Abstract
Five Chardonnay vineyards in the Niagara Peninsula were delineated according to soil texture using global positioning systems and geographic information systems (GPS/GIS). For three vintages, sentinel vines were classified based on weight of cane prunings (vine size) and the soil texture in which they were grown. Wines were made from vine size × soil texture combinations for the five sites and three vintages. Musts were analyzed for soluble solids, titratable acidity (TA), and pH, and wines analyzed for ethanol, TA, pH, and total phenols. Several within-site differences between vine size or soil texture were observed for must and wine composition, but there were few consistent trends. Wines were analyzed sensorially by descriptive analysis. Soil texture effects were evident at three sites (1999), four sites (2000), and three sites (2001). Three sites in 1999 and two sites in 2000 and 2001 demonstrated vine size effects. No consistent soil texture or vine size effects were observed within any site. Principal component analysis was used to discern trends in sensory data. No clear trends with respect to soil texture or vine size were obvious in 1999, but 2000 wines separated on the basis of site (Lakeshore versus Lake Plain/Escarpment) and 2001 wines separated based on soil textures, whereby wines from Lakeshore sites (2000) or clay textures (2001) were associated with citrus, vegetal, earthy, and astringency descriptors. Vintage and site effects were evident when comparing the 2000 and 2001 vintages; 2000 wines were characterized by citrus, vegetal, and earthy descriptors, while 2001 wines were associated with floral and melon descriptors. Results suggest that although soil texture and vine size impact wine sensory attributes, site and vintage appear to play the major roles.
The term terroir has different implications depending on the language: in French it may mean simply “soil,” whereas in English it may also imply human and climatic factors that influence wine quality (Van Leeuwen 2010). Terroir began as a definition of the distinct and unique qualities of a wine imparted on it by the “total natural environment” in which it is grown, namely the soil, topography, and climate (Van Leeuwen 2010). Taken as a derivative of the soil itself, the question becomes whether or not it is appropriate to isolate a single variable in determining the quality and origin of a wine. No discernible differences in aroma and flavor attributes were found between Australian wines grown on different soil textures (Rankine et al. 1971). Soil had some influence during one vintage on sensory attributes of Chenin blanc and Cinsault in the Stellenbosch region of South Africa (Saayman 1977). California Chardonnay wines produced from vineyards with differing soil types were also not different (Noble 1979). The conflicting opinions that exist about terroir may very well stem from an inability to properly define the term. In the Franken region in Germany, seven soil types were assembled in concrete bins on each of 11 sites and the vines were grown therein, ensuring a climate consistent across all soil types throughout the test blocks (Wahl 1988). Yields varied somewhat with water and nutrient supply associated with soil texture, but wine quality was heavily influenced by climatic factors such as wind, soil temperature, and humidity much more than by soil type.
France pioneered the definition of terroir as a means of distinguishing among regions through the institution of its Appellations d’Origine Contrôlée in response to concerns over the authenticity of wines labeled as from its preeminent producers (Seguin 1986). France, Germany, California, New Zealand, and many other regions have developed the idea of terroir in order to classify regions within their winemaking industry, further differentiating the local wines from those produced elsewhere (Seguin 1986). The winegrowing region of the Niagara Peninsula has been divided into subappellations based on a combination of geology, soil, and mesoclimatic influences. At least three Niagara viticultural regions have been clearly defined based on sensory analysis of commercial Riesling (Douglas et al. 2001), Chardonnay (Schlosser et al. 2005), and red wines produced from Bordeaux cultivars (Kontkanen et al. 2005).
Global positioning systems (GPS) coupled with geographical information systems (GIS) are recently applied technologies in the wine industry. GIS technology has been used to define the suitability of various regions in New York State for grape production through digital climatic, geological, and topographical mapping (Magarey et al. 1997). Such technology has also been used to study vineyard site selection in Virginia (Wolf 2001) and to distinguish between high and low vigor “management zones” in Zinfandel vineyards in California (Greenspan and O’Donnell 2001). In Ontario, GPS and GIS were used to map vine size and soil texture zones in a Riesling vineyard, and it was found that varietal typicity varied in accordance to these factors (Reynolds et al. 2007).
Multivariate statistical procedures such as principal component analysis (PCA) are techniques for simplifying a complex array of results to determine relationships that may otherwise remain hidden (Heymann and Noble 1989). PCA is widespread in sensory evaluation of wines, such as in delineating two Catalan viticultural regions (Larrechi and Rius 1987). Cluster analysis and other statistical tools were useful in defining adjacent viticultural zones in Spain based on wine chemical composition (Latorre et al. 1992). However, although the effects of soil and mesoclimate on sensory properties of wines have been investigated widely in Europe, there have been fewer studies in North America. Noble and Shannon (1987) used PCA to draw comparisons between the chemical and sensory composition of Zinfandel wines from different regions and vintages in California. Similar techniques were applied to distinguish between commercial Pinot noir and Chardonnay wines produced from different parts of the Okanagan Valley, British Columbia (Cliff and Dever 1996).
There are relationships among canopy management, aroma compounds in the grape berries, and intensity of wine varietal character, as well as among mesoclimate, flavor compounds, and wine sensory attributes with soil type held constant (Reynolds et al. 1995). Despite considerable research, it is still not clear whether soil is a primary determinant of wine quality or whether soil is simply a medium that impacts vine growth and vigor, and therefore the skill by which this vigor is accommodated ultimately determines wine quality. This study was intended to address this question with assistance from geomatic tools such as GPS and GIS.
Ontario vineyards are typically located on sites that contain heterogeneous soil types. It was hypothesized that soil texture would play a minor role in the determination of wine sensory attributes and that vine vigor, crop size, and fruit environment would play the major roles. This hypothesis was previously tested in a study with Riesling in the Niagara Peninsula (Reynolds et al. 2007). The current study attempted to resolve the question of direct soil effects by testing independent effects of soil and vine size on must and wine composition and wine sensory attributes of Chardonnay. Vines were geolocated by GPS/GIS and defined with respect to soil texture and vine size. Using this information, grapes were harvested according to category and fermented separately. It was thereby possible to isolate the impact of these variables on chemical and sensory aspects of the wine. The objective was to assess relative influences of both vine vigor and soil texture to either support or refute the Old World soil-based terroir model. An improved understanding of the impact of soil and vine size on wine quality could have substantial consequences for choice of future vineyard sites, cultural practices, grape cultivars, and rootstocks. Elucidation of flavor profiles from specific vineyard blocks could lead to site-specific cultural practices.
Materials and Methods
Site descriptions.
Five vineyard cooperators were selected: two in the Lakeshore region of Niagara-on-the-Lake (Buis, Falk), two in the Lakeshore Plain region (Chateau des Charmes [CDC], Lambert), and one on the Niagara Escarpment near Vineland (Wismer). All vineyard blocks had heterogeneous soil types, particularly with respect to soil texture (Kingston and Presant 1989) and hence were quite variable in terms of yield and vine size. In each vineyard block, a grid-style sampling pattern was established with a sentinel vine at each grid intersection point. These sampling sites (sentinel vines: 72 to 162 per vineyard block) were conspicuously marked. A GPS at <1 m accuracy was used in May 1998 to delineate the shape and size of the vineyard blocks, to provide detailed soil texture maps, and to geolocate each of the sentinel vines used for data collection. A GBX-12R GPS unit (CSI-Wireless, Calgary, Canada) was used in conjunction with an MGL-3 antenna, receiving beacon differential on frequency 322 from Youngstown, NY. Soil samples (~200 g) were collected using a 3 cm × 75 cm (diameter × length) soil probe near each sentinel vine in September 1998. Soil analyses, including elemental concentration, cation exchange capacity (CEC), base saturation (BS, as %Ca, Mg, and K), pH, and organic matter (OM) concentration, were performed on each soil sample. Proportions of sand, silt, and clay (mechanical analysis by hydrometer) were determined on subsamples, and soil texture and composition maps of each vineyard block were thereafter constructed from this information using GIS programs MapInfo and Vertical Mapper (Northwood GeoScience, Ottawa, Canada).
To minimize confounding variables, sites were chosen carefully based on common rootstocks, clones, training systems, canopy management, and overall management excellence and skill. Sites were planted to similar clones (ENTAV 75, 95, 96); three exclusively had ENTAV clone 96. Rootstocks were either SO4 (Buis, CDC, Falk, Lambert) or Couderc 3309 (Wismer). Crop load was controlled entirely by pruning; it is unusual for growers in Niagara to cluster-thin Chardonnay. Crop loads (Ravaz indices; RI) were all low (<11 RI) throughout the various sites across all years. Canopy management was likewise very similar across producers; the vigorous sites (Falk, Buis, Lambert, Wismer) used Scott Henry while CDC used standard Guyot training with vertical shoot-positioning. All growers performed midsummer hedging and basal leaf removal. Details of vines sampled, soil types, and vineyard management are found elsewhere (Reynolds and De Savigny 2001).
Climate data.
Climate data for each site, including rainfall and growing degree day (GDD) values for the veraison to harvest period (Table 1), were obtained from the Niagara Agricultural Weather Network weather stations situated as close to the respective sites as possible throughout the Niagara Peninsula. The Falk and Buis sites were adjacent and therefore calculations for both sites were based on data collected at Falk. CDC had highest rainfall in 1999 and 2001, while Wismer had the highest rainfall in 2000. Wismer had lowest rainfall in 1999 and 2001, while Lambert was lowest in 2000. CDC had the highest accumulation of GDD in all three vintages, while Wismer had the lowest accumulation in 1999 and 2000. In the 2001 vintage the lowest GDD accumulation occurred at Falk and Buis.
Vine size and soil texture categories.
In May 1998, GPS was used to delineate vine blocks and geolocate between 72 and 162 vines at each site. Soil and petiole samples were taken at each grid intersection. Soil analysis was undertaken to determine soil texture (sand, silt, clay) based on percent composition, and zones in each block were classified as either sand or clay based on relative percentages of each. Elemental analyses were also undertaken. In addition, vines were classified as to vine size (high vine size; low vine size) based on weight of cane prunings/vine (kg) during winter pruning (Table 2). Vines within one standard deviation of the mean were excluded for winemaking purposes. Using this information, a database of each vineyard site was created using MapInfo and Vertical Mapper software (Northwood GeoScience, Ottawa, Canada). The software was then used to create GIS maps, from which maps of different grapevine and wine components were constructed.
Within each vineyard block, sentinel vines were sorted based on vine size and identified accordingly on GIS-generated maps. Two vine size categories were established at three of the sites (Buis, CDC, Wismer) whereby vines 0.5 standard deviation above or below the mean weight of cane prunings were designated as high or low size, respectively (Table 2). For Falk and Lambert, the relatively low number of sentinel vines necessitated a single vine size category (medium) based upon those vines <0.5 standard deviation from the mean. Using soil texture maps derived from previously obtained data (Reynolds and De Savigny 2001), small and large vines located within high-clay and high-sand zones in each vineyard could be identified (Figure 1).
Winemaking.
At harvest, fruit from each vine size × soil texture category was combined into two replicates, and wines were made from each vine size × soil type × site replicate combination. Grapes were harvested for wines one to two days prior to commercial harvest in the 1999, 2000, and 2001 vintages, with the exception of Lambert in 2000 (due to powdery mildew). Grapes from each site were crushed and destemmed separately. A solution of 50 mg/L potassium metabisulfite was added to each batch and then crushed grapes were placed at 2°C for skin contact for 12 hr. Each batch was pressed using a bladder basket press to 2 bars and the juice transferred to carboys for settling overnight. The juice was racked into clean carboys and inoculated at 25 g/hL with Saccharomyces cerevisiae, strain Bourgoblanc (Lallemand, Montreal, Canada). The wines were fermented to dryness between 15 to 18°C. After completion the wines were racked off the lees and sulfited to 50 mg/L. Each carboy was cold-stabilized at −2°C for at least 48 hr, wines were racked off the tartrate crystals, sulfited with SO2, and stored for 4 months at −2°C. Prior to bottling, each batch was sulfited to 50 mg/L, membrane-filtered to 0.22 μ, and bottled. The bottles were stored at 11°C and 75% relative humidity. Bottling took place ~9 months after harvest in April to May of the year following each vintage.
Must analysis.
A 250 mL must sample was taken at time of pressing from each site × vine size × soil texture fermentation replicate. Must analysis was carried out for each sample in terms of total soluble solids (Brix), titratable acidity (TA), and pH using standard protocols (Zoecklein et al. 1995). Must Brix was measured using a temperature-compensated Abbé bench refractometer (model 10450; American Optical, Buffalo, NY). Must TA was evaluated using a PC-Titrate autotitrator (model PC-1300-475; Man-Tech Associates, Guelph, Canada). Must pH was measured using a Fisher Accumet pH meter (model 25; Fisher Scientific, Ottawa, Canada).
Wine chemical analysis.
Wine TA was evaluated using the PC-Titrate autotitrato (Man-Tech). Wine pH was measured using the same pH meter used for the musts. To measure ethanol, a 5% ethanol standard was initially prepared by transferring 5 mL absolute ethanol into a 100 mL volumetric flask and diluting to volume. Exactly 10 mL of the dilute solution and 1 mL 1-propanol internal standard were transferred to a second 100 mL volumetric flask and diluted to volume. This was repeated with 10 mL and 15 mL volumes of absolute ethanol. The gas chromatograph (GC; HP5700; Hewlett-Packard, Palo Alto, CA) was calibrated using the prepared ethanol standards and a calibration curve. The wine sample (40 μL) was injected into the GC and the peak areas of ethanol and 1-propanol were determined. The peak ratio was recorded and the ethanol concentration of the sample was determined using the equation of the line from the calibration curve. This was repeated in triplicate.
Spectrophotometric analysis of total phenols using the Folin-Ciocalteu reagent was undertaken (Zoecklein et al. 1995). A 5000 mg/L phenol stock solution was prepared by weighing 500 mg dry gallic acid and dissolving it in a 100 mL volumetric flask. Calibration concentrations of 0, 50, 100, 150, 250, and 500 mg/L gallic acid equivalents were thereafter prepared by diluting aliquots of the gallic acid stock solution in 100 mL volumetric flasks. From each of these, 1 mL gallic acid solution was transferred to a second 100 mL volumetric flask and filled with ~60 mL deionized water. Exactly 5 mL of Folin-Ciocalteu reagent was added to each, followed by ~15 mL 20% Na2CO3 solution. Each flask was subsequently brought to volume and left to incubate for 2 hr at room temperature (~20°C). Following incubation, the absorbance values of each standard were read at 765 nm using a Pharmacia Biotech Ultraspec 1000E spectrophotometer (Umea, Sweden). The wine samples were prepared by the same preparation procedure. The concentration of total phenols was interpolated from the gallic acid calibration curve.
Sensory analysis.
Sensory analysis took place 3 months (1999), 8 months (2000), and 3 months (2001) following bottling. For each vintage, five individuals initially assessed all wine samples for aroma to derive suitable descriptors for sensory analysis. A prototype scoresheet was then constructed. Thereafter, appropriate standards for each were prepared (Table 3). A tasting panel of 12 individuals evaluated the standards for compatibility with representative wine samples from each site. The panels consisted of Brock University faculty, staff, and students, and the membership of the panels was consistent across the three vintages. Panelist age varied between 22 and 48 years and all members had extensive experience tasting wines. Adjustments to the composition of the standards were made based on opinions of these 12 judges. No standards were provided for retronasal aroma (flavor), taste, or mouthfeel attributes. Six 1-hr sessions were conducted to ensure that the judges had complete familiarity with the standards and that they were using the scoresheet consistently.
Approximately equal volumes of each sample were poured into clear ISO glasses and covered with clear plastic petri dishes. Judges evaluated the intensity of aroma and flavor descriptors of each sample in duplicate under red light conditions using a 100-point line scale. Each tasting session consisted of two randomized flights of four wines presented in random order with each wine presented once in each flight. A 2-min break was implemented between flights to minimize palate fatigue. The intensity of the aroma and flavor attributes was assessed using CompuSense 4.2 software (CompuSense, Guelph, Canada).
Statistical analysis.
Data were analyzed by analysis of variance (ANOVA) and means were separated by Duncan’s multiple range test using the SAS statistical software package (SAS Institute, Cary, NC). The general linear models procedure was used for ANOVA, and proc corr was used to elucidate relationships among aroma and flavor descriptors. PCA was also used to further elucidate relationships among sensory variables, sites, soil textures, and vine vigor. PCA was performed on the entire data sets without transformation or rotation, using SAS procedures proc corr and proc factor data.
Results
Must composition.
Must Brix tended to be lowest at the Falk site and highest at CDC. Musts from clay soils had higher Brix at Buis (2000) and CDC (1999 and CDC (2000) but were lower in clay regions at Wismer (2000) and CDC (2001) (Table 4). High vine size zones were associated with higher Brix at Buis (1999 Buis (2001), Wismer (2001), and CDC (2001), but Brix was lower in high vine size zones in one case (Wismer 1999). Must TA at Lambert tended to be lowest across the three vintages, while that at Wismer was generally highest. Clay zones produced higher TA at Buis (2001), Lambert (2001), and Wismer (2000), but it was consistently lower in clay regions at CDC (1999–2001). High vine size zones were associated with higher TA at Buis (2000 Buis (2001) and CDC (1999 and CDC (2000) but TA was lower in high vine size zones at Wismer (2000) and CDC (2001). Must pH varied little across vintages and among sites, but tended to be lowest at Wismer and highest at CDC. Must pH was slightly higher in clay zones at Buis (2001) and Falk (2000), but consistently lower in clay zones at CDC (1999–2001). Must pH values were higher in low vine size zones at Buis (2001) but were higher in high vine size zones at Wismer (2000) and CDC (2001). Several soil texture × vine size interactions occurred, but they were not consistent across sites and vintages with respect to effects of the two factors (data not shown).
Wine composition.
Soil texture and vine size differences had little consistent effects on wine ethanol, TA, pH, and total phenols in the 1999 to 2001 vintages (Table 5). Wine TA was lower in samples from sandy soils at Wismer (2000, but higher in sandy regions at CDC (1999). High vine size zones were associated with higher wine TA at Buis (1999), Wismer (1999), and CDC (1999). Wine pH values were highest in wines from Falk, CDC, and Lambert. Wine pH was associated with clay zones at Buis (1999) and CDC (2001), but sandy zones were associated with higher pH at CDC (1999 CDC (2000). High vine size reduced pH at Buis (1999) and Wismer (1999) but slightly increased pH at CDC (2001). Wines from Wismer tended to have lowest mean ethanol levels (most <12%), while Buis and Lambert had wines with the highest ethanol (many >13%). Higher ethanol concentrations occurred in sand zones at Wismer (1999) and in high vine size zones at CDC (1999). No differences between soil texture or vine size were observed in the 2000 and 2001 vintages for total phenols. Lambert wines tended to have highest total phenols (clay, 2001), whereas Wismer wines were lowest overall. As with must composition, a few soil texture × vine size interactions occurred, but no pattern was evident across sites and vintages with respect to effects of the two factors (data not shown).
Sensory analysis.
No aroma differences were detected in the wines in terms of site or in terms of either soil texture or vine size/yield with all sites pooled (data not shown). No interactions were found, suggesting that the magnitude of impact of either soil texture or vine size did not vary substantially from site to site. Acidity was highest at one of the Lakeshore sites (Buis) but the four other sites did not differ from each other (data not shown). Lowest astringency/bitterness was perceived in Falk wines, while a slight increase in intensity/length of finish was apparent as vine size increased (data not shown).
Soil texture 1999.
When sensory data were analyzed by site, several differences between the wines were apparent (Figure 2). Vines located in sandy textures appeared to lead to wines with higher fruity and apple aromas, less cedar aroma, higher fruity and apple flavors, and less cedar flavor. Two sites displayed an association between sand and reduced sweetness in the wines (Figure 2A, C), but one site showed an opposite trend (Figure 2B). Similar opposing trends were observed for astringency/bitterness and body/mouthfeel (Figure 2B, C). One site also had wines from sandy soils associated with higher perception of acidity (data not shown), while another had wines from sandy textures with a longer finish and greater color (Figure 2B).
Soil texture 2000.
Soil texture impacted sensory aspects of the wines at four sites in 2000 (Figure 3). Fine-textured soils at Buis generally produced wines with more intense aroma and flavor attributes in 2000 than did sandy soils (Figure 3A). Clay soils had higher intensities of several attributes, including vegetal and earthy aromas and body/mouthfeel. The intensities of the remaining attributes were effectively the same for both soil types. Falk wines contained more melon and floral aromas as well as acidity when grown on clay textures, while wines from sandy textures had higher intensities of vegetal aromas (Figure 3B). Apple flavor, acidity, astringency/bitterness, and intensity/length of finish tended to be slightly higher in wines produced from clay textures. The wines from CDC differed between soil textures in terms of citrus aroma, apple flavor, and the intensity/length of finish (Figure 3C). Sandy textures were associated with higher intensities in all cases. Wines from both categories were characterized by low earthy aroma and flavor and a high acidity. Clay textures at Wismer in 2000 increased astringency/bitterness in the wines (Figure 3D). All remaining attributes were not impacted by soil texture.
Soil texture 2001.
No soil influence was present at either Buis or Falk in 2001 (data not shown). Lambert produced wines from grapevines grown on fine-textured soils with a higher level of floral flavor (Figure 4A). In addition, sand was associated with more body/mouthfeel as well as greater intensity/length of finish. In contrast, clay textures led to higher intensities of vegetal and earthy aromas. All Lambert wines had relatively high acidity levels. At CDC, fine-textured soils led to more intense earthy aroma and flavor (Figure 4B). Relatively high levels of acidity, astringency/bitterness, body/mouthfeel, and intensity/length of finish characterized all CDC wines. Apple, melon, and floral flavors were also very prominent in both sand and clay categories. In 2001, Wismer wines produced from clay textures contained more melon flavors than those from sandy textures (Figure 4C). Astringency/bitterness of the wines was more pronounced in coarse-textured soil zones. Beyond these attributes, Wismer wines appeared to have similar aroma and flavor profiles.
Vine size and interactions 1999.
High vine size was associated with greater sweetness and a longer, more intense finish (two sites; Figure 5A, C), higher fruity flavor, more body/mouthfeel, and more color (one site; Figure 5C), and less color (two sites) (Figure 5B, C). Several soil texture × vine size interactions were also observed for two sites (data not shown): Buis (cedar and earthy aroma; fruity and earthy flavor) and CDC (apple, fruity, and cedar flavors; sweetness; color). At Buis, cedar aroma and both earthy aroma and flavor were decreased in sandy textures in the case of high vine size, but decreased when vine size was lower (data not shown). Wines from high vine size on clay textures had lowest cedar aroma and earthy flavor. Fruity flavor increased in sandy textures as vine size increased, but decreased in intensity with increased vine size on clay. At CDC, wine color decreased with vine size in sandy textures but was unresponsive to vine size in wines from clay zones. Apple and fruity flavors increased and cedar flavor decreased with vine size in sandy textures but the opposite response was found in wines from grapes grown in clay textures.
Vine size and interactions 2000.
Vine size in 2000 at Buis influenced both aroma and flavor attributes of Chardonnay (Figure 6A). High vine size led to higher intensities of vegetal and earthy aromas, while floral aromas were higher in low size vines. Melon aromas and flavors and acidity were effectively the same for both high and low size vines. Three soil texture × vine size interactions occurred at the Buis site in 2000 for citrus and floral aroma as well as body. Citrus aroma increased for low vine size on clay soils, while floral aroma was highest for low vine size on sand (data not shown). In 2000, CDC vine size impacted melon flavor, with high vine size more intense (Figure 6B). Acidity was high in both vine size categories in 2000, as well as intensity of apple, melon, and floral flavors. No soil texture × vine size interactions were apparent at CDC and no vine size effects or interactions occurred at Wismer in 2000 (data not shown).
Vine size and interactions 2001.
At Buis, high vine size produced wines with higher acidity, astringency/bitterness, and body/mouthfeel (Figure 7A). By contrast, the wines produced from low vine size demonstrated higher apple flavors. The same overall aroma/flavor profile was consistent for wines from both vine size categories, with noticeable acidity and very low levels of earthy flavor. The apple, melon, and floral flavors were all noticeably pronounced, although no difference between the two vine sizes was observed. Only one vine size × soil texture interaction was apparent, whereby apple flavor increased in wines from low vine size derived from clay soils (data not shown). Wines from high vine size at CDC in 2001 had more intense citrus aroma (Figure 7B). The low vine size category produced wines that had intense aromas and flavors of earthiness. All other attributes were consistent between vine sizes. One vine size × soil texture interaction indicated that vegetal flavor increased in wines from low vine size vines on sand (data not shown). There were no vine size effects at Wismer (data not shown); however, there were vine size × soil texture interactions for apple, citrus, and vegetal flavors, and intensity/length of finish. Apple flavor and intensity of finish were highest in both descriptors in high vine size/sand combinations, while citrus and vegetal flavors were highest in high vine size/clay (data not shown).
PCA and correlations.
1999 PCA.
PCA of the 1999 wines revealed that 50.7% and 34.3% of the variation in the data set was associated with PC1 and PC2, respectively (Figure 8). Apple, fruity, color, sweetness, body/mouthfeel, and length/intensity of finish were correlated and found to the right of PC2 and were inversely related to vegetal, cedar, earthy, and astringency. Wines from Falk and CDC wines from sandy textures were associated with the right two quadrants, while three Buis wines plus CDC wines from clay textures were located to the left of PC2. These correlations were also noted following standard correlation analysis (data not shown).
2000 PCA.
PCA of the 2000 wines revealed that 42.6% and 33.4% of the variation was associated with PC1 and PC2, respectively (Figure 9). Apple, citrus, vegetal, earthy (aromas and flavors), astringency/bitterness, and body/mouthfeel were closely related and located on the right side of PC2, while floral and melon (aromas and flavors) and intensity/length of finish were inversely correlated and located to the left. Seven of the eight Buis wines and Falk wines from sandy textures were found in the right quadrants, while all CDC and six of eight Wismer wines were associated with the left quadrants. As the PCA suggested, earthy, vegetal, apple, and citrus aroma and flavor, body/mouthfeel, and astringency/bitterness were highly correlated (Supplemental Table 1). Melon and floral flavors, acidity, and intensity/length of finish were also correlated with one another but inversely correlated with the above attributes. Earthy aroma was correlated with vegetal and earthy flavors, astringency/bitterness, and body/mouthfeel. Melon aroma was positively correlated with floral aroma, melon flavor, and acidity. By contrast, vegetal aroma was negatively correlated with melon and floral flavors but positively correlated with vegetal flavor. Intensity/length of finish was positively correlated with astringency/bitterness. Apple flavor was not correlated with any attributes.
2001 PCA.
PCA of the 2001 wines revealed that 43.6% and 27.9% of the variation in the data set were associated with PC1 and PC2, respectively (Figure 10). Citrus, vegetal, earthy, astringency/bitterness, and intensity/length of finish were correlated and located to the right of PC2, while apple, floral, melon were inversely correlated and located to the left. Right of PC2 were 14 wines, nine of which were grown on clay textures, including Buis (five of eight wines), CDC (three of four wines, including both clay wines), Falk (all wines), Lambert (both clay wines). Left of PC2 were all wines from Wismer and those Lambert wines grown on sandy textures. Several correlations were observed between aroma and flavor attributes (Supplemental Table 2). As PCA suggested, floral aroma was positively correlated with apple, citrus, and floral flavors. Vegetal aroma was also positively correlated with vegetal flavor and earthy aroma and flavor. Melon flavor was negatively correlated with vegetal flavor, acidity, astringency/bitterness, and intensity/length of finish. Citrus aroma was not correlated with any other attribute.
Combined 2000 and 2001 vintages PCA.
The substantial differences between the 2000 and 2001 vintages in terms of GDD and precipitation presented an opportunity to explore vintage effects and the consistency of the sensory data across vintages. PCA of the combined 2000 and 2001 wines revealed that 49.8% and 30.4% of the variation in the data set were associated with PC1 and PC2, respectively (Figure 11). Earthy, vegetal, and citrus (aromas and flavors) and acidity were highly correlated and, as in 2000 and 2001, were located to the right of PC2. These attributes were inversely correlated with floral, melon, and apple (aromas and flavors), which were oriented to the left of PC2. Vegetal and citrus flavors appeared to contribute the majority of the variability to the sensory analysis. The 2000 vintage was mostly sorted into the right quadrants and correlated with citrus, vegetal, and earthy flavors and aromas, and acidity. By contrast, the 2001 vintage mostly fell into the left of PC2 and was aligned with floral, melon, and apple aroma and flavors. While less intense, Falk wines were most consistent with regard to overall sensory profile, remaining in the upper right quadrant in both vintages. Additionally, four 2000 Buis wines were located to the right of PC2. Six of eight CDC wines (including all 2001 wines) and five of eight Wismer wines (all 2001 wines) were located to the left of PC2, as well as all 2001 Buis wines.
There were many apparent correlations between aroma and flavor attributes upon amalgamation of results from both vintages (data not shown). Both apple and vegetal aromas were positively correlated with five other attributes, and vegetal aroma was negatively correlated with floral flavor. Each had positive correlations with citrus and vegetal flavors in common. Apple and vegetal flavors were both positively correlated with citrus flavor and earthy flavor, respectively. Citrus flavor was negatively correlated with two attributes (vegetal flavor, acidity), and positively correlated with two descriptors (body/mouthfeel, intensity/length of finish). Positive correlations for melon versus floral aromas and floral aroma versus floral flavor were consistent between the 2000 and 2001 vintages. Other consistent correlations included vegetal aroma versus citrus and vegetal flavors, and earthy aroma versus vegetal and earthy flavors. There were no negative correlations present in the pooled sensory data that were consistent between the 2000 and 2001 vintages.
Discussion
At the time of its inception, this study sought applications for geomatic technologies to test two major hypotheses: that soil texture would play a minor role in the determination of yield components, fruit composition, and wine sensory attributes (the terroir effect) and that vine size, crop size, and associated fruit environment would play the major roles. This study was possible by the use of vineyard blocks with heterogeneous soil textures and use of geomatic technologies to locate and accurately map vines of various size and yield categories. Ultimately, GPS and GIS allowed the management of large spatially based soil texture and composition, tissue elemental composition, yield, and berry composition data sets that were used to examine direct soil effects on grape composition and wine sensory attributes by testing the independent effects of soil and vine vigor on yield components, berry, must and wine composition, and wine sensory attributes of Chardonnay. This hypothesis was also tested in a related study with Riesling (Reynolds et al. 2007).
Influence of soil texture.
There is much evidence against the role of soil texture on wine quality (Noble 1979, Rankine et al. 1971). Wahl (1988) isolated soil texture and microclimate and kept mesoclimate constant, ultimately finding that climatic factors contributed most to wine quality. However, coarse-textured soils frequently permit deep penetration of root systems, allowing them to access water and nutrients that would otherwise not be available in more limited root systems (Van Leeuwen 2010, Van Leeuwen et al. 2004). Several cultivars across five vintages produced higher sugar and lower malate on sites containing clay textures, primarily due to limited root systems and reduced vine water status during veraison and harvest, which led to smaller berries and greater concentration compared to gravel and sandy textures (Van Leeuwen et al. 2004).
Soil texture seemed to have the most wide-reaching influence on wines from the various sites (three sites in 1999 and 2001; four sites in 2000), although consistency from one vintage to the next was not entirely evident. There appeared to be a tendency for zones within sites with high clay textures to produce wines that were higher in vegetal, earthy, and citrus aromas, whereas sandy zones tended to produce wines with floral and melon aromas and flavors. This was most apparent in the PCA of the 2001 vintage. It is difficult to fully discern the impact of soil texture on wine quality in this study due to an occasional relationship between soil texture and vine size (for example, at CDC, vine size was lowest in high clay zones). Moreover, high clay zones tended to have lower berry weights (CDC only), but also displayed evidence of slightly delayed fruit maturity—lower berry Brix values (Buis, CDC, Wismer), higher TA (Wismer), and lower pH (CDC)—but none of these patterns were consistent across sites and vintages (Reynolds and De Savigny 2001). Moreover, these patterns in berry composition were not necessarily repeated exactly in the must and wine composition (Tables 4 and 5). Although not measured, it is possible that the high clay zones also had some degree of water stress due to a restricted root volume and depth of solum (Van Leeuwen 2010), particularly in vintages with low precipitation (such as 1999such as 2001).
A frequent interrelationship of soil texture with vine size can be observed (Smart 1985): vines on coarse-textured soils are often vigorous and contain shaded canopies. However, opinions are common that soil composition is a determining factor in the sensory profile of a wine (Jackson and Lombard 1993). The texture of the soil will influence root systems (Van Leeuwen 2010), water-holding capacity (Seguin 1986, Van Leeuwen 2010, Van Leeuwen et al. 2004), and mineral composition (Van Leeuwen 2010). Each of these factors functions in determining vine size through its impact on vine vigor. Under water-deficient conditions, soil texture has been shown to influence wine sensory attributes (Reynolds et al. 1995, Saayman 1977, Willwerth et al. 2010), which necessarily requires the consideration of water-holding capacity and root system morphology (Seguin 1986, Van Leeuwen 2010). The macro- and mesoclimate of an area will ultimately determine how much moisture is available to the vine during the growing season. Soil composition was variable among sites, so therefore root systems and corresponding vine sizes were also variable, which may have contributed to inconsistencies observed in sensory and chemical results between vintages and sites and to the difficulty in associating yield-based and berry composition-based variables with sensory attributes.
Influence of vine size.
Investigations of vine size effects on sensory attributes of wines are uncommon. Low vigor zone Pinot noir wines in Oregon were differentiated from high and medium vigor treatments based on differences in astringency, earthy, chemical, heat, sweet, sour, and bitter attributes (Cortell et al. 2008). These were correlated with skin tannin, total fruit tannin, and total wine tannin concentrations, among other variables. Vine size in this study did not substantially nor consistently influence chemical composition of the musts and wines with the exception of high vine size categories increasing must Brix at three sites in 2001. Similarly, its impact on sensory properties of wines was not as great as in the different soil texture categories, with two sites affected in 1999 and two sites impacted in 2000 and 2001. Results were to some degree inconsistent across vintages. Achieving balance between vine size and yield influences fruit and wine quality; overvigorous vines will result in shading of the fruit, resulting in an increased maturation time necessary to achieve optimum fruit composition (Smart 1985). Vine size can also be accommodated through the use of different trellising systems (Smart 1985), as it was in this trial; the vigorous lakeshore sites were Scott Henry-trained whereas the less vigorous site was Guyot-trained. Varying crop loads between vine size categories and sites for the three vintages indicate that factors other than simply vine size were extant throughout the study.
There did not appear to be any clear sensory link exclusively to vine size across the three years. Certain sites, particularly Buis and Falk, were associated with citrus, vegetal, and earthy descriptors, perhaps because they were located in the Lakeshore appellation and had high vine size. In the context of the current study, a sincere attempt was made at the outset to isolate mesoclimate, soil texture, and vine size variables; however, canopy microclimate was not entirely under strict regulation, since growers in the region typically use practices such as hedging and basal leaf removal. This may have played a role in the inconsistent sensory results, as vine microclimate has a significant impact on fruit composition (Smart 1985). Overvigorous vines may lead to vine microclimate conditions that produce pronounced vegetal characteristics in the wine, as in the case of Sauvignon blanc (Conradie et al. 2002) or Bordeaux red wine cultivars (Van Leeuwen et al. 2004). Such was the case in wines from both high vine size and clay-textured soils at Buis in 2000, whereby all categories except low vine size on sand were associated with citrus, vegetal, and earthy aromas. This fact, when taken together with reduced positive sensory attributes (such as reduced melon and floral, which were more evident at other sites in 2000), suggests that high vine size may have played a considerable role in vine imbalance and poor fruit quality.
Influence of vintage.
The three vintages studied varied substantially in terms of GDD—the warmest sites in the region accumulated 1690, 1439, and 1650 GDD, respectively (Ontario Grape Growers Marketing Board, Annual Report, 2002). Environmental conditions of a given vintage will have a profound effect on the resultant quality and characteristics of a wine. Seasonal temperatures, daily temperatures, precipitation and other factors will determine if a vintage is successful in terms of yield and quality. A related study involving Riesling implicated vintage very clearly (versus soil texture and vine size) as having the greatest impact on varietal typicity (Reynolds et al. 2007). Jackson and Lombard (1993) suggested that temperature is the single most important variable that determines both cultivar selection for a region and its wine quality. Vintages within viticultural regions are often described in terms of GDD, demonstrating the importance of temperature; in warm vintages, berry Brix and pH will be higher than in cooler vintages and TA will decrease. This was particularly noticeable at some sites in the 2001 vintage. Elsewhere, impacts of vintage have been well documented. Numerous aroma compounds and phenols differed among three Spanish white wine cultivars due to vintage (De la Presa-Owens 1995). Vintages in Jerez (Spain) were classified using multivariate analyses; warmer and drier vintages produced fruit with higher Brix and lower TA (Palacios et al. 1997). Bordeaux vintages over several decades were analyzed and in most circumstances quality (high Brix, low TA) was associated with high GDD, particularly during veraison to harvest (Jones and Davis 2000). Two vintages were clearly different when Zinfandel wines from several counties in California were profiled sensorially, with the warmer vintage displaying higher black pepper, raisin, and berry attributes (Noble and Shannon 1987). Wine quality of several Italian wine vintages (1970–2002) across five regions (Brunello di Montalcino, Chianti Classico, Nobile di Montepulciano, Barolo-Barbaresco, Amarone) linked quality to temperature and other meteorological factors (Grifoni 2006).
Similarly, precipitation is critical in determining the quality of the vintage, particularly between veraison and the time of harvest, when water accumulation in the berries may deleteriously influence the sugar, flavor, and color of a wine (Cacho et al. 1992, Grifoni et al. 2006, Jones and Davis 2000, Palacios 1997). Differences in anthocyanins in three red wine cultivars were greater between vintages than those between cultivars and locations, and this was largely attributed to precipitation (Cacho et al. 1992). Wine quality associated with vintage in Italy was more strongly related to precipitation than to temperature (Grifoni et al. 2006). Precipitation in this study ranged across the region from 420 to 569 (1999), 547 to 695 mm (2000), and 325 to 518 (2001) (Ontario Grape Growers Marketing Board, Annual Report, 2002). While three of the five sites actually had more precipitation in 2001 than in 2000 during the fruit maturation period (particularly in October), the impact on the wines is unknown because growers used viticultural practices (such as basal leaf removal) to minimize its deleterious potential on berry maturation and resultant wine quality. The chemical composition of wines showed substantial differences between vintages. TA was lower in all wines in 2001 compared to 1999 and 2000, as a result of increased level of berry maturation. Ethanol levels were also highest in wines from the 2001 vintage due to higher temperatures.
The effect of vintage on wine quality was also evident in the sensory results. There was a clear separation between the 2000 and the 2001 vintages, with the 2000 wines characterized by higher intensities of vegetal and earthy aromas and flavors as well as acidity. The lower temperatures in 2000, combined with consistent vine sizes with proportionally lower yields (hence lower crop load or Ravaz index; the ratio of crop size to vine size) may have resulted in an inability of the vines to ripen their respective crops due to imbalance and canopy shading. Vine balance is defined as a crop load between 5 and 10 (Smart 1985). Given environmental differences between the vintages, it is possible that the crop size in 2000 was proportionately too low for the vigor of the vines. At Buis, for instance, vines carried crop loads of 1.8 and 2.0 (kg fruit/kg wt of cane prunings; high versus low vine size categories) in 2000. In 2001 the vines had much more acceptable crop loads of 5.8 and 7.3, respectively. All crop loads were higher in 2001 with the exception of CDC (8.5 high vs. 11.5 low, 2000; 8.4 high vs. 11.1 low, 2001) (data not shown). Crop loads in 1999 varied between 3.3 (Falk) and 5.5 (Wismer). In 2001, temperatures were higher at all sites compared to 2000, and when the 2000 and 2001 vintages were compared, the 2001 wines were characterized by more intense melon and floral aromas and flavors. The intensity/length of finish in the 2001 vintage was also positively correlated with these attributes.
Influence of site.
The impact of vineyard site on wine composition and sensory attributes has also been widely reported. Site differences differentiated Australian white wines based on phenolic analyses (Hooper et al. 1985), Bordeaux red wines based on elemental analyses (Martin et al. 1999), and Loire Valley Cabernet franc wines based on flavonoid composition (Brossaud et al. 1999). The influence of site on the chemical composition and sensory profile was apparent in all vintages in this study. Wines from each site differed from one another, with each characterized by a general trend from one vintage to the next. There are several explanations for this phenomenon, one of which is site mesoclimate. As was evident in climatic data and physical location, the sites varied in several significant areas. With the exception of Falk and Buis, each site was subject to different GDD and precipitation levels over the course of the maturation period. This may result largely from each site’s position relative to Lake Ontario. Large bodies of water (such as Lake Ontario) will influence the prevailing weather patterns of specific areas (Bonnardot et al. 2002) such as the Niagara Peninsula. Sites closer to the lakeshore (Falk, Buis) are more prone to cooler growing season temperatures. By contrast, CDC, situated farther inland, received higher GDD over the course of each growing season, providing a significant source of variability in the results. This explanation is supported by Bonnardot et al. (2002), who found that vineyard location played an important role in determining the overall sensory characteristics of Sauvignon blanc wines in the Stellenbosch–Klein Drakenstein region of South Africa, and confirmed by Scrinzi et al. (1997) in Trentino, Italy. Significant interactions of season and canopy density on fruit composition and wine sensory attributes were observed for several white wine cultivars in British Columbia (Reynolds et al. 1995). Considering variability attributable to season, crop load, and other factors, it was difficult to discern consistent sensory and chemical patterns within the five sites with respect to soil texture and vine size.
The term terroir refers to the impact of environment or site upon the uniqueness of wines made from a particular site (Van Leeuwen 2010). The terroir effect is not a reference to the chemical composition but rather to the unique sensory profile of a wine from a specific site. Observance of the sensory profiles from these sites, whether from vines of two sizes or vines grown on two soil textures, revealed similar sensory patterns. The greatest evidence in support of the terroir concept is seen in the PCA, which showed clear groupings of the respective sites across all three vintages. Despite variable weather conditions, Falk and Buis wines (Niagara Lakeshore) were positioned toward high intensities of citrus, vegetal, and earthy aromas and flavors in all vintages. By contrast, Wismer wines (Escarpment Bench) were characterized by melon and floral aromas and flavors in all vintages, despite variable weather conditions between 1999 and 2001. These results suggest factors beyond the scope of this study determined the overall character of the wines. The CDC site (Lakeshore Plain) was the only one of the five that demonstrated both soil texture and vine size effects in all vintages. Upon examination of the sensory data, all profiles were similarly characterized by high intensities of all attributes with the exception of earthy aroma and flavor and pronounced acidity. This supports the regional terroir model (Douglas et al. 2001, Schlosser et al. 2005, Kontkanen et al. 2005) that suggested the establishment of three subappellations: Lakeshore, Lakeshore Plain, and Escarpment Bench. It is also consistent with Noble et al. (1984), who found sensory differences in wines within four Bordeaux communes that were greater than those between communes.
Conclusions
Sensory descriptive analysis of 1999–2001 wines produced from individual soil × vine size or soil × yield categories displayed few consistent differences among sites, soil types, or vine size. Overall, the relative roles of soil texture and vine size remain to some degree unresolved. Vintage and site (and their concomitant influence on climate) appeared to be the dominant factors in determining wine quality. Soil texture can be a determining factor in vine size. However, vine size can be manipulated on all soil textures using innovative techniques to counter inherent soil properties that impact vine vigor. This, when considering the variability of meso- and microclimates from one site to the next, suggests an interdependence of all conceivable factors in determining wine quality. The synthesis of factors will ultimately determine the metaphysical and often identifiable terroir of a wine.
There are many opportunities for continued research into the relative importance and apparent interactions between vine size and soil texture. Tying sensory analysis with soil analysis may shed some light on the roles of various minerals and nutrients on sensory profiles of wines in the Niagara Peninsula and elsewhere. With such knowledge, growers could take new approaches to maximize fruit quality. Also, an ongoing relationship between viticultural practices during the growing season and chemical and sensory analysis on the resultant wines may provide more insight into the effects of viticultural practices on the wines. This may help to define a more consistent relationship among soil texture, vine size, and terroir.
Acknowledgments
Acknowledgments: Funding from the Natural Sciences and Engineering Council of Canada is gratefully acknowledged. The authors thank the grapegrowers who provided fruit for winemaking over the course of this trial and the sensory panelists who volunteered many hours of their time.
Footnotes
-
Supplemental data is freely available with the online version of this article.
-
Publication costs of this article defrayed in part by page fees.
- Received January 2012.
- Revision received October 2012.
- Revision received December 2012.
- Accepted January 2013.
- Published online December 1969
- ©2013 by the American Society for Enology and Viticulture