SciELO - Scientific Electronic Library Online

vol.38 número1Comparación de la composición nutricional del maíz transgénico (gen quitinasa) con maíz no transgénico índice de autoresíndice de materiabúsqueda de artículos
Home Pagelista alfabética de revistas  

Servicios Personalizados




Links relacionados


Ciencia e investigación agraria

versión On-line ISSN 0718-1620

Cienc. Inv. Agr. vol.38 no.1 Santiago abr. 2011 

Cien. Inv. Agr. 38(1):155-164. 2011




Spatial variability of an Alfisol cultivated with sugarcane

Variabilidad espacial de un Alfisol cultivado con caña de azúcar


Joedna Silva Cruz1, Raimundo Nonato de Assis Júnior2, Sammy Sidney Rocha Matias3, and Jesús Hernán Camacho-Tamayo4

1Departamento de Fitotecnia, Universidade Federal do Ceará - UFC, Av. Mister Hull, s/n - Campus do Pici-CEP 60021-970, Brazil.
2Departamento de Ciencias do Solo, Universidade Federal do Ceará - UFC, Av. Mister Hull, s/n - Campus do Pici - CEP 60021-970, Brazil.
Universidade Estadual do Piauí - UESPI. Rua Tres, 11; Bairro Nova Corrente; CEP 64980-000, Corrente - PI, Brazil.
4Programa de Ingeniería Agrícola, Facultad de Ingeniería, Universidad Nacional de Colombia. Ciudad Universitaria, Ed. 214 - Of. 304, Bogotá, Colombia.


Soil variability occurs due to natural and anthropogenic factors acting at various spatial and temporal scales. The purpose of this study was to evaluate the spatial variability as well as the relation of some chemical attributes using different statistical techniques. The study was conducted in the municipality of Maracanaú (CE, Brazil) in the commercial area of sugar cane in a typic Hapludalf. Soils were sampled in a 100 x 100 m grid every 10 m and at a depth between 0 and 20 cm for determination of exchangeable acidity, pH (KCl), Ca, Na, K, Mg, P, N, sum of bases, cationic exchange capacity, base saturation and organic matter. Data were subjected to analysis of descriptive statistics, multivariate analysis and geostatistics. Attributes showed low and medium variability, except for pH that did not show spatial dependence. Different statistical tools helped to identify the soil spatial variability and to get a better understanding of relationships among different attributes. Techniques also allowed to better plan and control agricultural production, especially in carrying out site specific management of agricultural inputs.

Key words: Geostatistics, principal components, cluster analysis.


La variabilidad del suelo ocurre debido a factores que actúan en varias escalas espaciales y temporales, naturales o antrópicas. El objetivo del presente estudio fue evaluar la variabilidad espacial y la relación de algunos atributos químicos, mediante diferentes técnicas estadísticas. El trabajo fue realizado en el municipio de Maracanaú (CE, Brasil), en área comercial de caña de azúcar, en un Typic Hapludalfs. Se tomaron muestras de suelo cada 10 m, en una malla regular de 100 m x 100 m, en una capa entre 0 y 0,20 m, para la determinación de acidez intercambiable, pH (KCl), Ca, Na, K, Mg, P, N, suma de bases, capacidad de intercambio catiónica, saturación de bases y materia orgánica. Los datos fueron analizados mediante estadística descriptiva, multivariada y, geoestadística e interpolación por kriging ordinario. Los atributos presentaron variabilidad baja o media, donde el pH fue el único atributo que no presentó dependencia espacial. Las diferentes técnicas estadísticas ayudaron a caracterizar la variabilidad espacial del suelo y la identificación de las diferentes relaciones entre atributos, cuyos resultados permiten una mejor planeación y control en la producción agrícola, especialmente en la realización de prácticas de manejo localizado de insumos agrícolas.

Palabras clave: Geoestadística, componentes principales, agrupamiento jerárquico.



Natural soil variability occurs due to factors acting on various spatial and temporal scales, caused by complex pedological processes (Burrough, 1993), influenced by relief and humidity regimes (Rezaei and Gilkes, 2005).

However, crop establishment also modifies soil variability (Burgos et ah, 2006), especially by farming work and implementation of agricultural inputs, especially on the surface (Cama-cho-Tamayo et al, 2008). Mello et al. (2006) found heterogeneous attributes in different soil types and indicated that spatial dependence is conditioned by relief, which causes to determine different management areas from each evaluated attribute of the soil, evidencing the need of carrying out researches in different soil classes.

In general, there are different tools to identify the behavior of soil attributes, as well as to quantify and characterize them to determine a specific crop; the univariate statistics however, is the most commonly used. According to Fi-dalski et al. (2007), these statistics may compromise interpretations and conclusions, as the existent dependence among the attributes analyzed is not explored. Other tools, like multivar-iate statistics and geostatistics, help understand the behavior of soil attributes.

In multivariate statistics, the principal components and hierarchical cluster analysis are widely used in environmental studies for the evaluation, classification and construction of mathematical models for decision making. These are implemented in the study of soil properties to understand attributes variability, and the most important, their dependence between each other, which cannot be verified using univariate statistics (Fidalski et al., 2007). The principal components analysis reduces the dimensions, showing the different basic components, called principal components. Normally, a small number of components explain a high percentage of the total variance, i.e., the data cluster may be described in a smaller dimensional space (Ramos et al., 2007). On the other hand, the hierarchical cluster analysis identifies groups where the attributes present some similarities between each other. The use of these techniques has favored the variability identification of chemical attributes through a cluster of correlated soil attributes (Fidalski et al., 2007; Theodoro et al., 2003).

Geostatistics and especially kriging interpolation are useful tools for map construction and for understanding variability of soil properties in a specific area. This tool may be implemented at different scales in relation to specific data, depending on the resolution desired in the study (Webster, 2008). Characterization of soil spatial variability allows a better representation of the complex relations between the soil characteristics and the environment (Goovaerts, 1998), which helps determine the practices to be used on crops management (Bouma et al., 1999). The search of representative information on spatial soil variability has increased, resulting in more appropriate model developments and management systems for agricultural production (Godwin and Miller, 2003), identifying the relation among soil properties as well as defining homogeneous management areas, as reported by different authors of sugar cane cultivation in order to adopt agriculture precision (Berner et al., 2007; Corá et al, 2004).

The objective of this study was to evaluate the spatial variability of some chemical attributes and their relations, in a Typic Hapludalf cultivated with sugar cane, using univariate statistics, multivariate analysis and geostatistics techniques.

Materials and methods

The work was carried out in Hacienda Jaganaú, located in the municipality of Maracanaú (CE, Brazil), with geographical coordenates 03°49'88'' South latitude, 38°38'19'' West longitude and a mean altitude of 47 m. According to the Koppen classification (1936), the climate in the region is AW', hot and humid, with temperatures higher than 18 °C in the coldest month, mean precipitation of 1,200 mm, with rains concentrated from February to April. The relief is soft, undulating, with a slope between 2 and 5%. The experiment was made in a sugar cane commercial area, cultivated more than 20 years ago, fertirrigated with manual harvest and without hey burning during the last five years. The soil was classified as Typic Hapludalf of sandy texture, with mean contents of 81% of sand, 9 % of clay and 10 % of silt.

The samples were collected in points with a regular distance of 10 m, in a grid of 100 x 100 m, in two perpendicular directions between each other, with a total of 100 points. Samples in the 0.0 - 0.20 m layer were collected in each point. The pH, organic matter content, P, K, Ca, Mg, Na, H+Al and N were determined with these samples, following the methodology proposed by Raij et al. (2001). P and K were extracted by the resin method. P was determined by pho-tocolorimeter, K and Na by flame emission spectrophotometry. Ca and Mg were extracted with solution 1 N of KCl and were determined by atomic absorption spectrophotometry. H + Al3+ were extracted with calcium acetate at pH 7. The sum of bases (SB) was determined from Ca, Mg and K, which along Al define the capacity of cationic exchange (CEC). The saturation by bases (V) was determined based on the relation between SB and CEC.

Initially, a descriptive data study was made for the statistical analysis, calculating mean, median, minimum and maximum, variation coefficient (CV), and also skewness and kurtosis. Therefore, the attribute normality was evaluated, which is not essential, but provides better predictions when it is associated to geostatisti-cal techniques (Diggle and Ribeiro, 2000). The Warrick and Nielsen classification (1980) was used for the CV analysis, with low variability for values lower than 12%, mean between 12 and 60%, and high for values higher than 60%.

Subsequently, a multivariate analysis was carried out, performing a factor analysis by principal components (AFCP) and hierarchical cluster (AAH), in order to determine the relation among the chemical attributes. The correlation matrix was used for this analysis, where the data from each attribute were previously standardized, with mean 0 and variance 1. For AFCP, the Varimax rotation was applied. The Euclidian distance was used for AAH, so as to separate an attribute cluster into groups, and the results were represented graphically (dendrogram) in order to facilitate the identification of the groups formed by the attributes analyzed, using the Ward algorithm (Manly, 1997). The analysis of descriptive statistics and multivari-ate analysis were made with the program SPSS v.17.

The theory of regionalized variables was considered for the experimental semivariograms fit to theoretical models (Vieira, 2000). These models were estimated by the program GS+ v.7 (Robertson, 1998), which adopts the highest value of the determination coefficient (R2), the lower residual sum of squares (SQR) and the closest value to one of the correlation coefficient obtained by the method of cross-validation (CV) as criteria for the model selection. The prediction in unsampled zones from each attribute by ordinary kriging was obtained by the resulting models, represented in contour maps, using the program Surfer v.7 (Golden Software Inc., 1999). The correlation among the soil attributes was verified along the maps and the Pearson Linear Correlation Analysis. The classification of the degree of spatial dependence (GDE) was made based on the relation between the nugget effect and sill (C/Co+C), assigning strong for GDE higher than 75%, moderate for GDE between 25 and 75% and low for GDE smaller than 25% (Cambardella et al., 1994).

Results and discussion

The results obtained from the descriptive analysis showed that the attributes are close to the normal distribution, as the similar values of the media and the median for each attribute indicated, in addition to the values close to zero of skewness and kurtosis, which was also verified by the Kolmogorov-Smirnov normality test (Table 1). The fit for normality for these attributes, in different soil types, is reported by various authors (Silva et al., 2001; Yemefack et al., 2005; Camacho-Tamayo et al., 2008).

Phosphorous was the only attribute that was not close to normal distribution, with distant media and median values, in addition to presenting a significant normality test. According to Cama-cho-Tamayo et al. (2008), this behavior with higher values than the natural soil contents and high variability shown by maximum and minimum values, is due to the distribution of this agricultural input through the years, modifying soil contents, especially on the surface, because of the scarce mobility and solubility of this element. According to Cressie (1993), it is convenient to verify that not very long tails are present in the normal distribution, more than in normality, in order to avoid a result compromise, especially when kriging is performed using estimations based on mean values (Warrick and Nielsen, 1980). Another important fact is the occurrence of the proportional effect between the mean and the data variance, across a surface, which allows estimating well defined sill in the semivariograms.

pH, saturation by bases and the Na and N contents presented low variability, following the CV classification by Warrick and Nielsen (1980). The other attributes presented mean variability, where the highest value of the CV of the P outstands. This variability observed in the attributes has been reported by different authors (Silva and Chaves, 2001; Yemefack et al., 2005; Camacho-Tamayo et al., 2008), who also emphasized the behavior of P, element that is distributed superficially or incorporated in the first centimeters of soil.

It was observed that the attributes formed two groups in the factor analysis by principal components. The first group is formed by pH, H+Al, MO, P and Mg. The second group is formed by N, Na, K, SB and CEC (Figure 1). It is evident that the Na and K contents presented a higher influence on the SB and CEC behavior, compared to the other cations (Figure 1). The attributes Ca and V were presented isolated; the exchangeable acidity presented an inverse relation with the bases saturation, and the pH did with the sum of bases.

The first four components were considered in the factors analysis, with eigenvalues higher than one, which in this case, showed an acceptable interval (Kaiser and Rice, 1974). These principal components explain approximately 75% of the total variance (Table 2). It was observed that the lowest communality values were P, MO and N, indicating that these attributes are slightly representative in the principal components analyzed. PC1 represented 32.43% of the total variance influenced by the attributes Na, K, SB and CEC, showing a strong relation among these attributes, in addition to an inverse correlation with pH. PC2 represented 21.58% of the total variance, evidencing an increasing influence of H+Al and SB, attributes that are inversely correlated, as observed in Figure 1. PC3 represented 11.29% of the total variance, showing the correlation existing between pH and Mg, possibly due to the superficial distribution of dolomite lime (30.4% of CaO and 21.95% of MgO). PC4 presented 9.50% of the total variance, revealing Ca as the attribute better explaining the behavior of this component.

The analysis of hyerarchical cluster identified two groups, similar to the groups found in the factor analysis. The first group is related to the presence of cations, formed by CEC, SB, K, Na, V, N and Ca. The second group is composed mainly by attributes helping the description of soil acidity, formed by P, MO, H+Al, Mg and pH. The strong influence that K in SB presented in this soil was also verified, as well as in CEC, along with Na, where these attributes formed a subgroup. Likewise, the influence of Mg on the pH behavior was confirmed.

The results of the theoretical model fit of isotro-pic semivariograms showed that the predominant model was spherical, followed by the exponential and Gaussian models (Table 3). The pH was the only attribute that presented an undefined spatial dependence, as the distribution of this attribute in soil is randomized, presenting a pure nugget effect (EPP). The determination coefficient (R2) was always higher than 0.80 for the attributes presenting spatial dependence, where Mg, H+Al, P, MO and N were the attributes showing the best fit. These R2 values, along the cross validation for all the attributes, indicate adequate data reliability. Studies developed by different authors obtained theoretical semivariograms models fitting for chemical attributes (Silva and Chaves, 2001; Gallardo and Paramá, 2007 and Camacho-Tamayo et al, 2008).

The attributes K, H+Al and MO presented the highest ranges, with distances higher than 80 m and a mean dependence degree. Lower ranges were found for SB, Ca and Na. Most of the attributes presented a moderate GDE, except SB, P and V, which presented a strong GDE; these attributes were fitted to theoretical exponential models.

Phosphorous was the only attribute presenting non significant correlations with the other attributes (Table 4), showing an inverse correlation of this attribute with pH, which indicates that the higher pH, the lower the P presence. However, it is observed a positive correlation between phosphorous and MO. A contrary situation was observed for CEC that presented the largest cluster of significant correlations with the other attributes, especially with SB. The pH presented a significant and positive correlation with Mg and Ca, which was stronger with Mg, as also observed in the hierarchical cluster. An inverse correlation was observed between pH and K, P, exchangeable acidity and MO.

The strong influence of K on SB and CEC is also confirmed by the high positive correlation among these attributes, which was verified by the factor analysis (Figure 1) and the hierarchical cluster analysis (Figure 2). This influence is also observed in the contour maps (Figures 3 and 4), where areas with high K content correspond to zones with higher CEC and SB values. In the same way, the similarity observed between MO and the exchangeable acidity in the hierarchical cluster analysis is confirmed by the positive correlation among these attributes, in addition to the similarity among the contour maps obtained by kriging. The influence of Ca on SB is also verified, where Mg is the cation with lower influence on SB and CEC.

The resulting contour maps confirmed the relation among the different attributes (Figures 3 and 4), as well as the variability existing in the soil, which indicates that a future application using variable rate techniques of different fertilizers and amendments, based on the requirements of the crop, is a convenient way to improve their use, as well as to diminish production costs and environmental impact, due to the excesses or deficiencies on the input distribution when conventionally distributed on the soil with a uniform rate.

The different attributes presented low or mean variability, where pH was the only attribute that did not show spatial dependence. The other attributes studied presented spatial dependence, identifying different zones using the elaboration of contour maps, which are basic for the specific management of agricultural input distributions, especially fertilizers.

The use of different statistical tools helped identify the spatial soil variability, as well as in the comprehension and determination of the different relations among attributes characterizing the soil, techniques that may help in decision making for a better organization and control of agricultural production.


Masters thesis presented to the Department of Soil Sciences of Universidad Federal de Ceará, Av. Mister Hull, s/n - Campus do Pici - CEP 60021-970, Brazil. A Proiect financed by the program CAPES/PROCAD/ESALQ/UFC 0095/00-1.


Berner, P.G.M., S.R. Vieira, E. Lima, and L.H.C. Anjos. 2007. Variabilidade espacial de propriedades físicas e químicas de um Cambissolo sob dois sistemas de manejo de cana-de-afúcar. Revista Brasileira de Ciencia do Solo 31:837-844.         [ Links ]

Bouma, J., B.J. Van Alphen Stoorvogel, and H.W.G. Booltink. 1999. Pedology, precision agriculture, and the changing paradigm of agricultural research. Soil Science of Society American Journal 63:1763-1768.         [ Links ]

Burgos, P., E. Madejón, A. Pérez-de-Mora, and F. Cabrera. 2006. Spatial variability of the chemical characteristics of a trace-element-contaminated soil before and after remediation. Geoderma 130:157-175.         [ Links ]

Burrough, P. A. 1993. Soil variability: a late 20th century view. Soils and Fertilizers 56:529-562.         [ Links ]

Camacho-Tamayo, J.H, C.A. Luengas, and F.R. Leiva. 2008. Effect of agricultural intervention on the spatial variability of some soils chemical properties in the eastern plains of Colombia. Chilean Journal of Agricultural Research 68:42-55.         [ Links ]

Cambardella, C.A., T.B. Moorman, J.M. Novak, T.B. Parkin, D.L. Karlen, R.F. Turco, and A.E. Konopka. 1994. Field-scale variability of soil properties in Central Iowa Soils. Soil Science Society American Journal 58:1501-1511.         [ Links ]

Corá, J.E., A.V. Araujo, G.T. Pereira, and J.M.G. Beraldo. 2004. Variabilidade espacial de atributos do solo para adofao do sistema de agricultura de precisao na cultura de cana-de-afúcar. Revista Brasileira de Ciencia do Solo 28:1013-1021 .         [ Links ]

Cressie, N. 1993. Statistics for spatial data. Revised Edition. John Wiley & Sons, New York, USA. 928 pp.         [ Links ]

Diggle, P.J., and J.R. Ribeiro. 2000. Model Based Geostatistics. 1. ed. Sao Paulo: Associafao Brasileira de Estatística. 129 pp.         [ Links ]

Fidalski, J., C.A. Tormena, and C.A. Scapim. 2007. Espacializafao vertical e horizontal dos indicadores de qualidade para um Latossolo Vermelho cultivado com citros. Revista Brasileira de Ciencia do Solo 31:9-19.         [ Links ]

Gallardo, A., and R. Paramá. 2007. Spatial variability of soil elements in two plant communities of NW Spain. Geoderma 139:199-208.         [ Links ]

Godwin, R.J., and P.C.H. Miller. 2003. A review of the technologies for mapping within-field variability. Biosystems Engeneering 84:393-407.         [ Links ]

Golden Software. 1999. Surface mapping system Inc. Surfer. Surfer version 7.00. Golden Software, Inc. Colorado. 619 pp.         [ Links ]

Goovaerts, P. 1998. Geostatistical tools for characterizing the spatial variability of microbiological and physico-chemical soil properties. Biology and Fertility of Soils 27:315-34.         [ Links ]

Kaiser, H.F., and J. Rice. 1974. Little jiffy, mark IV. Educational and Psychological Measurement 34:111-117.         [ Links ]

Köppen, W. 1936. Das geographisca System der Klimate, in: Handbuch der Klimatologie. Köppen, W. and Geiger, G. (eds.), 1. C.Gebr, Borntraeger. p. 1-44.         [ Links ]

Manly, B.F.J. 1997. Multivariate statistical methods. A primer. Second Ed. Chapman & Hall, London. 216 pp.         [ Links ]

Mello, G., C.R.P. Bueno, and G.T. Pereira, G. T. 2006. Variabilidade espacial das propriedades físicas e químicas do solo em áreas intensamente cultivadas. Revista Brasileira de Engenharia Agrícola e Ambiental 10:294-305.         [ Links ]

Ramos, M.C., R. Cots-Folch, and J.A. Martínez-Casasnovas. 2007. Effects of land terracing on soil properties in the Priorat region in Northeastern Spain: A multivariate analysis. Geoderma 142:51-261.         [ Links ]

Rezaei, S.A., and R.J. Gilkes. 2005. The effects of landscape attributes and plant community on soil physical properties in rangelands. Geoderma 125:145-154.         [ Links ]

Robertson, G.P. 1998. GS+: Geostatistics for the environmental sciences - GS+ users guide. Plain-well: Gamma Desing Software. 152 pp.         [ Links ]

Silva, P.C.M. da, and L.H.G. Chaves. 2001. Avaliafao e variabilidade espacial de fósforo, potássio e matéria orgánica em Alissolos. Revista Brasileira de Engenharia Agrícola e Ambiental 5:431-436.         [ Links ]

Theodoro, V.C.A., M.I.N. Alvarenga, R.J. Guimaraes, and C.A.S. Souza. 2003. Alterafòes químicas em solo submetido a diferentes formas de manejo do cafeeiro. Revista Brasileira de Ciencia do Solo 27:1039-1047.         [ Links ]

Vieira, S.R. 2000. Geoestatística em estudos de variabilidade espacial do solo. In: P.F. Novais, V.H. Alvarez and C.E.G.R. Schaefer (eds.). Tópicos em ciencia do solo. Vifosa: Sociedade Brasileira de Ciencia do Solo. 1:1-54.         [ Links ]

Warrick, A.W., and D.R. Nielsen. 1980. Spatial variability of soil physical properties in the field. In: D. Hillel (ed.) Applications of soil physics. New York: Academic Press. p. 319344.         [ Links ]

Webster, R. 2008. Soil science and geostatistics. In: P. Krasilnikov, F. Carré and L. Montanarella (eds.). Soil geography and geostatistics - concepts and applications. European Commission, Joint Research Centre, Institute for Environment and Sustainability. p. 1-11.         [ Links ]

Yemefack, M., D.G. Rossiter, and R. Njomgang. 2005. Multiscale characterization of soil variability within an agricultural landscape mosaic system in southern Cameroon. Geoderma 125:117-143.         [ Links ]

Received August 12, 2009. Accepted January 27, 2011.

Corresponding author:

Creative Commons License Todo el contenido de esta revista, excepto dónde está identificado, está bajo una Licencia Creative Commons