Differentiation of Commercial PDO Wines Produced in Istria ( Croatia ) According to Variety and Harvest Year Based on HS-SPME-GC / MS Volatile Aroma Compound Pro fi ling

Aroma is probably the most important organoleptic characteristic that defi nes the typicity and quality of wine. It is one of the crucial factors that determine its market value and price, and a key att ribute for consumer preference. Wine aroma results from the occurrence of several hundreds of odouriferous volatile compounds originating from grapes (primary or varietal aromas), compounds pro duced in fermentation (secondary or fermentation aromas), and, in aged wines, compounds produced during ma turation (tertiary aromas) (1). Volatile aroma compounds found in wine pertain to diff erent chemical classes, such as monoterpenes, norisoprenoids, higher alcohols, fatt y acids, esters, aldehydes, ketones, benzenoids, etc. (2), and occur in concentrations ranging from ng/L to a few hundreds of mg/L (3). The fi nal aroma of a wine is a result of complex interactions between several factors, such as geographical location of the vineyard (4), which is linked to soil and climatic conditions (5), harvest year (6,7), yeast strain (8), production parameters (9), etc. Particular att ention has been devoted to the infl uence of varietal origin, and several studies have focused on the identifi cation of volatile compounds typical for diff erent varieties, important for the expression of varietal characteristics in wine (10–20). The knowledge of the chemical (volatile aroma) composition of varietal wines may give opportunities to producers to deeper understand the phenomena they observe in practice and control the production with greater effi ciency to obtain wines with a more pronounced varietal typicity and higher quality. It may enable a proper characterisation and diff erentiation of varietal wines, attributing them an added marketing value. Wine is globally consumed, and it is a food commodity of relatively high commercial value and importance to the economy of many world countries. Therefore, wine authenticity control, among others, in terms of varietal origin characterisaISSN 1330-9862 preliminary communication


Introduction
Aroma is probably the most important organoleptic characteristic that defi nes the typicity and quality of wine.It is one of the crucial factors that determine its market value and price, and a key att ribute for consumer preference.Wine aroma results from the occurrence of several hundreds of odouriferous volatile compounds originating from grapes (primary or varietal aromas), compounds pro duced in fermentation (secondary or fermentation aromas), and, in aged wines, compounds produced during ma turation (tertiary aromas) (1).Volatile aroma compounds found in wine pertain to diff erent chemical classes, such as monoterpenes, norisoprenoids, higher alcohols, fatt y acids, esters, aldehydes, ketones, benzenoids, etc. (2), and occur in concentrations ranging from ng/L to a few hundreds of mg/L (3).The fi nal aroma of a wine is a result of complex interactions between several factors, such as geographical location of the vineyard (4), which is linked to soil and climatic conditions (5), harvest year (6,7), yeast strain (8), production parameters (9), etc. Particular att ention has been devoted to the infl uence of varietal origin, and several studies have focused on the identifi cation of volatile compounds typical for diff erent varieties, important for the expression of varietal characteristics in wine (10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20).The knowledge of the chemical (volatile aroma) composition of varietal wines may give opportunities to producers to deeper understand the phenomena they observe in practice and control the production with greater effi ciency to obtain wines with a more pronounced varietal typicity and higher quality.It may enable a proper characterisation and diff erentiation of varietal wines, attributing them an added marketing value.Wine is globally consumed, and it is a food commodity of relatively high commercial value and importance to the economy of many world countries.Therefore, wine authenticity control, among others, in terms of varietal origin characterisa- tion and diff erentiation, is continuously required to detect adulteration and to improve wine quality (21).
At this moment, the Republic of Croatia is the latest state that has joined the European Union.Like many other Mediterranean countries, it has a vitivinicultural tradition that is centuries long, with today a relatively developed wine industry and rather interesting domestic grape varieties.The quality of Croatian wines in recent years is rapidly and constantly improving, which resulted in signifi cant success on the national and international markets and quality competitions.Their reputation has recently been strengthened by gaining European Protected Designations of Origin (PDO), which certify their authenticity and the connection of their quality with varietal and territorial origin.In the last few decades, Croatian domestic varieties, linked to specifi c terroirs and adapted to the local environmental conditions, have been used to produce original and high-quality wines and compete on the market with widely spread international varieties to att ract consumers, nowadays more and more motivated by marketing att ributes and new wine types rather than just pleasant aroma and taste (7).
From a scientifi c point of view, Croatian wines are poorly characterised when compared to other European and world wines.Because of the lack of objective scientific information, knowledge on the varietal typicity of wines from native Croatian grape varieties is still on an informal level.For example, there is published data on the composition of wines made from Malvazĳ a istarska, the most spread and important native white grape variety in Croatia grown principally in the region of Istria (9,(22)(23)(24)(25), but the volatile compounds and their particular concentrations, indicators of its varietal origin and drivers of its varietal typicity, on the basis of which it might be distinguished and diff erentiated from other monovarietal white wines, are still unknown.On the other hand, Chardonnay is a globally spread and known variety, and its wine aroma has been investigated extensively (26)(27)(28)(29)(30).Despite many characterisation studies, Chardonnay has rarely been directly confronted with, compared to, and diff erentiated from other white wines of similar typology, especially from this part of Europe.The compounds responsible for typical Chardonnay aroma have been identifi ed (26)(27)(28)(29)(30), but it is still not known if the amounts found are specifi c for this variety and can diff erentiate it from other monovarietal wines.A small number of previous diff erentiation att empts were limited in that they were based only on fermentation-derived compounds (31), comparison with mostly red wines (32), or used only m/z fragments aft er direct injection in an electrospray ionisation Fourier transform mass spectrometer (ESI-FT-MS), without the identifi cation of volatile compounds as discriminating variables (33).Chardonnay wines in question were produced in South Africa, Brasil and Chile, respectively.Wines made from Muscat yellow grapes, another important variety in Croatia especially in the region of Istria, were globally studied extremely rarely, except those produced in Italy, which were investigated extensively, but relatively long ago (34)(35)(36).
The main objective of this study is to determine the differences in the composition of volatile aroma compounds between wines made from three important white grape varieties in the Istria region of Croatia (PDO Croatian Istria), through the combined use of HS-SPME-GC/MS profi ling with univariate and multivariate statistics.The aim is to contribute to the knowledge on the varietal typicity of the domestic Malvazĳ a istarska wine, as well as its differentiation from wines obtained from internationally known Chardonnay and Muscat yellow wines produced in the same area.Comparison between native Malvazĳ a istarska and introduced Chardonnay is especially interesting and important because these are competing varieties used in Istria to produce typologically similar wines.Further, among the local professional wine community, Malvazĳ a istarska wines are oft en linked to a so-called subtle 'muscat-like character'.The comparison with Muscat yellow wines in this study might help to clarify this att ribute.Interaction of varietal origin with other factors of infl uence, namely harvest year, may complicate the characterisation, diff erentiation and authentication of mo no varietal wines (37).For this reason, wines from two consecutive harvest years, signifi cantly diff erent with respect to climatic conditions, were included in this study.

Wine samples and harvest years
Samples of typical fresh, young, unoaked wine Malvazĳ a istarska, Chardonnay and Muscat yellow, with Protected Designation of Origin (PDO Croatian Istria, traditional term: Quality wines), produced by standard wine making technology (destemmed, crushed and mashed grapes, without or up to 24-hour skin contact, inoculation with commercial Saccharomyces cerevisiae yeast, and fermentation in stainless steel tanks at temperatures lower than 18 °C) were voluntarily consigned by the local producers in the Istria region of Croatia.Aft er an informal preliminary sensory assessment of a larger number of wines, the most typical samples were chosen by a consensus of professional, highly experienced and certifi ed wine tasters from the Institute of Agriculture and Tourism, Poreč, Croatia, on the basis of their personal varietal typicity concept.Five samples of each variety, from each of the two consecutive harvest seasons (2013 and 2014), were collected (total of 30 wines).The two studied harvests were rather diff erent considering climatic conditions.According to the Croatian Meteorological and Hydrological State Institute, in the period from April to September 2013, the sum of the eff ective temperatures was 1806 °C, while the total rainfall amounted to 359 mm.Most of the rainfall was recorded in August (112 mm).In the same period in 2014, the sum of the eff ective temperatures was lower, 1683 °C, while the total rainfall was much higher: 546 mm.Most of the rainfall was recorded in July (154 mm) and September (128 mm).Because of the lower temperatures and lots of rain, 2014 was characterised by problematic and late ripening.

Chemical standards and standard solutions of volatile aroma compounds
Pure standards of individual volatile aroma compounds were purchased from Merck (Darmstadt, Germany), Sigma-Aldrich (St. Louis, MO, USA), and Fluka (Buchs, Switzerland).Stock standard solutions were prepared in ethanol.Working standard solutions were prepared by dilution of stock standard solutions in synthetic wine containing 12 % of ethanol, 5 g/L of tartaric acid, 50 mg/L of each acetaldehyde, methanol, ethyl acetate, 1-propanol and isobutanol, and 150 mg/L of isoamyl alcohol.Working solutions were adjusted to pH=3.2 with 0.1 M NaOH.

Analysis of volatile aroma compounds by headspace solid-phase microextraction with gas chromatography/ mass spectrometry
Volatile aroma compounds were isolated using headspace solid-phase microextraction (HS-SPME) according to the modifi ed method of Noguerol-Pato et al. (38), and analysed by gas chromatography/mass spectrometry (GC/MS).SPME fi bre holder and 50/30 nm divinylbenzene-carboxen-polydimethylsiloxane (DVB-CAR-PDMS) fi bres were purchased from Supelco (Bellefonte, PA, USA).Wine sample was diluted fourfold, and 4 mL of the solution were placed in a 10-mL glass vial.A volume of 50 μL of internal standard solution (0.84 mg of 2-octanol per L of wine for determination of terpenes, norisoprenoids, alcohols and miscellaneous compounds, 0.82 mg of methyl nonanoate per L of wine for determination of esters, and 2.57 mg of heptanoic acid per L of wine for determination of acids) and 1 g of ammonium sulphate were added.The vial was sealed with a Tefl on-faced septum cap, and the sample was pre-conditioned at 40 °C for 15 min.Microextraction lasted for 40 min at 40 °C with stirring (800 rpm).For desorption, the fi bre was inserted into the GC/MS injector port at 248 °C for 5 min (3 min in splitless mode).
Identifi cation and quantifi cation of minor volatile compounds was performed using a Varian 3900 gas chromato graph coupled with a Varian Saturn 2100T ion trap mass spectrometer (Varian Inc., Harbour City, CA, USA).The column used was an Rtx-WAX (60 m×0.25 mm i.d., 0.25 μm fi lm thickness; Restek, Bellefonte, PA, USA).Initial oven temperature was 40 °C, then increased at 2 °C/ min to 240 °C, and then kept at 240 °C for 10 min.Injector, transfer line and ion trap temperatures were 245, 180 and 120 °C, respecti vely.Mass spectra were acquired in electron impact mode (70 eV) at 1 s/scan, full scan with a range of m/z=30-450.The carrier gas was helium (1 mL/ min).Identifi cation was performed by comparing retention times and mass spectra with those of pure standards when available, and with mass spectra from NIST05 library (National Institute of Standards and Technology, Gaithersburg, MD, USA).Identifi cation by comparison with mass spectra was considered successful for compounds with the MS spectra reverse match numbers higher than 800.Linear retention indices (relative to n-alkanes from C 10 to C 28 ) were calculated and compared to those from literature (22,(39)(40)(41)(42)(43)(44).When standards were available, standard calibration curves were constructed.For other compounds semi-quantitative analysis was carried out, and their concentrations were expressed as equivalents of compounds with similar chemical structure for which standards were available, assuming a response factor equal to one: monoterpenes were quantifi ed as linalool equivalents (response factor (RF) of linalool vs.

Odour activity values and aroma compound groups
Odour activity values (OAV) of volatile aroma compounds were calculated as the quotients of their concentration and the corresponding odour perception threshold from literature (45)(46)(47)(48).The OAVs of the compounds that exhibit similar olfactory sensation were grouped based on their odour description, as suggested by Moyano et al. (49).In this work eight groups (also known as series) were established: varietal terpenic, varietal fruity, sweet, fermentative fruity, berry fruit, fl oral, fatt y and green.Such a presentation of wine aroma profi le is an approximation and may diff er from the results of sensory analysis performed by a panel of trained tasters.However, it is certainly valid for establishing the potential olfactory impact of particular groups and individual compounds.In addition, it greatly reduces the number of variables to be considered, and facilitates the interpretation of results (49).

Statistical analysis
All experiments were performed in duplicate, and average values were used in further data analysis.Mean values of concentration and standard deviations were calculated from fi ve replicates, i.e. fi ve samples of each investi gated variety per harvest year.One-and two-way analyses of variance (ANOVA) and Fisher's least significant diff erence (LSD) test were used to compare the mean values at the level of signifi cance of p<0.05.To diff erentiate wines according to varietal origin and harvest year, stepwise linear discriminant analysis (SLDA) was applied, using Wilks' lambda as a selection criterion and F-statistic factor to establish the signifi cance of the changes in lambda when a new variable is tested.The prediction capacity of the discriminant model was estimated by cross--validation.Statistical elaboration was carried out using Statistica v. 8.0 soft ware (StatSoft Inc., Tulsa, OK, USA).

Results and Discussion
The standard physicochemical parameters determined in wines made from Malvazĳ a istarska, Chardonnay and Muscat yellow varieties in 2013 and 2014 are presented in Table 1.One-way analysis of variance (ANOVA) and Fisher's least signifi cant diff erence (LSD) test were applied to compare the diff erences between the mean values for each harvest year separately.Two-way ANOVA with factor variety and harvest year was applied to establish if these diff erences were consistent in the two harvest years, as well as to determine the eff ect of harvest year.When two years were considered separately, no significant diff erences were observed, except for the higher concentration of reducing sugars in Muscat yellow group, which was expected since it consisted of dry, semi-dry, and semi-sweet wines.Two-way ANOVA revealed significant diff erences, with higher concentrations of extract and ash, and higher total acidity found in wines from 2014.
The concentrations of volatile aroma compounds determined by HS-SPME-GC/MS analysis in wines made from Malvazĳ a istarska, Chardonnay and Muscat yellow varieties in 2013 and 2014 are presented in Table 2.

Varietal aroma compounds
Terpenic compounds originate from grapes, both as free volatile molecules and released from glycosidic precursors.As expected, Muscat yellow wines were the most abundant in terpenes, both qualitatively and quantitatively, with the domination of exceptionally high linalool concentration (Table 2).Such a composition is relatively in agreement with that previously determined in Muscat yellow must (34,35) and wine (36), where linalool concentrations were higher than 2500 μg/L in some cases.In this work, terpenic diols and particular other terpenes (hydroxyl forms) were not found probably because of their weaker volatility, which is also the reason why the concentration of linalool oxides was probably underestimated during semi-quantitative analysis relative to highly volatile linalool.On the other hand, the occurrence of many other monoterpenes in Muscat yellow wines, such as epoxylinalool, trans-β-ocimene, 6,10-dihydromyrcenol, menthol and trans-nerolidol has, to our knowledge, been confi rmed for the fi rst time.
In 2013, several monoterpenes, such as epoxylinalool, β-pinene, nerol oxide, and major monoterpenols linalool, hotrienol, α-terpineol, nerol and geraniol, were found in signifi cantly higher concentrations in Malvazĳ a istarska than in Chardonnay wines.Nerol emerged as a consistent diff erentiator of Malvazĳ a and Chardonnay, with significant diff erence determined in both years.The concentrations and the composition of monoterpenes in Malvazĳ a wines were generally in fair agreement with those report-ed previously: their content was moderate but signifi cant, with linalool followed by geraniol as the most abundant (9,(22)(23)(24)(25).Although relatively high concentrations of linalool were found in particular Chardonnay wines from other world regions (up to 142 μg/L), wines from this variety are mostly defi cient in monoterpenes (30), and the results of this study confi rmed it.
The eff ect of variety was found to be signifi cant for the majority of terpenes by two-way ANOVA, mostly due to signifi cantly higher levels in Muscat wines (Table 2).The eff ect of year was also established for many terpenic compounds, with higher amounts generally found in wines produced in harvest 2013, which was characterised by more favourable climatic conditions.The response of each variety to climatic conditions of the two harvest years was diff erent; the eff ect of year was more evident in Muscat and Malvazĳ a wines, and less in Chardonnay.This was confi rmed by signifi cant interaction eff ects on several terpenes (Table 2).Interestingly, the eff ect of harvest year was not signifi cant for the majority of the most important, major monoterpenols.C 13 norisoprenoid β-damascenone derives from the degradation of carotenoid molecules, such as β-carotene, lutein, neoxanthin and violaxanthin, during fermentation (50), and has an important positive role in wine varietal aroma because of its low odour perception threshold and pleasant odour reminiscent of honey, dried plum, and stewed apple.It was previously reported among the important contributors of both Malvazĳ a istarska (22,23) and Chardonnay aroma (26,30).Signifi cant diff erences between its content in the investigated monovarietal wines were not found.Such a result confi rmed that the levels of β-damascenone are more dependent on viticultural and winemaking conditions than on variety (50).C 6 compounds are mostly formed during prefermentation production steps by the enzymatic degradation of unsaturated fatt y acids and from glycosidic precursors.They contribute to wine aroma with vegetal and herba-     ceous odours, and may have a negative eff ect when present in high concentration.Some authors consider C 6 compounds to be varietal aromas (13), and their ratios were shown to be useful for varietal diff erentiation of certain wines (19).In this work, signifi cantly higher concentrations of cis-3-hexenyl derivatives (in both years) and lower concentrations of trans-3-hexen-1-ol and 1-hexanol (in 2013 and 2014, respectively) were noted in Muscat yellow than in other investigated monovarietal wines (Table 2).
Other compounds originating from grapes, polyfunctional thiols, have also been identifi ed in Chardonnay wines in earlier investigations, but their impact was not found to be as important as in the case of Sauvignon blanc, where they are crucial for typical varietal aroma (30).Analysis of thiols is not a trivial task because of their very low abundance and reactivity, and this may account for the fact that their occurrence in Malvazĳ a istarska and Muscat yellow has not been confi rmed up to date.Although thiols were targeted compounds in this work (standards were available), they were not identifi ed in any of the wines by the HS-SPME-GC/MS method.

Fermentation aroma compounds
Concentrations and the composition of wine major aroma compounds produced during fermentation, such as alcohols, straight-chain acids, and ethyl and acetate esters, were in a fair agreement with those found in the three varieties investigated earlier (22,23,30,31,51), and it was confi rmed that they make up a basis of the aroma profi le of unoaked young white wines.They depend mostly on fermentation parameters and conditions, although physicochemical composition of must may have a certain infl uence (52).Several studies reported that fermentation aroma compound composition signifi cantly depends on varietal origin and harvest year, and can be useful in diff erentiating wines on the basis of these two criteria (7,37).This especially refers to medium-chain fatty acids and their ethyl esters which, unlike acetates, depend more on the availability of substrates, and therefore agricultural conditions and variety, than on enzymatic activity of yeast (31).For example, it was shown that South African unoaked young Chardonnay wines can be discriminated from other varieties (correct classifi cation of 74 %) on the basis of 2-phenylethanol, diethyl succinate, ethyl hexanoate, ethyl decanoate, ethyl octanoate, hexyl acetate and 1-propanol concentrations (31).
The concentrations of major straight-chain acids (C6, C8 and C10) were notable in all wines, surpassing their odour perception thresholds of 420, 500 and 1000 μg/L, respectively (45).Fatt y acid production is determined in part by the initial composition of must (53).Signifi cant diff erences among varieties were found in hexanoic and decanoic acid, but in diff erent harvests, implying the effect of harvest year.Low levels of nonanoic acid emerged as a potentially typical feature of Muscat yellow wines (Table 2).The eff ect of harvest year (two-way ANOVA) was also signifi cant for this acid, with higher concentrations found in wines from 2014.
In 2013, Malvazĳ a istarska wines stood out with higher concentrations of short, branched-chain ethyl esters, but contained lower ethyl octanoate concentration than other investigated wines.In 2014, Muscat yellow wines had the highest concentration of major middle-chain ethyl esters (hexanoate, octanoate and decanoate).Ethyl esters with odd number of carbon atoms (heptanoate and nonanoate) were found useful in diff erentiating Chardonnay from Muscat yellow wines, being more abundant in the former.Two-way ANOVA results showed a signifi cant eff ect of variety on the concentrations of major straight-chain ethyl esters, while the eff ect of harvest year was observed on the majority of branched-chain ethyl esters, with higher concentrations noted in 2013.These compounds, together with higher alcohol acetates, derive mainly from the yeast amino acid metabolism, and it is well known that the concentration of amino acids in grape depends on climatic conditions (7).The eff ect of year was also signifi cant on odd-chain ethyl esters, with signifi cantly higher concentrations in wines from the less favourable harvest of 2014.Odd-chain ethyl esters basically followed the same patt ern observed for the odd-chain fatt y acids (Table 2).
Chardonnay wines were more abundant in acetate esters, with signifi cant diff erences in some cases, more pronounced in 2013.In both harvest years, signifi cant differences were found in a few other esters, in some cases corroborated by two-way ANOVA results.Ethyl cinnamate emerged as a consistent diff erentiator of Malvazĳ a istarska and Chardonnay wines, with higher concentration in the latt er.Such a result partly confi rmed previous fi ndings in which ethyl cinnamate was among the key compounds responsible for the typical aroma of Chardonnay (26,27).It is worth mentioning that the chromatographic peak of ethyl cinnamate in Muscat yellow wines interfered with a much larger signal belonging to an unknown compound, which obstructed its identifi cation and quantifi cation in the majority of samples.
Among other compounds, particular benzenoids were found to be characteristic for Chardonnay, especially in 2013 when they were able to diff erentiate all three wines according to varietal origin.Like ethyl cinnamate, a tentatively identifi ed dimethylbenzaldehyde isomer, with mass spectra showing a rather high degree of similarity with that of 2,4-dimethylbenzaldehyde (characteristic ions with m/z (relative intensity): 133 (100), 134 (41), 105 (35), 77 (16), 75 (15), with mass spectra forward and reverse match number of 873), for which the standard was available, turned out to be a potentially consistent diff erentiator of Chardonnay wines in both years.Muscat yellow wines had higher levels of tentatively identifi ed 4'-ethoxy-2'-hydroxyoctanophenone and 2-(phenylmethylene)-octanal in 2014.A signifi cant eff ect of year was observed in the case of benzaldehyde, ethyl benzeneacetate, and γ-nonalactone with higher amounts found in wines from the rainy 2014.On the other hand, Malvazĳ a istarska and Chardonnay wines from 2013 contained more dimethylbenzaldehyde than in 2014.

Stepwise linear discriminant analysis
Aromatic Muscat yellow wines were clearly diff erentiated from relatively neutral Malvazĳ a istarska and Chardonnay by (mono)terpene concentrations higher by an order of magnitude (Table 2).For this reason, further investigation was focused on determining the diff erence between the latt er two, and stepwise linear discriminant analysis (SLDA) was applied only on the Malvazĳ a istarska and Chardonnay data set.The number of groups was four, since wines from the same variety from diff erent harvest year were considered as separate groups.SLDA model extracted 12 compounds according to Wilks' lambda criterion, and was successful in classifying wines according to both varietal origin and harvest year.A 100 % correct classifi cation was achieved aft er the inclusion of only four compounds in the following order: dimethylbenzaldehyde, isobutyl acetate, β-pinene, and trans-3--hexen-1-ol.The projection of monovarietal Malvazĳ a istarska and Chardonnay wine samples classifi ed according to variety and harvest year in two-dimensional space defi ned by the fi rst two discriminant functions, as well as the compounds included in the model are shown in Fig. 1.The samples were grouped according to variety along the direction of the second, and according to harvest year along the direction of the fi rst discriminant function.The prediction capacity of the SLDA model was evaluated by 'leave one-out' cross-validation, where each wine was removed from the model and classifi ed by the functions derived from all cases other than that case.The percentage of correct prediction by cross-validation was also 100 %.
When SLDA was applied to the set of Malvazĳ a istarska and Chardonnay samples divided into two groups based on variety, the obtained model included 16 compounds in total.Ethyl cinnamate entered the model as the fi rst, emerging as the most potent diff erentiator, and classifi ed correctly all Malvazĳ a wines.The inclusion of butyl acetate and 1,2-benzenedimethanol resulted in a 100 % correct classifi cation of all wines.When two groups were formed with harvest year as a criterion, only two compounds were suffi cient for a 100 % correct classifi cation: dimethylbenzaldehyde and isobutyl acetate.Another 13 compounds entered and additionally improved the differentiation capacity of the model.

Impact odourants and aroma groups
One-and two-way ANOVA followed by SLDA analysis extracted many possible markers of varietal origin of Malvazĳ a istarska, Chardonnay and Muscat yellow wines among the analysed volatile compounds.The diff erentiation was rather clear when considering Muscat wines, clearly distinguished from others by high concentrations of odouriferous monoterpenols (Table 2), which significantly surpassed the corresponding odour perception thresholds in the majority of samples from both years (linalool threshold 6, citronellol 18, nerol 15 and geraniol 30 μg/L) (46)(47)(48).It can be stated with certitude that the aroma of Muscat yellow wines was typically muscat.To explain the diff erences between aroma profi les of Malvazĳ a istarska and Chardonnay, two typologically similar wines, represented a more challenging task.Many of the compounds found to discriminate those wines in this work are of doubtful or unknown sensory impact.In an att empt to approximate how the diff erences in chemical composition between Malvazĳ a istarska and Chardonnay wines possibly refl ect on their sensory quality, the quantifi ed main odourants occurring in peri-and suprathreshold con centrations, and therefore exhibiting odour activity val ues (OAV) near and higher than 1, were selected and grouped into main aroma groups based on the similarity of their odours.Selected odourants are listed in Table 3 together with their odour descriptors and affi liation to corresponding aroma groups, while the constructed aroma group profi les are presented in Fig. 2.
The dominance of fruity aroma compounds formed in fermentation with ethyl octanoate as the most powerful odourant was determined in all wines (Fig. 2), and corresponded to a profi le of a standard young white wine and previous fi ndings on Malvazĳ a and Chardonnay (22,23,51).Many of the esters formed in fermentation contributed to the formation of sweet aroma group, which was also among the dominant aromas.In Chardonnay wines, these two groups had higher values, suggesting that Chardonnay wine aroma was characterised by higher intensities of fruity and sweet odours than that of Malvazĳ a.On the other hand, the profi le of Malvazĳ a istarska wines was somewhat more complex, with higher levels of the varietal terpenic group in both years, berry fruit group in 2013, and fatt y group in 2014.It is possible that the sub-tle, so-called 'muscat character' of Malvazĳ a istarska wines, which is oft en encountered in practice, derives from the sensory activity of linalool supported by other major monoterpenols through synergistic and additive effects.The eff ect of harvest year was also evident: varietal terpenic, fruity and sweet groups prevailed in wines from 2013 in relation to those from 2014, implying the former were more aromatic.Values of particular aroma group were multiplied or divided by a factor in order to obtain a more uniform display

Conclusions
By using volatile aroma profi les obtained by HS--SPME-GC/MS analysis, elaborated by univariate and multivariate statistics, a detailed characterisation of three monovarietal wines, Malvazĳ a istarska, Chardonnay and Muscat yellow produced in Istria (Croatia) was achieved.Varietal diff erentiation of wines produced in climatically more favourable harvest of 2013 was more successful, and many compounds emerged as potential discriminators.Colder and humid 2014 exerted a signifi cant eff ect by partially or totally annulling their diff erentiating ability, but extracted new markers specifi c for 2014.It is clear that variety and harvest year did not aff ect the variability in wine composition independently, but interacted with each other.Nevertheless, particular compounds emerged as statistically consistent markers of varietal origin through both vintages: Malvazĳ a had specifi c nerol concentration, ethyl cinnamate and a dimethylbenzaldehyde isomer were characteristic for Chardonnay, while Muscat yellow wines were clearly discriminated by the highest concentration of terpenes.It is worth emphasising that several neglected compounds with small or no sensory signifi cance, commonly not listed among wine origin differentiators, such as particular odd-chain acids and esters and benzenoids, were found to have interesting discrimination capacity.Apart from that, it was shown that not only varietal aromas, but several fermentation aroma compounds were signifi cantly aff ected by variety.Wines from the more favourable harvest of 2013 contained higher concentrations of the majority of important volatile compounds than in 2014.A special att ention was given to the comparison of Malvazĳ a istarska and Chardonnay wines, two wines similar by typology, the former being a domestic, and the latt er an introduced variety in Istria and Croatia.Malvazĳ a wines were more abundant in monoterpenes, while Chardonnay aroma was characterised by higher concentrations of fruity esters.Such diff erences were pronounced and statistically signifi cant in wines from 2013, but were partially overpowered by the eff ect of unfavourable 2014.Despite that, a 100 % correct classifi cation of Malvazĳ a istarska and Chardonnay wines according to both variety and harvest year was achieved by stepwise linear discriminant analysis, confi rming that detailed HS-SPME-GC/MS aroma profi ling accompanied by multivariate statistics is a powerful tool for diff erentiation of wines based on various criteria.The concept of odour activity values and aroma groups pointed to the potential diff erences between sensory profi les of Malvazija istarska and Chardonnay wines.
It is worth emphasising that this study represents the fi rst successful att empt to compare and diff erentiate Malvazĳ a istarska from other monovarietal white wines, as an important step in determination of its typicity and uniqueness.The study showed that many compounds which had been linked to typical Chardonnay aroma in earlier works have not been confi rmed to be Chardonnay--specifi c, which confi rms the importance of varietal differentiation and discrimination analysis, in order to determine the unique varietal characteristics of wines.
It was demonstrated that wine from Malvazĳ a istarska variety is an interesting and viable alternative to common globally known varieties, such as Chardonnay, thus increasing the off er for consumers, and favouring diff erentiation of Croatian wines on the national and international market.The results obtained are of interest to the industry, and represent useful guidelines for the characterisation and diff erentiation of Istrian and Croatian monovarietal wines.However, further study is needed, comprising larger sample sets from several vintages.Such an investigation is currently being performed under the framework of a national scientifi c project.

Fig. 1 .
Fig. 1.Projection of Malvazĳ a istarska (M) and Chardonnay (CH) monovarietal wines produced in 2013 and 2014, classifi ed according to varietal origin and harvest year, along the directions of two discriminant functions by stepwise linear discriminant analysis (SLDA) (a), and standardised coeffi cients of compounds selected by the SLDA model (b)

Fig. 2 .
Fig. 2. Aroma profi les of Malvazĳ a istarska (M) and Chardonnay (CH) monovarietal wines produced in 2013 (a) and 2014 (b), obtained on the basis of aroma group values calculated from the odour activity values of the corresponding aroma compounds.Asterisks denote statistically signifi cant diff erences at p<0.05 obtained by one-way ANOVA and Fisher's least signifi cant diff erence (LSD).Values of particular aroma group were multiplied or divided by a factor in order to obtain a more uniform display

Table 1 .
Standard physicochemical parameters of Malvazĳ a istarska, Chardonnay and Muscat yellow monovarietal wines produced in 2013 and 2014 Results are expressed as mean value±standard deviation, N=5.Diff erent lowercase lett ers in superscript in a row represent statistically signifi cant diff erences between mean values at p<0.05 obtained by one-way ANOVA and Fisher's least signifi cant diff erence (LSD) test for each harvest year separately.
Two-way ANOVA factors: V=variety, Y=harvest year, I=interaction (V×Y); asterisk represents a statistically signifi cant eff ect at p<0.05 by 2-way ANOVA; n.s.=not signifi cant

Table 2 .
HS-SPME-GC/MS profi les of volatile aroma compounds in Malvazĳ a istarska, Chardonnay and Muscat yellow monovarietal wines produced in 2013 and 2014 (39)(40)(41)(42)(43)(44)mpounds: S=retention time and mass spectrum consistent with that of the pure standard and with NIST05 mass spectra electronic library, MS=mass spectra consistent with those from NIST05 mass spectra electronic library, LRI=linear retention index consistent with that found in literature(39)(40)(41)(42)(43)(44).The compounds for which pure standards were not available (without lett er S in the ID column) were quantifi ed semi-quantitatively, and their concentrations were expressed as equivalents of compounds with similar chemical structure assuming a response factor=1.The compounds with only MS symbol in the ID column were tentatively identifi ed.RM=mass spectra reverse match number; FM=mass spectra forward match number; n.i.=not identifi ed; n.d.=not detected.Diff erent lowercase superscript lett ers in a row represent statistically signifi cant diff erences between mean values at p<0.05 obtained by one-way ANOVA and Fisher's least signifi cant diff erence (LSD) test for each harvest year separately.