Effects of Frozen Materials Other Than Grapes on Red Wine Aroma Compounds. Impacts of Harvest Technologies

An undesirable sensory attribute (“floral taint”) has been detected in red wines in North America, caused by leaves and petioles (materials other than grapes [MOG]) introduced during mechanical harvest after killing frosts. From 2017 to 2019, several harvest strategies were evaluated on Ontario Cabernet franc: hand harvest (HH), conventional machine harvesting (MECH), Braud-New Holland Opti (OPTI), Gregoire GL8, MECH + optical sorting (MECH+OS), and MECH with preharvest leaf removal (MECH+BLR). Concentrations of 41 odor-active compounds were quantified by gas chromatography-mass spectrometry. Harvest treatment effects varied by season. In 2017, HH resulted in lowest ethyl isobutyrate (MECH+BLR), ethyl nonoate, cis-linalool oxide (plus MECH and OPTI), trans-linalool oxide (plus MECH+OS), β-citral, and cis- and trans-rose oxide (plus MECH and OPTI). Ethyl hexanoate was lowest in MECH, and MECH+BLR, isoamyl hexanoate was lowest in all treatments except HH, and α-ionone was lowest in MECH and MECH+BLR. In 2018, HH resulted in the lowest β-damascenone, ethyl salicylate (plus OPTI and Gregoire), citronellol (plus Gregoire), cis- and trans-rose oxide (plus Gregoire), and eugenol (plus Gregoire). Isobutyl acetate, isoamyl hexanoate, and nerol were additionally reduced by Gregoire, and isopropylmethoxypyrazine was reduced by all treatments except HH. In 2019, harvest strategy affected 27 of 41 compounds, including 11 esters and 12 terpenes. Treatments leading to lowest concentrations were HH (nine compounds), MECH (eight compounds), MECH+BLR (10 compounds), OPTI (21 compounds), Gregoire (10 compounds), and MECH+OS (22 compounds). Wines from fruit that had undergone a killing frost contained different concentrations of 14 and eight compounds (2018), and 17 and 13 compounds (2019) for Cabernet franc and Cabernet Sauvignon, respectively. Results suggest that specific harvest technologies can reduce MOG and associated increases in aroma compounds, although seasonal differences may occur.

grapes (MOG) in harvest loads. Presence of MOG, e.g., leaves and petioles, can substantially diminish the quality and composition of wines, particularly red wines (Noble et al. 1975, Huang et al. 1988, Guerrini et al. 2018. Many studies were conducted worldwide with the advent of mechanical harvesting in the 1970s (e.g., Christensen et al. 1973, Petrucci and Siegfried 1976, Johnson 1977, Peterson 1979. Normally, mechanized harvest was considered equal to hand harvesting with respect to wine quality; however, most early mechanical harvesters were unable to expel all the MOG accumulated during harvest. Current harvesting technology has largely eliminated this problem except under unusual circumstances, such as with frost-damaged canopies (Parenti et al. 2015, Hendrickson et al. 2016). Many new mechanical harvesting technologies have optical sorting capabilities that will eject leaves, petioles, and unripe fruit Oberholster 2017, Kilmartin andOberholster 2022). However, this equipment may not be the best financial option for grapegrowers to avoid sensory taints. Modifying vineyard management strategies such as time or method of harvest, or hand-picking specific grape varieties, may be suitable alternatives that can mitigate adverse effects of late-season mechanized harvest. This may encourage equipment companies to tailor products more toward individual grape varieties and the level of product quality, as well as the purchasing ability, anticipated by wineries. In some situations, a grapegrower with a modest acreage may choose custom harvesting using new equipment with sorting capabilities rather than purchasing a new harvester. Some grape cultivars (e.g., Cabernet franc) are especially prone to a unique sensory defect, locally referred to as "floral taint," introduced by MOG following killing frosts. Warmer autumns, perhaps due to climate change, have permitted harvests of red wine cultivars such as Cabernet franc as late as mid-to late-November, when foliage had heretofore been killed by frost. It is highly likely that undesirable aroma compounds, e.g., monoterpenes and norisoprenoids, can be introduced by frozen leaves and petioles, and postfrost machine harvesting is a major contributor (Wang et al. 2020). These sensory taints, along with enabling grapegrowers to adapt to a changing climate, must both be addressed; therefore, the wine industry must understand the effects of frozen MOG on basic wine composition and wine aroma compounds, and identify harvest technologies that will minimize its incorporation into harvest loads.
Mechanical harvesting can adversely affect physical characteristics of fruit including berry damage, premature juice release, and increased MOG content (Pezzi et al. 2008, Kilmartin andOberholster 2022). Differences in one study were largely attributed to oxidation during prolonged transport, more than differences between hand-harvested versus machine-harvested wines (Pezzi et al. 2008). Mechanical harvest results in the release of specific enzymes from both fruit and MOG, and these enzymes may come into contact with substrates (Kilmartin and Oberholster 2022). Conventional must and wine composition can consequently be adversely affected by MOG incorporation including ethanol, titratable acidity (TA), pH, malic acid, total phenol concentration, and color (Arfelli et al. 2010, Ward et al. 2015, Guerrini et al. 2018. Wine ethanol concentration can be substantially reduced with high MOG levels (Ward et al. 2015, Guerrini et al. 2018, and the presence of MOG during fermentation can additionally result in wines with higher pH and lower TA, albeit with higher malic acid concentrations (Arfelli et al. 2010, Ward et al. 2015, Guerrini et al. 2018. Incorporation of MOG into the must also has an adverse effect on color and phenolic concentration of red winegrape varieties (Wagener 1980, Huang et al. 1988, Guerrini et al. 2018. Phenolic compounds, e.g., flavanoids and tannins, can transfer from MOG into fermenting wines (Spranger et al. 2004, Suriano et al. 2016. MOG, especially petioles that are incorporated into fermentations, can also absorb anthocyanins from the fermenting must, resulting in decreased color intensity (Pascual et al. 2016, Suriano et al. 2016. However, increased MOG is frequently associated with an increase in total phenols, mainly due to elevated flavonoid concentrations in MOGaffected wines (Huang et al. 1988, Arfelli et al. 2010, Guerrini et al. 2018. High MOG levels can consequently also lead to increases in the tannin and anthocyanin concentrations in wines, and as a result, inclusion of MOG has an adverse effect on the hue and color intensity (Wagener 1980, Huang et al. 1988, Guerrini et al. 2018, often in fact leading to increases in both color intensity and hue (Guerrini et al. 2018). These alterations in the chemical composition can potentially influence the overall wine sensory profile and quality (Noble et al. 1975, Guerrini et al. 2018. MOG may likewise affect concentrations of wine aroma compounds (Kilmartin and Oberholster 2022). The concentration and composition of volatile compounds vary depending on grapevine organ, especially among the vegetative tissues (e.g., stems, tendrils, rachis, peduncles), in addition to the berries themselves (Matarese et al. 2014). The volatile compound composition associated with the rachis, peduncles, stems, and tendrils are composed mainly of monoterpenes such as geraniol, linalool, and β-citronellol (Matarese et al. 2014). Remaining organs including leaves are composed mostly of aliphatic C6 compounds such as hexanal; 2-hexenal; 2,4-heptadienal; 2,4-hexadienal; plus 1-octen-3-ol; which are characterized by a green odor (Matarese et al. 2014). Other volatile compounds contributed by MOG are present at lower concentrations, e.g., methyl salicylate, benzyl alcohol, benzaldehyde, 2-phenylethanol, various norisoprenoids, and eugenol (Gunata et al. 1986, Matarese et al. 2014. Harvested fruit can likewise contribute to high concentrations of C6 compounds as a result of lipoxygenase activity (Kilmartin and Oberholster 2022).
Early studies found few sensory or aroma compound differences among hand-harvested and mechanically-harvested treatments (Noble et al. 1975). No increases were reported in the concentrations of leaf volatiles, such as trans-3-hexenal and cis-3-hexenol in machine-harvested must, and no sensory differences were found between hand-harvested and machine-harvested wines (Joslin and Ough 1978). However, more recent work with Pinot noir using optical berry sorting in combination with mechanized harvest demonstrated positive effects of the mechanized harvest method as well as postharvest optical sorting (Hendrickson et al. 2016, Hendrickson andOberholster 2017). Hand-harvested fruit that was subsequently sorted produced wines with the highest dark-fruit aroma, while mechanical harvest that included the Selectiv' process (onboard MOG removal) led to higher tropical fruit aromas (Hendrickson et al. 2016, Kilmartin andOberholster 2022). Mechanically-harvested grapes had higher concentrations of linalool, β-myrcene, α-terpinene, and β-damascenone, presumably due to glycosidic hydrolysis triggered by berry damage during harvest (Hendrickson et al. 2016). Differences in wine composition attributable to harvest treatment were reduced by postharvest optical sorting.
Floral sensory taints, observed in Ontario's commercial red wines, have been associated with mechanically-harvested grapes, so therefore it was hypothesized that increased concentrations of undesirable aroma compounds such as monoterpenes and norisoprenoids are being introduced by frozen leaves and petioles. Despite being present in very small concentrations, terpenes can have a considerable effect on the sensory properties of grapes and wines. Floral monoterpenebased aromas are typical and desirable in white wines such as Muscat, Riesling, and Gewürztraminer (Marais 1983), however, they are atypical in red table wines (Ferreira et al. 2000). It was estimated that >5% of petioles were required to substantially alter sensory attributes, particularly increasing terpene-based floral aromas in Cabernet Sauvignon (Ward et al. 2015). It is possible that prolonged autumn seasons associated with climate change, while permitting harvests later in the season, also allow for odor-active compounds such as terpenes to develop to undesirable concentrations in mature fruit. These compounds may likewise be introduced through processes such as postfrost machine harvesting. It was also hypothesized that methoxypyrazines would be reduced because of light freezing of grapes that could occur along with late harvesting, consistent with local industry anecdotal evidence. The objectives of this study were to investigate the effect of several harvesting strategies on Cabernet franc for their ability to reduce frozen MOG that negatively affects wine composition and quality. The second objective was to specifically assess the effect of freezing itself on the concentrations of aroma compounds in Cabernet franc and Cabernet Sauvignon.
A preliminary analysis of the effect of frozen MOG has appeared (Wang et al. 2020). This work extends that study to three vintages, 2017, 2018, and 2019, with a more robust data analysis.

Materials and Methods
Harvesting treatments. All Cabernet franc (Vitis vinifera L.) grapes were harvested following a hard frost in the 2017 to 2019 seasons from the Andrew Peller Ltd. Carlton St. vineyards, located in the Niagara Peninsula Vintners Quality Alliance (VQA) subappellation of Four Mile Creek in Niagara-On-The-Lake, Ontario, Canada. Entire rows (≈750 kg of fruit) were designated for each treatment. Treatments were hand harvest (HH), conventional mechanical harvest (MECH), MECH preceded by full canopy mechanical leaf removal (MECH+BLR), Braud-New Holland 9060L MECH with Opti-Grape system (OPTI; CHN Industrial), Gregoire GL8 MECH with EasyClean Destemmer Sorter (2018, 2019 only; SDF Group), and MECH followed by in-winery optical sorting (MECH+OS; 2017, 2019 only; Selectiv' Process Vision 2, Pellenc ST). Conventional MECH and MECH+BLR were performed in 2017 with a Braud-New Holland BRAUD 9060L without the Opti-Grape system implemented; in 2018 and 2019, MECH was performed with a Gregoire GL8 without the EasyClean Destemmer Sorter option implemented. Following harvest and/or in-winery optical sorting, grapes from each treatment were divided equally into three replicates for subsequent winemaking. Harvest dates were 14 Nov 2017, 11 Nov 2018, and 13 Nov 2019.
Frost treatments. Cabernet franc and Cabernet Sauvignon grapes were harvested prior to and following a hard frost in the 2018 to 2019 seasons from the Andrew Peller Carlton St. vineyards. Entire rows (≈750 kg of fruit) were designated for each treatment. Following harvest, grapes from each treatment were divided equally into three replicates for subsequent winemaking. Pre-and postfrost harvest dates were 4 versus 13 Nov 2018, and 5 versus 16 Nov 2019 for both cultivars).
Winemaking. Following destemming, must was treated with 50 mg/L potassium metabisulfite. Fermentations were performed in triplicates of 40 kg in 46-L plastic fermentation buckets. The fermentation vessels were then placed in a 24°C fermentation chamber and allowed to warm up overnight. Juice samples were taken immediately prior to inoculation and frozen at -25°C for future analysis. Fermentations were inoculated with 350 mg/L of yeast strain CSM (Lallemand) one day after harvest. An addition of 200 mg/L diammonium phosphate was made 24 hrs after inoculation. Fermentations were hand plunged twice daily, and fermentation kinetics (sugar concentration and temperature) were monitored daily. Seven days after destemming, the must was pressed and the wine was inoculated with the malolactic bacteria strain LAC-TOENOS SB3 Direct (Laffort).
Conventional analysis. Conventional chemical analysis (e.g., ethanol, TA, pH, total anthocyanins, total phenols) was also performed using standard methods. Wine pH was obtained using standard methods (Reynolds and Wardle 1989). Wine TA was measured with a PC-Titrate autotitrator (Man-Tech Associates) to a pH 8.2 end point. Color intensity and hue were determined using a modified method provided by Mazza et al. (1999) and were calculated from absorbance values measured at 420 and 520 nm on an Ultrospec 2100 Pro UV-vis spectrophotometer (Biochrom Ltd.). Total anthocyanins were measured by the pH shift method (Fuleki and Francis 1968). Total phenols were determined according to the Folin-Ciocalteu micro method (Singleton andRossi 1965, Waterhouse 2001).
Gas chromatography-mass spectrometry (GC-MS). Wines were analyzed using GC-MS with Gerstel thermal desorption technology according to previous methods Reynolds 2012, Moreno Luna et al. 2018). A 30-mL sample was taken from each wine treatment replicate immediately prior to bottling and was kept at 4°C in the presence of N2 inert gas until analysis. In duplicate, 100 μL of an internal standard, prepared with 10 μL of 98% 1-dodecanol (Sigma-Aldrich) in 10 mL of 100% ethyl alcohol (Commercial Alcohols, Inc.), was poured to the mark into a 10-mL volumetric flask followed by the addition of wine and then mixed. The prepared sample was transferred into a 10-mL Gerstel extraction vial followed by the addition of 10 mL of a saturated sodium chloride solution. A 10-mm stir bar (Twister, Gerstel, Inc.) coated with polydimethylsiloxane (0.5 mm film thickness) was added to the sample and stirred for 1 hr at 1000 g for extraction at room temperature. After extraction, the stir bar was removed, rinsed with Milli-Q water (Millipore) and dried out with lint free tissue, then placed in a 4-mL amber vial at 4°C until analysis during the same day. The stir bar was then inserted inside the extraction glass tube, which was thereafter placed inside the thermal desorption unit (TDU) attached to the GC-MS.
GC-MS conditions and conditioning of materials. An Agilent 6890N/5975B GC-MS equipped with a Gerstel TDU, cooled injection system, and programmable temperature vaporization was used. Columns were Agilent 19091S-433 HP-5MS 5% phenyl methylsiloxane, nominal length 30.0 m, nominal diameter 250.00 μm, nominal film thickness 0.25 μm; and J&W 122-7032 DB-WAX nominal length 30.0 m, nominal diameter 250.00 μm, nominal film thickness 0.25 μm. Instrument conditions were identical to Bowen and Reynolds (2012) and Moreno Luna et al. (2018). MS information included solvent delay: 3 min; scan acquisition method for identification compounds, low mass: 30; high mass: 400; threshold: 150; and selective ion monitoring/scan mode for quantification of aroma compounds. Stir bars used for extraction were conditioned every time before use to avoid any cross contamination. After analysis, each stir bar was kept overnight in a solution of 80:20 acetonitrile/methanol, allowed to dry, and then placed at 250°C for 2 hrs with a constant flow of nitrogen (N2) inert gas. All glass material was washed with Milli-Q water and methanol, then dried at 250°C for 1 hr.
Model wine was used for calibration curves and prepared based on Bowen and Reynolds (2012), using 12% (v/v) of pure anhydrous ethanol (Commercial Alcohols, Inc.) diluted in Milli-Q water and 5 g/L of tartaric acid. The pH of the model wine was adjusted to 3.6 with 1N sodium hydroxide. Each aroma standard was diluted first in pure anhydrous ethanol at 1000 mg/L and kept at 4°C until analysis, then diluted at different concentrations in the model wine. Calibration samples were analyzed in SIM/SCAN mode using the same conditions as described previously with the use of the same internal standard. Odor activity values (OAVs) were calculated as a ratio between each concentration obtained by calibration versus their respective threshold. Thresholds were obtained from literature (Buttery et al. 1968, 1982, Ruth 1986, Takeoka et al. 1990, Etiévant 1991, Guth 1997, Ferreira et al. 2000, Plotto et al. 2004. These data were used to generate concentrations of aroma compounds for comparative analysis. A list of these compounds and their aroma descriptors is found in Table 1. Methoxypyrazines were quantified by GC-MS using the methods of Kotseridis et al. (2008).
Statistical analysis. All data were analyzed using XL-Stat (Addinsoft). Data were also subjected to analysis of variance with mean separation by Duncan's multiple range test, p < 0.05. All bar plot figures were prepared using the 'ggplot' function in the 'ggplot2' package in an R environment (http://www.r-project.org/).

Results and Discussion
Effect of harvest strategy. 2017 season. In 2017, hand harvesting resulted in lowest concentrations of ethyl isobutyrate (plus MECH+BLR), ethyl nonoate, cis-linalool oxide (plus MECH and OPTI), trans-linalool oxide (plus MECH+OS), β-citral (plus MECH and OPTI), and cis-and trans-rose oxide (plus MECH, OPTI, and MECH+OS) (Figure 1 and Supplemental Table 1). Additionally, ethyl hexanoate was the lowest in MECH and MECH+BLR, isoamyl hexanoate was lowest in all treatments except HH, and α-ionone was lowest in MECH and MECH+BLR. Overall, the treatments led to the lowest concentrations as follows: HH (seven compounds), MECH (six compounds), MECH+BLR (four compounds), OPTI (five compounds), and MECH+OS (three compounds).
β-damascenone (OAV= 433, 23, 291) and α-ionone (OAV= 9.2,11,19) were also responsive to harvest treatments and might therefore be assumed to play some role in floral taint. Basic wine composition was affected to a limited degree by harvest technologies, and these strategies also varied between seasons (Supplemental Table 4). The highest pH was measured in the MECH+BLR wines (2017,2018) and in HH (2019) The high potential of grape leaves being integrated into the fermentation following mechanical harvesting poses a Figure 1 Effect of harvest technology on Ontario Cabernet franc aroma compounds, 2017. Control, hand harvested; OPTI, mechanically harvested using a Braud-New Holland optical sorting harvester; MECH+OPTI, conventional mechanical harvest followed by optical sorting; MECH+BLR, conventional mechanical harvest preceded by mechanical leaf removal; MECH, conventional mechanical harvest. Adapted with permission from Wang et al. (2020). Bars denoted by different letters are significantly different at p < 0.05, Duncan's multiple range test. particular risk to the final wine composition (Noble et al. 1975, Wildenradt et al. 1975, Ward et al. 2015. Leaf blades (laminae) contain several volatile compounds that could potentially be extracted into wine, primarily C6 aldehydes and alcohols (2-hexanal; 2-hexen-1-ol; n-hexanol; 2-hexenol-1-ol; and hexa-2,4-dienal), terpenes (linalool, geraniol, citral, nerol, citronellol, α-terpineol, and myrcene), and norisoprenoids (β-damascenone, α-, and β-ionone) (Wildenradt et al. 1975, Gunata et al. 1986, Wang et al. 2020, although as shown in this study, many of these compounds are not necessarily detected at concentrations that exceed sensory threshold. The C6 compounds responsible for the "grassy" characteristics associated with grape leaves likely originate from lipoxygenase activity on the fatty acids in leaf cellular structures (Wildenradt et al. 1975). Petioles have high concentrations of monoterpenes, especially citronellol and geraniol (Gunata et al. 1986). The petiole may act as a storage instrument for free terpenes prior to transportation to other parts of the vine, or for utilization in metabolic pathways such as the mevalonate pathway, hence the high concentrations of terpenes in petioles compared to the laminae (Gunata et al. 1986).

Figure 2
Effect of harvest technology on Ontario Cabernet franc aroma compounds, 2018. Control, hand harvested; Gregoire, mechanically harvested by a Gregoire GL8 harvester with optical sorting; OPTI, mechanically harvested using a Braud-New Holland optical sorting harvester; MECH+BLR, conventional mechanical harvest preceded by mechanical leaf removal; MECH, conventional mechanical harvest; IBMP, isobutyl methoxypyrazine; IPMP, isopropyl methoxypyrazine; SBPM, sec-butyl methoxypyrazine. Bars denoted by different letters are significantly different at p < 0.05, Duncan's multiple range test.
Juice contact with MOG during fermentation results in extraction of several volatile grapevine compounds into the final wine (Ward et al. 2015, Suriano et al. 2016, Guerrini et al. 2018, Wang et al. 2020. The presence of increasing MOG during fermentation leads to wine with higher concentrations of several monoterpenes, including geraniol, linalool, nerol, β-citronellol, and α-terpineol, as well as other aroma compounds (Ward et al. 2015, Suriano et al. 2016, Guerrini et al. 2018, Wang et al. 2020. In the presence of high MOG concentrations, geraniol, linalool, and β-citronellol may be found at concentrations above their sensory detection thresholds, suggesting a potential effect on the perceived sensory profile (Ward et al. 2015). However, only two terpenes (geraniol and cis-rose-oxide) were found at OAV values >1 in this study. Higher alcohols and esters increase in the presence of stems (Suriano et al. 2016, Guerrini et al. 2018. Benzyl alcohol, eugenol, 1-hexanol, methyl salicylate, and ethyl salicylate can also increase as a result of MOG incorporation (Ward et al. 2015, Guerrini et al. 2018. However, contrary to other aroma compounds found in MOG treatments, methoxypyrazines can be reduced in the presence of high petiole concentrations (Ward et al. 2015). Ward et al. (2015) hypothesized Figure 3 Effect of harvest technology on Ontario Cabernet franc aroma compounds, 2019. Control, hand harvested; Gregoire, mechanically harvested by a Gregoire GL8 harvester with optical sorting; OPTI, mechanically harvested using a Braud-New Holland optical sorting harvester; MECH+OPTI, conventional mechanical harvest followed by optical sorting; MECH+BLR, conventional mechanical harvest preceded by mechanical leaf removal; MECH, conventional mechanical harvest. Bars denoted by different letters are significantly different at p < 0.05, Duncan's multiple range test. that petioles may act to adsorb 3-isobutyl-2-methoxypyrazine from the fermenting wine.
It is noteworthy that very few studies have focused on the effects of mechanical harvesters with optical sorting capabilities, or optical sorting in general, and none have addressed the effects of incorporating frozen MOG into harvest loads on ultimate wine quality. Mechanical harvesting has been documented as affecting concentrations of numerous wine aroma compounds. Esters, including isoamyl acetate and hexyl acetate, were increased in mechanically harvested Montuni grapes (Arfelli et al. 2010). Several esters (ethyl 2-methylbutyrate, ethyl acetate, ethyl hexanoate, ethyl oc-tanoate, isoamyl acetate) differed between harvest treatments in Pinot noir in California (Hendrickson et al. 2016), although as with this study, these compounds were not the main drivers of differences between hand harvest, mechanical harvest, and optical sorting treatment combinations. Higher alcohols have likewise been affected, including increases in hexanol and butanediol, plus reductions in isoamyl alcohol and phenylethyl alcohol (Arfelli et al. 2010). As with esters, Hendrickson et al. (2016) reported differences in several alcohols (benzyl alcohol, cis-2-hexen-1-ol, cis-3-hexen-1-ol, trans-2-hexen-1-ol, and trans-3-hexen-1-ol) between harvest treatments, but again as with the current study, none were the main foundations of Figure 3 continued Effect of harvest technology on Ontario Cabernet franc aroma compounds, 2019. Control, hand harvested; Gregoire, mechanically harvested by a Gregoire GL8 harvester with optical sorting; OPTI, mechanically harvested using a Braud-New Holland optical sorting harvester; MECH+OPTI, conventional mechanical harvest followed by optical sorting; MECH+BLR, conventional mechanical harvest preceded by mechanical leaf removal; MECH, conventional mechanical harvest. Bars denoted by different letters are significantly different at p < 0.05, Duncan's multiple range test.  treatment differences. Several organic acids have been reduced by mechanical harvesting, including isobutyric, isovaleric, and phenylacetic acids, whereas hexanoic, octanoic, and decanoic acids have been reported to increase, but their sensorial significance was not discussed (Arfelli et al. 2010).
Of greatest concern in this study were those compounds with low sensory thresholds-terpenes and norisoprenoids in particular-which were likely to be the main source of floral taint in wines produced from fruit mechanically harvested subsequent to killing frosts. In Australia, Cabernet Sauvignon wines produced from fruit with high MOG concentrations had geraniol, linalool, and β-citronellol concentrations above their detection thresholds, suggesting a potential sensory effect (Ward et al. 2015). Machine-harvested Pinot noir grapes in the Russian River Valley in California had higher concentrations of linalool, β-myrcene, α-terpinene, and β-damascenone, which were attributed to glycosidic hydrolysis due to berry damage during harvest or from wounding response-induced synthesis (Hendrickson et al. 2016). Other terpenes affected by mechanical pruning treatments included geraniol, nerol, nerol oxide, citronellol, and two sesquiterpenes (Hendrickson et al. 2016).
The effect of frost on the ultimate basic wine composition was limited for both cultivars (Supplemental Table 5). For Cabernet franc, prefrost wines had slightly higher pH in 2018, but there were no other effects in that season and none in 2019. For Cabernet Sauvignon, there were no effects in 2018; freezing increased pH and reduced TA in 2019. No other effects were observed.
Studies into preharvest freezing of grapes intended for table wines are uncommon. Lan et al. (2016) reported that Beibinghong berries sampled in the last stages of maturity (which included freezing) were characterized by several terpenes (e.g., cis-and trans-linalool oxides and cis-and transrose oxides), norisoprenoids (cis-and trans-theaspirane and vitispirane A and B), and higher alcohols (e.g., heptenol, 2-octanol, and 1-butanol). In a related study involving comparisons with table wines produced from nonfrozen grapes, Beibinghong icewines made from frozen grapes were characterized by higher concentrations of volatile phenols, lactones, β-damascenone, phenylacetaldehyde, and diacetyl, versus corresponding dry wines made from fresh grapes (Lan et al. 2019). These differences were ascribed to "complex reactions induced by water loss and freezing-thawing events" during the on-vine dehydration and freezing processes. Delayed harvest of four cultivars in Ontario (Pinot gris, Riesling, Cabernet franc, and Cabernet Sauvignon), which involved brief freezing episodes for the red cultivars, resulted in substantial changes in most wine aroma compounds (Moreno Luna et al. 2018). Pinot gris and Riesling displayed increases in monoterpenes, norisoprenoids, esters, aldehydes, and alcohols in wines from late harvested fruit. Reduced concentrations of volatile acids and green odor compounds (e.g., 1-hexanol) with delayed harvest were also evident. Delayed harvest was also associated with the production of benzaldehyde, diethyl acetal, and higher concentrations of higher alcohols (e.g., isoamyl alcohol and nonanol), which could have been associated with preharvest desiccation rather than freezing. Cold maceration (without freezing) with automatic pump-over of Cabernet Sauvignon resulted in decreases in isobutanol and isopentanol and an increase in some esters (especially acetate esters) (Cai et al. 2014). Ethyl 2-hexenoate and diethyl succinate were decreased, and geranyl acetone was increased in both pump-over and punch-down fermentations, whereas β-damascenone increased by cold maceration in pump-over fermentations but decreased in punch-down fermentations.
Studies with harvest dates for icewine grapes suggest effects of multiple freezing events on aroma compounds. In an investigation of Ontario Vidal blanc and Riesling icewines, the highest concentrations for most aroma compounds were found in the latest of three harvest dates (16 of 24 for Vidal; 17 of 23 for Riesling) (Bowen and Reynolds 2015). The latest harvest date had the highest ethyl isobutyrate, ethyl 3-methylbutyrate, 1-hexanol, 1-octen-3-ol, 1-octanol, cis-rose oxide, nerol oxide, ethyl benzoate, ethyl phenylacetate, γ-nonalactone, and β-damascenone. The earliest harvest, which still had experienced at least one freezing episode, had the highest ethyl butyrate, ethyl hexanoate, linalool, 4-vinylguaiacol, and ethyl octanoate. However, there were no table wines produced in this study with which to compare the icewines to ascertain the effects of freezing. Similarly, Lutskova and Martirosyan (2021) compared early and later icewine harvests for four cultivars (Rkatsiteli, Telti kuruk, Marselan, and Moldova) and reported differences in aroma compounds; for example, Rkatsiteli icewine obtained from an early icewine harvest contained highest concentrations of geraniol, 1-octanol, and 2-phenyl acetate, while highest concentrations of ethyl hexanoate and ethyl octanoate were detected in the icewines from red cultivars Moldova and Marselan from a later harvest. However, once again, no wines produced from nonfrozen grapes were included for comparison.

Conclusions
These results suggest that specific harvest technologies can reduce MOG and its associated increases in undesirable aroma compounds, although seasonal differences may occur. It is apparent that mechanical harvesters with sorting capabilities can reduce MOG and consequently lower concentrations of compounds, such as terpenes, that may be associated with floral taints. In-winery optical sorting following conventional mechanical harvesting is likewise efficacious. A limited number of ethyl esters (ethyl isobutyrate, ethyl hexanoate, ethyl heptanoate, and ethyl octanoate); three acetate esters (isoamyl acetate, isobutyl acetate, and hexyl acetate); one terpene (cis-rose oxide), and two norisoprenoids (β-damascenone and α-ionone) were responsive to harvest treatments and also found at concentrations above sensory thresholds. Moreover, several esters, terpenes, and norisoprenoids increased following killing frost in hand-harvested grapes, suggesting that enhanced harvest technologies may not entirely overcome issues of floral taints in red wine cultivars.