2.1. Chemical Analysis
The basic chemical composition of the wines at bottling is shown in
Table 1. There were small but significant differences among treatments in all measured parameters (α = 0.05), but only the differences in alcohol content and residual sugar are deemed important to have a sensory impact [
21]. In general, differences in ethanol content were driven by sugar content differences in the grapes coming from the three different vineyards (
Supplementary data, Table S1).
In
Table 2, the total (free and bound) volatile phenol composition of the different wine treatments area shown. Free to total volatile phenol ratios were consistent and showed similar trends within wine treatments. In panel 1 wines, >85% of the VPs were in the free form, potentially due to late-season exposure to smoke. Additionally, the smoke impact in grapes from the Oakville AVA was relatively low, and the VP composition of the wines was only determined at the time of sensory three to six months after bottling. However, panel 2 wines showed a higher contribution from bound VPs, and although PCAs indicated a similar correlation between total and free VPs, the individual bound and free VP data are provided in
Supplementary Tables (Tables S3 and S4). Most volatile phenols measured were significantly different among the various wine treatments. In general, the yeast treatments did not result in large changes among the treatments. D80 wine treatments contained the lowest amounts of volatile phenols, with D254 containing the highest. Overall, the impact of yeast was low when only evaluating the volatile phenol composition, and no significant impact due to yeast treatment was observed in the amount of free VPs released from bound precursors. Similarly, the addition of oak alternatives did not have a large impact on the volatile phenol composition of the wines. Nobile Base (NB) contributed slightly more volatile phenols to the ferment than Nobile Fresh (NF), although syringol concentrations were in fact lower than the control treatment (EC1118). The Quertanin addition (ellagic tannin) did not contribute significant amounts of volatile phenols to the wines.
Two vineyard blocks at the Oakville Experimental Station in Oakville, AVA were harvested (OFV6, block 6, and EC1118 control treatment from block 9). These grapes were exposed to similar environmental conditions, but contained significantly different amounts of volatile phenols, indicating the potential impact of vine age and plant vigor (clone, rootstock, and trellis systems were similar). Block 6 was planted in 1993, whereas block 9 was planted in 2014. Although the wines made from non-smoke-exposed grapes contained the lowest amounts of volatile phenols, the comparative increase in the smoke-exposed grapes was relatively modest, potentially indicating a low impact due to smoke exposure. Although wildfires move within a mile of the vineyard, the wind direction was mostly away from the vineyard.
The different fermentation temperatures at the seven-day skin contact did not have a significant impact on the final volatile phenol content. Only for the 10 °C treatment, where grapes were cold soaked for seven days, pressed, and then fermented, were there volatile phenols (guaiacol and syringol) significantly lower compared to the control treatment (AV_25 °C). The extraction of VPs at different temperatures was followed as shown in
Table 3. Free and bound VPs were extracted at a constant rate irrespective of temperature (data not shown). Furthermore, our data indicated that after three days of skin contact, independent of temperature, between 78 to 91% of the VPs were extracted compared to the final wines and that pressing did not have a significant impact on the amount of VPs in the wine after at least 7 days of skin contact. Furthermore, the treatment applied to the Stag’s Leap District grapes were also unsuccessful, with the wines treated with pectolytic enzymes and ellagic tannin (ST_E+T) having, in general, similar or higher levels of all the measured volatile phenols compared to the control wine made using standard experimental winemaking protocols (ST_C). The reference wines for panel 2 were from Oakville, AVA, harvested before and after smoke exposure, with an increase in volatile phenol concentrations for all measured compounds in the wines made from smoke-exposed versus not-smoke-exposed grapes.
2.2. Sensory Analysis
The trained judges from panel 1 evaluated 30 attributes corresponding to aroma, taste, and mouthfeel for each of the wines made from grapes from Oakville, AVA in triplicate (unsmoked, EC1118, BDX, D80, D254, NB, NF, and QT) (
Supplementary Table S5). Multivariate analysis of variance (MANOVA) indicated that there were no significant differences among the wine treatments (α = 0.05) for the 30 sensory attributes evaluated and, therefore, the subsequent ANOVAs need to be evaluated with caution. ANOVA showed that out of those thirty attributes, only nine were considered significantly different among the wine treatments when analyzed individually. The attribute means and Fisher’s least significant difference (LSD) for the significant attributes are shown in
Table 4. The ‘smoke’ aroma attribute was surprisingly not significantly different among the wine treatments, even though a non-smoke-exposed control treatment was included in the evaluation. All wines were rated smoky at a low level (mean of 1.3) indicating that a smoky aroma was not a major attribute for any of the wines evaluated, but also that there was a small amount of carryover among the samples, even though a 1 min wait time was enforced between samples. In subsequent studies, we determined that longer wait times are needed when evaluating smoke-impacted wines. This was confirmed by a recent study by Fryer et al. [
22] that found a 2 min waiting time to be optimal. Our own investigation indicated that a 90 s wait time using a 0.5% pectin rinse was sufficient for low to medium smoke-impacted wines. Nevertheless, an ashy aftertaste was found to be one of the nine significantly different attributes.
A principal component analysis (PCA) was performed to better visualize differences among the wines (
Figure 1). The first two components (PCs) were able to explain approximately 74% of the data. ‘Fig/dried fruit’ and ‘honey’ aromas, ‘sweet’ tastes, and ‘viscous’ mouthfeel correlated positively with each other and correlated negatively to attributes such as ‘ashy aftertaste’ aroma and ‘astringent’ mouthfeel (
Supplementary Table S6). The non-smoke-exposed control wines (Unsmoked) were rated lower for ‘ashy aftertaste’, while they were also considered to be much higher in ‘sweet’ taste and ‘fig/dried fruit’ aroma. The low ‘ashy aftertaste’ rating does indicate some carryover; however, the panelists were still able to distinguish among the wines based on this attribute.
Wines made from fruit not exposed to smoke were seen as very different from the rest of the wine treatments, despite being made from grapes from the same vineyard block. Thus, another PCA was created without the unsmoked controls to better evaluate the differences perceived by the judges for the treatments applied to the wines from smoke-exposed fruit. The first two components of the PCA shown in
Figure 2 explain 63% of the variance.
The third dimension was constructed, and although it resulted in a better representation of the attributes ‘petrol’ and ‘astringent’, the first two dimensions were chosen as it explained more of the variance and the rest of the attributes were better represented. This third dimension (not shown) explains another 13% of the variance in the data and was able to show that wines fermented with the yeast D254 were more closely correlated with the ‘astringent’ and ‘petrol’ aroma characteristics. As mentioned previously, the wines fermented with EC1118 are considered the reference treatment (standard winemaking practices) for the purpose of this study. Variability between fermentation replicates does not allow clear treatment interpretation. Wines from the Oakville Experimental Station block 6 (OFV6), which were made under the standard experimental winemaking conditions with the yeast EC1118, can be compared with OFV9 as this block came from the same vineyard and received a similar amount of smoke exposure. OFV6 wines were perceived to be sweeter and fruitier, with a decreased ‘ashy aftertaste’. The addition of oak chips (NB, NF) and an ellagic tannin (QT) during fermentation did not have a clear impact on the sensory perception of the wine compared to the control (EC1118).
The trained judges from panel 2 evaluated a total of 27 attributes corresponding to aroma, taste, and mouthfeel for each of the wines made from grapes from the Stag’s Leap District and Alexander Valley AVAs in triplicate. MANOVA revealed significant differences among wine treatments (α = 0.05) for the sensory attributes evaluated (
Supplementary Table S7). ANOVAs showed that 15 attributes were significantly different among all wine treatments. For some of these attributes, there was a significant wine-by-judge and wine-by-replicate interaction and, therefore, a pseudo-mixed model was applied (
Supplementary Table S8). Adjusted F values indicated that for ‘dark fruit’ and ‘leather’ aromas and ‘astringency’ mouthfeel, the effect of the wine treatment is more important than the individual interactions. ‘Dark fruit’, ‘vanilla’, and ‘raisin/prune’ aromas were highly correlated, while mouthfeel attributes ‘hot’ and ‘astringent’ were negatively correlated to the ‘smoky’ aroma character (
Supplementary Table S9). The attribute means and Fisher’s least significant difference (LSD) values for the significant attributes are shown in
Table 5.
Although the non-smoke-exposed wines were rated as ‘smoky’ by panelists, they were rated lowest for the ‘smoky’ aroma attribute while also considered to be on the higher end of attributes such as ‘dark fruit’, ‘vanilla’, and ‘raisin/prune’. To visualize the differences among the wines for the eight significantly different attributes, PCA with a covariance matrix was performed, where the first two principal components (PC) were able to explain 80% of the data (
Figure 3). The first component of the PCA is mainly driven by astringency on the horizontal axis while the second component is driven positively by the fruity aromas and negatively by green and leather characteristics.
There is a clear clustering of the wines made from grapes from different vineyards, as
Figure 3 shows the wines from fruit from Alexander Valley (AV) on the left side and wines made from fruit from Stag’s Leap District (ST) on the right side of the PCA. The non-smoked-exposed (NS_Control) and the smoke-exposed (S_Control) control wines were made from fruit from Oakville, AVA. It appears that wines made from the Stag’s Leap District, AVA were less ‘smoky’, suggesting that the smoke impact in Alexander Valley was greater. The Pocket Fire in Sonoma County was closer to the vineyard in Alexander Valley [
23] than the fires were to the Napa County vineyards. Unfortunately, ‘ashy aftertaste’ was not significantly different among the wine treatments due to the significant wine-by-judge interaction. This could be a result of carryover due to only 90 s in wait time with a water rinse between wines. Different fermentation temperatures (15, 20, and 25 °C, and a cold soak at 10 °C) did not have a significant impact on smoke expression in the resulting wines, while the enzyme and tannin addition treatment seemed to enhance the perception of astringency of the wines.
Figure 4 shows the first and third components, explaining another 9% of the data. Wines from the Stag’s Leap District grapes, especially the treatment with an enzyme and tannin addition (ST_E+T), were rated as being more astringent and not very smoky, while the non-smoke-exposed control wines (NS-Control) were rated as ‘hot’ and the lowest values for ‘smoky’ character. The wines that were cold soaked at 10 °C then pressed and fermented following a white winemaking protocol (AV_10) proved to be better balanced, with some fruit, a mild green character, and lower astringency, although they were still perceived as smoky. Wines fermented at 15 °C correlated with green character, while higher fermentation temperatures did not, and, in general, wines from the Alexander Valley were rated as less astringent than the rest of the wines evaluated. Overall, the wines used as controls for both smoke- and non-smoke-exposed fruit from the Oakville Experimental Station (S-Control and NS-Control) revealed a more intense vanilla character, while the control wines from Stag’s Leap District (ST_C) were perceived to have a higher dark fruit aroma when plotting the first against the fourth component of the PCA (results not shown).