RANDOM AND SYSTEMATIC SPATIAL VARIABILITY OF 137Cs INVENTORIES AT REFERENCE SITES IN SOUTH-CENTRAL BRAZIL

ABSTRACT: The precision of the Cs fallout redistribution technique for the evaluation of soil erosion rates is strongly dependent on the quality of an average inventory taken at a representative reference site. The knowledge of the sources and of the degree of variation of the Cs fallout spatial distribution plays an important role on its use. Four reference sites were selected in the South-Central region of Brazil which were characterized in terms of soil chemical, physical and mineralogical aspects as well as the spatial variability of Cs inventories. Some important differences in the patterns of Cs depth distribution in the soil profiles of the different sites were found. They are probably associated to chemical, physical, mineralogical and biological differences of the soils but many questions still remain open for future investigation, mainly those regarding the adsorption and dynamics of the Cs ions in soil profiles under tropical conditions. The random spatial variability (inside each reference site) was higher than the systematic spatial variability (between reference sites) but their causes were not clearly identified as possible consequences of chemical, physical, mineralogical variability, and/or precipitation.


INTRODUCTION
Atomic-bomb derived 137 Cs fallout has been widely used for the evaluation of soil erosion rates and patterns of soil redistribution within the landscape (Ritchie & McHenry, 1990;Walling, 1995;Basher et al., 1995).A basic premise for the use of the 137 Cs technique in soil erosion evaluation is that the 137 Cs inventories at the investigated points in the study site can be directly compared with the reference inventory (taken at a refer-ence site) and that the residual values indicate the relative magnitude of erosion or deposition (Walling & Quine, 1990).
The 137 Cs inventory at the reference site (local, flat, uneroded, undisturbed, stable site) represents the accumulated atmospheric input per unit surface area, adjusted for radioactive decay.Therefore, the value attributed to the reference inventory plays a crucial role in converting 137 Cs measurements to estimates of rates of soil erosion or sediment deposition.The use of inaccurate ref- erence inventories will result in biased and unreliable estimates of soil erosion and sediment deposition (Owens & Walling, 1996;Sutherland, 1996).Despite its crucial importance, there is very little information for the Southern Hemisphere on the spatial variability of 137 Cs within areas taken as reference sites.
This paper presents preliminary results from a study that is being carried out in the South-Central region of Brazil with the objective of evaluating the random and systematic spatial variabilities in reference 137 Cs inventories.Different potential reference sites, close to areas where the 137 Cs technique is being applied for erosion studies, were selected and sampled for a larger investigation.The study is of special interest for Brazilian conditions, where the 137 Cs activity in soils is very low.In such condition of very low 137 Cs activity, the variability in reference inventories can have even a larger contribution in the uncertainties associated to 137 Cs evaluations.

General characteristics of the selected reference sites
Four reference sites, in Piracicaba, SP, were initially selected for this study.It is intended to extend the number of sites, as well as to expand the number of soil profiles per area, to improve the methodology in future studies.All four sites are under sub-tropical climate conditions, situated in a radius of about 50 km from Piracicaba, with a Cwa climate according to Köppen, which is dry during the winter and rainy during summer.Figure 1 shows a map of the relative positions of the four reference sites.
Reference site R 1 -The first reference site R 1 was taken in Piracicaba, SP (22°42'S, 47°38'W and 560 m of altitude.)The site is a 20,000 m 2 grass field of an old garden.More than 40 years ago the area was embanked in order to establish a very flat grass (Paspalum notatum Flugge) field to be used only for recreation.The precipitation is 1,253 mm per year and the soil is an embankment with 180 g kg -1 sand, 390 g kg -1 clay and 430 g kg -1 silt.From this reference site, five soil profiles were taken and analyzed.The sampling points are 25 m distant from each other in a transect that crosses the field in its flat portion.
Reference site R2 -The second reference site R2 is situated at coordinates 22°47' S, 47°19'W and 600 m of altitude in the municipality of Nova Odessa, SP.The sampling place, with an area of about 30,000 m 2 , is a pasture (Brachiaria decumbens Stapf) since 1979 after the deforestation and presents an average slope of 0.5%.The total annual precipitation in the region is 1,317 mm.The soil is an Arenic Paleudult with 520 g kg -1 sand, 140 g kg -1 clay and 340 g kg -1 silt.Six soil profiles from this site, taken along a transect, were sampled and analyzed.The distance between sampling points in the transect is 5 m.
Reference site R3 -The third reference site R3 is located at coordinates 22°09' S, 47°01'W and 700 m of altitude in the municipality of Campinas, SP.The total annual precipitation in the region is 1,400 mm.The selected sampling site is in a very flat area of 2,400 m 2 covered with a perennial legume, called "tropical kudzu" [Pueraria phaseoloides (Roxb) Benth.].The area was seeded only once in 1943 and the local soil remained uncultivated since that time.The soil is classified as a Typic Hapludox with 380 g kg -1 sand, 380 g kg -1 clay and 240 g kg -1 silt.Five profiles from this area were sampled and analyzed.The sampling points were taken on a transect with a distance of 5 m between points.
Reference site R4 -The fourth reference site R4 is located in the municipality of Anhembi, SP.It is located at coordinates 22°40'S, 48°10'W and 460 m of altitude.The local annual precipitation is 1,100 mm.The select area of 20,000 m 2 has an average slope of 0.5% and the soil is a Typic Quartzipsamments with 850 g kg -1 sand, 60 g kg -1 clay and 90 g kg -1 silt.The site was deforested more than 40 years ago for the establishment of a pasture (Brachiaria decumbens Stapf).Six soil profiles, selected in a randomized way, in a distance of about 10 m from each other, were taken and analyzed.

Soil sampling and 137 Cs analysis
The sampling device (Figure 2) used allows the collection of soil samples in 5 cm thickness layers, down to the 50 cm depth in the soil profile, covering a surface area of 672 cm 2 , which corresponds to a sampling volume of 3,360 cm 3 for each 5 cm layer.
Soil samples were analyzed for 137 Cs activity using a HPGE Coaxial Detector (GEM-20180P, PopTop) with a detection efficiency of 0.7% 1 for the adopted geometry (1 litter Marinelli Beakers) and a minimum de-1 Detector efficiency determined using a IAEA standard soil sample presenting a known activity of 2.97 Bq kg -1 .tectable activity of the order of 0.2 Bq kg -1 .Due to the very low 137 Cs activity of the soil samples and the very low detection efficiency, the counting time varied from 24 to 56 hours according to soil sample activities in order to assure an acceptable integration error (<10%) during the spectrum analysis.
The soils of the four sites studied were analyzed for chemical and physical properties in depth in order to look for correlations between 137 Cs distribution and other soil property, related to the dynamics of the cesium.

RESULT AND DISCUSSION
The average inventories (Table 1) were compared using the statistical t test.At site R1 the inventory was higher (P < 0.01) than, the other three sites.No differences were found between average inventories of the three other sites (R2, R3 and R4).In R1 inventories varied from 277 to 367 Bq m -2 .The 137 Cs activities decreased from 302 Bq m -2 in the upper 15 cm of the soil profile to 11 Bq m -2 in the 15-25 cm layer.No 137 Cs activity was detected below the 20-25 cm layer.
At site R2 the inventories varied from 163 to 393 Bq m -2 and 91% of the 137 Cs was found in the first upper 20 cm of the soil profile.At site R3 inventories varied from 177 to 291 Bq m -2 with more than 98% of the total 137 Cs activity in the upper 20 cm of the soil profile.For R4 the inventories varied from 205 to 290 Bq m -2 .The 137 Cs activities presented a smooth reduction in depth from the 15-20 cm layer down to 45 cm depth below which no radioisotope was detected.
The conditions found on the site R1, both the 137 Cs distribution in the soil profiles and smaller variation between the inventories (very low CV), are closer to the expected and desired conditions of an adequate reference site.In the sandy soil of site R4 137 Cs activity decrease in depth was also observed but the 137 Cs distribution in the soil profile was more uniform when compared to the distribution of the other sites.
The average 137 Cs activity distribution found in the soil profiles of the four sites is in accordance to reference sites described in the literature (Owens & Walling, 1996;Walling & Quine, 1992;Walling, 1995).In non disturbed soil profiles the higher 137 Cs activity is found in the upper 20 to 25 cm of the soil profile (Basher et al., 1995).
The organic matter content of the soil at site R3 is higher (P < 0.001) than those of the other sites (Table 2).The same behavior was observed for soil pH.The  amount of potassium of the soil of R3 was similar to the soil of R1 but much higher than that of the soils of R2 and R4.The amounts of oxides in the soil of R3 were also higher in relation to the other soils.Table 3 shows the difference of silt content of the soils (P < 0.001).
The amount of organic matter is positively correlated with 137 Cs activity for soil sites R1 (R 2 = 0.94; P < 0.01), R2 (R 2 = 0.93; P < 0.001) and R4 (R 2 = 0.86; P < 0.001), while for site R3 the correlation was not good (R 2 = 0.52; P < 0.1) (Figure 3).Another correlation found was between 137 Cs activity and Ca +2 concentrations (P < 0.001) for site R4.Although some correlations have been identified, it is not sure that there are cause/effect relationships between them.Roque (2003) reports that tropical soils present specific exchange sites for 137 Cs, mainly in the silt fraction (Roque et al., that the humic organic matter does not present this kind of adsorption sites.Wässerman et al. (2002) and Sheppard & Thibault (1992) observed that most of the 137 Cs of fallout seems to be associated to soil oxides.Unfortunately in the present study the concentration of oxides was determined only for the upper 0-20 cm soil layer and therefore it was not possible to analyze its possible correlation with 137 Cs activities.
The recommendation that reference sites should be taken in places that have not been cultivated or mechanically disturbed since the 137 Cs fallout is well known.In such condition, it is expected that the plant nutrients as well as the 137 Table 2 -Chemical and mineralogical analysis of the soils from the reference sites (average for profiles).p = number of soil profiles analyzed.nd = not determined.
Table 3 -Physical analysis of the soils from the reference sites (average for profiles).per first centimeters of the soil profile.It is also expected that the concentrations of such elements decrease exponentially in depth.Such behavior was observed in the soils of the sites R1, R2 and R4 but not in R3.Some facts can explain these results.At areas R1, R2 and R4 the soils were in grass whose root system is composed basically of fine roots that are concentrated in the upper 20 cm of the soil profile.The depth organic matter distribution in these three sites (Figure 3) is very different from that found for site R3, which has been covered for 60 years with a perennial legume (Pueraria phaseoloides) that presents a root system composed of a main deep tap-root and many secondary roots.The presence of 137 Cs in deeper soil layers in the site R3 could be associated to internal cm -1 ), organic matter (OM, g kg -1 cm -1 ), potassium (K, mmol c kg -1 cm -1 ) and silt fraction (Silt, g kg -1 cm -1 ) in the four reference sites.
soil redistribution in the soil profile which could have been occurring in large and continuous pores left after the decomposition of these large deep roots, as well as by the action of large amount of soil organisms present in the area.
The relatively uniform depth distribution of organic matter at site R3 explains the lack of a significant correlation between this variable and the 137 Cs activity.The presence of 137 Cs in deeper soil layers in the sandy soil profile of site R4 is probably associated with soil texture and structure of the Typic Quartzipsamments.This soil presents high permeability and fast drainage due to its high macro-porosity (Resende et al., 1988) and as a consequence they present higher nutrient leaching susceptibility as compared to clay soils (Raij, 1991).
This study showed some important differences in the pattern of 137 Cs depth distribution in soil profiles of reference sites.They are probably associated with the chemical, physical, mineralogical and biological differences of the studied soils but many questions still remain open for future investigation, mainly those regarding the adsorption and dynamics of the 137 Cs in soil profiles under tropical conditions.
Taking into account that the sampling and analytical procedures were the same for all soil profiles and that the sample 137 Cs activities were of the same order of magnitude, the observed variability can be attributed to: a) random spatial variability (inside each reference site) -this variability, expressed by the coefficient of variation (CV%) of the average of the inventories of each site, varied from 11%, observed for site R1, to 31%, for site R2.Using the arbitrary categories established by Wilding & Drees (1983), mentioned by Sutherland (1996), these CV values would place the 137 Cs random spatial variability in a category of moderate variability, in the same class of other soil characteristics such as the clay and organic matter content (Wilding & Drees, 1983;Dahiya et al., 1984).Therefore the results observed in the present study are in accordance with other authors (Owens & Walling, 1996).These results indicate that the common simple assumption that 137 Cs is uniformly distributed in undisturbed landscapes (Ritchie & McHenry, 1990) needs detailed examination.b) systematic spatial variability (between reference sites) -this variability, expressed by the coefficient of variation (CV%) of the average of the inventories of the four sites, was of 11%.The average inventory of the four sites was 268 ± 30 Bq m -2 .The amplitude of variation was from 250 Bq m -2 (site R2) to 314 Bq m -2 (site R1).Flores et al. (2001) found an average of 491 ± 15 Bq m -2 (CV = 3%).The statistically higher average inventory found in the site R1, could be attributed to systematic differences of precipitation, soil and vegetation types, as mentioned by Owens & Walling (1996).It is well known that in tropical regions the precipitation can present a high variability at short distances when they are analyzed for short periods of time and that this variability can be very small when analyzed over the long run (Reichardt et al., 1995).Taking into account that the annual precipitation values for the studied sites are very close, it is difficult to interpret the real effect of the rainfall variability on the observed 137 Cs fallout systematic spatial distribution.
The variability of the soil physical and chemical characteristics was not able to explain the random 137 Cs spatial variability which was higher than the systematic spatial variability.Since the use of the 137 Cs technique for erosion rate evaluation is based on average reference inventories it is clear that all data interpretation will de- pend strongly on the spatial distribution pattern of the 137 Cs fallout in the selected reference site.Based on the observed systematic spatial variability, it is also clear that the reference site must be as close as possible of the study area.

Figure 1 -
Figure 1 -Relative location of the four reference sites.

Figure 2 -
Figure 2 -Equipment used for sampling soil profiles at reference sites.
Cs have higher concentrations in the up-