SODIUM-AFFECTED ALFISOLS OF THE AGRESTE REGION , STATE OF PARAÍBA , BRAZIL , AS KNOWN BY POTTER-FARMERS AND AGRONOMISTS

Decisions and practical attitudes of land use and management adopted by rural populations are, to a great extent, based on local soil knowledge. This study was performed to describe and analyze chemical, physical and morphological characteristics of sodium-affected Alfisols employed in traditional pottery; and to compare, through statistical analysis, pedogenetical horizons and soil layers, as recognized in soil profiles by professional researchers and potterfarmers, respectively. Fieldwork was performed in Chã da Pia, a rural village in the municipality of Areia, in the Agreste region of the State of Paraíba, Northeast Brazil. Ethnoscientific techniques were used to describe and analyze soil knowledge of potter-farmers who played the role of informants. A comparison of soil categories recognized by farmers and researchers was performed through discriminant-canonical analysis. Soils from which local people obtain a material named “barro de loiça” (pottery clay) were classified as Typic Natraqualf and Vertic Albaqualf, according to Soil Taxonomy. Potter-farmers were able to identify differences between the tillable topsoil and underlying layers from which they extract ceramic clay. Some soil layers, recognized by potterfarmers, were similar to pedogenetic horizons of the same soils, recognized by researchers, in terms of physical, chemical and morphological characteristics. Discriminant-canonical analysis was a useful tool to articulate and compare information related to soil knowledge held by distinct social groups.


INTRODUCTION
The need of fitting land use strategies to local reality has determined the quest for efficient mechanisms for communication among researchers with formal education and rural populations, mainly small scale farmers.Ethnopedological studies attempt to support this pursue since they may provide more in-depth understanding of the relationship among humans, soils and other constituents of the ecosystems (Alves & Marques, 2005;Alves et al., 2006).The rationale is that social groups tend to assign different meanings to soils (Brady, 1989;Krasilnikov & Tabor, 2003;Cooper et al., 2005), and that decisions and practical attitudes of use and management adopted by these groups are based on their specific knowledge and cosmovision.These are generally transmitted through generations without necessarily using written language (Toledo, 2000).
Statistical multivariate analysis has shown to be useful in pedological studies, facilitating the comparison and integration between soil data sets, even when they are related to different social groups (e.g.soil scientists and peasants), as it happens in ethnopedological research (Williams & Ortiz-Solorio, 1981;Queiroz & Norton, 1992).In this study, "formal" soil knowledge is that one which is shared by researchers with academic education (soil scientists, in this case), while the knowledge shared by social groups which use and manage soil resources in their daily activities (peasant potters, in this case) is assumed to be "local" (1) .
The majority of ethnopedological studies have focused local knowledge on the tillable topsoil, with special emphasis on the spatial distribution of soil categories over the land surface as recognized by the populations under study.On the other hand, Barrera-Bassols & Zinck (2003) have shown the existence and importance of peasant knowledge of soil variations with depth.In the Agreste region of the Paraíba State (Brazil), Alves et al. (2005) observed that some potterfarmers recognized differences between tillable topsoil and underlying soil in Albaqualfs and Natraqualfs, and were also capable of distinguishing, identifying and naming, on their own language and criteria, some of the soil materials stratified throughout the profile.
This study had the objective of describing and analyzing chemical, physical and morphological features of some soils used in traditional pottery and to establish, through statistical multivariate analysis, com-parisons between pedogenetical horizons and the layers recognized in sodium-affected Alfisols by local potter-farmers in Areia, State of Paraíba, Brazil.

Data Collection and Analysis
Fieldwork was performed in Chã da Pia, which is located in the rural zone of the municipality of Areia (Agreste region of Paraíba State), Brazil.Approximately forty people were working directly and regularly in the making of pottery clay ("loiça de barro") for commerce and family usage in the area, and they call themselves "loiceiros" (potters) (2) .Local altitude varies between 484 and 552 m, latitude between 6°54'15' ' and 6°55'6" S, and longitude between 35°46'39'' and 35°47'41" W. Predominant relief is undulated or gently undulated.The climate is As' (hot and dry, with rain during autumn and winter), according to the Köppen's classification, with annual rainfall averaging 700 mm (Jacomine et al., 1972).
During the first phase of the fieldwork, 35 local potter-farmers worked as informants through openended interviews (Albuquerque & Lucena, 2004).Afterwards the interviews were intercalated with tours (Spradley, 1979) in which potter-farmers guided the authors to the sources of "barro de loiça" (local denomination of the soil materials that provide plasticity to the ceramic paste).The information obtained was recorded in tape and registered on field notebooks for later transcription, systematization and analysis.In addition to the information offered by the peasant craftsmen, the authors obtained data related to soils used in local pottery through direct observation.Sampling and soils description were performed through conventional methods (Santos et al., 2005) in five sites (pottery clay sources) by four agronomists (the authors of this paper) who had previously accomplished graduate studies in soil science, thus with formal knowledge in Pedology.
For each of the five profiles, following conventional description, complementary soil sampling was done with the help of an informant who was asked to name and describe, according to local knowledge, the soils (or soil materials) of the location.Pointing to the soil profile, the authors would ask questions such as: "show me where the pottery clay is" and "what else can you show me here?" (1) Santos (2006) has shown that some local features may be found in formal or scientific knowledge.So, the adjectives "formal" and "local", as used in this paper, do not mean the opposite of each other. (2) The term "loiça" is widely used in colloquial language in Northeast Brazil, and refers to traditional pottery.It appears in the "Aurélio" Dictionary [http://www.uol.com.br/Aurelio] as being parallel to "louça".Access in June, 2004.Part of the chemical, physical and morphological data used in the statistical analysis (those from potter-farmers' layers) were published in Alves et al. (2005), whilst this paper brings a different data set that represents the pedogenetic horizons of the same soils.
Linear functions (canonical roots) were estimated through DCA, based on selected independent variables data, enabling the distinction and listing of individuals, maximizing variance among the groups and minimizing variance within the groups (Silva et al., 2001).The program "Statistica" (Statsoft, 1995) was used to perform the analysis.Based on a combination of data from original variables, DCA makes it possible to obtain compound variables named canonical roots or discriminating functions.Each canonical root consists of a linear combination (Z) of the independent variables (Yi), in order to maximize the correlation between Z and Yi.The linear combination of i variables Y, forming a discriminant function Z may be represented by the following model: In this study, a discriminant model was built step by step using a "forward stepwise" analytical procedure.In each step, all variables were reevaluated and those that would contribute the most to discriminate among the groups were detected.This variable was then included in the model and the procedure continued with the other variables, included one by one, until completion.During this procedure, the variables that did not significantly contribute to group distinction were excluded (Statsoft, 1995).
A total of twenty-eight independent variables were inserted in the analysis, six of them related to morphological features, 11 to physical features and the other 11 to chemical features of layers and horizons.Variables with a high potential degree of intercorrelation with others were excluded.Thus, values of silt and clay were inserted, but the relation silt/clay was excluded, despite being used as a criterion to classify specific soils.In the same way, exchangeable cation values were inserted, but the sum of basic cations was excluded, otherwise it would not be possible to perform the test.The problem of "constant sum" that occurs, for example, with the sum of granulometric fractions, was considered as well.In this case, one of the fractions (coarse sand) was excluded, as performed by Vaselli et al. (1997).
Qualitative data from morphological variables were codified into numeric values (Table 1), similar to those described by Queiroz (1985) and Queiroz & Norton (1992).Hue was codified from 1 (redder) to Sci.Agric.(Piracicaba, Braz.), v.64, n.5, p.495-505, September/October 2007 8 (more yellow); chroma was codified from 1 (lower intensity of hue) to 8 (higher intensity of hue); value was codified from 2 (higher contribution of black) to 8 (higher contribution of white); texture was codified from 1 (higher influence of the sand fraction) to 7 (higher influence of clay and/or expansive mineral fractions); plasticity and stickiness were both codified from 1 (non plastic, non sticky) to 4 (very plastic and very sticky).For statistical purposes, the color of the wet samples was considered, while formal classification was based on the color of moist and dry samples.After thoroughly mixed with water, the samples presented uniform color, while the moist soil sometimes had a variegated color pattern.Having a single color for each (wet) sample made it possible to consider color aspects (hue, chroma and value) as variables in the statistical procedure.

RESULTS AND DISCUSSION
Soils described in places where pottery clay was normally extracted by local potters were classified as Typic Natraqualf e Vertic Albaqualf (Soil Survey Staff, 1999), which correspond to Planossolo Nátrico e Planossolo Háplico in the Brazilian Soil Classification System (Embrapa, 2006), respectively (Tables 2 to 4).
Pottery clay was normally extracted from Bt horizons (Alves, 2004;Alves et al., 2005).These horizons generally present a clayey texture.Dry samples were extremely hard and extremely firm, while wet samples ranged from sticky to very sticky and, from plastic to very plastic (Tables 2 to 4).Brownish colors were predominant with the exception of profile 3 (with vertic properties), where grayish colors were prevalent.From a chemical viewpoint the hyper- Table 1 -Morphological soil attributes and codes used in the discriminant canonical analysis, based on Queiroz (1985) and Queiroz & Norton (1992).
*Clay content higher than 60 g kg -1 .This class is admitted in Brazil (Santos et al., 2005), but not in Soil Taxonomy (Soil Survey Staff, 1999).
eutrophic feature was remarkable, as well as the relatively high sodium saturation levels in the exchange sites (from 7 to 25%).
Local slope ranged between 4% (3 rd profile) and 24% (5 th profile).Natraqualfs and Albaqualfs are most commonly associated to plain and gently undulated relief, but may occur in sites with relatively high slope in the Brazilian Northeast (Sampaio et al., 1976) and in other regions (Faivre, 1977).Sodium-affected Argids locally named "barro de louça" (pottery clay, in reference to the Bt horizon) occur in the Acaraú River Valley (Ceará State), under a large slope range, not being restricted only to low slope sites near streams and rivers (Queiroz, 1985).

Discriminant-Canonical Analysis
The canonical roots analysis contributed to distinguish local and formal categories (layers and horizons, respectively) related to the soils under study (Wilk's lambda = 0.00001 and P < 0.0001) (Table 5).In this case, lower partial lambda values indicated higher contribution of the respective variables for group distinction.Therefore, the variable contributing the most was depth, followed by extractable phosphorus, pH in 1mol L -1 KC1, chroma, gravel content and electrolytic conductivity.Because depth was the most discriminating variable, it ratified a trend going back to the beginning of the 20 th century, when soil horizons were identified mainly by their relative position in the profile (Nikiforoff, 1931).
Judging by the distance between centroids (Table 6) there were similarities between certain local and formal soil categories, as follows: "terra" was particularly similar to the A horizons and, to a lesser degree, to E horizons.On the other hand, "piçarro" was particularly similar to E horizons and, to a lesser de-         gree, to A horizons."Cabeça do barro" and "barro de loiça" were similar to each other and also to the B horizons."Pedra mole" was similar to BC horizons, but the respective centroids were not as close in this particular comparison as in the other ones.The most remarkable differences were obtained when comparing surface-or-elluvial categories (A, E, "terra" and "piçarro") with subsurface-or-illuvial categories (B, BC "cabeça do barro", "barro de loiça" and "pedra mole").Therefore, the "individuals" under study were divided in two groups (Figure 1) as related to their position in the profile: an upper group formed by surface-andelluvial categories, and a lower one formed by subsurface-and-illuvial categories.
Table 4 -Chemical attributes of soils that are used as a source of pottery clay by peasant potters in Chã da Pia, Paraíba, Brazil.
Three canonical roots were significant (P < 0.01).Nevertheless, the following discussion is based on the first two roots, disregarding the others (Table 7), since the first canonical root explained most of the variation (68%), while the second root added 22% resulting in an accumulated total of 90%.Canonical correlation R was very high for the first two roots, demonstrating the high degree of correlation between these two roots and the other variables.Averages of each category in the multivariate space that was defined by these two main canonical roots (Figure 1) were useful to indicate those groups that were distinguished by a root.The first root made distinctions between the categories situated in the upper positions of the profile ("terra", "piçarro", A and E) and the lower ones ("cabeça do barro", "barro de loiça", "pedra Sci.Agric.(Piracicaba, Braz.), v.64, n.5, p.495-505, September/October 2007 mole", Bt and BC).The second root, on the other hand, enabled a distinction between the categories more influenced by the proximity of the parent material ("pedra mole" and BC) from the others.
Structural coefficients (Table 8) expressed correlations between the variables in the model and the discriminant functions, making it possible to assign meaning to the discriminant functions (roots).In the first root, negative values of the following variables are outstanding: water-dispersed clay, soil texture, total clay, exchangeable magnesium, stickiness, plasticity, depth, and exchangeable sodium.In the second root, positive values obtained for soil depth, electrolytic conductivity, extractable phosphorus and exchangeable sodium were outstanding.
Through the trend revealed in the first root, it was seen that the categories with lower position ("barro de loiça", "cabeça do barro", "pedra mole", Bt and, BC) stood out from the upper ones, the first ones having higher values for the variables previously depicted in table 8 (water-dispersed clay, texture, total clay, exchangeable magnesium, stickiness, plasticity, depth and exchangeable sodium).Similarly, it was demonstrated through the second root that "pedra mole" and BC were similar within themselves and distinguished from the other categories, since they showed higher values for the variables previously cited (depth, electrolytic conductivity, extractable phosphorus and exchangeable sodium).
This comparison between formal and local soil categories made it possible to accept the hypothesis of morphological, physical and chemical similarities between the layers recognized by potter-farmers and the pedogenetical horizons from which these materi-      als are normally collected.On the other hand, the evident distinction among categories located in upper and lower positions of the soil profile was coherent with certain characteristics of Albaqualfs and Natraqualfs, such as the high textural gradient and the abrupt transition from horizons A or E to the underlying Bt.The difficulty in distinguishing between "barro de loiça" and "cabeça do barro", through the chosen statistical procedure seemed to be related to field situations, since the potters not always reject the materials from the top of Bt ("cabeça do barro") when they collected the clay for pottery.
Researchers with formal pedological training might also have doubts and divergences, on the field, related to a possible subdivision of the Bt horizon in sub-horizons, when studying Albaqualfs and Natraqualfs, and might depend on laboratory analysis and further evaluations to define the horizons or subhorizons to be accepted in each case.Sometimes it was possible to visualize a strong darkness in the external face of the aggregates, on the top of Bt ("cabeça do barro"), but the structure could be formed by blocks or prisms comprehending all of this horizon extension, making a subdivision unfeasible.Williams & Ortiz-Solorio (1981) obtained a low degree of correspondence between soil categories recognized by farmers and researchers, mainly because in that situation, the local classification was based on the features of a bi-dimensional taxonomic unit (tillable topsoil), rather than a three-dimensional one (soil profile).Nevertheless, their data demonstrated that local categories reflected discontinuities on the soil surface, which originated measurable and statistically valid categories.Queiroz & Norton (1992) demonstrated that the cluster arrangement of soils from the Acaraú Valley (Ceará State), based on morphological data from formal soil surveys, was strongly coincident with the soil categories indicated by peasants, although there were many differences in nomenclature between the two data sets.Further analysis demonstrated that these clusters differed among themselves in relation to water holding capacity and pH.As a consequence, the local classification was considered to be valid for distinguishing soils as related to non-morphological characteristics as well.For each soil profile, Queiroz & Norton (1992) inserted morphological data of two horizons (a surface and a subsurface one), so that the "individuals" that they submitted to the clustering test were "soils", each of them represented by a pair of horizons.
The practical applications and epistemological implications of statistically analyzing data from local categories of soils, attempting to validate local knowledge through laboratorial analysis, have been submitted to divergent interpretations and criticism.Thus, Winkler-Prins (1999) evaluated the quest of a "scientific rationale" underlying local soil knowledge.Her understanding is that these studies aiming at validating local knowledge through formal procedures have the merit of clearly demonstrating that this knowledge can be "scientifically valid" and that consulting with local people is a beneficial activity for development practitioners.On the other hand, she emphasized that some validation attempts could presume "that scientific knowledge is superior to local knowledge, and that the later needs to be proven in order to be used by scientists".This would be contradictory with the idea that local knowledge has its intrinsic value.
Considering these restrictions, statistical and laboratorial studies not only pursue the validation of local knowledge, but also represent opportunities for the emergence of an integrating language, attempting to reduce the distance separating local knowledge from that one practiced and accepted in the academy.These studies may, within certain contexts, revalidate and feedback on the knowledge and practices of people conducting formal scientific research, reconnecting them with the peasant's experience.

Table 2 -
Morphological attributes of soils that are used as a source of pottery clay by peasant potters in Chã da Pia, Paraíba, Brazil.

Table 3 -
Physical attributes of soils that are used as a source of pottery clay by peasant potters in Chã da Pia, Paraíba, Brazil.

Table 5 -
Variables included and excluded from the with respective F, and P 1 variables excluded from the model.

Table 7 -
Significance indexes for the canonical roots.

Table 8 -
Indexes of correlation between the variables and the canonical roots.