Canonical Correlation between the Leaf Quality Indicators of＂Moderate Aroma＂Flue-cured Tobacco

Lin MENG ，Yuangang DAI，Chengdong WANG，Shusheng WANG，W enjing SONG ，Yuanhua WU，Yim in XU*

1.Tobacco Research Institute of Chinese Academy of Agricultural Sciences，Qingdao 266101，China；2.Technological Center of China Tobacco Hunan Industrial Co.，Ltd.，Changsha 410007，China

1 Introduction

The tobacco quality evaluation indicator includes appearance quality，physical properties，chemical component，smoking quality and safety，and it is a complex and comprehensive concept.Various quality evaluation indicators consistof different interrelated individual evaluation indicators，and the level of balance and coordination determines the use value of tobacco in the cigarette industry［1-2］.The relationship between tobacco quality evaluation indicators has been great concern among tobacco researchers.Di Huihui etal.［3］and Wang Yujun etal.［4］study the correlation between the chemical component and physical properties of fluecured tobacco；Wei Chunyang et al.［5］and Huang Qingfen et al.［6］study the correlation between appearance quality and smoking quality of tobacco；Gao Jiahe et al.［7］，Bi Shufeng et al.［8］and Chang Aixia et al.［9］study the influence of chemical component of flue-cured tobacco on smoking quality；Yang Yingming et al.［10］and Cai Xianjie et al.［11］study the correlation between tobacco appearance quality and chemical component.Although there aremany results of research in thisarea，most researchersonly analyze the relationship between two or three groups，and the researchmethodsare limited to simple correlation［12-13］or regression analysis［14］. When researching the relationship between two groups of variables with different nature，we can employ the canonical correlation analysis which can fully reflect the intrinsic link between the variables.Bao Zichao et al.［15］study the relationship between different tobacco indicators（appearance quality，physical properties，chemical component and smoking quality），but the simple correlation analysis is still used as the method for research.Deng Xiaohua et al.［16］conduct the canonical correlation analysisof four quality evaluation indicators concerning the tobacco samples inmain flue-cured tobacco growing areas of Hunan Province，and the results show that it is feasible to use canonical correlation analysis for simplification of quality evaluation indicators.Although there aremany results of research in thisarea，few of people conduct comprehensive and systematic analysis of traditional"moderate aroma"flue-cured tobacco in main tobacco producing areas.In this study，with the 143 tobacco samples from China's typical"moderate aroma"tobacco producing areas as the materials，we evaluate the appearance quality and smoking quality and measure the physical properties and chemical component.Using canonical correlation analysis，we study themain indicators that affect the evaluation of flue-cured tobacco quality，in order to provide a reference for improving flue-cured tobacco quality evaluation system and revealing the relationship between fluecured tobacco quality evaluation indicators.

2 Materials and methods

2.1 M aterialsWe collected 143 C3F flue-cured tobacco samples in the typical"moderate aroma"flue-cured tobacco producing areas from 26 counties（cities）in Hunan Province，Hubei Prov-ince，Shandong Province，Liaoning Province，Guizhou Province，Shaanxi Province，Chongqing Province and Jilin Province in 2011.There were 4 to 7 samples in each county，and the weight of each samplewas about5 kg.Themain varietiesare Yunyan 87，K326.

2.2 Experimentalmethods

2.2.1Determination of physical indicators.The determination of physical indicators in this experiment is based on the method of Bao Zichao et al.［15］，including single leaf quality，leaf length，leafwidth，leaf thickness，leaf density，equilibrium moisture content and percentage of stem in tobacco leaf.

2.2.2Determination of appearance quality.The determination of tobacco appearance quality indicators is based on themethod of Bao Zichao et al.［15］，includingmaturity，oil content，structure，color，identity，and color intensity.

2.2.3Determination of chemical components.The determination of chlorogenic acid，rutin and scopoletin is based on themethod of Zhuang Yadong et al.［17］；the determination of total alkaloids is based on YC／T160-2002；the determination of total sugar and reducing sugar is based on YC／T159-2002；the determination of total nitrogen is based on YC／T161-2002；the determination of potassium ion is based on YC／T173-2002；the determination of chloride ion is based on YC／T162-2002；the determination of organic acids（oxalic acid，citric acid，palmitic acid，linoleic acid，oleic acid and hard fatty acid）isbased on themethod of Jin Yongming et al.，［18］.

2.2.4Evaluation of sensory quality.The evaluation standard for the sensory quality of flue-cured tobacco is based on Evaluation Methods for Sensory Quality and Style of Tobacco（Trial Version），developed by Zhengzhou Tobacco Research Institute of China National Tobacco Corporation.The specific scoring standard isbased on themethod of Deng Xiaohua et al.［19］.It is completed by the national judges organized by China Tobacco Hunan Industrial Co.，Ltd.，and the 0-5 equidistant scale scoring method is used.Themain evaluation indicators in this study include aroma status（5 points），smoke concentration（5 points），smoking strength（5 points），aroma quality（5 points），aroma volume（5 points），penetration（5 points），refinement degree（5 points），softness degree（5 points），mellowness（5 points），irritation（5 points），dryness（5 points），aftertaste（5 points）and miscellaneous odors（total scores of various odors）.

2.3 Data processingThe canonical correlation analysis is performed using SAS 9.2 statistical analysis software，and it is analyzed according to the linear expression of standardized canonical variables.

3 Results and analysis

3.1 Canonical correlation analysis of physical traits and smoking qualityThe canonical correlation analysis resultsof tobacco physical traits and smoking quality and the composition of canonical variables can be shown in Table1.The correlation coefficient of the first two canonical variables of physical characteristics and smoking quality reaches a significant level（P＜0.01），and the cumulative variance contribution rate reaches 64.13%，suggesting that there is a highly significant correlation between tobacco smoking quality and physical traits.From the linear expression of canonical variable composition，the amount of load on leaf weight（P1），leaf length（P5），aroma quality（S4）and aroma volume S5）in the first set of canonical variables is relatively large，mainly showing that the score of aroma quality and aroma volume is negatively correlated with leafweight but positively correlated with leaf length；the amount of load on leaf length（P5），smoke concentration（S2），penetration（S6）and softness degree（S8）in the second set of canonical variables is relatively large，mainly showing that the leaf length is negatively correlated with the score of smoke concentration，penetration and softness degree.

Table 1 Canonical correlation between physical traits and smoking quality of tobacco

3.2 Canonical correlation analysis of appearance qualityand smoking qualityThe correlation coefficients between appearance quality and the firstand second setof canonical variables of smoking quality reach a very significant level（P＜0.01）and a significant level（P＜0.05）respectively（Table 2），and the cumulative variance contribution rate of the two setsof canonical variables reaches68.76%，showing that there isa significant canonical correlation between tobacco appearance quality and sensory taste.As can be seen from the linear expression of canonical variables，there isa large load on smoke concentration（S2），smoking strength（S3），aroma volume（S5），color（A5）in the first set of canonical variables，mainly suggesting that color is negatively correlated with the score of smoke concentration but positively correlated with the score of smoking strength and aroma volume；there is a large load on maturity（A1），color（A6），smoking strength（S3）and irritation（S10）in the second set of canonical variables，suggesting thatmaturity is positively correlated with the score of irritation but negatively correlated with the score of smoking strength，and color intensity is negatively correlated with the score of irritation but positively correlated with the score of smoking strength.

Table 2 Canonical correlation between tobacco appearance quality and smoking quality

3.3 Canonical correlation analysis of appearance quality and physical traitsAs shown in Table 3，there is a significant correlation between physical traits and appearance quality of tobacco.The correlation coefficientof the first two sets of canonical variables reachesa very significant level（P＜0.01），and the correlation coefficient of the third setof canonical variables reaches a significant level（P＜0.05）.The cumulative variance contribution rate of three sets of canonical variables is up to 93.14%.As can be seen from the linear expression of canonical variables，there is a large load on oil content（A3）and leaf length（P5）in the first set of canonical variables，reflecting the positive correlation between oil contentand leaf length；there isa large load on identity（A4），structure（A5），color intensity（A6），leaf weight（P1）and leaf length（P5）in the second set of canonical variables，mainly suggesting that leaf weight is positively correlated with identity and color intensity but negatively correlated with organizational structure，and leaf length isnegatively correlated with identity and color intensity but positively correlated with organizational structure；there is a large load on maturity（A1），oil content（A3），color intensity（A6），leaf density（P2）and leaf width（P6）in the third set of canonical variables，mainly showing that the leaf density is positively correlated with maturity and oil content but negatively correlated with color intensity，and leaf width is positively correlated withmaturity and oil contentbutnegatively correlated with color intensity.

Table3 Canonical correlation between appearance quality and physical traits of tobacco

3.4 Canonical correlation analysis of chem ical componentand smoking qualityThe correlation coefficient of chemical component and the first two sets of canonical variables of smoking quality reaches a very significant level（P＜0.01），and the correlation coefficient of chemical component and the third set of canonical variables reaches a significant level（P＜0.05）.The cumulative variance contribution rate of the first three sets of canonical variables is69.17%，indicating that there is a significant correlation between smoking quality and chemical component of tobacco.From the linear expression of canonical variables（Table 4），it can be found that there is a large load on refinement degree（S7），irritation（S10），dryness（S11），total sugar（C1），total alkali（C4），ratio of total sugar to betaine（C7）and stearic acid（C17）in the firstsetof canonical variables，mainly suggesting that the score of smoke refinement degree and irritation is positively correlated with stearic acid and total sugar content but negatively correlated with total alkali content and ratio of total sugar to betaine，the score of dryness is negatively correlated with total sugar content and stearic acid content but positively correlated with total alkali content and ratio of total sugar to betaine；there is a large load on softness degree（S8），reducing sugar（C2）and potassium（C6）in the second set of canonical variables，show that the score of softness degree is negatively correlated with potassium content but positively correlated with reducing sugar content；there is a large load on irritation（S10）and ratio of total sugar to betaine（C7）in the third set of canonical variables，mainly showing the negative correlation between irritation score and ratio of total sugar to betaine.

Table 4 Canonical correlation between chem ical component and smoking quality of tobacco

3.5 Canonical correlation analysis of physical traits and chem ical componentThe correlation coefficient of physical traits and the first three sets of canonical variables of chemical component reaches a very significant level（P＜0.01），and the correlation coefficientof physical traitsand the fourth setof canonical variables reaches a significant level（P＜0.05）.The cumulative variance contribution rate of the first four sets of canonical variables reaches 88.72%，indicating that there is a significant canonical correlation between physical traits and chemical component of tobacco.From the linear expression of canonical variable composition，it can be found that there is a large load on percentage of stem in tobacco leaf（P3）and oxalic acid（C11）in the first setof canonical variables，mainly reflecting the positive correlation between percentage of stem in tobacco leaf and oxalic acid content；there is a large load on leaf density（P2），palmitic acid（C14）and oleic acid（C16）in the second set of canonical variables，indicating that leaf density is positively correlated with palmitic acid but negatively correlated with oleic acid；there is a large load on leaf density（P2），leaf length（P5），total sugar（C1），reducing sugar（C2）and palmitic acid（C14）in the third set of canonical variables，suggesting that leaf density and leaf length are positively correlated with total sugar content but negatively correlated with reducing sugar and palmitic acid content；there is a large load on leafweight（P1），leaf length（P5），total nitrogen（C3），ratio of total sugar to betaine（C7），scopoletin（C10）and palmitic acid（C14）in the fourth set of canonical variables，showing that leafweight is positively correlated with total nitrogen and ratio of total sugar to betaine but negatively correlated with scopoletin and palmitic acid content，and leaf length is negatively correlated with total nitrogen and ratio of total sugar to betaine butpositively correlated with scopoletin and palmitic acid content.

Table 5 Canonical correlation of physical traits and chem ical component of tobacco

3.6 Canonical correlation analysis of appearance quality and chem ical componentThe correlation coefficient of appearance quality and the first two canonical variables of chemical component reachesa very significant level（P＜0.01），and the cumulative variance contribution rate of the two is63.66%，indicating that there isa significant canonical correlation between appearance quality and chemical component of tobacco.From the linear expression of canonical variable composition（Table 6），it can be found that there is a large load on identity（A4），structure（A5），total sugar（C1），potassium（C6）and ratio of total sugar to betaine（C7）in the first set of canonical variables，mainly suggesting that the score of identity is negatively correlated with total sugar and potassium content but positively correlated with ratio of total sugar to betaine，and leaf organizational structure is positively correlated with total sugar and potassium contentbutnegatively correlated with ratio of total sugar to betaine；there is a large load on structure（A5），color（A6），total sugar（C1），total alkali（C4），palmitic acid（C14），linoleic acid（C15），oleic acid（C16）and stearic acid（C17）in the second setof canonical variables，mainly showing that leaf organizational structure is negatively correlated with total sugar，palmitic acid，total alkali and linoleic acid but positively correlated with oleic acid and stearic acid.

Table 6 Canonical correlation analysis of appearance quality and chem ical component of tobacco

4 Conclusions and discussions

In this paper，we analyze 43 individual indicators of four groups of tobacco quality evaluation indicators，and the results show that there are varying degrees of linkages between the quality indicators which affect the quality of tobacco.According to the cumulative contribution rate of significant canonical variables，we can judge the degree of correlation between them.From the cumulative variance contribution rate of significant canonical variables of smoking quality and the other three groups of quality evaluation indicators，chemical component（69.17%）＞appearance quality（68.76%）＞physical traits（64.13%），showing that chemical component is most closely related to smoking quality.Results indicate that physical traits are alsomost closely related to appearance quality（93.84%），and this result is consistentwith the findings of Deng Xiaohua et al.［19］in the producing areas of Hunan.Thus，in the same production area or different production areas，it is feasible to use the physical traitsof flue-cured tobacco leaf to reflect the appearance quality and use chemical component to reflect the intrinsic quality of tobacco.In addition，the individual indicators in various quality evaluation indicators also directly or indirectly affect the quality of tobacco to varying degrees.Leaf length and single leaf quality in tobacco physical traits are closely related to sensory taste，mainly affecting smoke concentration，penetration，softness degree，aroma quality and aroma volume in sensory taste indicators.Maturity and color in tobacco appearance quality indicators are strongly correlated with sensory taste，mainly affecting aroma volume，smoking strength，irritation and smoke concentration in smoking quality indicators.Tang Ruoyun et al.［20］study themiddle leaf of tobacco in tobacco producing areas of South Hunan，and also find the correlation between color and maturity，and aroma volume，consistentwith the findings in this paper.Total sugar，total alkali，ratio of total sugar to betaine，stearic acid，reducing sugar，potassium，ratio of total sugar to betaine in tobacco chemical component evaluation indicators have a great impact on sensory taste，mainly affecting irritation，softness degree，dryness and refinement degree in smoking quality indicators.Tissue structure，maturity，oil content，identity and color in tobacco appearance quality indicators are closely related to leaf length，leaf weight and leaf density in physical trait indicators.Percentage of stem in tobacco leaf，leaf density，leaf length and leaf weight in tobacco physical trait indicators are closely related to total sugar，reducing sugar，total nitrogen，ratio of total sugar to betaine，palmitic acid，oxalic acid，oleic acid and scopoletin in tobacco chemical component indicators.Structure and identity in tobacco appearance quality indicators are closely related to potassium，total sugar，ratio of total sugar to betaine，palmitic acid，linoleic acid，oleic acid and stearic acid in tobacco chemical component.In the evaluation of the quality of tobacco，the comprehensive analysis of all individual indicators concerning tobacco quality is time-consuming and not conducive to grasping the essence of the problem.Therefore，when testing the quality of tobacco，it is necessary to select some indicators that can reflect the quality of tobacco from a number of tobacco quality evaluation indicators，and the canonical correlation analysis in this paper just solves this problem.

［1］ZHU ZQ.On the usability of tobacco leaf and the safety of cigarette［J］.Tobacco Science＆Technology，2000（8）：3-6.（in Chinese）.

［2］ZHOU JH.Physiology and biochemistry of tobacco plants［M］.Hefei：Press of University of Science and Technology of China，1996.（in Chinese）.

［3］DIHH，SHIHZ.Differences in alkaloid contents and smoking quality among different leaf traits for flue-cured tobacco［J］.Chinese Agricultural Science Bulletin，2011，27（29）：85-91.（in Chinese）.

［4］WANGYJ，XIE SL，XING SH，et al.Study on correlation between flue cured tobacco leaf thickness and main chemical composition［J］.Chinese Tobacco Science，1997，1（1）：12-14.（in Chinese）.

［5］WEICY，LUO ZP，LIF，et al.Grey incidence analysis onmajor appearance traits and smoking quality of flue-cured tobacco［J］.Tobacco Science＆Technology，2010（10）：48-51.（in Chinese）.

［6］HUANG QF，ZHANG YJ.Analysis of correlation between smoking quality and appearance quality of tobacco leaf［J］.Acta Agriculturae Jiangxi，2011，23（11）：89-90.（in Chinese）.

［7］GAO JH，QINXY，TAN ZX，etal.Influenceofmain chemical components on smoking quality of flue-cured tobacco［J］.Journal of Mountain Agriculture＆Biology，2004，23（6）：497-501.（in Chinese）.

［8］BISF.Study on the corrections between the chemical components and the smoking quality of Yunnan flue-cured tobacco［J］.Journalof Huangshan U-niversity，2005，7（3）：61-63.（in Chinese）.

［9］CHANG AX，DU YM，FUQJ，etal.Correlationship betweenmain chemical components and sensory quality of flue-cured tobacco［J］.Chinese Tobacco Science，2009，30（6）：9-12.（in Chinese）.

［10］YANG YM，LOUHY，WANGC，etal.Canonical correlation analysis between appearance quality and chemical components［J］.Journal of Yuxi Teachers College，2009，25（8）：23-29.（in Chinese）.

［11］CAIXJ，WANGXM，YINQS.Study on the quantitative relationship betweenmaturity and quality of tobacco leaf［J］.Acta Tabacaria Sinica，2005，11（4）：42-46.（in Chinese）.

［12］DU YM，GUOCF，ZHANGHB，etal.Study on relationship between content ofwater soluble sugar，alkaloid，total nitrogen and taste quality of flue cured tobacco［J］.Chinese Tobacco Science，2000（1）：9-12.（in Chinese）.

［13］DENG XH，ZHOU JH，CHEN XL，et al.Correlation analysis on quality evaluating indexes in tobacco leaf［J］.Acta Tabacaria Sinica，2008，14（2）：1-8.（in Chinese）.

［14］WANG XQ，DENGXH，LIXZ，etal.Correlation，path and regression analysis between chemical components and tar in flue-cured tobacco leaves in Hunan［J］.Crops，2010（2）：32-35.（in Chinese）.

［15］BAO ZC，SONGWJ，XU YM，et al.Correlation analysis among quality indices in upper leaves of flue-cured tobacco［J］.Chinese Tobacco Science，2013，34（5）：23-27.（in Chinese）.

［16］DENG XH，ZHOUQM，ZHOU JH，et al.Canonical correlation analysis between quality evaluating indices of tobacco leaf［J］.Acta Tabacaria Sinica，2011，17（3）：17-22.（in Chinese）.

［17］ZHUANG YD，ZHANG Y，WANG F，et al.Analysis of polyphenols in cigarettes［J］.Tobacco Science＆Technology，2004（1）：23-26.（in Chinese）.

［18］JIN YM，ZHANG MF，LIU BZ.Analysis of polybasic and higher fatty acids in tobacco［J］.Tobacco Science＆Technology，2002（4）：21-24.（in Chinese）.

［19］DENG XH，YANG LL，LU ZS，et al.Sensory evaluation on style and characteristics of tobacco leaves from Xiangxi［J］.Acta Tabacaria Sinica，2013，19（5）：22-27.（in Chinese）.

［20］TANG RY，ZHAO AJ，DENG X.Grey incidence analysis on major appearance traits and concentration and quality of aroma in flue-cured tobacco in Southern Hunan Province［J］.Crop Research，2013，27（6）：561-563.（in Chinese）.