Water Science and Engineering, 2009, 2(2): 79-86

doi:10.3882/j.issn.1674-2370.2009.02.009
http://kkb.hhu.edu.cn
e-mail: [email protected]

Study on soil water characteristics of tobacco fields based

on canonical correlation analysis
Xiao-hou SHAO1, 2, Yu WANG*1, 2, Li-dong BI1, 2, You-bo YUAN3,
Xian-kun SU3, Jian-guo MO4
1. College of Agricultural Engineering, Hohai University, Nanjing 210098, P. R. China
2. Key Laboratory of Efficient Irrigation-Drainage and Agricultural Soil-Water Environment in Southern
China, Ministry of Education, Nanjing 210098, P. R. China
3. Guizhou Tobacco Science Research Institute, Guiyang 550003, P. R. China
4. Guizhou Institute of Mountainous Environment and Climate, Guiyang 550002, P. R. China

Abstract: In order to identify the principal factors influencing soil water characteristics (SWC) and
evaluate SWC effectively, the multivariate-statistical canonical correlation analysis (CCA) method
was used to study and analyze the correlation between SWC and soil physical and chemical
properties. Twenty-two soil samples were taken from 11 main tobacco-growing areas in Guizhou
Province in China and the soil water characteristic curves (SWCC) and basic physical and chemical
properties of the soil samples were determined. The results show that: (1) The soil bulk density, soil
total porosity and soil capillary porosity have significant effects on SWC of tobacco fiels. Bulk
density and total porosity are positively correlated with soil water retention characteristics (SWRC),
and soil capillary porosity is positively correlated with soil water supply characteristics (SWSC). (2)
Soil samples from different soil layers at the same soil sampling point show similarity or consistency
in SWC. Inadequate soil water supply capability and imbalance between SWRC and SWSC are
problems of tobacco soil. (3) The SWC of loamy clay are generally superior to those of silty clay loam.
Key words: canonical correlation analysis; tobacco soils; soil water characteristics; soil texture

1 Introduction
China now has the largest cultivated area and yield of tobacco in the world. As one of the
main cash crops, tobacco plays a very important role in China’s national economy. In Guizhou
Province, which is the second largest high-quality tobacco-producing province in China,
flue-cured tobacco production has become an important industrial mainstay and is significant
to the development of the regional economy. SWC result from combinations of physical and
chemical properties of soil (Saxton et al. 1986; Kern 1995; Rajkai et al. 1996; Keller et al.
2007; Lipiec et al. 2007). They affect soil productivity (Sifola and Postiglione 2002; Çakir and
Çebi 2006), yield, and the quality of flue-cured tobacco (Tonello et al. 1993; Maw et al. 1997;
Bilalis et al. 2009), and are the premise and basis of management of irrigation and drainage for
üüüüüüüüüüüüü
This work was supported by the National Key High-Tech Program (863) of China (Grant No. 2006AA10Z271)
and the Key Project of the Guizhou Tobacco Monopoly Administration (2007-7).
*Corresponding author (e-mail: [email protected])
Received Mar. 18, 2009; accepted May 9, 2009
flue-cured tobacco (Tonello et al. 1993; Maw et al. 1997). It is therefore important to
understand the correlation between SWC and soil physical and chemical properties. In the
present study, basic soil physical and chemical factors influencing SWC were identified using
the CCA method, and comprehensive SWC evaluation of tobacco soils was carried out
according to the canonical variables. The study aims to provide a basis for improving soil
water conditions, formulating reasonable plans and measures for irrigation and draught, and
further enhancing economic and environmental benefits of flue-cured tobacco production in
Guizhou Province in China.

2 Materials and methods

2.1 Basic soil properties determination
According to on-site survey and study, 11 tobacco-growing areas where flue-cured
tobacco had been widely planted for many years and the yield and quality of flue-cured
tobacco was relatively high were selected as the soil sampling points. Undisturbed soils from
layers at 0-30 cm and 30-60 cm depths were used as experimental soil. Basic physical and
chemical properties of soil samples were measured and the soil texture was analyzed (Table 1)
according to the conventional methods (Sparks et al. 1996; Dane and Topp 2002).
Table 1 Selected soil properties of 11 tobacco-growing areas in Guizhou Province, China
0-30 cm 30-60 cm
Sampling Organic Bulk Total Capillary
Clay Organic
Bulk Total Capillary
Clay
No.
point matter density porosity porosity content Soil matter
density porosity porosity
content Soil
content (<2μm) texture content (<2μm) texture
(g/cm3) (m3/m3) (m3/m3) (g/cm3) (m3/m3) (m3/m3)
(g/kg) (g/g) (g/kg) (g/g)
Silty Silty
1 Tongzi 33.7 1.33 0.51 0.20 0.22 clay 30.6 1.38 0.49 0.18 0.24 clay
loam loam
Silty Silty
2 Sinan 9.0 1.42 0.48 0.15 0.34 3.5 1.31 0.51 0.16 0.40
clay clay
Loamy Loamy
3 Kaili 18.5 1.39 0.48 0.14 0.44 11.3 1.43 0.48 0.18 0.44
clay clay

Loamy Silty
4 Hezhang 37.3 1.27 0.53 0.17 0.36 29.3 1.46 0.46 0.17 0.24 clay
clay
loam
Clay Loamy
5 Shuicheng 49.9 1.04 0.61 0.21 0.24 43.6 1.35 0.51 0.21 0.32
loam clay
Silty Silty
6 Fenggang 25.5 1.10 0.58 0.29 0.24 clay 13.7 1.45 0.49 0.24 0.15 clay
loam loam
Loamy Loamy
7 Xiuwen 21.5 1.18 0.55 0.19 0.40 12.3 1.46 0.49 0.19 0.40
clay clay
Sandy Sandy
8 Fuquan 19.6 1.52 0.43 0.17 0.20 clay 11.3 1.46 0.45 0.19 0.16 clay
loam loam
Loamy Loamy
9 Anshun 30.0 1.26 0.53 0.18 0.40 17.0 1.18 0.56 0.17 0.36
clay clay
Clay Silty
10 Dafang 41.0 1.23 0.54 0.15 0.24 38.4 1.35 0.50 0.17 0.24 clay
loam
loam
Silty Silty
11 Zunyi 26.5 1.46 0.48 0.18 0.24 clay 24.0 1.35 0.49 0.17 0.24 clay
loam loam

80 Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86
2.2 SWCC determination
A soil water characteristic curve is the curve of soil matrix suction against soil water
content. It reflects internal relations between energy and quantity of soil water and is an
important tool for studying SWRC and SWSC. In the present study, the drying SWCC of 22
soil samples from tobacco fields in the suction range of 0.1-15 bar were measured in the
laboratory with the pressure membrane method. Experimental data were fitted using the
Gardner model (Gardner et al. 1970a, 1970b):
T =AS  B (1)
where T is the soil volumetric moisture content, S is the soil matrix suction, and A and B are
the fitting parameters (Table 2). Table 2 shows that determination coefficients (r2) are almost
greater than 0.940. The Gardner model provides a good fit with the measured data.
Table 2 Fitting results of SWCC
0-30 cm 30-60 cm
No. Sampling point
A B AB r2 A B AB r2
1 Tongzi 0.270 0.126 0.034 0 0.977 0.265 0.115 0.030 5 0.975
2 Sinan 0.352 0.072 0.025 3 0.945 0.348 0.076 0.026 4 0.955
3 Kaili 0.332 0.070 0.023 2 0.943 0.341 0.090 0.030 7 0.950
4 Hezhang 0.314 0.090 0.028 3 0.934 0.345 0.083 0.028 6 0.940
5 Shuicheng 0.264 0.128 0.033 8 0.971 0.381 0.870 0.331 5 0.955
6 Fengang 0.301 0.160 0.048 2 0.977 0.331 0.117 0.038 7 0.963
7 Xiuwen 0.287 0.108 0.031 0 0.954 0.366 0.086 0.031 5 0.932
8 Fuquan 0.233 0.128 0.029 8 0.992 0.214 0.156 0.033 4 0.967
9 Anshun 0.364 0.080 0.029 1 0.989 0.442 0.064 0.028 3 0.981
10 Defang 0.280 0.095 0.026 6 0.964 0.328 0.088 0.028 9 0.957
11 Zunyi 0.336 0.088 0.029 6 0.977 0.346 0.078 0.027 0 0.960

2.3 Data analysis method

In natural science and social science research, the correlation between two sets of
variables is very complex and variable information often overlaps and intersects. Assuming
that x1 , x2 , x3 , , x p and y1 , y2 , y3 , yq are two sets of variables correlated with each other,
p×q correlation coefficients are calculated in p×q analytical processes of paired variables
using conventional correlation analysis methods. Obviously, it is difficult to get to the essence
of things. CCA is similar to principal component analysis, namely, the study of correlation
between representative composite indicators (linear combinations of variables) instead of two
sets of variables. This is more effective in illustrating the complex and diverse relationships
between two sets of variables (Barnett and Preisendorfer 1987; Barnston and Ropelewski 1992;
Melzer et al. 2003; Yang et al. 2006). In the present study, CCA was conducted with the
computer program SPSS.
3 Results and analysis
3.1 Canonical variable extraction
The SWCC are dynamic. When using the SWCC for research on SWC, people seek some

Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86 81
reasonable static parameters to reflect this dynamic process (Zhao et al. 1989). At present, soil
water content and specific water capacity at 1 bar (soil suction threshold value for normal crop
growth) are often used as static indices to evaluate SWRC and SWSC (Shen and Xi 1990; Liu
and Dong 1997).
Two sets of variables were constructed according to the data shown in Table 1 and
Table 2. The first set represents SWC: a static index of SWRC (X1), and a static index of
SWSC (X2). The second set represents basic soil physical and chemical properties: organic
matter content (Y1), bulk density (Y2), total porosity (Y3), capillary porosity (Y4), and clay
content (< 2μm) (Y5). CCA of the two sets was carried out. The results showed that the two
pairs of canonical variables extracted from the two sets of variables were all statistically
significant (P < 0.05). Canonical correlation coefficients (R) of the first and second pair of
canonical variables were 0.990 (R1) and 0.669 (R2), respectively (Table 3). Two pairs of
canonical variables were constructed according to the standardized coefficients of variables
shown in Table 4:
U1 =0.088 X 1 +1.108X 2 (2)
V1 0.033Y1  0.015Y2  0.022Y3  0.988Y4  0.051Y5 (3)
U2 1.027 X 1  0.166 X 2 (4)
V2 0.312Y1  2.618Y2  2.916Y3  0.023Y4  0.525Y5 (5)
where U1 and U2 represent the first set of canonical variables of SWC, and V1 and V2 represent
the second set of canonical variables of basic soil physical and chemical properties.
Table 3 R and significance test

F
2
Pair number R Degree of freedom Probability value P
1 0.990 76.274 10 0
2 0.669 10.082 4 0.039

Table 4 Standardized coefficients of variables

Pair number X1 X2 Y1 Y2 Y3 Y4 Y5
1 0.088 1.108 0.033 0.015 0.022 0.988 0.051
2 1.027 0.166 0.312 2.618 2.916 0.023 0.525

3.2 Structure analysis of canonical variables

For the first pair of canonical variables, the variables that have the largest absolute values
of coefficients are X2 in U1 and Y4 in V1. This indicates that U1 is mainly dependent on the
factor X2 and V1 is mainly dependent on the factor Y4. Besides, the coefficients of X2 and Y4 are
both positive. This shows that there is a positive correlation between SWSC and capillary
porosity. Coefficients of Y1, Y2, Y3, and Y5 are all negative and their absolute values are very
small. This indicates that soil organic matter content, bulk density, total porosity, and clay
content (< 2μm) are negatively correlated with SWSC and have little influence on SWSC.
In the second pair of canonical variables, the coefficient of X1 has the maximum absolute

82 Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86
value in U2. This indicates that U2 is mainly dependent on the factor X1. Absolute values of the
coefficients of Y2 and Y3 are relatively larger in V2. This illustrates that V2 is chiefly dependent
on the factors of soil bulk density and total porosity. Moreover, the coefficients of X1, Y2 and
Y3 are all positive. This indicates that bulk density and total porosity are positively correlated
with SWRC. Similarly, SWRC have a negative relation with soil organic matter content and
capillary porosity, but a positive relation with clay content (< 2μm). Y1, Y4 and Y5 all have little
influence on SWRC.
In addition, as shown in Table 3, R1 (0.990) > R2 (0.669), and R1 (P = 0) is more statistically
significant than R2 (P = 0.039). This indicates that SWSC are more closely related with the
influencing factor Y4, but some factors influencing SWRC other than Y2 and Y3 still play a role
that cannot be ignored.

3.3 SWC evaluation of tobacco soils

V1 and V2 are canonical variables of basic soil physical and chemical properties that
comprehensively reflect the SWC. Thus, scores of V1 and V2 can be used to evaluate SWC of
tobacco soils. A scatter diagram of 22 soil samples was made with the score of V1 being the
abscissa and the score of V2 the vertical coordinate (Fig. 1, in which ai  i 1, 2, ,11
represents soil samples from the layer at 0-30 cm depth and bi  i 1, 2, ,11 represents soil
samples from the layer at 30-60 cm depth). Due to the positive correlation between the
determinants (Y4, Y2 and Y3) of V1 and V2 and SWC, points located to the right of the V2-axis
are considered to be favorable SWSC, while points located to the left of the V2-axis are
considered to be poor SWSC. Similarly, points above the V1-axis are considered to be
favorable SWRC, but points below the V1-axis are considered to be poor SWRC.

Fig. 1 Scatter diagram of 22 soil samples

As shown in Fig. 1, of the total 22 points, 12 (more than half) are located in the second
and forth quadrants and only four are located in the first quadrant. In other words, there are
more soil samples with relatively favorable SWRC (points in the second quadrant) or SWSC
(points in the fourth quadrant) than with both relatively favorable SWRC and SWSC (points in
the first quadrant). This suggests that there is a serious imbalance of SWC in the main
tobacco-growing areas in Guizhou Province in China. Further observation shows that two

Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86 83
points ( ai and bi ) from the same sampling location are in adjacent quadrants (both points are
located to one side of the V1-axis or V2-axis) or in the same quadrant, but not in diagonally
situated quadrants. This indicates that the SWC of soil samples from the layers at 0-30 cm and
30-60 cm depths at the same sampling point are alike or consistent. In addition, there are more
points located to the left side of the V2-axis than to the right side of the V2-axis, with a ratio of
1.75/1. This means that most tobacco soils in Guizhou Province in China have inadequate soil
water supply capability.
With regard to the soil texture, loamy clay soil samples are located in each quadrant
except for the third (with a proportion of 3/4 in the first quadrant, 5/8 in the second quadrant,
and 1/4 in the fourth quadrant), but silty clay loam soil samples are located in all four
quadrants (with a proportion of 1/4 in the first quadrant, 1/8 in the second quadrant, 4/6 in the
third quadrant, and 2/4 in the fourth quadrant). Conjoint analysis of the proportion in each
quadrant shows that loamy clay has favorable SWRC (loamy clay sampling points make up a
large proportion of points above the V1-axis), but silty clay loam has poor SWRC (silty clay
loam points make up a large proportion of points below the V1-axis). In addition, the great
majority of soil samples (16 out of all 22 sampling points) are loamy clay and silty clay loam,
so there are great differences in the SWRC of the two types of soil texture. Further analysis of
the proportion of the two main types of soil texture in each quadrant shows that loamy clay
soil samples make up the largest proportion in the first quadrant (3/4) and silty clay loam soil
samples make up the largest proportion in the third quadrant (4/6). In general, loamy clay is
better than silty clay loam in terms of balance between SWRC and SWSC. Other types of soil
texture make up small proportions of all the sampling points (clay loam: 2/22; silty clay: 2/22;
and sandy clay loam: 2/22). They are not typical or representative, so we need no further
analysis of them.

4 Conclusions
(1) The Gardner model can be used to describe the drying SWCC of tobacco soils in the
suction range of 0.1-15 bar in Guizhou Province in China. Generally, the suction range of
0.1-15 bar is also the suction interval of total available soil water, so the Gardner model
constructed in this paper can provide a theoretical support for rational utilization of soil water
resources and management of irrigation and drainage for flue-cured tobacco.
(2) Through the CCA, the internal relations between SWC and basic physical and
chemical properties of soil samples were effectively presented, and soil bulk density, total
porosity and capillary porosity were determined to be principal factors influencing SWC. Soils
in the layers at 0-30 cm and 30-60 cm depths in the same tobacco-growing area were found to
have similar or consistent SWC, which accords with the actual conditions of tobacco soils.
(3) There are serious problems related to an imbalance between SWRC and SWSC and
inadequate soil water supply capability in the main tobacco-growing areas in Guizhou

84 Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86
Province in China. Texture analysis shows that loamy clay and silty clay loam are the main
soil textures in the main tobacco-growing areas, but the SWC of these two soil textures are
very different from one another. The SWC of loamy clay are generally superior to those of
silty clay loam. This research can provide a theoretical support for development and
evaluation of high-quality tobacco production areas.
(4) Because of the complexity of the factors influencing SWC, this evaluation model did
not take into consideration some other factors of SWRC. This may influence the precision of
the results. Improvements should be made in future studies.

References
Barnett, T. P., and Preisendorfer, R. 1987. Origins and levels of monthly and seasonal forecast skill for United
States surface air temperatures determined by Canonical correlation analysis. Monthly Weather Review,
115(9), 1825-1850. [doi:10.1175/1520-0493(1987)115<1825:OALOMA>2.0.CO;2]
Barnston, A. G., and Ropelewski, C. F. 1992. Prediction of ENSO episodes using canonical correlation
analysis. Journal of Climate, 5(11), 1316-1345. [doi:10.1175/1520-0442(1992)005<1316:POEEUC>
2.0.CO;2]
Bilalis, D., Karkanis, A., Efthimiadou, A., Konstantas, Ar., and Triantafyllidis, V. 2009. Effects of irrigation
system and green manure on yield and nicotine content of Virginia (flue-cured) Organic tobacco
(Nicotiana tabaccum), under Mediterranean conditions. Industrial Crops and Products, 29(2), 388-394.
[doi:10.1016/ j.indcrop.2008.07.007]
Çakir, R., and Çebi, U. 2006. Growth and dry matter accumulation dynamics of flue-cured tobacco under
different soil moisture regimes. Journal of Agronomy, 5(1), 79-86. [doi: 10.3923/ja.2006.79.86]
Dane, J. H., and Topp, G. C. 2002. Methods of Soils Analysis. Part 4: Physical Methods. Madison: Soil
Science Society of America.
Gardner, W. R., Hillel, D., and Benyamini, Y. 1970a. Post-irrigation of soil water: I. Redistribution. Water
Resources Research, 6(3), 851-861.
Gardner, W. R., Hillel, D., and Benyamini, Y. 1970b. Post-irrigation of soil water: II. Simultaneous
redistribution and evaporation. Water Resources Research, 6(4), 1148-1153.
Keller, T., Arvidsson, J., and Dexter, A. R. 2007. Soil structures produced by tillage as affected by soil water
content and the physical quality of soil. Soil & Tillage Research, 92(1-2), 45-52. [doi:10.1016/j.still.
2006.01.001]
Kern, J. S. 1995. Evaluation of soil water retention models based on basic soil physical properties. Soil
Science Society of America Journal, 59(4), 1134-1141.
Lipiec, J., Walczak, R., Witkowska-Walczak, B., Nosalewicz, A., SáowiĔska-Jurkiewicz, A., and SáawiĔski, C.
2007. The effect of aggregate size on water retention and pore structure of two silt loam soils of different
genesis. Soil and Tillage Research, 97(2), 239-246. [doi:10.1016/j.still.2007.10.001]
Liu, E. B., and Dong, S. L. 1997. Evaluation on water capacity and antidrought of main soils in Loess Plateau.
Bulletin of Soil and Water Conservation, 17(7), 20-26. (in Chinese)
Maw, B. W., Stansell, J. R., and Mullinix, B. G. 1997. Seasonal soil water use of flue-cured tobacco. Tobacco
Science, 41(4), 91-93.
Melzer, T., Reiter, M., and Bischof, H. 2003. Appearance models based on kernel canonical correlation
analysis. Pattern Recognition, 36(9), 1961-1971. [doi:10.1016/S0031-3203(03)00058-X]
Rajkai, K., Kabos, S., van Genuchten, M. Th., and Jansson, P.-E. 1996. Estimation of water-retention
characteristics from the bulk density and particle-size distribution of Swedish soils. Soil Science, 161(12),
832-845.

Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86 85
Saxton, K. E., Rawls, W. J., Romberger, J. S., and Papendick, R. I. 1986. Estimating generalized soil-water
characteristics from texture. Soil Science Society of America Journal, 50(4), 1031-1036.
Shen, S. Y., and Xi, C. F. 1990. Soil water-retention and its relationship with particle size distribution in north
Anhui Province. Acta Pedologica Sinica, 27(1), 34-42. (in Chinese)
Sifola, M. I., and Postiglione, L. 2002. The effect of increasing NaCl in irrigation water on growth, gas
exchange and yield of tobacco Burley type. Field Crops Research, 74(1), 81-91. [doi:10.1016
/S0378-4290(01)00202-7]
Sparks, D., Page, A., Helmke, P., and Loeppert, R. 1996. Methods of Soil Analysis. Part 3: Chemical Methods.
Madison: Soil Science Society of America.
Tonello, P. E., Pregno, L. M., and Weeden, B. R. 1993. Factors influencing the irrigation efficiency of
flue-cured tobacco grown in north Queensland, Australia. Tobacco Science, 37, 69-77.
Yang, Y., Mekki, D. M., Lü, S. J., Yu, J. H., Wang, L. Y., Wang, J. Y., Xie, K. Z, and Dai, G. J. 2006. Canonical
correlation analysis of body weight, body measurement and carcass characteristics of Jinghai yellow
chicken. Journal of Animal and Veterinary Advances, 5(11), 980-984.
Zhao, W. S., Meng, C. F., and Shen, J. G. 1989. Study on the water-retention characters of six red soils in
Jinhua Prefecture of Zhejiang Province. Journal of Zhejiang University (Agriculture and Life Science),
15(4), 377-382. (in Chinese)

86 Xiao-hou SHAO et al. Water Science and Engineering, Jun. 2009, Vol. 2, No. 2, 79-86