© The Authors, published by EDP Sciences, 2023

1 Introduction

Boğazkere grapes are one of the primary indigenous red wine varieties. Turkey is one of the earliest regions where Vitis vinifera was domesticated and is the origin of numerous grape varieties [1]. Boğazkere (BG) varieties produce one of the most widely consumed and well-known red wines in Turkey. It is a medium-sized, round grape with thick skin that naturally originates from Eastern Anatolia such as Diyarbakır and Elazığ plateaus [2]. However, the variety has also lately begun to be planted in Western Anatolia's Güney and Pendora districts on the Denizli and Manisa plateaus, respectively.

The term of wine quality is complex and hard to evaluate and varies between vineyard locations. Still, numerous studies have been conducted to identify the factors that influence quality [3,4]. Aroma is a crucial factor that influences the character of wine. The presence of numerous aroma compounds contributes to the complexity of wine. Due to this complexity, the wine's quality increases [5]. The importance of aroma in sensory perception, which plays an essential part in wine consumer preference, cannot be overstated. One of the most notable characteristics of aroma compounds is that they may have a significant impact on quality despite being perceptible even in small amounts [6].

Numerous research [5,7,8] have been conducted to figure out the specific aromas components that give each wine its unique flavour. In some instances, a single aroma compound can impart an odour to the entire wine, while in other instances, a single odour can be caused by multiple aroma compounds [9]. The aroma components of wines are identified and quantified using gas chromatography-mas spectrophotometry [10]. Furthermore, descriptive analysis is widely used for defining the sensory attributes of wine [10-13]. Therefore, the wine's distinctive characteristics are determined. Due to the association between aroma compounds and sensory attributes, characterization of the wine's aroma provides insight into which aroma compounds are associated with particular attributes. The results of instrumental analysis should be associated with those of sensory evaluation and supported by chemometric methods. So, chemometric approaches may be very beneficial in defining the quality category or geographical indication of a wine by proving its distinctive aroma composition and sensory descriptors [14]. Numerous studies [15-18] have utilized chemometric techniques such as PCA (principal component analysis) and PLSR (Partial Least Squares Regression) to investigate the relationship between analytical results and sensory perception and to enhance modelling.

In recent years, Boğazkere wines have evolved into the trademark of Turkey internationally. But, there are a limited number of studies that have linked the aroma characterization and sensory evaluation of Boğazkere wines [2]. The objectives of the research were aimed at (1) characterize the aroma compounds, aroma-active compounds of Boğazkere wines throughout different grape growing locations, and (2) describe the relationship between the sensory attributes and aroma components of these wines.

2 Material and methods

2.1 Samples

Twelve Boğazkere red wines from three distinct locations were collected. They were selected from the foremost wine company in Turkey based on their premium quality, single variety, and limited grape yields, free-run and aging potential. The sampling consisted of three replications. All of the supplied wine samples are included in Table 1, along with additional information.

2.2 Chemical analysis

Alcohol, density, total acidity, pH, dry matter, total phenolic compounds, antosiyanin, tannin, and total and free sulphur dioxide were analysed [19]. Table 1 provides the details of the composition of the wines.

2.3 Aroma Compounds Analysis by Solid Phase Extraction (SPE)

The solid phase extraction (SPE) techniques provided by [20]. were used for analysing aroma compounds. This analysis was performed three times. 50 mL of wine containing an internal standard (3000 g/mL 4-nanonol) was run through a 200 mg LiChrolut EN (Merck, Darmstadt, Germany) cartridge at a flow rate of approximately 2 mL/min. After the flow of 2 millilitres of water, the absorbent was dried out under air (-0.6 bar, 10 minutes). The components were recovered through elution via 1.6 mL of dichloromethane 1% methanol. The extract was subsequently analysed with GC/MS/FID. GC/MS/FID conditions have been completely described in a previous study [13]. The aroma compounds were identified by comparing their retention index and mass spectra on the DB-Wax column with those of a commercial spectra database (W10N14, NIST11, NBS75k) and the instrument's internal library, which was compiled from previous laboratory studies. The injection of chemical standards into the GC-MS system confirmed a number of the identified compounds. Utilizing an n-alkane series, the retention indices of the compounds were determined. After identifying aroma compounds, the internal standard method was used to quantify the aroma compounds [13,21]. The ratio of peak area was corrected using response factors for each compound, which were calculated using the intensity ratio of each compound to the internal standard. Then, the mean values of GC analyses conducted in triplicate were calculated.

2.4 Aroma-active compounds analysis by Olfactometry and mass spectrometry

Purge-trap systems applying the dynamic headspace approach [22,23]. were used to extract the aroma-active compounds of wine. In the extraction, a cartridge containing 400 mg of LiChrolut EN resin with an inner diameter of 0.8 cm and an inner volume of 3 ml was used. This resin was selected because it can effectively extract aroma compounds from wine [20]. 80 ml of wine sample was placed in the 37 °C-heated glass container and whirled with a magnetic stirrer prior to the cartridge being set on top of the glass container. 500 ml/min of nitrogen gas was then passed through the system for 100 minutes. In this procedure, the cartridge retains the aroma-active compounds released in the headspace of the glass container. 3.2 ml of dichloromethane containing 5% methanol was then passed through the cartridge to elute the retained aroma compounds. The extract was then concentrated to 200 µL under a flow of purified nitrogen gas. The extract was then utilized to generate the olfactometric profile. The extracted samples were analysed by GC/MS-Olfactometry. GC/MS-O conditions have been entirely described in a previous study [13].

2.4.1 Olfactory Sniffing

Six trained assessors who were a part of the research staff and ranged in age from 23 to 55 conducted the olfactometric analysis. To prevent fatigue, each olfactometry analysis was carried out in two sessions. The sniffing duration for each assessor was maintained at approximately 20 minutes per session. The assessors rated the overall intensity of each odour on a seven-point scale (including half values: 0 = undetectable; 1 = weak, scarcely discernible odour; 2 = clean but not intense odour; and 3 = intense odour). Ferreira et al. [24] have validated the quantitative sufficiency of this technique. Modified Frequency (%) values were calculated from the scores (detection intensity and frequency) to determine the aroma-active compounds. The modified frequency value was computed utilizing the formula suggested by Dravnieks [25].

$$MF(\%)\text{\hspace{0.17em}}=\text{}\sqrt{F(\%)xI(\%)}$$(1)

In this formula, F (%) represents the percentage of an odorant's detection frequency, and I (%) represents the average intensity as a percentage of the maximum intensity. Subsequently, these compounds were identified by comparing their retention index and mass spectra on the DB-Wax column with those of a commercial spectra database (W10N14, NIST11, NBS 75k) and the instrument's internal library, which was created during the preceding research. The injection of chemical standards into the GC-MS system confirmed a number of the identified compounds.

2.5 Quantitative descriptive analysis (QDA)

Quantitative Descriptive Analysis (QDA), as described by [26], were evaluated by research staff, consisting of eight females and four males aged 25-55. Academic personnel have a minimum of 200 hours of training and experience in descriptive analysis, and they routinely participate in the university's sensory evaluation of research. The assessors participated in seven 2-hour sessions. Assessors participated in an open session to discuss and develop descriptors for the wines during the first and second sessions. According to the open session, the reference standards for each attribute were developed for the following sessions. Assessors presented prepared reference standards during the third and fourth sessions, and a discussion on the reference standards took place during the entire session. In the fifth session, one visual, three taste, four mouthfeel, five aroma and six flavour attributes were selected using a consensus-based approach to define the attributes for descriptive analysis. The sixth, and seventh sessions, assessors used a 15-cm scale to evaluate the intensity of each attribute. Wine (30 mL at 20 °C) was served in random order in International Standard Organization (ISO) wine glasses covered with petri dishes during sessions.

2.6 Statistical analysis

The aromatic constituents that are quantified in the wines were subjected to a two-way Anova analysis using XLSTAT software (Addinsoft version 2021.2, France) to determine the effect of both location and year. Year and location interactions were also evaluated. The Tukey’s HSD pairwise comparison test was then used to determine the statistical significance of the differences. A principal component analysis was performed on the sensory data to show the outcome. To determine the relationship between chemical and sensory evaluation results, partial least squares regression (PLSR) analysis was performed with the XLSTAT software (Addinsoft version 2021.2, France) and the sensory descriptors associated with these compounds revealing the characteristics of the wines were identified.

3 Results and Discussion

3.1 Aroma compounds of Boğazkere wines

The aroma characteristics of Boğazkere wines from three different locations and different vintages were evaluated. Table 2 compares the minimum, maximum, effect of the location, and location-year interaction of the aroma compositions of Boğazkere wines from the locations of Diyarbakır, Manisa, and Denizli. In generally, 93 aroma compounds were identified in Boğazkere wines, including 19 higher alcohols, 27 ester, 7 lactones, 20 volatile phenols, 1 norisoprenoids, 2 terpenes, 9 carbonyl compounds, and 8 volatile acids. The average total amount of aroma substances is 162.7 mg/L in Diyarbakır, 148.3 mg/L in Manisa, and 142.67 mg/L in Denizli, based on the locations. Comparing the total number and amount of aroma compounds identified and quantified in the locations, it was discovered that Diyarbakır wines contained a greater number and amount of aroma compounds. Cabaroglu et al. [2] identified 40 aroma constituents in Boğazkere wines and reported the total aroma substance concentration to be 164.3 mg/L.

The esters are the most significant group of aroma substances in wine. Esters, which are the secondary products of yeasts produced during the fermentation of ethyl alcohol, are generally responsible for the wine's fruity aroma and play a vital role in determining the sensory characteristics of young wines. Although esters are hydrolysed over time, the amount of major esters in aged wines exceeds the detection thresholds [27,28]. Boğazkere wines have been identified with 27 ester compounds. Maximum ester amounts have been determined at 55.7 mg/L in Diyarbakır, 46.9 mg/L in Manisa, and 40.5 mg/L in Denizli. Ethyl octanoate, ethyl hexanoate, ethyl butanoate, and ethyl decanoate were the most abundant ethyl esters of fatty acids that affect the fruit aroma of red wines in Boğazkere wines from the all locations. It has been reported that the detection threshold of these compounds in wine matrix is 14 µg/L for ethyl hexanoate, 5 µg/L for ethyl octanoate, 20 µg/L for ethyl butanoate, and 200 µg/L for ethyl decanoate [15,9,30] Table 2 indicates that the concentrations of these compounds in Boğazkere wines are well above the detection threshold. There was a statistically significant difference between the amounts of ethyl octanoate, ethyl hexanoate, and ethyl butanoate in the wines from the locations (p<0.001 and p<0.01). However, a location-year interaction was observed for these compounds (Table 2). It has been reported that the amounts of ethyl hexanoate, ethyl octanoate, and ethyl butanoate compounds in Boğazkere wines reach the detection threshold [2]. Other significant esters found in Boğazkere wines include those with branched chain structures, such as ethyl-2-methyl-propanoate, ethyl-2-methyl-butanoate, ethyl-3-methyl-butanoate, ethyl-3-hydroxy-butanoate, ethyl-4-hydroxy-butanoate, and ethyl-2-hydroxy-4-methyl-pentanoate. It has been reported that the detection threshold value of these compounds in wine matrix is 15 µg/L for ethyl-2-methyl propanoate, 18 µg/L for ethyl-2-methyl-butanoate, 3 µg/L for ethyl-3-methyl-butanoate and ethyl-2-hydroxy-4-methyl-pentanoate for 300 µg/L [7,31] Table 2 shows that the amounts of ethyl-2-methyl propanoate, ethyl-2-methyl butanoate, and ethyl-3-methyl butanoate in Boğazkere wines are well above the detection threshold. According to Ugliano and Henschke [27] and Falcao et al. [31], branched chain esters such as ethyl-2-methyl-propanoate, ethyl-2-methyl-butanoate, ethyl-3-methyl-butanoate, and ethyl-2-hydroxy-4-methyl-pentanoate contribute red and blackberry fruit aromas to wine. It has been reported that ethyl-2-hydroxy-4-methyl pentanoate, in particular, gives a fresh blackberry aroma, has a significant effect on the aroma of wines despite being below the detection threshold value, and exhibits a synergistic perception interaction with ethyl butanoate. The amounts of ethyl-2-methyl-butanoate, ethyl-3-methyl-butanoate, and ethyl-4-hydroxy-butanoate varied significantly among the wines (p<0.05, p<0.01, p<0.001). Higher amounts of ethyl-2-methyl-butanoate and ethyl-3-methyl-butanoate have been detected in wines coming from Manisa. Table 2 indicates that only ethyl-3-methyl-butanoate and ethyl-4-hydroxybutanoate did not exhibit location-year interactions. According to Camara et al. [32] and Reboredo et al. [33], diprotic acid esters such as monoethyl succinate, diethyl succinate, and ethyl lactate play a significant role in the aroma of aged wine, and these compounds increase with age and oxidation. Diethyl succinate, mono-ethyl succinate, and ethyl lactate are three of the most prominent esters found in Boğazkere wines. Diethyl succinate was the only compound for which the difference was statistically significant (p<0.01). These compounds were found in higher concentrations in Manisa wines than in wines from other locations.

The majority of lactones are generated by yeast activity. As the most important lactone group in wines, gamma lactones are typically present in all wines [6]. In Diyarbakır, Manisa, and Denizli Boğazkere wines, seven and five lactone compounds have been identified, respectively. The total amount of lactones in Diyarbakır, Manisa, and Denizli was determined to be 1 mg/L, 0.79 mg/L, and 0.94 mg/L, respectively. In all locations, 4-ethoxycarbonyl-gamma-butyrolactone and gamma-butyrolactone were the predominant lactone compounds. Although gamma-butyrolactone is the most well-known lactone, it does not appear to contribute significantly to the organoleptic properties of wine [6]. The compound 4-ethoxycarbonyl-gamma-butyrolactone, which does not show a year-location interaction between the different locations, was found to have a higher range in Diyarbakır. According to Rocha et al. [34], the 4-ethoxycarbonyl-gamma-butyrolactone compound gives a red fruit aroma to wines. Cabaroglu et al. [2] determined that Boğazkere wines consist of 4-ethoxycarbonyl-gamma-butyrolactone and gamma-butyrolactone compounds.

Boğazkere wines from Diyarbakır contained 20 volatile phenol compounds, whereas wines from Denizli and Manisa contained 12 volatile phenol compounds. compounds, Manisa 5.0 mg/L, and Denizli 6.7 mg/L.

Diyarbakır contained 8.6 mg/L of volatile phenol Comparing the locations revealed that Diyarbakır wines had a higher total amount of volatile phenol, with the difference being statistically significant without any effect by location-year interaction

3-oxo-ionol is one of the norisopronoids that are varietal aroma compounds present in wines. In grapes and wines, the amount of the compounds present is measured in micrograms [35]. When the locations were compared, it was determined that 3-oxo-ionol was statistically significantly higher in the Diyarbakır location. Cabaroglu et al. [2] observed seven norisopronoid components in Boğazkere wine and characterised the 3-oxo-alpha-ionol compound as a bonded aroma compound.

3.2 Aroma-active compounds of Boğazkere wines

The aroma-active compounds, linear retention indices (LRI), modified frequency (MF%) values, and sensory descriptors are presented in Table 3. Boğazkere wines contained 29 aroma-active compounds. A total of 13 esters, 7 higher alcohols, 1 C6 compound, 1 carbonyl compound, 6 volatile acids, and 3 lactones were identified in Boğazkere wines. The MF% values of the compounds detected in wines range from 8% to 89%. As is well-known, when the MF% value of aroma-active compounds rises, the compounds' perception also improves [23].

It was determined that compounds and odours with MF% values greater than 75% were, in order: ethyl-2-methyl propanoate associated with red apple odour; ethyl-3-methyl butanoate/ethyl-2-methyl butanoate associated with strawberry odour; isoamyl acetate associated with banana odour; isoamyl alcohol associated with whey/alcohol-like odour; It is observed that the majority of active compounds in Boğazkere wines arise from esters, which provide potent odours of red and black berries. As previously mentioned, ethyl esters of fatty acids and branched-chain esters give red wines their fresh red and blackberry odours [31, 36].

Upon comparing the locations, the compounds isoamyl acetate, isoamyl alcohol, 2-phenylethyl alcohol, ethyl 2-hydroxy-4-methyl-pentanoate, and mono-ethyl succinate displayed significant differences. Specifically, blackberry-associated ethyl 2-hydroxy-4-methyl-pentanoate had significantly higher value in the wines of Diyarbakır and Manisa. Moreover, Manisa and Denizli wines had significantly higher mono-ethyl succinate associated with dried fruit odour, a compound known to increase with ageing and oxidation [32,33]

3.3 Sensory evaluation of Boğazkere wines

Descriptive Analysis (DA) was used in order to evaluate the sensory characteristics of Boğazkere wines. Twelve assessors performed sensory analyses of the wines. The assessors evaluated the wines with 23 distinct attributes using a 15-cm scale to determine the intensity of each attribute in DA. One visual attribute (colour), three taste attributes (sweetness, sourness, bitterness), four mouthfeel attributes (astringency, body, complexity, aftertaste, overall flavour intensity), five aroma attributes (N-red fruit, N-black fruit, N-dry fruit, N-confectionery, N-woody, N-chocolate, N-spicy), and six flavour attributes (P-red fruit, P-black fruit, P-dry fruit, P-confectionery, P-woody, P-chocolate, P-spicy) were assessed. Figure 1 displays the principal component analysis (PCA) of sensory evaluation data for Boğazkere wines with two dimensions. According to PCA, PC1 explains 38.7% of the variance, PC2 explains 21.2% of the variance, PC3 explains 12.6% of the variance, and PC4 explains 8.9% of the variance, for a total of 4 components explaining 81.4% of the variance. The PC3 and PC4 have not been included for the purpose of simplicity, as they have little impact on the outcome.

Overall flavour intensity descriptors display a positive correlation with aftertaste (r = 0.92), body (r = 0.90), and complexity (r = 0.89) according to the PCA correlation matrix. Descriptors of red fruit on the palate were positively correlated with descriptors of confectionery on the palate (r = 0.84) and red fruit on the nose (r = 0.72), whereas sourness and dry fruit on the palate were negatively correlated (r = -0.73). The first component (PC1) explains the descriptors of dried fruit, astringency, spicy, black fruit, chocolate, and woody, while Diyarbakır wines show a positive correlation in terms of dry fruit, black fruit, spicy, woody, and astringency attributes. The 2nd component (PC2), on the other hand, explains the descriptors of red fruit, confectionery, body, aftertaste, complexity, and overall flavour intensity, whereas Manisa wines demonstrate a positive correlation with the descriptors of red fruit and confectionery. The PC2 component distinguished Diyarbakır wines from those of Manisa and Denizli, whereas the PC1 component separated Manisa and Denizli wines.

PC1 and PC2 components were divided into three groups, and the majority of Diyarbakır wines were gathered in the upper right section. Black fruit, dry fruit, red fruit, confectionary, and spicy have been considered to be the most prominent aroma descriptors of Boğazkere wines. Although dried fruit, black fruit, and chocolate aromas are prominent in Diyarbakır wines, Denizli and Manisa wines have more red fruit and confectionery odours. Assessors referenced the odours of strawberry, raspberry, and red cherry as descriptors of red fruit; black mulberry, blackberry, and black cherry as descriptors of black fruit; and prune, raisin, and dried figs as descriptors of dried fruit. Yıldırım et al. [37] reported that Boğazkere wines are characterised by raisin, woody, and astringency descriptors.

Figure 1 Principal component analysis of the dataset featuring descriptive analyses of wines. PC1 and PC2 dimensions. C, indicates wines from Çermik/Diyarbakır, P, indicates wines from Pendore/Manisa, and G, indicates wines from Güney/Denizli.

3.4 Relating between aroma compounds and sensory attributes of Boğazkere wines

The relationship between aroma compounds, chemical composition, and sensory attributes has been further investigated using partial least squares regression (PLSR) analysis. According to Jaeger et al. [38], the PLSR method is frequently used to correlate different data groups in sensory analysis studies. In PLSR analysis, the aroma compounds (mostly aroma-active) determined in the wines and the general composition of the wines were defined as X data, whereas sensory attributes were defined as Y data with standardised results. In order to provide a better explanation of the model, some aroma compounds were eliminated from the X data as these did not fit well with the model as determined by VIPs (lower than 0.8). The PLSR analysis consisted of two-dimensional components. The regression model contained 21 aroma-active compounds, 12 aroma compounds, 5 general compositions, and 19 sensory descriptors. Figure 2 indicates the component charts for PLS1 and PLS2. The PLS1 component (X: 31.3%, Y: 31.5%), PLS2 component (X: 34.8%, Y: 39.5%) collectively explained 71.0% of the variance in the Y data.

PLS1 and PLS2 components separated Diyarbakır wines (C1 to C8) from Manisa and Denizli wines (P1, P2, G1, G2), which were grouped together on the top left side of the coordinate system based on the PLSR analysis of Boğazkere wines (Fig. 1). Figure 1 indicates a positive correlation between the red fruit and confectionary attributes of the wines and ethyl butanoate, ethyl octanate, ethly decanoate, diethyl dl-malate, and ethyl-2-methylbutanoate. In addition, these compounds were associated with the wines P1, P2, G1, and G2. On the contrary, ethyl-2-hydroxy-methyl-pentanoate, 4-ethoxycarbonyl-gamma butyrolactone, gamma nonalactone, pantolactone, and ethyl-2-methyl-propanoate were found to be positively correlated with black fruit and dry fruit attributes found in wines from Diyarbakır. Black fruit (nose and palate) had a significant positive correlation with gamma-nonalactone, whereas P-black fruit had a significant negative correlation with ethyl-2-methyl butanoate. Dry fruit has a significant positive correlation with pantolactone and isomaly alcohol on the nose. In addition, the spicy attribute on the nose correlated positively with propiovanillone, and the spicy attribute on the palate correlated positively with 2-methoxy-4-vinylphenol and propiovanillone.

In Bordeaux wines, it has been reported that blackberry fruit odour is associated with ethyl propanoate, ethyl-2-methylpropanoate, and ethyl-2-methyl butanoate compounds, and redberry fruit odour is associated with ethyl butanoate, ethyl hexanoate, ethyl octanoate, and ethyl 3-hydroxybutanoate compounds [39]. Robinson et al. [40] reported that the compounds ethyl-2-methyl propanoate, ethyl-2-hydroxy 3-methyl butanoate, and benzene propyl acetate were positively correlated with red berry fruit odours in Cabernet Sauvignon wines. According to studies conducted by Falcao et al. [31], ethyl-2-hydroxy-4-methylpentanoate has a substantial impact on the overall odour of red wines and has been associated with the aroma of fresh blackberries. In addition, it has been reported that ethyl-2-hydroxy-4-methylpentanoate has a considerable effect on aroma even at levels below the detection threshold (300 g/L in model wine) and that it has a synergistic perception effect with ethyl butanoate [31]. Sensory features may exhibit synergistic formation as a consequence of the interaction of various aroma compounds with one another [41]. According to the olfactory threshold, the concentration of different ethyl esters had no direct effect on the fruit aroma of red wines. However, an overall sensory effect of the red and black berry odours was clearly determined because of the cumulative effect of these esters (ethyl propanoate, ethyl 2-methylpropanoate, and ethyl 2-methylbutanoate for black-berry aromas; ethyl butanoate, ethyl hexanoate, ethyl octanoate, and ethyl 3-hydroxybutanoate for red-berry aromas) [39]. In addition, it is known that the non-volatile wine matrix plays an important role in the perception of aroma compounds [40,42].

Figure 2 PLS-R loading for sensory attributes (Y variables) and aroma and aroma-active compounds with general composition of Boğazkere wines (X variables) from three locations.

4 Conclusion

This study used chemometrics to highlight the relationships between the chemical analyses and the sensory evaluation results for Boğazkere wines produced in Turkey's various viticultural locations and in two different vintages. Esters, higher alcohols, and lactones were identified as three of the majority of significant compounds responsible for the distinctive wine aroma across both vintages and locations. Oenological factors such as maceration techniques, fermentation activators, yeast, and barrel ageing can be investigated further to learn how wine odour evolves over time. In addition, the further research aims to study the microclimate effect on the vineyards of the original locations for the Boğazkere such as sub region of Diyarbakır and Elazığ.

The authors are thankful to the Scientific Projects Unit of Cukurova University for the financial support provided (Research Project No: FBA-2016-5214 and FYL-2022-15628). Furthermore, we would like to thank the Diren, Doluca, Kavaklıdere, and Mey-Diageo wineries who supported the study by providing samples.

References

All Tables

All Figures

	Figure 1 Principal component analysis of the dataset featuring descriptive analyses of wines. PC1 and PC2 dimensions. C, indicates wines from Çermik/Diyarbakır, P, indicates wines from Pendore/Manisa, and G, indicates wines from Güney/Denizli.
In the text

	Figure 2 PLS-R loading for sensory attributes (Y variables) and aroma and aroma-active compounds with general composition of Boğazkere wines (X variables) from three locations.
In the text