confirmed a differentiated exploration of the soil between tussocks and under the tussocks, probably connected to the heterogeneous distribution of the BB root system as observed in the soil profiles.The proliferation of roots and the increase in forage-absorbed P occurs as a response to microlocations rich in P supply (Jackson et al., 1990).In this work, the nine-year-old plant-tussock system contributed to increas-  1 sd = standard deviation, results between parentheses in rows; 2 na = not analyzed.
Table 1 -Soil buffer power (R/r 1 ), concentration of phosphate ions in soil solution (Cp), linear slope (n), total P (Pt), and organic P (Po) of compound samples per treatment.
pling results of the 2.5 m grid (Figure 7) are lower than the median (15.8 mg kg -1 ) and the mean (19.7 mg kg -1 ) as compared to the sampling of the 10 m grid.This reflects a localized condition that might have been influenced by a non-uniform application of P in the experimental area, which also favored greater sampling variability in the entire area represented by the 10 m grid.This also reflects the non-stationarity of the mean because of the previously verified tendency, in which a decrease in P r values occurs under tussocks in the ascending direction of the slope. A

Figure 1
Figure 1 -a) Schematics of the experimental area with treatments and corresponding paddock subdivisions.b) Spatial distribution of the 98 sampling points in the experiment area.Selected at random in T6 P2.

Figure 2 -Figure 3 -
Figure 2 -Box-plots of P contents evaluated by the resin method in the treatments divisions.